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/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86 #include <linux/random.h>
  87 #include <linux/uaccess.h>
  88 #include <asm/unaligned.h>
  89
  90 #include <linux/proc_fs.h>
  91 #include <linux/seq_file.h>
  92 #include <linux/export.h>
  93
  94 /* Set to 3 to get tracing... */
  95 #define ACONF_DEBUG 2
  96
  97 #if ACONF_DEBUG >= 3
  98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
  99 #else
 100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 101 #endif
 102
 103 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 104
 105 static inline u32 cstamp_delta(unsigned long cstamp)
 106 {
 107         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 108 }
 109
 110 #ifdef CONFIG_SYSCTL
 111 static int addrconf_sysctl_register(struct inet6_dev *idev);
 112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 113 #else
 114 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 115 {
 116         return 0;
 117 }
 118
 119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 120 {
 121 }
 122 #endif
 123
 124 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 125 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 126 static void ipv6_regen_rndid(unsigned long data);
 127
 128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 129 static int ipv6_count_addresses(struct inet6_dev *idev);
 130
 131 /*
 132  *      Configured unicast address hash table
 133  */
 134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 135 static DEFINE_SPINLOCK(addrconf_hash_lock);
 136
 137 static void addrconf_verify(void);
 138 static void addrconf_verify_rtnl(void);
 139 static void addrconf_verify_work(struct work_struct *);
 140
 141 static struct workqueue_struct *addrconf_wq;
 142 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 143
 144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 146
 147 static void addrconf_type_change(struct net_device *dev,
 148                                  unsigned long event);
 149 static int addrconf_ifdown(struct net_device *dev, int how);
 150
 151 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 152                                                   int plen,
 153                                                   const struct net_device *dev,
 154                                                   u32 flags, u32 noflags);
 155
 156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 157 static void addrconf_dad_work(struct work_struct *w);
 158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 159 static void addrconf_dad_run(struct inet6_dev *idev);
 160 static void addrconf_rs_timer(unsigned long data);
 161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 163
 164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 165                                 struct prefix_info *pinfo);
 166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 167                                struct net_device *dev);
 168
 169 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 170         .forwarding             = 0,
 171         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 172         .mtu6                   = IPV6_MIN_MTU,
 173         .accept_ra              = 1,
 174         .accept_redirects       = 1,
 175         .autoconf               = 1,
 176         .force_mld_version      = 0,
 177         .mldv1_unsolicited_report_interval = 10 * HZ,
 178         .mldv2_unsolicited_report_interval = HZ,
 179         .dad_transmits          = 1,
 180         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 181         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 182         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 183         .use_tempaddr           = 0,
 184         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 185         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 186         .regen_max_retry        = REGEN_MAX_RETRY,
 187         .max_desync_factor      = MAX_DESYNC_FACTOR,
 188         .max_addresses          = IPV6_MAX_ADDRESSES,
 189         .accept_ra_defrtr       = 1,
 190         .accept_ra_from_local   = 0,
 191         .accept_ra_pinfo        = 1,
 192 #ifdef CONFIG_IPV6_ROUTER_PREF
 193         .accept_ra_rtr_pref     = 1,
 194         .rtr_probe_interval     = 60 * HZ,
 195 #ifdef CONFIG_IPV6_ROUTE_INFO
 196         .accept_ra_rt_info_max_plen = 0,
 197 #endif
 198 #endif
 199         .proxy_ndp              = 0,
 200         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 201         .disable_ipv6           = 0,
 202         .accept_dad             = 1,
 203         .suppress_frag_ndisc    = 1,
 204 };
 205
 206 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 207         .forwarding             = 0,
 208         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 209         .mtu6                   = IPV6_MIN_MTU,
 210         .accept_ra              = 1,
 211         .accept_redirects       = 1,
 212         .autoconf               = 1,
 213         .force_mld_version      = 0,
 214         .mldv1_unsolicited_report_interval = 10 * HZ,
 215         .mldv2_unsolicited_report_interval = HZ,
 216         .dad_transmits          = 1,
 217         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 218         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 219         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 220         .use_tempaddr           = 0,
 221         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 222         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 223         .regen_max_retry        = REGEN_MAX_RETRY,
 224         .max_desync_factor      = MAX_DESYNC_FACTOR,
 225         .max_addresses          = IPV6_MAX_ADDRESSES,
 226         .accept_ra_defrtr       = 1,
 227         .accept_ra_from_local   = 0,
 228         .accept_ra_pinfo        = 1,
 229 #ifdef CONFIG_IPV6_ROUTER_PREF
 230         .accept_ra_rtr_pref     = 1,
 231         .rtr_probe_interval     = 60 * HZ,
 232 #ifdef CONFIG_IPV6_ROUTE_INFO
 233         .accept_ra_rt_info_max_plen = 0,
 234 #endif
 235 #endif
 236         .proxy_ndp              = 0,
 237         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 238         .disable_ipv6           = 0,
 239         .accept_dad             = 1,
 240         .suppress_frag_ndisc    = 1,
 241 };
 242
 243 /* Check if a valid qdisc is available */
 244 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 245 {
 246         return !qdisc_tx_is_noop(dev);
 247 }
 248
 249 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 250 {
 251         if (del_timer(&idev->rs_timer))
 252                 __in6_dev_put(idev);
 253 }
 254
 255 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 256 {
 257         if (cancel_delayed_work(&ifp->dad_work))
 258                 __in6_ifa_put(ifp);
 259 }
 260
 261 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 262                                   unsigned long when)
 263 {
 264         if (!timer_pending(&idev->rs_timer))
 265                 in6_dev_hold(idev);
 266         mod_timer(&idev->rs_timer, jiffies + when);
 267 }
 268
 269 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 270                                    unsigned long delay)
 271 {
 272         if (!delayed_work_pending(&ifp->dad_work))
 273                 in6_ifa_hold(ifp);
 274         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 275 }
 276
 277 static int snmp6_alloc_dev(struct inet6_dev *idev)
 278 {
 279         int i;
 280
 281         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 282         if (!idev->stats.ipv6)
 283                 goto err_ip;
 284
 285         for_each_possible_cpu(i) {
 286                 struct ipstats_mib *addrconf_stats;
 287                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 288                 u64_stats_init(&addrconf_stats->syncp);
 289         }
 290
 291
 292         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 293                                         GFP_KERNEL);
 294         if (!idev->stats.icmpv6dev)
 295                 goto err_icmp;
 296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 297                                            GFP_KERNEL);
 298         if (!idev->stats.icmpv6msgdev)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         kfree(idev->stats.icmpv6dev);
 305 err_icmp:
 306         free_percpu(idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 312 {
 313         struct inet6_dev *ndev;
 314         int err = -ENOMEM;
 315
 316         ASSERT_RTNL();
 317
 318         if (dev->mtu < IPV6_MIN_MTU)
 319                 return ERR_PTR(-EINVAL);
 320
 321         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 322         if (ndev == NULL)
 323                 return ERR_PTR(err);
 324
 325         rwlock_init(&ndev->lock);
 326         ndev->dev = dev;
 327         INIT_LIST_HEAD(&ndev->addr_list);
 328         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 329                     (unsigned long)ndev);
 330         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 331         ndev->cnf.mtu6 = dev->mtu;
 332         ndev->cnf.sysctl = NULL;
 333         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 334         if (ndev->nd_parms == NULL) {
 335                 kfree(ndev);
 336                 return ERR_PTR(err);
 337         }
 338         if (ndev->cnf.forwarding)
 339                 dev_disable_lro(dev);
 340         /* We refer to the device */
 341         dev_hold(dev);
 342
 343         if (snmp6_alloc_dev(ndev) < 0) {
 344                 ADBG(KERN_WARNING
 345                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 346                         __func__, dev->name);
 347                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 348                 dev_put(dev);
 349                 kfree(ndev);
 350                 return ERR_PTR(err);
 351         }
 352
 353         if (snmp6_register_dev(ndev) < 0) {
 354                 ADBG(KERN_WARNING
 355                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 356                         __func__, dev->name);
 357                 goto err_release;
 358         }
 359
 360         /* One reference from device.  We must do this before
 361          * we invoke __ipv6_regen_rndid().
 362          */
 363         in6_dev_hold(ndev);
 364
 365         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 366                 ndev->cnf.accept_dad = -1;
 367
 368 #if IS_ENABLED(CONFIG_IPV6_SIT)
 369         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 370                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 371                 ndev->cnf.rtr_solicits = 0;
 372         }
 373 #endif
 374
 375         INIT_LIST_HEAD(&ndev->tempaddr_list);
 376         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 377         if ((dev->flags&IFF_LOOPBACK) ||
 378             dev->type == ARPHRD_TUNNEL ||
 379             dev->type == ARPHRD_TUNNEL6 ||
 380             dev->type == ARPHRD_SIT ||
 381             dev->type == ARPHRD_NONE) {
 382                 ndev->cnf.use_tempaddr = -1;
 383         } else {
 384                 in6_dev_hold(ndev);
 385                 ipv6_regen_rndid((unsigned long) ndev);
 386         }
 387
 388         ndev->token = in6addr_any;
 389
 390         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 391                 ndev->if_flags |= IF_READY;
 392
 393         ipv6_mc_init_dev(ndev);
 394         ndev->tstamp = jiffies;
 395         err = addrconf_sysctl_register(ndev);
 396         if (err) {
 397                 ipv6_mc_destroy_dev(ndev);
 398                 del_timer(&ndev->regen_timer);
 399                 goto err_release;
 400         }
 401         /* protected by rtnl_lock */
 402         rcu_assign_pointer(dev->ip6_ptr, ndev);
 403
 404         /* Join interface-local all-node multicast group */
 405         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 406
 407         /* Join all-node multicast group */
 408         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 409
 410         /* Join all-router multicast group if forwarding is set */
 411         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 412                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 413
 414         return ndev;
 415
 416 err_release:
 417         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 418         ndev->dead = 1;
 419         in6_dev_finish_destroy(ndev);
 420         return ERR_PTR(err);
 421 }
 422
 423 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 424 {
 425         struct inet6_dev *idev;
 426
 427         ASSERT_RTNL();
 428
 429         idev = __in6_dev_get(dev);
 430         if (!idev) {
 431                 idev = ipv6_add_dev(dev);
 432                 if (IS_ERR(idev))
 433                         return NULL;
 434         }
 435
 436         if (dev->flags&IFF_UP)
 437                 ipv6_mc_up(idev);
 438         return idev;
 439 }
 440
 441 static int inet6_netconf_msgsize_devconf(int type)
 442 {
 443         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 444                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 445
 446         /* type -1 is used for ALL */
 447         if (type == -1 || type == NETCONFA_FORWARDING)
 448                 size += nla_total_size(4);
 449 #ifdef CONFIG_IPV6_MROUTE
 450         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 451                 size += nla_total_size(4);
 452 #endif
 453         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 454                 size += nla_total_size(4);
 455
 456         return size;
 457 }
 458
 459 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 460                                       struct ipv6_devconf *devconf, u32 portid,
 461                                       u32 seq, int event, unsigned int flags,
 462                                       int type)
 463 {
 464         struct nlmsghdr  *nlh;
 465         struct netconfmsg *ncm;
 466
 467         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 468                         flags);
 469         if (nlh == NULL)
 470                 return -EMSGSIZE;
 471
 472         ncm = nlmsg_data(nlh);
 473         ncm->ncm_family = AF_INET6;
 474
 475         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 476                 goto nla_put_failure;
 477
 478         /* type -1 is used for ALL */
 479         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 480             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 481                 goto nla_put_failure;
 482 #ifdef CONFIG_IPV6_MROUTE
 483         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 484             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 485                         devconf->mc_forwarding) < 0)
 486                 goto nla_put_failure;
 487 #endif
 488         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 489             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 490                 goto nla_put_failure;
 491
 492         return nlmsg_end(skb, nlh);
 493
 494 nla_put_failure:
 495         nlmsg_cancel(skb, nlh);
 496         return -EMSGSIZE;
 497 }
 498
 499 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 500                                   struct ipv6_devconf *devconf)
 501 {
 502         struct sk_buff *skb;
 503         int err = -ENOBUFS;
 504
 505         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 506         if (skb == NULL)
 507                 goto errout;
 508
 509         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 510                                          RTM_NEWNETCONF, 0, type);
 511         if (err < 0) {
 512                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 513                 WARN_ON(err == -EMSGSIZE);
 514                 kfree_skb(skb);
 515                 goto errout;
 516         }
 517         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 518         return;
 519 errout:
 520         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 521 }
 522
 523 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 524         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 525         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 526         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 527 };
 528
 529 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 530                                      struct nlmsghdr *nlh)
 531 {
 532         struct net *net = sock_net(in_skb->sk);
 533         struct nlattr *tb[NETCONFA_MAX+1];
 534         struct netconfmsg *ncm;
 535         struct sk_buff *skb;
 536         struct ipv6_devconf *devconf;
 537         struct inet6_dev *in6_dev;
 538         struct net_device *dev;
 539         int ifindex;
 540         int err;
 541
 542         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 543                           devconf_ipv6_policy);
 544         if (err < 0)
 545                 goto errout;
 546
 547         err = EINVAL;
 548         if (!tb[NETCONFA_IFINDEX])
 549                 goto errout;
 550
 551         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 552         switch (ifindex) {
 553         case NETCONFA_IFINDEX_ALL:
 554                 devconf = net->ipv6.devconf_all;
 555                 break;
 556         case NETCONFA_IFINDEX_DEFAULT:
 557                 devconf = net->ipv6.devconf_dflt;
 558                 break;
 559         default:
 560                 dev = __dev_get_by_index(net, ifindex);
 561                 if (dev == NULL)
 562                         goto errout;
 563                 in6_dev = __in6_dev_get(dev);
 564                 if (in6_dev == NULL)
 565                         goto errout;
 566                 devconf = &in6_dev->cnf;
 567                 break;
 568         }
 569
 570         err = -ENOBUFS;
 571         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 572         if (skb == NULL)
 573                 goto errout;
 574
 575         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 576                                          NETLINK_CB(in_skb).portid,
 577                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 578                                          -1);
 579         if (err < 0) {
 580                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 581                 WARN_ON(err == -EMSGSIZE);
 582                 kfree_skb(skb);
 583                 goto errout;
 584         }
 585         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 586 errout:
 587         return err;
 588 }
 589
 590 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 591                                       struct netlink_callback *cb)
 592 {
 593         struct net *net = sock_net(skb->sk);
 594         int h, s_h;
 595         int idx, s_idx;
 596         struct net_device *dev;
 597         struct inet6_dev *idev;
 598         struct hlist_head *head;
 599
 600         s_h = cb->args[0];
 601         s_idx = idx = cb->args[1];
 602
 603         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 604                 idx = 0;
 605                 head = &net->dev_index_head[h];
 606                 rcu_read_lock();
 607                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 608                           net->dev_base_seq;
 609                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 610                         if (idx < s_idx)
 611                                 goto cont;
 612                         idev = __in6_dev_get(dev);
 613                         if (!idev)
 614                                 goto cont;
 615
 616                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 617                                                        &idev->cnf,
 618                                                        NETLINK_CB(cb->skb).portid,
 619                                                        cb->nlh->nlmsg_seq,
 620                                                        RTM_NEWNETCONF,
 621                                                        NLM_F_MULTI,
 622                                                        -1) <= 0) {
 623                                 rcu_read_unlock();
 624                                 goto done;
 625                         }
 626                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 627 cont:
 628                         idx++;
 629                 }
 630                 rcu_read_unlock();
 631         }
 632         if (h == NETDEV_HASHENTRIES) {
 633                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 634                                                net->ipv6.devconf_all,
 635                                                NETLINK_CB(cb->skb).portid,
 636                                                cb->nlh->nlmsg_seq,
 637                                                RTM_NEWNETCONF, NLM_F_MULTI,
 638                                                -1) <= 0)
 639                         goto done;
 640                 else
 641                         h++;
 642         }
 643         if (h == NETDEV_HASHENTRIES + 1) {
 644                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 645                                                net->ipv6.devconf_dflt,
 646                                                NETLINK_CB(cb->skb).portid,
 647                                                cb->nlh->nlmsg_seq,
 648                                                RTM_NEWNETCONF, NLM_F_MULTI,
 649                                                -1) <= 0)
 650                         goto done;
 651                 else
 652                         h++;
 653         }
 654 done:
 655         cb->args[0] = h;
 656         cb->args[1] = idx;
 657
 658         return skb->len;
 659 }
 660
 661 #ifdef CONFIG_SYSCTL
 662 static void dev_forward_change(struct inet6_dev *idev)
 663 {
 664         struct net_device *dev;
 665         struct inet6_ifaddr *ifa;
 666
 667         if (!idev)
 668                 return;
 669         dev = idev->dev;
 670         if (idev->cnf.forwarding)
 671                 dev_disable_lro(dev);
 672         if (dev->flags & IFF_MULTICAST) {
 673                 if (idev->cnf.forwarding) {
 674                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 675                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 676                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 677                 } else {
 678                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 679                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 680                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 681                 }
 682         }
 683
 684         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 685                 if (ifa->flags&IFA_F_TENTATIVE)
 686                         continue;
 687                 if (idev->cnf.forwarding)
 688                         addrconf_join_anycast(ifa);
 689                 else
 690                         addrconf_leave_anycast(ifa);
 691         }
 692         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 693                                      dev->ifindex, &idev->cnf);
 694 }
 695
 696
 697 static void addrconf_forward_change(struct net *net, __s32 newf)
 698 {
 699         struct net_device *dev;
 700         struct inet6_dev *idev;
 701
 702         for_each_netdev(net, dev) {
 703                 idev = __in6_dev_get(dev);
 704                 if (idev) {
 705                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 706                         idev->cnf.forwarding = newf;
 707                         if (changed)
 708                                 dev_forward_change(idev);
 709                 }
 710         }
 711 }
 712
 713 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 714 {
 715         struct net *net;
 716         int old;
 717
 718         if (!rtnl_trylock())
 719                 return restart_syscall();
 720
 721         net = (struct net *)table->extra2;
 722         old = *p;
 723         *p = newf;
 724
 725         if (p == &net->ipv6.devconf_dflt->forwarding) {
 726                 if ((!newf) ^ (!old))
 727                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 728                                                      NETCONFA_IFINDEX_DEFAULT,
 729                                                      net->ipv6.devconf_dflt);
 730                 rtnl_unlock();
 731                 return 0;
 732         }
 733
 734         if (p == &net->ipv6.devconf_all->forwarding) {
 735                 net->ipv6.devconf_dflt->forwarding = newf;
 736                 addrconf_forward_change(net, newf);
 737                 if ((!newf) ^ (!old))
 738                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 739                                                      NETCONFA_IFINDEX_ALL,
 740                                                      net->ipv6.devconf_all);
 741         } else if ((!newf) ^ (!old))
 742                 dev_forward_change((struct inet6_dev *)table->extra1);
 743         rtnl_unlock();
 744
 745         if (newf)
 746                 rt6_purge_dflt_routers(net);
 747         return 1;
 748 }
 749 #endif
 750
 751 /* Nobody refers to this ifaddr, destroy it */
 752 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 753 {
 754         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 755
 756 #ifdef NET_REFCNT_DEBUG
 757         pr_debug("%s\n", __func__);
 758 #endif
 759
 760         in6_dev_put(ifp->idev);
 761
 762         if (cancel_delayed_work(&ifp->dad_work))
 763                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 764                           ifp);
 765
 766         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 767                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 768                 return;
 769         }
 770         ip6_rt_put(ifp->rt);
 771
 772         kfree_rcu(ifp, rcu);
 773 }
 774
 775 static void
 776 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 777 {
 778         struct list_head *p;
 779         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 780
 781         /*
 782          * Each device address list is sorted in order of scope -
 783          * global before linklocal.
 784          */
 785         list_for_each(p, &idev->addr_list) {
 786                 struct inet6_ifaddr *ifa
 787                         = list_entry(p, struct inet6_ifaddr, if_list);
 788                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 789                         break;
 790         }
 791
 792         list_add_tail(&ifp->if_list, p);
 793 }
 794
 795 static u32 inet6_addr_hash(const struct in6_addr *addr)
 796 {
 797         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 798 }
 799
 800 /* On success it returns ifp with increased reference count */
 801
 802 static struct inet6_ifaddr *
 803 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 804               const struct in6_addr *peer_addr, int pfxlen,
 805               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 806 {
 807         struct inet6_ifaddr *ifa = NULL;
 808         struct rt6_info *rt;
 809         unsigned int hash;
 810         int err = 0;
 811         int addr_type = ipv6_addr_type(addr);
 812
 813         if (addr_type == IPV6_ADDR_ANY ||
 814             addr_type & IPV6_ADDR_MULTICAST ||
 815             (!(idev->dev->flags & IFF_LOOPBACK) &&
 816              addr_type & IPV6_ADDR_LOOPBACK))
 817                 return ERR_PTR(-EADDRNOTAVAIL);
 818
 819         rcu_read_lock_bh();
 820         if (idev->dead) {
 821                 err = -ENODEV;                  /*XXX*/
 822                 goto out2;
 823         }
 824
 825         if (idev->cnf.disable_ipv6) {
 826                 err = -EACCES;
 827                 goto out2;
 828         }
 829
 830         spin_lock(&addrconf_hash_lock);
 831
 832         /* Ignore adding duplicate addresses on an interface */
 833         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 834                 ADBG("ipv6_add_addr: already assigned\n");
 835                 err = -EEXIST;
 836                 goto out;
 837         }
 838
 839         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 840
 841         if (ifa == NULL) {
 842                 ADBG("ipv6_add_addr: malloc failed\n");
 843                 err = -ENOBUFS;
 844                 goto out;
 845         }
 846
 847         rt = addrconf_dst_alloc(idev, addr, false);
 848         if (IS_ERR(rt)) {
 849                 err = PTR_ERR(rt);
 850                 goto out;
 851         }
 852
 853         neigh_parms_data_state_setall(idev->nd_parms);
 854
 855         ifa->addr = *addr;
 856         if (peer_addr)
 857                 ifa->peer_addr = *peer_addr;
 858
 859         spin_lock_init(&ifa->lock);
 860         spin_lock_init(&ifa->state_lock);
 861         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 862         INIT_HLIST_NODE(&ifa->addr_lst);
 863         ifa->scope = scope;
 864         ifa->prefix_len = pfxlen;
 865         ifa->flags = flags | IFA_F_TENTATIVE;
 866         ifa->valid_lft = valid_lft;
 867         ifa->prefered_lft = prefered_lft;
 868         ifa->cstamp = ifa->tstamp = jiffies;
 869         ifa->tokenized = false;
 870
 871         ifa->rt = rt;
 872
 873         ifa->idev = idev;
 874         in6_dev_hold(idev);
 875         /* For caller */
 876         in6_ifa_hold(ifa);
 877
 878         /* Add to big hash table */
 879         hash = inet6_addr_hash(addr);
 880
 881         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 882         spin_unlock(&addrconf_hash_lock);
 883
 884         write_lock(&idev->lock);
 885         /* Add to inet6_dev unicast addr list. */
 886         ipv6_link_dev_addr(idev, ifa);
 887
 888         if (ifa->flags&IFA_F_TEMPORARY) {
 889                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 890                 in6_ifa_hold(ifa);
 891         }
 892
 893         in6_ifa_hold(ifa);
 894         write_unlock(&idev->lock);
 895 out2:
 896         rcu_read_unlock_bh();
 897
 898         if (likely(err == 0))
 899                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 900         else {
 901                 kfree(ifa);
 902                 ifa = ERR_PTR(err);
 903         }
 904
 905         return ifa;
 906 out:
 907         spin_unlock(&addrconf_hash_lock);
 908         goto out2;
 909 }
 910
 911 enum cleanup_prefix_rt_t {
 912         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
 913         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
 914         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
 915 };
 916
 917 /*
 918  * Check, whether the prefix for ifp would still need a prefix route
 919  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
 920  * constants.
 921  *
 922  * 1) we don't purge prefix if address was not permanent.
 923  *    prefix is managed by its own lifetime.
 924  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
 925  * 3) if there are no addresses, delete prefix.
 926  * 4) if there are still other permanent address(es),
 927  *    corresponding prefix is still permanent.
 928  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
 929  *    don't purge the prefix, assume user space is managing it.
 930  * 6) otherwise, update prefix lifetime to the
 931  *    longest valid lifetime among the corresponding
 932  *    addresses on the device.
 933  *    Note: subsequent RA will update lifetime.
 934  **/
 935 static enum cleanup_prefix_rt_t
 936 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
 937 {
 938         struct inet6_ifaddr *ifa;
 939         struct inet6_dev *idev = ifp->idev;
 940         unsigned long lifetime;
 941         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
 942
 943         *expires = jiffies;
 944
 945         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 946                 if (ifa == ifp)
 947                         continue;
 948                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 949                                        ifp->prefix_len))
 950                         continue;
 951                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
 952                         return CLEANUP_PREFIX_RT_NOP;
 953
 954                 action = CLEANUP_PREFIX_RT_EXPIRE;
 955
 956                 spin_lock(&ifa->lock);
 957
 958                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 959                 /*
 960                  * Note: Because this address is
 961                  * not permanent, lifetime <
 962                  * LONG_MAX / HZ here.
 963                  */
 964                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
 965                         *expires = ifa->tstamp + lifetime * HZ;
 966                 spin_unlock(&ifa->lock);
 967         }
 968
 969         return action;
 970 }
 971
 972 static void
 973 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
 974 {
 975         struct rt6_info *rt;
 976
 977         rt = addrconf_get_prefix_route(&ifp->addr,
 978                                        ifp->prefix_len,
 979                                        ifp->idev->dev,
 980                                        0, RTF_GATEWAY | RTF_DEFAULT);
 981         if (rt) {
 982                 if (del_rt)
 983                         ip6_del_rt(rt);
 984                 else {
 985                         if (!(rt->rt6i_flags & RTF_EXPIRES))
 986                                 rt6_set_expires(rt, expires);
 987                         ip6_rt_put(rt);
 988                 }
 989         }
 990 }
 991
 992
 993 /* This function wants to get referenced ifp and releases it before return */
 994
 995 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 996 {
 997         int state;
 998         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
 999         unsigned long expires;
1000
1001         ASSERT_RTNL();
1002
1003         spin_lock_bh(&ifp->state_lock);
1004         state = ifp->state;
1005         ifp->state = INET6_IFADDR_STATE_DEAD;
1006         spin_unlock_bh(&ifp->state_lock);
1007
1008         if (state == INET6_IFADDR_STATE_DEAD)
1009                 goto out;
1010
1011         spin_lock_bh(&addrconf_hash_lock);
1012         hlist_del_init_rcu(&ifp->addr_lst);
1013         spin_unlock_bh(&addrconf_hash_lock);
1014
1015         write_lock_bh(&ifp->idev->lock);
1016
1017         if (ifp->flags&IFA_F_TEMPORARY) {
1018                 list_del(&ifp->tmp_list);
1019                 if (ifp->ifpub) {
1020                         in6_ifa_put(ifp->ifpub);
1021                         ifp->ifpub = NULL;
1022                 }
1023                 __in6_ifa_put(ifp);
1024         }
1025
1026         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1027                 action = check_cleanup_prefix_route(ifp, &expires);
1028
1029         list_del_init(&ifp->if_list);
1030         __in6_ifa_put(ifp);
1031
1032         write_unlock_bh(&ifp->idev->lock);
1033
1034         addrconf_del_dad_work(ifp);
1035
1036         ipv6_ifa_notify(RTM_DELADDR, ifp);
1037
1038         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1039
1040         if (action != CLEANUP_PREFIX_RT_NOP) {
1041                 cleanup_prefix_route(ifp, expires,
1042                         action == CLEANUP_PREFIX_RT_DEL);
1043         }
1044
1045         /* clean up prefsrc entries */
1046         rt6_remove_prefsrc(ifp);
1047 out:
1048         in6_ifa_put(ifp);
1049 }
1050
1051 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1052 {
1053         struct inet6_dev *idev = ifp->idev;
1054         struct in6_addr addr, *tmpaddr;
1055         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1056         unsigned long regen_advance;
1057         int tmp_plen;
1058         int ret = 0;
1059         u32 addr_flags;
1060         unsigned long now = jiffies;
1061
1062         write_lock_bh(&idev->lock);
1063         if (ift) {
1064                 spin_lock_bh(&ift->lock);
1065                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1066                 spin_unlock_bh(&ift->lock);
1067                 tmpaddr = &addr;
1068         } else {
1069                 tmpaddr = NULL;
1070         }
1071 retry:
1072         in6_dev_hold(idev);
1073         if (idev->cnf.use_tempaddr <= 0) {
1074                 write_unlock_bh(&idev->lock);
1075                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1076                 in6_dev_put(idev);
1077                 ret = -1;
1078                 goto out;
1079         }
1080         spin_lock_bh(&ifp->lock);
1081         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1082                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1083                 spin_unlock_bh(&ifp->lock);
1084                 write_unlock_bh(&idev->lock);
1085                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1086                         __func__);
1087                 in6_dev_put(idev);
1088                 ret = -1;
1089                 goto out;
1090         }
1091         in6_ifa_hold(ifp);
1092         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1093         __ipv6_try_regen_rndid(idev, tmpaddr);
1094         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1095         age = (now - ifp->tstamp) / HZ;
1096         tmp_valid_lft = min_t(__u32,
1097                               ifp->valid_lft,
1098                               idev->cnf.temp_valid_lft + age);
1099         tmp_prefered_lft = min_t(__u32,
1100                                  ifp->prefered_lft,
1101                                  idev->cnf.temp_prefered_lft + age -
1102                                  idev->cnf.max_desync_factor);
1103         tmp_plen = ifp->prefix_len;
1104         tmp_tstamp = ifp->tstamp;
1105         spin_unlock_bh(&ifp->lock);
1106
1107         regen_advance = idev->cnf.regen_max_retry *
1108                         idev->cnf.dad_transmits *
1109                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1110         write_unlock_bh(&idev->lock);
1111
1112         /* A temporary address is created only if this calculated Preferred
1113          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1114          * an implementation must not create a temporary address with a zero
1115          * Preferred Lifetime.
1116          * Use age calculation as in addrconf_verify to avoid unnecessary
1117          * temporary addresses being generated.
1118          */
1119         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1120         if (tmp_prefered_lft <= regen_advance + age) {
1121                 in6_ifa_put(ifp);
1122                 in6_dev_put(idev);
1123                 ret = -1;
1124                 goto out;
1125         }
1126
1127         addr_flags = IFA_F_TEMPORARY;
1128         /* set in addrconf_prefix_rcv() */
1129         if (ifp->flags & IFA_F_OPTIMISTIC)
1130                 addr_flags |= IFA_F_OPTIMISTIC;
1131
1132         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1133                             ipv6_addr_scope(&addr), addr_flags,
1134                             tmp_valid_lft, tmp_prefered_lft);
1135         if (IS_ERR(ift)) {
1136                 in6_ifa_put(ifp);
1137                 in6_dev_put(idev);
1138                 pr_info("%s: retry temporary address regeneration\n", __func__);
1139                 tmpaddr = &addr;
1140                 write_lock_bh(&idev->lock);
1141                 goto retry;
1142         }
1143
1144         spin_lock_bh(&ift->lock);
1145         ift->ifpub = ifp;
1146         ift->cstamp = now;
1147         ift->tstamp = tmp_tstamp;
1148         spin_unlock_bh(&ift->lock);
1149
1150         addrconf_dad_start(ift);
1151         in6_ifa_put(ift);
1152         in6_dev_put(idev);
1153 out:
1154         return ret;
1155 }
1156
1157 /*
1158  *      Choose an appropriate source address (RFC3484)
1159  */
1160 enum {
1161         IPV6_SADDR_RULE_INIT = 0,
1162         IPV6_SADDR_RULE_LOCAL,
1163         IPV6_SADDR_RULE_SCOPE,
1164         IPV6_SADDR_RULE_PREFERRED,
1165 #ifdef CONFIG_IPV6_MIP6
1166         IPV6_SADDR_RULE_HOA,
1167 #endif
1168         IPV6_SADDR_RULE_OIF,
1169         IPV6_SADDR_RULE_LABEL,
1170         IPV6_SADDR_RULE_PRIVACY,
1171         IPV6_SADDR_RULE_ORCHID,
1172         IPV6_SADDR_RULE_PREFIX,
1173         IPV6_SADDR_RULE_MAX
1174 };
1175
1176 struct ipv6_saddr_score {
1177         int                     rule;
1178         int                     addr_type;
1179         struct inet6_ifaddr     *ifa;
1180         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1181         int                     scopedist;
1182         int                     matchlen;
1183 };
1184
1185 struct ipv6_saddr_dst {
1186         const struct in6_addr *addr;
1187         int ifindex;
1188         int scope;
1189         int label;
1190         unsigned int prefs;
1191 };
1192
1193 static inline int ipv6_saddr_preferred(int type)
1194 {
1195         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1196                 return 1;
1197         return 0;
1198 }
1199
1200 static int ipv6_get_saddr_eval(struct net *net,
1201                                struct ipv6_saddr_score *score,
1202                                struct ipv6_saddr_dst *dst,
1203                                int i)
1204 {
1205         int ret;
1206
1207         if (i <= score->rule) {
1208                 switch (i) {
1209                 case IPV6_SADDR_RULE_SCOPE:
1210                         ret = score->scopedist;
1211                         break;
1212                 case IPV6_SADDR_RULE_PREFIX:
1213                         ret = score->matchlen;
1214                         break;
1215                 default:
1216                         ret = !!test_bit(i, score->scorebits);
1217                 }
1218                 goto out;
1219         }
1220
1221         switch (i) {
1222         case IPV6_SADDR_RULE_INIT:
1223                 /* Rule 0: remember if hiscore is not ready yet */
1224                 ret = !!score->ifa;
1225                 break;
1226         case IPV6_SADDR_RULE_LOCAL:
1227                 /* Rule 1: Prefer same address */
1228                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1229                 break;
1230         case IPV6_SADDR_RULE_SCOPE:
1231                 /* Rule 2: Prefer appropriate scope
1232                  *
1233                  *      ret
1234                  *       ^
1235                  *    -1 |  d 15
1236                  *    ---+--+-+---> scope
1237                  *       |
1238                  *       |             d is scope of the destination.
1239                  *  B-d  |  \
1240                  *       |   \      <- smaller scope is better if
1241                  *  B-15 |    \        if scope is enough for destination.
1242                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1243                  * d-C-1 | /
1244                  *       |/         <- greater is better
1245                  *   -C  /             if scope is not enough for destination.
1246                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1247                  *
1248                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1249                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1250                  * Assume B = 0 and we get C > 29.
1251                  */
1252                 ret = __ipv6_addr_src_scope(score->addr_type);
1253                 if (ret >= dst->scope)
1254                         ret = -ret;
1255                 else
1256                         ret -= 128;     /* 30 is enough */
1257                 score->scopedist = ret;
1258                 break;
1259         case IPV6_SADDR_RULE_PREFERRED:
1260                 /* Rule 3: Avoid deprecated and optimistic addresses */
1261                 ret = ipv6_saddr_preferred(score->addr_type) ||
1262                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1263                 break;
1264 #ifdef CONFIG_IPV6_MIP6
1265         case IPV6_SADDR_RULE_HOA:
1266             {
1267                 /* Rule 4: Prefer home address */
1268                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1269                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1270                 break;
1271             }
1272 #endif
1273         case IPV6_SADDR_RULE_OIF:
1274                 /* Rule 5: Prefer outgoing interface */
1275                 ret = (!dst->ifindex ||
1276                        dst->ifindex == score->ifa->idev->dev->ifindex);
1277                 break;
1278         case IPV6_SADDR_RULE_LABEL:
1279                 /* Rule 6: Prefer matching label */
1280                 ret = ipv6_addr_label(net,
1281                                       &score->ifa->addr, score->addr_type,
1282                                       score->ifa->idev->dev->ifindex) == dst->label;
1283                 break;
1284         case IPV6_SADDR_RULE_PRIVACY:
1285             {
1286                 /* Rule 7: Prefer public address
1287                  * Note: prefer temporary address if use_tempaddr >= 2
1288                  */
1289                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1290                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1291                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1292                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1293                 break;
1294             }
1295         case IPV6_SADDR_RULE_ORCHID:
1296                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1297                  *          non-ORCHID vs non-ORCHID
1298                  */
1299                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1300                         ipv6_addr_orchid(dst->addr));
1301                 break;
1302         case IPV6_SADDR_RULE_PREFIX:
1303                 /* Rule 8: Use longest matching prefix */
1304                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1305                 if (ret > score->ifa->prefix_len)
1306                         ret = score->ifa->prefix_len;
1307                 score->matchlen = ret;
1308                 break;
1309         default:
1310                 ret = 0;
1311         }
1312
1313         if (ret)
1314                 __set_bit(i, score->scorebits);
1315         score->rule = i;
1316 out:
1317         return ret;
1318 }
1319
1320 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1321                        const struct in6_addr *daddr, unsigned int prefs,
1322                        struct in6_addr *saddr)
1323 {
1324         struct ipv6_saddr_score scores[2],
1325                                 *score = &scores[0], *hiscore = &scores[1];
1326         struct ipv6_saddr_dst dst;
1327         struct net_device *dev;
1328         int dst_type;
1329
1330         dst_type = __ipv6_addr_type(daddr);
1331         dst.addr = daddr;
1332         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1333         dst.scope = __ipv6_addr_src_scope(dst_type);
1334         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1335         dst.prefs = prefs;
1336
1337         hiscore->rule = -1;
1338         hiscore->ifa = NULL;
1339
1340         rcu_read_lock();
1341
1342         for_each_netdev_rcu(net, dev) {
1343                 struct inet6_dev *idev;
1344
1345                 /* Candidate Source Address (section 4)
1346                  *  - multicast and link-local destination address,
1347                  *    the set of candidate source address MUST only
1348                  *    include addresses assigned to interfaces
1349                  *    belonging to the same link as the outgoing
1350                  *    interface.
1351                  * (- For site-local destination addresses, the
1352                  *    set of candidate source addresses MUST only
1353                  *    include addresses assigned to interfaces
1354                  *    belonging to the same site as the outgoing
1355                  *    interface.)
1356                  */
1357                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1358                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1359                     dst.ifindex && dev->ifindex != dst.ifindex)
1360                         continue;
1361
1362                 idev = __in6_dev_get(dev);
1363                 if (!idev)
1364                         continue;
1365
1366                 read_lock_bh(&idev->lock);
1367                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1368                         int i;
1369
1370                         /*
1371                          * - Tentative Address (RFC2462 section 5.4)
1372                          *  - A tentative address is not considered
1373                          *    "assigned to an interface" in the traditional
1374                          *    sense, unless it is also flagged as optimistic.
1375                          * - Candidate Source Address (section 4)
1376                          *  - In any case, anycast addresses, multicast
1377                          *    addresses, and the unspecified address MUST
1378                          *    NOT be included in a candidate set.
1379                          */
1380                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1381                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1382                                 continue;
1383
1384                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1385
1386                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1387                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1388                                 LIMIT_NETDEBUG(KERN_DEBUG
1389                                                "ADDRCONF: unspecified / multicast address "
1390                                                "assigned as unicast address on %s",
1391                                                dev->name);
1392                                 continue;
1393                         }
1394
1395                         score->rule = -1;
1396                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1397
1398                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1399                                 int minihiscore, miniscore;
1400
1401                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1402                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1403
1404                                 if (minihiscore > miniscore) {
1405                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1406                                             score->scopedist > 0) {
1407                                                 /*
1408                                                  * special case:
1409                                                  * each remaining entry
1410                                                  * has too small (not enough)
1411                                                  * scope, because ifa entries
1412                                                  * are sorted by their scope
1413                                                  * values.
1414                                                  */
1415                                                 goto try_nextdev;
1416                                         }
1417                                         break;
1418                                 } else if (minihiscore < miniscore) {
1419                                         if (hiscore->ifa)
1420                                                 in6_ifa_put(hiscore->ifa);
1421
1422                                         in6_ifa_hold(score->ifa);
1423
1424                                         swap(hiscore, score);
1425
1426                                         /* restore our iterator */
1427                                         score->ifa = hiscore->ifa;
1428
1429                                         break;
1430                                 }
1431                         }
1432                 }
1433 try_nextdev:
1434                 read_unlock_bh(&idev->lock);
1435         }
1436         rcu_read_unlock();
1437
1438         if (!hiscore->ifa)
1439                 return -EADDRNOTAVAIL;
1440
1441         *saddr = hiscore->ifa->addr;
1442         in6_ifa_put(hiscore->ifa);
1443         return 0;
1444 }
1445 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1446
1447 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1448                       u32 banned_flags)
1449 {
1450         struct inet6_ifaddr *ifp;
1451         int err = -EADDRNOTAVAIL;
1452
1453         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1454                 if (ifp->scope > IFA_LINK)
1455                         break;
1456                 if (ifp->scope == IFA_LINK &&
1457                     !(ifp->flags & banned_flags)) {
1458                         *addr = ifp->addr;
1459                         err = 0;
1460                         break;
1461                 }
1462         }
1463         return err;
1464 }
1465
1466 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1467                     u32 banned_flags)
1468 {
1469         struct inet6_dev *idev;
1470         int err = -EADDRNOTAVAIL;
1471
1472         rcu_read_lock();
1473         idev = __in6_dev_get(dev);
1474         if (idev) {
1475                 read_lock_bh(&idev->lock);
1476                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1477                 read_unlock_bh(&idev->lock);
1478         }
1479         rcu_read_unlock();
1480         return err;
1481 }
1482
1483 static int ipv6_count_addresses(struct inet6_dev *idev)
1484 {
1485         int cnt = 0;
1486         struct inet6_ifaddr *ifp;
1487
1488         read_lock_bh(&idev->lock);
1489         list_for_each_entry(ifp, &idev->addr_list, if_list)
1490                 cnt++;
1491         read_unlock_bh(&idev->lock);
1492         return cnt;
1493 }
1494
1495 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1496                   const struct net_device *dev, int strict)
1497 {
1498         struct inet6_ifaddr *ifp;
1499         unsigned int hash = inet6_addr_hash(addr);
1500
1501         rcu_read_lock_bh();
1502         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1503                 if (!net_eq(dev_net(ifp->idev->dev), net))
1504                         continue;
1505                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1506                     !(ifp->flags&IFA_F_TENTATIVE) &&
1507                     (dev == NULL || ifp->idev->dev == dev ||
1508                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1509                         rcu_read_unlock_bh();
1510                         return 1;
1511                 }
1512         }
1513
1514         rcu_read_unlock_bh();
1515         return 0;
1516 }
1517 EXPORT_SYMBOL(ipv6_chk_addr);
1518
1519 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1520                                struct net_device *dev)
1521 {
1522         unsigned int hash = inet6_addr_hash(addr);
1523         struct inet6_ifaddr *ifp;
1524
1525         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1526                 if (!net_eq(dev_net(ifp->idev->dev), net))
1527                         continue;
1528                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1529                         if (dev == NULL || ifp->idev->dev == dev)
1530                                 return true;
1531                 }
1532         }
1533         return false;
1534 }
1535
1536 /* Compares an address/prefix_len with addresses on device @dev.
1537  * If one is found it returns true.
1538  */
1539 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1540         const unsigned int prefix_len, struct net_device *dev)
1541 {
1542         struct inet6_dev *idev;
1543         struct inet6_ifaddr *ifa;
1544         bool ret = false;
1545
1546         rcu_read_lock();
1547         idev = __in6_dev_get(dev);
1548         if (idev) {
1549                 read_lock_bh(&idev->lock);
1550                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1551                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1552                         if (ret)
1553                                 break;
1554                 }
1555                 read_unlock_bh(&idev->lock);
1556         }
1557         rcu_read_unlock();
1558
1559         return ret;
1560 }
1561 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1562
1563 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1564 {
1565         struct inet6_dev *idev;
1566         struct inet6_ifaddr *ifa;
1567         int     onlink;
1568
1569         onlink = 0;
1570         rcu_read_lock();
1571         idev = __in6_dev_get(dev);
1572         if (idev) {
1573                 read_lock_bh(&idev->lock);
1574                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1575                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1576                                                    ifa->prefix_len);
1577                         if (onlink)
1578                                 break;
1579                 }
1580                 read_unlock_bh(&idev->lock);
1581         }
1582         rcu_read_unlock();
1583         return onlink;
1584 }
1585 EXPORT_SYMBOL(ipv6_chk_prefix);
1586
1587 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1588                                      struct net_device *dev, int strict)
1589 {
1590         struct inet6_ifaddr *ifp, *result = NULL;
1591         unsigned int hash = inet6_addr_hash(addr);
1592
1593         rcu_read_lock_bh();
1594         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1595                 if (!net_eq(dev_net(ifp->idev->dev), net))
1596                         continue;
1597                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1598                         if (dev == NULL || ifp->idev->dev == dev ||
1599                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1600                                 result = ifp;
1601                                 in6_ifa_hold(ifp);
1602                                 break;
1603                         }
1604                 }
1605         }
1606         rcu_read_unlock_bh();
1607
1608         return result;
1609 }
1610
1611 /* Gets referenced address, destroys ifaddr */
1612
1613 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1614 {
1615         if (ifp->flags&IFA_F_PERMANENT) {
1616                 spin_lock_bh(&ifp->lock);
1617                 addrconf_del_dad_work(ifp);
1618                 ifp->flags |= IFA_F_TENTATIVE;
1619                 if (dad_failed)
1620                         ifp->flags |= IFA_F_DADFAILED;
1621                 spin_unlock_bh(&ifp->lock);
1622                 if (dad_failed)
1623                         ipv6_ifa_notify(0, ifp);
1624                 in6_ifa_put(ifp);
1625         } else if (ifp->flags&IFA_F_TEMPORARY) {
1626                 struct inet6_ifaddr *ifpub;
1627                 spin_lock_bh(&ifp->lock);
1628                 ifpub = ifp->ifpub;
1629                 if (ifpub) {
1630                         in6_ifa_hold(ifpub);
1631                         spin_unlock_bh(&ifp->lock);
1632                         ipv6_create_tempaddr(ifpub, ifp);
1633                         in6_ifa_put(ifpub);
1634                 } else {
1635                         spin_unlock_bh(&ifp->lock);
1636                 }
1637                 ipv6_del_addr(ifp);
1638         } else {
1639                 ipv6_del_addr(ifp);
1640         }
1641 }
1642
1643 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1644 {
1645         int err = -ENOENT;
1646
1647         spin_lock_bh(&ifp->state_lock);
1648         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1649                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1650                 err = 0;
1651         }
1652         spin_unlock_bh(&ifp->state_lock);
1653
1654         return err;
1655 }
1656
1657 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1658 {
1659         struct inet6_dev *idev = ifp->idev;
1660
1661         if (addrconf_dad_end(ifp)) {
1662                 in6_ifa_put(ifp);
1663                 return;
1664         }
1665
1666         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1667                              ifp->idev->dev->name, &ifp->addr);
1668
1669         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1670                 struct in6_addr addr;
1671
1672                 addr.s6_addr32[0] = htonl(0xfe800000);
1673                 addr.s6_addr32[1] = 0;
1674
1675                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1676                     ipv6_addr_equal(&ifp->addr, &addr)) {
1677                         /* DAD failed for link-local based on MAC address */
1678                         idev->cnf.disable_ipv6 = 1;
1679
1680                         pr_info("%s: IPv6 being disabled!\n",
1681                                 ifp->idev->dev->name);
1682                 }
1683         }
1684
1685         spin_lock_bh(&ifp->state_lock);
1686         /* transition from _POSTDAD to _ERRDAD */
1687         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1688         spin_unlock_bh(&ifp->state_lock);
1689
1690         addrconf_mod_dad_work(ifp, 0);
1691 }
1692
1693 /* Join to solicited addr multicast group. */
1694
1695 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1696 {
1697         struct in6_addr maddr;
1698
1699         ASSERT_RTNL();
1700
1701         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1702                 return;
1703
1704         addrconf_addr_solict_mult(addr, &maddr);
1705         ipv6_dev_mc_inc(dev, &maddr);
1706 }
1707
1708 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1709 {
1710         struct in6_addr maddr;
1711
1712         ASSERT_RTNL();
1713
1714         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1715                 return;
1716
1717         addrconf_addr_solict_mult(addr, &maddr);
1718         __ipv6_dev_mc_dec(idev, &maddr);
1719 }
1720
1721 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1722 {
1723         struct in6_addr addr;
1724
1725         ASSERT_RTNL();
1726
1727         if (ifp->prefix_len >= 127) /* RFC 6164 */
1728                 return;
1729         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1730         if (ipv6_addr_any(&addr))
1731                 return;
1732         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1733 }
1734
1735 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1736 {
1737         struct in6_addr addr;
1738
1739         ASSERT_RTNL();
1740
1741         if (ifp->prefix_len >= 127) /* RFC 6164 */
1742                 return;
1743         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1744         if (ipv6_addr_any(&addr))
1745                 return;
1746         __ipv6_dev_ac_dec(ifp->idev, &addr);
1747 }
1748
1749 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1750 {
1751         if (dev->addr_len != ETH_ALEN)
1752                 return -1;
1753         memcpy(eui, dev->dev_addr, 3);
1754         memcpy(eui + 5, dev->dev_addr + 3, 3);
1755
1756         /*
1757          * The zSeries OSA network cards can be shared among various
1758          * OS instances, but the OSA cards have only one MAC address.
1759          * This leads to duplicate address conflicts in conjunction
1760          * with IPv6 if more than one instance uses the same card.
1761          *
1762          * The driver for these cards can deliver a unique 16-bit
1763          * identifier for each instance sharing the same card.  It is
1764          * placed instead of 0xFFFE in the interface identifier.  The
1765          * "u" bit of the interface identifier is not inverted in this
1766          * case.  Hence the resulting interface identifier has local
1767          * scope according to RFC2373.
1768          */
1769         if (dev->dev_id) {
1770                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1771                 eui[4] = dev->dev_id & 0xFF;
1772         } else {
1773                 eui[3] = 0xFF;
1774                 eui[4] = 0xFE;
1775                 eui[0] ^= 2;
1776         }
1777         return 0;
1778 }
1779
1780 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1781 {
1782         if (dev->addr_len != IEEE802154_ADDR_LEN)
1783                 return -1;
1784         memcpy(eui, dev->dev_addr, 8);
1785         eui[0] ^= 2;
1786         return 0;
1787 }
1788
1789 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1790 {
1791         union fwnet_hwaddr *ha;
1792
1793         if (dev->addr_len != FWNET_ALEN)
1794                 return -1;
1795
1796         ha = (union fwnet_hwaddr *)dev->dev_addr;
1797
1798         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1799         eui[0] ^= 2;
1800         return 0;
1801 }
1802
1803 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1804 {
1805         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1806         if (dev->addr_len != ARCNET_ALEN)
1807                 return -1;
1808         memset(eui, 0, 7);
1809         eui[7] = *(u8 *)dev->dev_addr;
1810         return 0;
1811 }
1812
1813 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1814 {
1815         if (dev->addr_len != INFINIBAND_ALEN)
1816                 return -1;
1817         memcpy(eui, dev->dev_addr + 12, 8);
1818         eui[0] |= 2;
1819         return 0;
1820 }
1821
1822 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1823 {
1824         if (addr == 0)
1825                 return -1;
1826         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1827                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1828                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1829                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1830                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1831                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1832         eui[1] = 0;
1833         eui[2] = 0x5E;
1834         eui[3] = 0xFE;
1835         memcpy(eui + 4, &addr, 4);
1836         return 0;
1837 }
1838
1839 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1840 {
1841         if (dev->priv_flags & IFF_ISATAP)
1842                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1843         return -1;
1844 }
1845
1846 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1847 {
1848         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1849 }
1850
1851 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1852 {
1853         memcpy(eui, dev->perm_addr, 3);
1854         memcpy(eui + 5, dev->perm_addr + 3, 3);
1855         eui[3] = 0xFF;
1856         eui[4] = 0xFE;
1857         eui[0] ^= 2;
1858         return 0;
1859 }
1860
1861 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1862 {
1863         switch (dev->type) {
1864         case ARPHRD_ETHER:
1865         case ARPHRD_FDDI:
1866                 return addrconf_ifid_eui48(eui, dev);
1867         case ARPHRD_ARCNET:
1868                 return addrconf_ifid_arcnet(eui, dev);
1869         case ARPHRD_INFINIBAND:
1870                 return addrconf_ifid_infiniband(eui, dev);
1871         case ARPHRD_SIT:
1872                 return addrconf_ifid_sit(eui, dev);
1873         case ARPHRD_IPGRE:
1874                 return addrconf_ifid_gre(eui, dev);
1875         case ARPHRD_6LOWPAN:
1876         case ARPHRD_IEEE802154:
1877                 return addrconf_ifid_eui64(eui, dev);
1878         case ARPHRD_IEEE1394:
1879                 return addrconf_ifid_ieee1394(eui, dev);
1880         case ARPHRD_TUNNEL6:
1881                 return addrconf_ifid_ip6tnl(eui, dev);
1882         }
1883         return -1;
1884 }
1885
1886 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1887 {
1888         int err = -1;
1889         struct inet6_ifaddr *ifp;
1890
1891         read_lock_bh(&idev->lock);
1892         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1893                 if (ifp->scope > IFA_LINK)
1894                         break;
1895                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1896                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1897                         err = 0;
1898                         break;
1899                 }
1900         }
1901         read_unlock_bh(&idev->lock);
1902         return err;
1903 }
1904
1905 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1906 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1907 {
1908 regen:
1909         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1910         idev->rndid[0] &= ~0x02;
1911
1912         /*
1913          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1914          * check if generated address is not inappropriate
1915          *
1916          *  - Reserved subnet anycast (RFC 2526)
1917          *      11111101 11....11 1xxxxxxx
1918          *  - ISATAP (RFC4214) 6.1
1919          *      00-00-5E-FE-xx-xx-xx-xx
1920          *  - value 0
1921          *  - XXX: already assigned to an address on the device
1922          */
1923         if (idev->rndid[0] == 0xfd &&
1924             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1925             (idev->rndid[7]&0x80))
1926                 goto regen;
1927         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1928                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1929                         goto regen;
1930                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1931                         goto regen;
1932         }
1933 }
1934
1935 static void ipv6_regen_rndid(unsigned long data)
1936 {
1937         struct inet6_dev *idev = (struct inet6_dev *) data;
1938         unsigned long expires;
1939
1940         rcu_read_lock_bh();
1941         write_lock_bh(&idev->lock);
1942
1943         if (idev->dead)
1944                 goto out;
1945
1946         __ipv6_regen_rndid(idev);
1947
1948         expires = jiffies +
1949                 idev->cnf.temp_prefered_lft * HZ -
1950                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1951                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1952                 idev->cnf.max_desync_factor * HZ;
1953         if (time_before(expires, jiffies)) {
1954                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1955                         __func__, idev->dev->name);
1956                 goto out;
1957         }
1958
1959         if (!mod_timer(&idev->regen_timer, expires))
1960                 in6_dev_hold(idev);
1961
1962 out:
1963         write_unlock_bh(&idev->lock);
1964         rcu_read_unlock_bh();
1965         in6_dev_put(idev);
1966 }
1967
1968 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1969 {
1970         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1971                 __ipv6_regen_rndid(idev);
1972 }
1973
1974 /*
1975  *      Add prefix route.
1976  */
1977
1978 static void
1979 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1980                       unsigned long expires, u32 flags)
1981 {
1982         struct fib6_config cfg = {
1983                 .fc_table = RT6_TABLE_PREFIX,
1984                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1985                 .fc_ifindex = dev->ifindex,
1986                 .fc_expires = expires,
1987                 .fc_dst_len = plen,
1988                 .fc_flags = RTF_UP | flags,
1989                 .fc_nlinfo.nl_net = dev_net(dev),
1990                 .fc_protocol = RTPROT_KERNEL,
1991         };
1992
1993         cfg.fc_dst = *pfx;
1994
1995         /* Prevent useless cloning on PtP SIT.
1996            This thing is done here expecting that the whole
1997            class of non-broadcast devices need not cloning.
1998          */
1999 #if IS_ENABLED(CONFIG_IPV6_SIT)
2000         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2001                 cfg.fc_flags |= RTF_NONEXTHOP;
2002 #endif
2003
2004         ip6_route_add(&cfg);
2005 }
2006
2007
2008 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2009                                                   int plen,
2010                                                   const struct net_device *dev,
2011                                                   u32 flags, u32 noflags)
2012 {
2013         struct fib6_node *fn;
2014         struct rt6_info *rt = NULL;
2015         struct fib6_table *table;
2016
2017         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2018         if (table == NULL)
2019                 return NULL;
2020
2021         read_lock_bh(&table->tb6_lock);
2022         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2023         if (!fn)
2024                 goto out;
2025         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2026                 if (rt->dst.dev->ifindex != dev->ifindex)
2027                         continue;
2028                 if ((rt->rt6i_flags & flags) != flags)
2029                         continue;
2030                 if ((rt->rt6i_flags & noflags) != 0)
2031                         continue;
2032                 dst_hold(&rt->dst);
2033                 break;
2034         }
2035 out:
2036         read_unlock_bh(&table->tb6_lock);
2037         return rt;
2038 }
2039
2040
2041 /* Create "default" multicast route to the interface */
2042
2043 static void addrconf_add_mroute(struct net_device *dev)
2044 {
2045         struct fib6_config cfg = {
2046                 .fc_table = RT6_TABLE_LOCAL,
2047                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2048                 .fc_ifindex = dev->ifindex,
2049                 .fc_dst_len = 8,
2050                 .fc_flags = RTF_UP,
2051                 .fc_nlinfo.nl_net = dev_net(dev),
2052         };
2053
2054         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2055
2056         ip6_route_add(&cfg);
2057 }
2058
2059 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2060 {
2061         struct inet6_dev *idev;
2062
2063         ASSERT_RTNL();
2064
2065         idev = ipv6_find_idev(dev);
2066         if (!idev)
2067                 return ERR_PTR(-ENOBUFS);
2068
2069         if (idev->cnf.disable_ipv6)
2070                 return ERR_PTR(-EACCES);
2071
2072         /* Add default multicast route */
2073         if (!(dev->flags & IFF_LOOPBACK))
2074                 addrconf_add_mroute(dev);
2075
2076         return idev;
2077 }
2078
2079 static void manage_tempaddrs(struct inet6_dev *idev,
2080                              struct inet6_ifaddr *ifp,
2081                              __u32 valid_lft, __u32 prefered_lft,
2082                              bool create, unsigned long now)
2083 {
2084         u32 flags;
2085         struct inet6_ifaddr *ift;
2086
2087         read_lock_bh(&idev->lock);
2088         /* update all temporary addresses in the list */
2089         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2090                 int age, max_valid, max_prefered;
2091
2092                 if (ifp != ift->ifpub)
2093                         continue;
2094
2095                 /* RFC 4941 section 3.3:
2096                  * If a received option will extend the lifetime of a public
2097                  * address, the lifetimes of temporary addresses should
2098                  * be extended, subject to the overall constraint that no
2099                  * temporary addresses should ever remain "valid" or "preferred"
2100                  * for a time longer than (TEMP_VALID_LIFETIME) or
2101                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2102                  */
2103                 age = (now - ift->cstamp) / HZ;
2104                 max_valid = idev->cnf.temp_valid_lft - age;
2105                 if (max_valid < 0)
2106                         max_valid = 0;
2107
2108                 max_prefered = idev->cnf.temp_prefered_lft -
2109                                idev->cnf.max_desync_factor - age;
2110                 if (max_prefered < 0)
2111                         max_prefered = 0;
2112
2113                 if (valid_lft > max_valid)
2114                         valid_lft = max_valid;
2115
2116                 if (prefered_lft > max_prefered)
2117                         prefered_lft = max_prefered;
2118
2119                 spin_lock(&ift->lock);
2120                 flags = ift->flags;
2121                 ift->valid_lft = valid_lft;
2122                 ift->prefered_lft = prefered_lft;
2123                 ift->tstamp = now;
2124                 if (prefered_lft > 0)
2125                         ift->flags &= ~IFA_F_DEPRECATED;
2126
2127                 spin_unlock(&ift->lock);
2128                 if (!(flags&IFA_F_TENTATIVE))
2129                         ipv6_ifa_notify(0, ift);
2130         }
2131
2132         if ((create || list_empty(&idev->tempaddr_list)) &&
2133             idev->cnf.use_tempaddr > 0) {
2134                 /* When a new public address is created as described
2135                  * in [ADDRCONF], also create a new temporary address.
2136                  * Also create a temporary address if it's enabled but
2137                  * no temporary address currently exists.
2138                  */
2139                 read_unlock_bh(&idev->lock);
2140                 ipv6_create_tempaddr(ifp, NULL);
2141         } else {
2142                 read_unlock_bh(&idev->lock);
2143         }
2144 }
2145
2146 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2147 {
2148         struct prefix_info *pinfo;
2149         __u32 valid_lft;
2150         __u32 prefered_lft;
2151         int addr_type;
2152         struct inet6_dev *in6_dev;
2153         struct net *net = dev_net(dev);
2154
2155         pinfo = (struct prefix_info *) opt;
2156
2157         if (len < sizeof(struct prefix_info)) {
2158                 ADBG("addrconf: prefix option too short\n");
2159                 return;
2160         }
2161
2162         /*
2163          *      Validation checks ([ADDRCONF], page 19)
2164          */
2165
2166         addr_type = ipv6_addr_type(&pinfo->prefix);
2167
2168         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2169                 return;
2170
2171         valid_lft = ntohl(pinfo->valid);
2172         prefered_lft = ntohl(pinfo->prefered);
2173
2174         if (prefered_lft > valid_lft) {
2175                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2176                 return;
2177         }
2178
2179         in6_dev = in6_dev_get(dev);
2180
2181         if (in6_dev == NULL) {
2182                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2183                                     dev->name);
2184                 return;
2185         }
2186
2187         /*
2188          *      Two things going on here:
2189          *      1) Add routes for on-link prefixes
2190          *      2) Configure prefixes with the auto flag set
2191          */
2192
2193         if (pinfo->onlink) {
2194                 struct rt6_info *rt;
2195                 unsigned long rt_expires;
2196
2197                 /* Avoid arithmetic overflow. Really, we could
2198                  * save rt_expires in seconds, likely valid_lft,
2199                  * but it would require division in fib gc, that it
2200                  * not good.
2201                  */
2202                 if (HZ > USER_HZ)
2203                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2204                 else
2205                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2206
2207                 if (addrconf_finite_timeout(rt_expires))
2208                         rt_expires *= HZ;
2209
2210                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2211                                                pinfo->prefix_len,
2212                                                dev,
2213                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2214                                                RTF_GATEWAY | RTF_DEFAULT);
2215
2216                 if (rt) {
2217                         /* Autoconf prefix route */
2218                         if (valid_lft == 0) {
2219                                 ip6_del_rt(rt);
2220                                 rt = NULL;
2221                         } else if (addrconf_finite_timeout(rt_expires)) {
2222                                 /* not infinity */
2223                                 rt6_set_expires(rt, jiffies + rt_expires);
2224                         } else {
2225                                 rt6_clean_expires(rt);
2226                         }
2227                 } else if (valid_lft) {
2228                         clock_t expires = 0;
2229                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2230                         if (addrconf_finite_timeout(rt_expires)) {
2231                                 /* not infinity */
2232                                 flags |= RTF_EXPIRES;
2233                                 expires = jiffies_to_clock_t(rt_expires);
2234                         }
2235                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2236                                               dev, expires, flags);
2237                 }
2238                 ip6_rt_put(rt);
2239         }
2240
2241         /* Try to figure out our local address for this prefix */
2242
2243         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2244                 struct inet6_ifaddr *ifp;
2245                 struct in6_addr addr;
2246                 int create = 0, update_lft = 0;
2247                 bool tokenized = false;
2248
2249                 if (pinfo->prefix_len == 64) {
2250                         memcpy(&addr, &pinfo->prefix, 8);
2251
2252                         if (!ipv6_addr_any(&in6_dev->token)) {
2253                                 read_lock_bh(&in6_dev->lock);
2254                                 memcpy(addr.s6_addr + 8,
2255                                        in6_dev->token.s6_addr + 8, 8);
2256                                 read_unlock_bh(&in6_dev->lock);
2257                                 tokenized = true;
2258                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2259                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2260                                 in6_dev_put(in6_dev);
2261                                 return;
2262                         }
2263                         goto ok;
2264                 }
2265                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2266                                     pinfo->prefix_len);
2267                 in6_dev_put(in6_dev);
2268                 return;
2269
2270 ok:
2271
2272                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2273
2274                 if (ifp == NULL && valid_lft) {
2275                         int max_addresses = in6_dev->cnf.max_addresses;
2276                         u32 addr_flags = 0;
2277
2278 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2279                         if (in6_dev->cnf.optimistic_dad &&
2280                             !net->ipv6.devconf_all->forwarding && sllao)
2281                                 addr_flags = IFA_F_OPTIMISTIC;
2282 #endif
2283
2284                         /* Do not allow to create too much of autoconfigured
2285                          * addresses; this would be too easy way to crash kernel.
2286                          */
2287                         if (!max_addresses ||
2288                             ipv6_count_addresses(in6_dev) < max_addresses)
2289                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2290                                                     pinfo->prefix_len,
2291                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2292                                                     addr_flags, valid_lft,
2293                                                     prefered_lft);
2294
2295                         if (IS_ERR_OR_NULL(ifp)) {
2296                                 in6_dev_put(in6_dev);
2297                                 return;
2298                         }
2299
2300                         update_lft = 0;
2301                         create = 1;
2302                         spin_lock_bh(&ifp->lock);
2303                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2304                         ifp->cstamp = jiffies;
2305                         ifp->tokenized = tokenized;
2306                         spin_unlock_bh(&ifp->lock);
2307                         addrconf_dad_start(ifp);
2308                 }
2309
2310                 if (ifp) {
2311                         u32 flags;
2312                         unsigned long now;
2313                         u32 stored_lft;
2314
2315                         /* update lifetime (RFC2462 5.5.3 e) */
2316                         spin_lock(&ifp->lock);
2317                         now = jiffies;
2318                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2319                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2320                         else
2321                                 stored_lft = 0;
2322                         if (!update_lft && !create && stored_lft) {
2323                                 const u32 minimum_lft = min(
2324                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2325                                 valid_lft = max(valid_lft, minimum_lft);
2326
2327                                 /* RFC4862 Section 5.5.3e:
2328                                  * "Note that the preferred lifetime of the
2329                                  *  corresponding address is always reset to
2330                                  *  the Preferred Lifetime in the received
2331                                  *  Prefix Information option, regardless of
2332                                  *  whether the valid lifetime is also reset or
2333                                  *  ignored."
2334                                  *
2335                                  * So we should always update prefered_lft here.
2336                                  */
2337                                 update_lft = 1;
2338                         }
2339
2340                         if (update_lft) {
2341                                 ifp->valid_lft = valid_lft;
2342                                 ifp->prefered_lft = prefered_lft;
2343                                 ifp->tstamp = now;
2344                                 flags = ifp->flags;
2345                                 ifp->flags &= ~IFA_F_DEPRECATED;
2346                                 spin_unlock(&ifp->lock);
2347
2348                                 if (!(flags&IFA_F_TENTATIVE))
2349                                         ipv6_ifa_notify(0, ifp);
2350                         } else
2351                                 spin_unlock(&ifp->lock);
2352
2353                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2354                                          create, now);
2355
2356                         in6_ifa_put(ifp);
2357                         addrconf_verify();
2358                 }
2359         }
2360         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2361         in6_dev_put(in6_dev);
2362 }
2363
2364 /*
2365  *      Set destination address.
2366  *      Special case for SIT interfaces where we create a new "virtual"
2367  *      device.
2368  */
2369 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2370 {
2371         struct in6_ifreq ireq;
2372         struct net_device *dev;
2373         int err = -EINVAL;
2374
2375         rtnl_lock();
2376
2377         err = -EFAULT;
2378         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2379                 goto err_exit;
2380
2381         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2382
2383         err = -ENODEV;
2384         if (dev == NULL)
2385                 goto err_exit;
2386
2387 #if IS_ENABLED(CONFIG_IPV6_SIT)
2388         if (dev->type == ARPHRD_SIT) {
2389                 const struct net_device_ops *ops = dev->netdev_ops;
2390                 struct ifreq ifr;
2391                 struct ip_tunnel_parm p;
2392
2393                 err = -EADDRNOTAVAIL;
2394                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2395                         goto err_exit;
2396
2397                 memset(&p, 0, sizeof(p));
2398                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2399                 p.iph.saddr = 0;
2400                 p.iph.version = 4;
2401                 p.iph.ihl = 5;
2402                 p.iph.protocol = IPPROTO_IPV6;
2403                 p.iph.ttl = 64;
2404                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2405
2406                 if (ops->ndo_do_ioctl) {
2407                         mm_segment_t oldfs = get_fs();
2408
2409                         set_fs(KERNEL_DS);
2410                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2411                         set_fs(oldfs);
2412                 } else
2413                         err = -EOPNOTSUPP;
2414
2415                 if (err == 0) {
2416                         err = -ENOBUFS;
2417                         dev = __dev_get_by_name(net, p.name);
2418                         if (!dev)
2419                                 goto err_exit;
2420                         err = dev_open(dev);
2421                 }
2422         }
2423 #endif
2424
2425 err_exit:
2426         rtnl_unlock();
2427         return err;
2428 }
2429
2430 /*
2431  *      Manual configuration of address on an interface
2432  */
2433 static int inet6_addr_add(struct net *net, int ifindex,
2434                           const struct in6_addr *pfx,
2435                           const struct in6_addr *peer_pfx,
2436                           unsigned int plen, __u32 ifa_flags,
2437                           __u32 prefered_lft, __u32 valid_lft)
2438 {
2439         struct inet6_ifaddr *ifp;
2440         struct inet6_dev *idev;
2441         struct net_device *dev;
2442         int scope;
2443         u32 flags;
2444         clock_t expires;
2445         unsigned long timeout;
2446
2447         ASSERT_RTNL();
2448
2449         if (plen > 128)
2450                 return -EINVAL;
2451
2452         /* check the lifetime */
2453         if (!valid_lft || prefered_lft > valid_lft)
2454                 return -EINVAL;
2455
2456         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2457                 return -EINVAL;
2458
2459         dev = __dev_get_by_index(net, ifindex);
2460         if (!dev)
2461                 return -ENODEV;
2462
2463         idev = addrconf_add_dev(dev);
2464         if (IS_ERR(idev))
2465                 return PTR_ERR(idev);
2466
2467         scope = ipv6_addr_scope(pfx);
2468
2469         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2470         if (addrconf_finite_timeout(timeout)) {
2471                 expires = jiffies_to_clock_t(timeout * HZ);
2472                 valid_lft = timeout;
2473                 flags = RTF_EXPIRES;
2474         } else {
2475                 expires = 0;
2476                 flags = 0;
2477                 ifa_flags |= IFA_F_PERMANENT;
2478         }
2479
2480         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2481         if (addrconf_finite_timeout(timeout)) {
2482                 if (timeout == 0)
2483                         ifa_flags |= IFA_F_DEPRECATED;
2484                 prefered_lft = timeout;
2485         }
2486
2487         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2488                             valid_lft, prefered_lft);
2489
2490         if (!IS_ERR(ifp)) {
2491                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2492                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2493                                               expires, flags);
2494                 }
2495
2496                 /*
2497                  * Note that section 3.1 of RFC 4429 indicates
2498                  * that the Optimistic flag should not be set for
2499                  * manually configured addresses
2500                  */
2501                 addrconf_dad_start(ifp);
2502                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2503                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2504                                          true, jiffies);
2505                 in6_ifa_put(ifp);
2506                 addrconf_verify_rtnl();
2507                 return 0;
2508         }
2509
2510         return PTR_ERR(ifp);
2511 }
2512
2513 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2514                           const struct in6_addr *pfx, unsigned int plen)
2515 {
2516         struct inet6_ifaddr *ifp;
2517         struct inet6_dev *idev;
2518         struct net_device *dev;
2519
2520         if (plen > 128)
2521                 return -EINVAL;
2522
2523         dev = __dev_get_by_index(net, ifindex);
2524         if (!dev)
2525                 return -ENODEV;
2526
2527         if ((idev = __in6_dev_get(dev)) == NULL)
2528                 return -ENXIO;
2529
2530         read_lock_bh(&idev->lock);
2531         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2532                 if (ifp->prefix_len == plen &&
2533                     ipv6_addr_equal(pfx, &ifp->addr)) {
2534                         in6_ifa_hold(ifp);
2535                         read_unlock_bh(&idev->lock);
2536
2537                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2538                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2539                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2540                                                  jiffies);
2541                         ipv6_del_addr(ifp);
2542                         addrconf_verify_rtnl();
2543                         return 0;
2544                 }
2545         }
2546         read_unlock_bh(&idev->lock);
2547         return -EADDRNOTAVAIL;
2548 }
2549
2550
2551 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2552 {
2553         struct in6_ifreq ireq;
2554         int err;
2555
2556         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2557                 return -EPERM;
2558
2559         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2560                 return -EFAULT;
2561
2562         rtnl_lock();
2563         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2564                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2565                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2566         rtnl_unlock();
2567         return err;
2568 }
2569
2570 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2571 {
2572         struct in6_ifreq ireq;
2573         int err;
2574
2575         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2576                 return -EPERM;
2577
2578         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2579                 return -EFAULT;
2580
2581         rtnl_lock();
2582         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2583                              ireq.ifr6_prefixlen);
2584         rtnl_unlock();
2585         return err;
2586 }
2587
2588 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2589                      int plen, int scope)
2590 {
2591         struct inet6_ifaddr *ifp;
2592
2593         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2594                             scope, IFA_F_PERMANENT,
2595                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2596         if (!IS_ERR(ifp)) {
2597                 spin_lock_bh(&ifp->lock);
2598                 ifp->flags &= ~IFA_F_TENTATIVE;
2599                 spin_unlock_bh(&ifp->lock);
2600                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2601                 in6_ifa_put(ifp);
2602         }
2603 }
2604
2605 #if IS_ENABLED(CONFIG_IPV6_SIT)
2606 static void sit_add_v4_addrs(struct inet6_dev *idev)
2607 {
2608         struct in6_addr addr;
2609         struct net_device *dev;
2610         struct net *net = dev_net(idev->dev);
2611         int scope, plen;
2612         u32 pflags = 0;
2613
2614         ASSERT_RTNL();
2615
2616         memset(&addr, 0, sizeof(struct in6_addr));
2617         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2618
2619         if (idev->dev->flags&IFF_POINTOPOINT) {
2620                 addr.s6_addr32[0] = htonl(0xfe800000);
2621                 scope = IFA_LINK;
2622                 plen = 64;
2623         } else {
2624                 scope = IPV6_ADDR_COMPATv4;
2625                 plen = 96;
2626                 pflags |= RTF_NONEXTHOP;
2627         }
2628
2629         if (addr.s6_addr32[3]) {
2630                 add_addr(idev, &addr, plen, scope);
2631                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2632                 return;
2633         }
2634
2635         for_each_netdev(net, dev) {
2636                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2637                 if (in_dev && (dev->flags & IFF_UP)) {
2638                         struct in_ifaddr *ifa;
2639
2640                         int flag = scope;
2641
2642                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2643
2644                                 addr.s6_addr32[3] = ifa->ifa_local;
2645
2646                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2647                                         continue;
2648                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2649                                         if (idev->dev->flags&IFF_POINTOPOINT)
2650                                                 continue;
2651                                         flag |= IFA_HOST;
2652                                 }
2653
2654                                 add_addr(idev, &addr, plen, flag);
2655                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2656                                                       pflags);
2657                         }
2658                 }
2659         }
2660 }
2661 #endif
2662
2663 static void init_loopback(struct net_device *dev)
2664 {
2665         struct inet6_dev  *idev;
2666         struct net_device *sp_dev;
2667         struct inet6_ifaddr *sp_ifa;
2668         struct rt6_info *sp_rt;
2669
2670         /* ::1 */
2671
2672         ASSERT_RTNL();
2673
2674         if ((idev = ipv6_find_idev(dev)) == NULL) {
2675                 pr_debug("%s: add_dev failed\n", __func__);
2676                 return;
2677         }
2678
2679         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2680
2681         /* Add routes to other interface's IPv6 addresses */
2682         for_each_netdev(dev_net(dev), sp_dev) {
2683                 if (!strcmp(sp_dev->name, dev->name))
2684                         continue;
2685
2686                 idev = __in6_dev_get(sp_dev);
2687                 if (!idev)
2688                         continue;
2689
2690                 read_lock_bh(&idev->lock);
2691                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2692
2693                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2694                                 continue;
2695
2696                         if (sp_ifa->rt) {
2697                                 /* This dst has been added to garbage list when
2698                                  * lo device down, release this obsolete dst and
2699                                  * reallocate a new router for ifa.
2700                                  */
2701                                 if (sp_ifa->rt->dst.obsolete > 0) {
2702                                         ip6_rt_put(sp_ifa->rt);
2703                                         sp_ifa->rt = NULL;
2704                                 } else {
2705                                         continue;
2706                                 }
2707                         }
2708
2709                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2710
2711                         /* Failure cases are ignored */
2712                         if (!IS_ERR(sp_rt)) {
2713                                 sp_ifa->rt = sp_rt;
2714                                 ip6_ins_rt(sp_rt);
2715                         }
2716                 }
2717                 read_unlock_bh(&idev->lock);
2718         }
2719 }
2720
2721 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2722 {
2723         struct inet6_ifaddr *ifp;
2724         u32 addr_flags = IFA_F_PERMANENT;
2725
2726 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2727         if (idev->cnf.optimistic_dad &&
2728             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2729                 addr_flags |= IFA_F_OPTIMISTIC;
2730 #endif
2731
2732
2733         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2734                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2735         if (!IS_ERR(ifp)) {
2736                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2737                 addrconf_dad_start(ifp);
2738                 in6_ifa_put(ifp);
2739         }
2740 }
2741
2742 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2743 {
2744         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2745                 struct in6_addr addr;
2746
2747                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2748                 /* addrconf_add_linklocal also adds a prefix_route and we
2749                  * only need to care about prefix routes if ipv6_generate_eui64
2750                  * couldn't generate one.
2751                  */
2752                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2753                         addrconf_add_linklocal(idev, &addr);
2754                 else if (prefix_route)
2755                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2756         }
2757 }
2758
2759 static void addrconf_dev_config(struct net_device *dev)
2760 {
2761         struct inet6_dev *idev;
2762
2763         ASSERT_RTNL();
2764
2765         if ((dev->type != ARPHRD_ETHER) &&
2766             (dev->type != ARPHRD_FDDI) &&
2767             (dev->type != ARPHRD_ARCNET) &&
2768             (dev->type != ARPHRD_INFINIBAND) &&
2769             (dev->type != ARPHRD_IEEE802154) &&
2770             (dev->type != ARPHRD_IEEE1394) &&
2771             (dev->type != ARPHRD_TUNNEL6) &&
2772             (dev->type != ARPHRD_6LOWPAN)) {
2773                 /* Alas, we support only Ethernet autoconfiguration. */
2774                 return;
2775         }
2776
2777         idev = addrconf_add_dev(dev);
2778         if (IS_ERR(idev))
2779                 return;
2780
2781         addrconf_addr_gen(idev, false);
2782 }
2783
2784 #if IS_ENABLED(CONFIG_IPV6_SIT)
2785 static void addrconf_sit_config(struct net_device *dev)
2786 {
2787         struct inet6_dev *idev;
2788
2789         ASSERT_RTNL();
2790
2791         /*
2792          * Configure the tunnel with one of our IPv4
2793          * addresses... we should configure all of
2794          * our v4 addrs in the tunnel
2795          */
2796
2797         if ((idev = ipv6_find_idev(dev)) == NULL) {
2798                 pr_debug("%s: add_dev failed\n", __func__);
2799                 return;
2800         }
2801
2802         if (dev->priv_flags & IFF_ISATAP) {
2803                 addrconf_addr_gen(idev, false);
2804                 return;
2805         }
2806
2807         sit_add_v4_addrs(idev);
2808
2809         if (dev->flags&IFF_POINTOPOINT)
2810                 addrconf_add_mroute(dev);
2811 }
2812 #endif
2813
2814 #if IS_ENABLED(CONFIG_NET_IPGRE)
2815 static void addrconf_gre_config(struct net_device *dev)
2816 {
2817         struct inet6_dev *idev;
2818
2819         ASSERT_RTNL();
2820
2821         if ((idev = ipv6_find_idev(dev)) == NULL) {
2822                 pr_debug("%s: add_dev failed\n", __func__);
2823                 return;
2824         }
2825
2826         addrconf_addr_gen(idev, true);
2827 }
2828 #endif
2829
2830 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2831                            void *ptr)
2832 {
2833         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2834         struct inet6_dev *idev = __in6_dev_get(dev);
2835         int run_pending = 0;
2836         int err;
2837
2838         switch (event) {
2839         case NETDEV_REGISTER:
2840                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2841                         idev = ipv6_add_dev(dev);
2842                         if (IS_ERR(idev))
2843                                 return notifier_from_errno(PTR_ERR(idev));
2844                 }
2845                 break;
2846
2847         case NETDEV_UP:
2848         case NETDEV_CHANGE:
2849                 if (dev->flags & IFF_SLAVE)
2850                         break;
2851
2852                 if (event == NETDEV_UP) {
2853                         if (!addrconf_qdisc_ok(dev)) {
2854                                 /* device is not ready yet. */
2855                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2856                                         dev->name);
2857                                 break;
2858                         }
2859
2860                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2861                                 idev = ipv6_add_dev(dev);
2862
2863                         if (!IS_ERR_OR_NULL(idev)) {
2864                                 idev->if_flags |= IF_READY;
2865                                 run_pending = 1;
2866                         }
2867                 } else {
2868                         if (!addrconf_qdisc_ok(dev)) {
2869                                 /* device is still not ready. */
2870                                 break;
2871                         }
2872
2873                         if (idev) {
2874                                 if (idev->if_flags & IF_READY)
2875                                         /* device is already configured. */
2876                                         break;
2877                                 idev->if_flags |= IF_READY;
2878                         }
2879
2880                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2881                                 dev->name);
2882
2883                         run_pending = 1;
2884                 }
2885
2886                 switch (dev->type) {
2887 #if IS_ENABLED(CONFIG_IPV6_SIT)
2888                 case ARPHRD_SIT:
2889                         addrconf_sit_config(dev);
2890                         break;
2891 #endif
2892 #if IS_ENABLED(CONFIG_NET_IPGRE)
2893                 case ARPHRD_IPGRE:
2894                         addrconf_gre_config(dev);
2895                         break;
2896 #endif
2897                 case ARPHRD_LOOPBACK:
2898                         init_loopback(dev);
2899                         break;
2900
2901                 default:
2902                         addrconf_dev_config(dev);
2903                         break;
2904                 }
2905
2906                 if (!IS_ERR_OR_NULL(idev)) {
2907                         if (run_pending)
2908                                 addrconf_dad_run(idev);
2909
2910                         /*
2911                          * If the MTU changed during the interface down,
2912                          * when the interface up, the changed MTU must be
2913                          * reflected in the idev as well as routers.
2914                          */
2915                         if (idev->cnf.mtu6 != dev->mtu &&
2916                             dev->mtu >= IPV6_MIN_MTU) {
2917                                 rt6_mtu_change(dev, dev->mtu);
2918                                 idev->cnf.mtu6 = dev->mtu;
2919                         }
2920                         idev->tstamp = jiffies;
2921                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2922
2923                         /*
2924                          * If the changed mtu during down is lower than
2925                          * IPV6_MIN_MTU stop IPv6 on this interface.
2926                          */
2927                         if (dev->mtu < IPV6_MIN_MTU)
2928                                 addrconf_ifdown(dev, 1);
2929                 }
2930                 break;
2931
2932         case NETDEV_CHANGEMTU:
2933                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2934                         rt6_mtu_change(dev, dev->mtu);
2935                         idev->cnf.mtu6 = dev->mtu;
2936                         break;
2937                 }
2938
2939                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2940                         idev = ipv6_add_dev(dev);
2941                         if (!IS_ERR(idev))
2942                                 break;
2943                 }
2944
2945                 /*
2946                  * if MTU under IPV6_MIN_MTU.
2947                  * Stop IPv6 on this interface.
2948                  */
2949
2950         case NETDEV_DOWN:
2951         case NETDEV_UNREGISTER:
2952                 /*
2953                  *      Remove all addresses from this interface.
2954                  */
2955                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2956                 break;
2957
2958         case NETDEV_CHANGENAME:
2959                 if (idev) {
2960                         snmp6_unregister_dev(idev);
2961                         addrconf_sysctl_unregister(idev);
2962                         err = addrconf_sysctl_register(idev);
2963                         if (err)
2964                                 return notifier_from_errno(err);
2965                         err = snmp6_register_dev(idev);
2966                         if (err) {
2967                                 addrconf_sysctl_unregister(idev);
2968                                 return notifier_from_errno(err);
2969                         }
2970                 }
2971                 break;
2972
2973         case NETDEV_PRE_TYPE_CHANGE:
2974         case NETDEV_POST_TYPE_CHANGE:
2975                 addrconf_type_change(dev, event);
2976                 break;
2977         }
2978
2979         return NOTIFY_OK;
2980 }
2981
2982 /*
2983  *      addrconf module should be notified of a device going up
2984  */
2985 static struct notifier_block ipv6_dev_notf = {
2986         .notifier_call = addrconf_notify,
2987 };
2988
2989 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2990 {
2991         struct inet6_dev *idev;
2992         ASSERT_RTNL();
2993
2994         idev = __in6_dev_get(dev);
2995
2996         if (event == NETDEV_POST_TYPE_CHANGE)
2997                 ipv6_mc_remap(idev);
2998         else if (event == NETDEV_PRE_TYPE_CHANGE)
2999                 ipv6_mc_unmap(idev);
3000 }
3001
3002 static int addrconf_ifdown(struct net_device *dev, int how)
3003 {
3004         struct net *net = dev_net(dev);
3005         struct inet6_dev *idev;
3006         struct inet6_ifaddr *ifa;
3007         int state, i;
3008
3009         ASSERT_RTNL();
3010
3011         rt6_ifdown(net, dev);
3012         neigh_ifdown(&nd_tbl, dev);
3013
3014         idev = __in6_dev_get(dev);
3015         if (idev == NULL)
3016                 return -ENODEV;
3017
3018         /*
3019          * Step 1: remove reference to ipv6 device from parent device.
3020          *         Do not dev_put!
3021          */
3022         if (how) {
3023                 idev->dead = 1;
3024
3025                 /* protected by rtnl_lock */
3026                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3027
3028                 /* Step 1.5: remove snmp6 entry */
3029                 snmp6_unregister_dev(idev);
3030
3031         }
3032
3033         /* Step 2: clear hash table */
3034         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3035                 struct hlist_head *h = &inet6_addr_lst[i];
3036
3037                 spin_lock_bh(&addrconf_hash_lock);
3038         restart:
3039                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3040                         if (ifa->idev == idev) {
3041                                 hlist_del_init_rcu(&ifa->addr_lst);
3042                                 addrconf_del_dad_work(ifa);
3043                                 goto restart;
3044                         }
3045                 }
3046                 spin_unlock_bh(&addrconf_hash_lock);
3047         }
3048
3049         write_lock_bh(&idev->lock);
3050
3051         addrconf_del_rs_timer(idev);
3052
3053         /* Step 2: clear flags for stateless addrconf */
3054         if (!how)
3055                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3056
3057         if (how && del_timer(&idev->regen_timer))
3058                 in6_dev_put(idev);
3059
3060         /* Step 3: clear tempaddr list */
3061         while (!list_empty(&idev->tempaddr_list)) {
3062                 ifa = list_first_entry(&idev->tempaddr_list,
3063                                        struct inet6_ifaddr, tmp_list);
3064                 list_del(&ifa->tmp_list);
3065                 write_unlock_bh(&idev->lock);
3066                 spin_lock_bh(&ifa->lock);
3067
3068                 if (ifa->ifpub) {
3069                         in6_ifa_put(ifa->ifpub);
3070                         ifa->ifpub = NULL;
3071                 }
3072                 spin_unlock_bh(&ifa->lock);
3073                 in6_ifa_put(ifa);
3074                 write_lock_bh(&idev->lock);
3075         }
3076
3077         while (!list_empty(&idev->addr_list)) {
3078                 ifa = list_first_entry(&idev->addr_list,
3079                                        struct inet6_ifaddr, if_list);
3080                 addrconf_del_dad_work(ifa);
3081
3082                 list_del(&ifa->if_list);
3083
3084                 write_unlock_bh(&idev->lock);
3085
3086                 spin_lock_bh(&ifa->state_lock);
3087                 state = ifa->state;
3088                 ifa->state = INET6_IFADDR_STATE_DEAD;
3089                 spin_unlock_bh(&ifa->state_lock);
3090
3091                 if (state != INET6_IFADDR_STATE_DEAD) {
3092                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3093                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3094                 }
3095                 in6_ifa_put(ifa);
3096
3097                 write_lock_bh(&idev->lock);
3098         }
3099
3100         write_unlock_bh(&idev->lock);
3101
3102         /* Step 5: Discard multicast list */
3103         if (how)
3104                 ipv6_mc_destroy_dev(idev);
3105         else
3106                 ipv6_mc_down(idev);
3107
3108         idev->tstamp = jiffies;
3109
3110         /* Last: Shot the device (if unregistered) */
3111         if (how) {
3112                 addrconf_sysctl_unregister(idev);
3113                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3114                 neigh_ifdown(&nd_tbl, dev);
3115                 in6_dev_put(idev);
3116         }
3117         return 0;
3118 }
3119
3120 static void addrconf_rs_timer(unsigned long data)
3121 {
3122         struct inet6_dev *idev = (struct inet6_dev *)data;
3123         struct net_device *dev = idev->dev;
3124         struct in6_addr lladdr;
3125
3126         write_lock(&idev->lock);
3127         if (idev->dead || !(idev->if_flags & IF_READY))
3128                 goto out;
3129
3130         if (!ipv6_accept_ra(idev))
3131                 goto out;
3132
3133         /* Announcement received after solicitation was sent */
3134         if (idev->if_flags & IF_RA_RCVD)
3135                 goto out;
3136
3137         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3138                 write_unlock(&idev->lock);
3139                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3140                         ndisc_send_rs(dev, &lladdr,
3141                                       &in6addr_linklocal_allrouters);
3142                 else
3143                         goto put;
3144
3145                 write_lock(&idev->lock);
3146                 /* The wait after the last probe can be shorter */
3147                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3148                                              idev->cnf.rtr_solicits) ?
3149                                       idev->cnf.rtr_solicit_delay :
3150                                       idev->cnf.rtr_solicit_interval);
3151         } else {
3152                 /*
3153                  * Note: we do not support deprecated "all on-link"
3154                  * assumption any longer.
3155                  */
3156                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3157         }
3158
3159 out:
3160         write_unlock(&idev->lock);
3161 put:
3162         in6_dev_put(idev);
3163 }
3164
3165 /*
3166  *      Duplicate Address Detection
3167  */
3168 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3169 {
3170         unsigned long rand_num;
3171         struct inet6_dev *idev = ifp->idev;
3172
3173         if (ifp->flags & IFA_F_OPTIMISTIC)
3174                 rand_num = 0;
3175         else
3176                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3177
3178         ifp->dad_probes = idev->cnf.dad_transmits;
3179         addrconf_mod_dad_work(ifp, rand_num);
3180 }
3181
3182 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3183 {
3184         struct inet6_dev *idev = ifp->idev;
3185         struct net_device *dev = idev->dev;
3186
3187         addrconf_join_solict(dev, &ifp->addr);
3188
3189         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3190
3191         read_lock_bh(&idev->lock);
3192         spin_lock(&ifp->lock);
3193         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3194                 goto out;
3195
3196         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3197             idev->cnf.accept_dad < 1 ||
3198             !(ifp->flags&IFA_F_TENTATIVE) ||
3199             ifp->flags & IFA_F_NODAD) {
3200                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3201                 spin_unlock(&ifp->lock);
3202                 read_unlock_bh(&idev->lock);
3203
3204                 addrconf_dad_completed(ifp);
3205                 return;
3206         }
3207
3208         if (!(idev->if_flags & IF_READY)) {
3209                 spin_unlock(&ifp->lock);
3210                 read_unlock_bh(&idev->lock);
3211                 /*
3212                  * If the device is not ready:
3213                  * - keep it tentative if it is a permanent address.
3214                  * - otherwise, kill it.
3215                  */
3216                 in6_ifa_hold(ifp);
3217                 addrconf_dad_stop(ifp, 0);
3218                 return;
3219         }
3220
3221         /*
3222          * Optimistic nodes can start receiving
3223          * Frames right away
3224          */
3225         if (ifp->flags & IFA_F_OPTIMISTIC)
3226                 ip6_ins_rt(ifp->rt);
3227
3228         addrconf_dad_kick(ifp);
3229 out:
3230         spin_unlock(&ifp->lock);
3231         read_unlock_bh(&idev->lock);
3232 }
3233
3234 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3235 {
3236         bool begin_dad = false;
3237
3238         spin_lock_bh(&ifp->state_lock);
3239         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3240                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3241                 begin_dad = true;
3242         }
3243         spin_unlock_bh(&ifp->state_lock);
3244
3245         if (begin_dad)
3246                 addrconf_mod_dad_work(ifp, 0);
3247 }
3248
3249 static void addrconf_dad_work(struct work_struct *w)
3250 {
3251         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3252                                                 struct inet6_ifaddr,
3253                                                 dad_work);
3254         struct inet6_dev *idev = ifp->idev;
3255         struct in6_addr mcaddr;
3256
3257         enum {
3258                 DAD_PROCESS,
3259                 DAD_BEGIN,
3260                 DAD_ABORT,
3261         } action = DAD_PROCESS;
3262
3263         rtnl_lock();
3264
3265         spin_lock_bh(&ifp->state_lock);
3266         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3267                 action = DAD_BEGIN;
3268                 ifp->state = INET6_IFADDR_STATE_DAD;
3269         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3270                 action = DAD_ABORT;
3271                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3272         }
3273         spin_unlock_bh(&ifp->state_lock);
3274
3275         if (action == DAD_BEGIN) {
3276                 addrconf_dad_begin(ifp);
3277                 goto out;
3278         } else if (action == DAD_ABORT) {
3279                 addrconf_dad_stop(ifp, 1);
3280                 goto out;
3281         }
3282
3283         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3284                 goto out;
3285
3286         write_lock_bh(&idev->lock);
3287         if (idev->dead || !(idev->if_flags & IF_READY)) {
3288                 write_unlock_bh(&idev->lock);
3289                 goto out;
3290         }
3291
3292         spin_lock(&ifp->lock);
3293         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3294                 spin_unlock(&ifp->lock);
3295                 write_unlock_bh(&idev->lock);
3296                 goto out;
3297         }
3298
3299         if (ifp->dad_probes == 0) {
3300                 /*
3301                  * DAD was successful
3302                  */
3303
3304                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3305                 spin_unlock(&ifp->lock);
3306                 write_unlock_bh(&idev->lock);
3307
3308                 addrconf_dad_completed(ifp);
3309
3310                 goto out;
3311         }
3312
3313         ifp->dad_probes--;
3314         addrconf_mod_dad_work(ifp,
3315                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3316         spin_unlock(&ifp->lock);
3317         write_unlock_bh(&idev->lock);
3318
3319         /* send a neighbour solicitation for our addr */
3320         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3321         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3322 out:
3323         in6_ifa_put(ifp);
3324         rtnl_unlock();
3325 }
3326
3327 /* ifp->idev must be at least read locked */
3328 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3329 {
3330         struct inet6_ifaddr *ifpiter;
3331         struct inet6_dev *idev = ifp->idev;
3332
3333         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3334                 if (ifpiter->scope > IFA_LINK)
3335                         break;
3336                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3337                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3338                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3339                     IFA_F_PERMANENT)
3340                         return false;
3341         }
3342         return true;
3343 }
3344
3345 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3346 {
3347         struct net_device *dev = ifp->idev->dev;
3348         struct in6_addr lladdr;
3349         bool send_rs, send_mld;
3350
3351         addrconf_del_dad_work(ifp);
3352
3353         /*
3354          *      Configure the address for reception. Now it is valid.
3355          */
3356
3357         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3358
3359         /* If added prefix is link local and we are prepared to process
3360            router advertisements, start sending router solicitations.
3361          */
3362
3363         read_lock_bh(&ifp->idev->lock);
3364         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3365         send_rs = send_mld &&
3366                   ipv6_accept_ra(ifp->idev) &&
3367                   ifp->idev->cnf.rtr_solicits > 0 &&
3368                   (dev->flags&IFF_LOOPBACK) == 0;
3369         read_unlock_bh(&ifp->idev->lock);
3370
3371         /* While dad is in progress mld report's source address is in6_addrany.
3372          * Resend with proper ll now.
3373          */
3374         if (send_mld)
3375                 ipv6_mc_dad_complete(ifp->idev);
3376
3377         if (send_rs) {
3378                 /*
3379                  *      If a host as already performed a random delay
3380                  *      [...] as part of DAD [...] there is no need
3381                  *      to delay again before sending the first RS
3382                  */
3383                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3384                         return;
3385                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3386
3387                 write_lock_bh(&ifp->idev->lock);
3388                 spin_lock(&ifp->lock);
3389                 ifp->idev->rs_probes = 1;
3390                 ifp->idev->if_flags |= IF_RS_SENT;
3391                 addrconf_mod_rs_timer(ifp->idev,
3392                                       ifp->idev->cnf.rtr_solicit_interval);
3393                 spin_unlock(&ifp->lock);
3394                 write_unlock_bh(&ifp->idev->lock);
3395         }
3396 }
3397
3398 static void addrconf_dad_run(struct inet6_dev *idev)
3399 {
3400         struct inet6_ifaddr *ifp;
3401
3402         read_lock_bh(&idev->lock);
3403         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3404                 spin_lock(&ifp->lock);
3405                 if (ifp->flags & IFA_F_TENTATIVE &&
3406                     ifp->state == INET6_IFADDR_STATE_DAD)
3407                         addrconf_dad_kick(ifp);
3408                 spin_unlock(&ifp->lock);
3409         }
3410         read_unlock_bh(&idev->lock);
3411 }
3412
3413 #ifdef CONFIG_PROC_FS
3414 struct if6_iter_state {
3415         struct seq_net_private p;
3416         int bucket;
3417         int offset;
3418 };
3419
3420 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3421 {
3422         struct inet6_ifaddr *ifa = NULL;
3423         struct if6_iter_state *state = seq->private;
3424         struct net *net = seq_file_net(seq);
3425         int p = 0;
3426
3427         /* initial bucket if pos is 0 */
3428         if (pos == 0) {
3429                 state->bucket = 0;
3430                 state->offset = 0;
3431         }
3432
3433         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3434                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3435                                          addr_lst) {
3436                         if (!net_eq(dev_net(ifa->idev->dev), net))
3437                                 continue;
3438                         /* sync with offset */
3439                         if (p < state->offset) {
3440                                 p++;
3441                                 continue;
3442                         }
3443                         state->offset++;
3444                         return ifa;
3445                 }
3446
3447                 /* prepare for next bucket */
3448                 state->offset = 0;
3449                 p = 0;
3450         }
3451         return NULL;
3452 }
3453
3454 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3455                                          struct inet6_ifaddr *ifa)
3456 {
3457         struct if6_iter_state *state = seq->private;
3458         struct net *net = seq_file_net(seq);
3459
3460         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3461                 if (!net_eq(dev_net(ifa->idev->dev), net))
3462                         continue;
3463                 state->offset++;
3464                 return ifa;
3465         }
3466
3467         while (++state->bucket < IN6_ADDR_HSIZE) {
3468                 state->offset = 0;
3469                 hlist_for_each_entry_rcu_bh(ifa,
3470                                      &inet6_addr_lst[state->bucket], addr_lst) {
3471                         if (!net_eq(dev_net(ifa->idev->dev), net))
3472                                 continue;
3473                         state->offset++;
3474                         return ifa;
3475                 }
3476         }
3477
3478         return NULL;
3479 }
3480
3481 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3482         __acquires(rcu_bh)
3483 {
3484         rcu_read_lock_bh();
3485         return if6_get_first(seq, *pos);
3486 }
3487
3488 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3489 {
3490         struct inet6_ifaddr *ifa;
3491
3492         ifa = if6_get_next(seq, v);
3493         ++*pos;
3494         return ifa;
3495 }
3496
3497 static void if6_seq_stop(struct seq_file *seq, void *v)
3498         __releases(rcu_bh)
3499 {
3500         rcu_read_unlock_bh();
3501 }
3502
3503 static int if6_seq_show(struct seq_file *seq, void *v)
3504 {
3505         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3506         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3507                    &ifp->addr,
3508                    ifp->idev->dev->ifindex,
3509                    ifp->prefix_len,
3510                    ifp->scope,
3511                    (u8) ifp->flags,
3512                    ifp->idev->dev->name);
3513         return 0;
3514 }
3515
3516 static const struct seq_operations if6_seq_ops = {
3517         .start  = if6_seq_start,
3518         .next   = if6_seq_next,
3519         .show   = if6_seq_show,
3520         .stop   = if6_seq_stop,
3521 };
3522
3523 static int if6_seq_open(struct inode *inode, struct file *file)
3524 {
3525         return seq_open_net(inode, file, &if6_seq_ops,
3526                             sizeof(struct if6_iter_state));
3527 }
3528
3529 static const struct file_operations if6_fops = {
3530         .owner          = THIS_MODULE,
3531         .open           = if6_seq_open,
3532         .read           = seq_read,
3533         .llseek         = seq_lseek,
3534         .release        = seq_release_net,
3535 };
3536
3537 static int __net_init if6_proc_net_init(struct net *net)
3538 {
3539         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3540                 return -ENOMEM;
3541         return 0;
3542 }
3543
3544 static void __net_exit if6_proc_net_exit(struct net *net)
3545 {
3546         remove_proc_entry("if_inet6", net->proc_net);
3547 }
3548
3549 static struct pernet_operations if6_proc_net_ops = {
3550        .init = if6_proc_net_init,
3551        .exit = if6_proc_net_exit,
3552 };
3553
3554 int __init if6_proc_init(void)
3555 {
3556         return register_pernet_subsys(&if6_proc_net_ops);
3557 }
3558
3559 void if6_proc_exit(void)
3560 {
3561         unregister_pernet_subsys(&if6_proc_net_ops);
3562 }
3563 #endif  /* CONFIG_PROC_FS */
3564
3565 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3566 /* Check if address is a home address configured on any interface. */
3567 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3568 {
3569         int ret = 0;
3570         struct inet6_ifaddr *ifp = NULL;
3571         unsigned int hash = inet6_addr_hash(addr);
3572
3573         rcu_read_lock_bh();
3574         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3575                 if (!net_eq(dev_net(ifp->idev->dev), net))
3576                         continue;
3577                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3578                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3579                         ret = 1;
3580                         break;
3581                 }
3582         }
3583         rcu_read_unlock_bh();
3584         return ret;
3585 }
3586 #endif
3587
3588 /*
3589  *      Periodic address status verification
3590  */
3591
3592 static void addrconf_verify_rtnl(void)
3593 {
3594         unsigned long now, next, next_sec, next_sched;
3595         struct inet6_ifaddr *ifp;
3596         int i;
3597
3598         ASSERT_RTNL();
3599
3600         rcu_read_lock_bh();
3601         now = jiffies;
3602         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3603
3604         cancel_delayed_work(&addr_chk_work);
3605
3606         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3607 restart:
3608                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3609                         unsigned long age;
3610
3611                         /* When setting preferred_lft to a value not zero or
3612                          * infinity, while valid_lft is infinity
3613                          * IFA_F_PERMANENT has a non-infinity life time.
3614                          */
3615                         if ((ifp->flags & IFA_F_PERMANENT) &&
3616                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3617                                 continue;
3618
3619                         spin_lock(&ifp->lock);
3620                         /* We try to batch several events at once. */
3621                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3622
3623                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3624                             age >= ifp->valid_lft) {
3625                                 spin_unlock(&ifp->lock);
3626                                 in6_ifa_hold(ifp);
3627                                 ipv6_del_addr(ifp);
3628                                 goto restart;
3629                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3630                                 spin_unlock(&ifp->lock);
3631                                 continue;
3632                         } else if (age >= ifp->prefered_lft) {
3633                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3634                                 int deprecate = 0;
3635
3636                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3637                                         deprecate = 1;
3638                                         ifp->flags |= IFA_F_DEPRECATED;
3639                                 }
3640
3641                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3642                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3643                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3644
3645                                 spin_unlock(&ifp->lock);
3646
3647                                 if (deprecate) {
3648                                         in6_ifa_hold(ifp);
3649
3650                                         ipv6_ifa_notify(0, ifp);
3651                                         in6_ifa_put(ifp);
3652                                         goto restart;
3653                                 }
3654                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3655                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3656                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3657                                         ifp->idev->cnf.dad_transmits *
3658                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3659
3660                                 if (age >= ifp->prefered_lft - regen_advance) {
3661                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3662                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3663                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3664                                         if (!ifp->regen_count && ifpub) {
3665                                                 ifp->regen_count++;
3666                                                 in6_ifa_hold(ifp);
3667                                                 in6_ifa_hold(ifpub);
3668                                                 spin_unlock(&ifp->lock);
3669
3670                                                 spin_lock(&ifpub->lock);
3671                                                 ifpub->regen_count = 0;
3672                                                 spin_unlock(&ifpub->lock);
3673                                                 ipv6_create_tempaddr(ifpub, ifp);
3674                                                 in6_ifa_put(ifpub);
3675                                                 in6_ifa_put(ifp);
3676                                                 goto restart;
3677                                         }
3678                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3679                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3680                                 spin_unlock(&ifp->lock);
3681                         } else {
3682                                 /* ifp->prefered_lft <= ifp->valid_lft */
3683                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3684                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3685                                 spin_unlock(&ifp->lock);
3686                         }
3687                 }
3688         }
3689
3690         next_sec = round_jiffies_up(next);
3691         next_sched = next;
3692
3693         /* If rounded timeout is accurate enough, accept it. */
3694         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3695                 next_sched = next_sec;
3696
3697         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3698         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3699                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3700
3701         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3702               now, next, next_sec, next_sched);
3703         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3704         rcu_read_unlock_bh();
3705 }
3706
3707 static void addrconf_verify_work(struct work_struct *w)
3708 {
3709         rtnl_lock();
3710         addrconf_verify_rtnl();
3711         rtnl_unlock();
3712 }
3713
3714 static void addrconf_verify(void)
3715 {
3716         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3717 }
3718
3719 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3720                                      struct in6_addr **peer_pfx)
3721 {
3722         struct in6_addr *pfx = NULL;
3723
3724         *peer_pfx = NULL;
3725
3726         if (addr)
3727                 pfx = nla_data(addr);
3728
3729         if (local) {
3730                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3731                         *peer_pfx = pfx;
3732                 pfx = nla_data(local);
3733         }
3734
3735         return pfx;
3736 }
3737
3738 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3739         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3740         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3741         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3742         [IFA_FLAGS]             = { .len = sizeof(u32) },
3743 };
3744
3745 static int
3746 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3747 {
3748         struct net *net = sock_net(skb->sk);
3749         struct ifaddrmsg *ifm;
3750         struct nlattr *tb[IFA_MAX+1];
3751         struct in6_addr *pfx, *peer_pfx;
3752         u32 ifa_flags;
3753         int err;
3754
3755         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3756         if (err < 0)
3757                 return err;
3758
3759         ifm = nlmsg_data(nlh);
3760         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3761         if (pfx == NULL)
3762                 return -EINVAL;
3763
3764         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3765
3766         /* We ignore other flags so far. */
3767         ifa_flags &= IFA_F_MANAGETEMPADDR;
3768
3769         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3770                               ifm->ifa_prefixlen);
3771 }
3772
3773 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3774                              u32 prefered_lft, u32 valid_lft)
3775 {
3776         u32 flags;
3777         clock_t expires;
3778         unsigned long timeout;
3779         bool was_managetempaddr;
3780         bool had_prefixroute;
3781
3782         ASSERT_RTNL();
3783
3784         if (!valid_lft || (prefered_lft > valid_lft))
3785                 return -EINVAL;
3786
3787         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3788             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3789                 return -EINVAL;
3790
3791         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3792         if (addrconf_finite_timeout(timeout)) {
3793                 expires = jiffies_to_clock_t(timeout * HZ);
3794                 valid_lft = timeout;
3795                 flags = RTF_EXPIRES;
3796         } else {
3797                 expires = 0;
3798                 flags = 0;
3799                 ifa_flags |= IFA_F_PERMANENT;
3800         }
3801
3802         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3803         if (addrconf_finite_timeout(timeout)) {
3804                 if (timeout == 0)
3805                         ifa_flags |= IFA_F_DEPRECATED;
3806                 prefered_lft = timeout;
3807         }
3808
3809         spin_lock_bh(&ifp->lock);
3810         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3811         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3812                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
3813         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3814                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3815                         IFA_F_NOPREFIXROUTE);
3816         ifp->flags |= ifa_flags;
3817         ifp->tstamp = jiffies;
3818         ifp->valid_lft = valid_lft;
3819         ifp->prefered_lft = prefered_lft;
3820
3821         spin_unlock_bh(&ifp->lock);
3822         if (!(ifp->flags&IFA_F_TENTATIVE))
3823                 ipv6_ifa_notify(0, ifp);
3824
3825         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3826                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3827                                       expires, flags);
3828         } else if (had_prefixroute) {
3829                 enum cleanup_prefix_rt_t action;
3830                 unsigned long rt_expires;
3831
3832                 write_lock_bh(&ifp->idev->lock);
3833                 action = check_cleanup_prefix_route(ifp, &rt_expires);
3834                 write_unlock_bh(&ifp->idev->lock);
3835
3836                 if (action != CLEANUP_PREFIX_RT_NOP) {
3837                         cleanup_prefix_route(ifp, rt_expires,
3838                                 action == CLEANUP_PREFIX_RT_DEL);
3839                 }
3840         }
3841
3842         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3843                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3844                         valid_lft = prefered_lft = 0;
3845                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3846                                  !was_managetempaddr, jiffies);
3847         }
3848
3849         addrconf_verify_rtnl();
3850
3851         return 0;
3852 }
3853
3854 static int
3855 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3856 {
3857         struct net *net = sock_net(skb->sk);
3858         struct ifaddrmsg *ifm;
3859         struct nlattr *tb[IFA_MAX+1];
3860         struct in6_addr *pfx, *peer_pfx;
3861         struct inet6_ifaddr *ifa;
3862         struct net_device *dev;
3863         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3864         u32 ifa_flags;
3865         int err;
3866
3867         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3868         if (err < 0)
3869                 return err;
3870
3871         ifm = nlmsg_data(nlh);
3872         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3873         if (pfx == NULL)
3874                 return -EINVAL;
3875
3876         if (tb[IFA_CACHEINFO]) {
3877                 struct ifa_cacheinfo *ci;
3878
3879                 ci = nla_data(tb[IFA_CACHEINFO]);
3880                 valid_lft = ci->ifa_valid;
3881                 preferred_lft = ci->ifa_prefered;
3882         } else {
3883                 preferred_lft = INFINITY_LIFE_TIME;
3884                 valid_lft = INFINITY_LIFE_TIME;
3885         }
3886
3887         dev =  __dev_get_by_index(net, ifm->ifa_index);
3888         if (dev == NULL)
3889                 return -ENODEV;
3890
3891         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3892
3893         /* We ignore other flags so far. */
3894         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3895                      IFA_F_NOPREFIXROUTE;
3896
3897         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3898         if (ifa == NULL) {
3899                 /*
3900                  * It would be best to check for !NLM_F_CREATE here but
3901                  * userspace already relies on not having to provide this.
3902                  */
3903                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3904                                       ifm->ifa_prefixlen, ifa_flags,
3905                                       preferred_lft, valid_lft);
3906         }
3907
3908         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3909             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3910                 err = -EEXIST;
3911         else
3912                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3913
3914         in6_ifa_put(ifa);
3915
3916         return err;
3917 }
3918
3919 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3920                           u8 scope, int ifindex)
3921 {
3922         struct ifaddrmsg *ifm;
3923
3924         ifm = nlmsg_data(nlh);
3925         ifm->ifa_family = AF_INET6;
3926         ifm->ifa_prefixlen = prefixlen;
3927         ifm->ifa_flags = flags;
3928         ifm->ifa_scope = scope;
3929         ifm->ifa_index = ifindex;
3930 }
3931
3932 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3933                          unsigned long tstamp, u32 preferred, u32 valid)
3934 {
3935         struct ifa_cacheinfo ci;
3936
3937         ci.cstamp = cstamp_delta(cstamp);
3938         ci.tstamp = cstamp_delta(tstamp);
3939         ci.ifa_prefered = preferred;
3940         ci.ifa_valid = valid;
3941
3942         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3943 }
3944
3945 static inline int rt_scope(int ifa_scope)
3946 {
3947         if (ifa_scope & IFA_HOST)
3948                 return RT_SCOPE_HOST;
3949         else if (ifa_scope & IFA_LINK)
3950                 return RT_SCOPE_LINK;
3951         else if (ifa_scope & IFA_SITE)
3952                 return RT_SCOPE_SITE;
3953         else
3954                 return RT_SCOPE_UNIVERSE;
3955 }
3956
3957 static inline int inet6_ifaddr_msgsize(void)
3958 {
3959         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3960                + nla_total_size(16) /* IFA_LOCAL */
3961                + nla_total_size(16) /* IFA_ADDRESS */
3962                + nla_total_size(sizeof(struct ifa_cacheinfo))
3963                + nla_total_size(4)  /* IFA_FLAGS */;
3964 }
3965
3966 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3967                              u32 portid, u32 seq, int event, unsigned int flags)
3968 {
3969         struct nlmsghdr  *nlh;
3970         u32 preferred, valid;
3971
3972         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3973         if (nlh == NULL)
3974                 return -EMSGSIZE;
3975
3976         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3977                       ifa->idev->dev->ifindex);
3978
3979         if (!((ifa->flags&IFA_F_PERMANENT) &&
3980               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3981                 preferred = ifa->prefered_lft;
3982                 valid = ifa->valid_lft;
3983                 if (preferred != INFINITY_LIFE_TIME) {
3984                         long tval = (jiffies - ifa->tstamp)/HZ;
3985                         if (preferred > tval)
3986                                 preferred -= tval;
3987                         else
3988                                 preferred = 0;
3989                         if (valid != INFINITY_LIFE_TIME) {
3990                                 if (valid > tval)
3991                                         valid -= tval;
3992                                 else
3993                                         valid = 0;
3994                         }
3995                 }
3996         } else {
3997                 preferred = INFINITY_LIFE_TIME;
3998                 valid = INFINITY_LIFE_TIME;
3999         }
4000
4001         if (!ipv6_addr_any(&ifa->peer_addr)) {
4002                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
4003                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
4004                         goto error;
4005         } else
4006                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4007                         goto error;
4008
4009         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4010                 goto error;
4011
4012         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4013                 goto error;
4014
4015         return nlmsg_end(skb, nlh);
4016
4017 error:
4018         nlmsg_cancel(skb, nlh);
4019         return -EMSGSIZE;
4020 }
4021
4022 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4023                                 u32 portid, u32 seq, int event, u16 flags)
4024 {
4025         struct nlmsghdr  *nlh;
4026         u8 scope = RT_SCOPE_UNIVERSE;
4027         int ifindex = ifmca->idev->dev->ifindex;
4028
4029         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4030                 scope = RT_SCOPE_SITE;
4031
4032         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4033         if (nlh == NULL)
4034                 return -EMSGSIZE;
4035
4036         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4037         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4038             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4039                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4040                 nlmsg_cancel(skb, nlh);
4041                 return -EMSGSIZE;
4042         }
4043
4044         return nlmsg_end(skb, nlh);
4045 }
4046
4047 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4048                                 u32 portid, u32 seq, int event, unsigned int flags)
4049 {
4050         struct nlmsghdr  *nlh;
4051         u8 scope = RT_SCOPE_UNIVERSE;
4052         int ifindex = ifaca->aca_idev->dev->ifindex;
4053
4054         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4055                 scope = RT_SCOPE_SITE;
4056
4057         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4058         if (nlh == NULL)
4059                 return -EMSGSIZE;
4060
4061         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4062         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4063             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4064                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4065                 nlmsg_cancel(skb, nlh);
4066                 return -EMSGSIZE;
4067         }
4068
4069         return nlmsg_end(skb, nlh);
4070 }
4071
4072 enum addr_type_t {
4073         UNICAST_ADDR,
4074         MULTICAST_ADDR,
4075         ANYCAST_ADDR,
4076 };
4077
4078 /* called with rcu_read_lock() */
4079 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4080                           struct netlink_callback *cb, enum addr_type_t type,
4081                           int s_ip_idx, int *p_ip_idx)
4082 {
4083         struct ifmcaddr6 *ifmca;
4084         struct ifacaddr6 *ifaca;
4085         int err = 1;
4086         int ip_idx = *p_ip_idx;
4087
4088         read_lock_bh(&idev->lock);
4089         switch (type) {
4090         case UNICAST_ADDR: {
4091                 struct inet6_ifaddr *ifa;
4092
4093                 /* unicast address incl. temp addr */
4094                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4095                         if (++ip_idx < s_ip_idx)
4096                                 continue;
4097                         err = inet6_fill_ifaddr(skb, ifa,
4098                                                 NETLINK_CB(cb->skb).portid,
4099                                                 cb->nlh->nlmsg_seq,
4100                                                 RTM_NEWADDR,
4101                                                 NLM_F_MULTI);
4102                         if (err <= 0)
4103                                 break;
4104                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4105                 }
4106                 break;
4107         }
4108         case MULTICAST_ADDR:
4109                 /* multicast address */
4110                 for (ifmca = idev->mc_list; ifmca;
4111                      ifmca = ifmca->next, ip_idx++) {
4112                         if (ip_idx < s_ip_idx)
4113                                 continue;
4114                         err = inet6_fill_ifmcaddr(skb, ifmca,
4115                                                   NETLINK_CB(cb->skb).portid,
4116                                                   cb->nlh->nlmsg_seq,
4117                                                   RTM_GETMULTICAST,
4118                                                   NLM_F_MULTI);
4119                         if (err <= 0)
4120                                 break;
4121                 }
4122                 break;
4123         case ANYCAST_ADDR:
4124                 /* anycast address */
4125                 for (ifaca = idev->ac_list; ifaca;
4126                      ifaca = ifaca->aca_next, ip_idx++) {
4127                         if (ip_idx < s_ip_idx)
4128                                 continue;
4129                         err = inet6_fill_ifacaddr(skb, ifaca,
4130                                                   NETLINK_CB(cb->skb).portid,
4131                                                   cb->nlh->nlmsg_seq,
4132                                                   RTM_GETANYCAST,
4133                                                   NLM_F_MULTI);
4134                         if (err <= 0)
4135                                 break;
4136                 }
4137                 break;
4138         default:
4139                 break;
4140         }
4141         read_unlock_bh(&idev->lock);
4142         *p_ip_idx = ip_idx;
4143         return err;
4144 }
4145
4146 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4147                            enum addr_type_t type)
4148 {
4149         struct net *net = sock_net(skb->sk);
4150         int h, s_h;
4151         int idx, ip_idx;
4152         int s_idx, s_ip_idx;
4153         struct net_device *dev;
4154         struct inet6_dev *idev;
4155         struct hlist_head *head;
4156
4157         s_h = cb->args[0];
4158         s_idx = idx = cb->args[1];
4159         s_ip_idx = ip_idx = cb->args[2];
4160
4161         rcu_read_lock();
4162         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4163         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4164                 idx = 0;
4165                 head = &net->dev_index_head[h];
4166                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4167                         if (idx < s_idx)
4168                                 goto cont;
4169                         if (h > s_h || idx > s_idx)
4170                                 s_ip_idx = 0;
4171                         ip_idx = 0;
4172                         idev = __in6_dev_get(dev);
4173                         if (!idev)
4174                                 goto cont;
4175
4176                         if (in6_dump_addrs(idev, skb, cb, type,
4177                                            s_ip_idx, &ip_idx) <= 0)
4178                                 goto done;
4179 cont:
4180                         idx++;
4181                 }
4182         }
4183 done:
4184         rcu_read_unlock();
4185         cb->args[0] = h;
4186         cb->args[1] = idx;
4187         cb->args[2] = ip_idx;
4188
4189         return skb->len;
4190 }
4191
4192 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4193 {
4194         enum addr_type_t type = UNICAST_ADDR;
4195
4196         return inet6_dump_addr(skb, cb, type);
4197 }
4198
4199 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4200 {
4201         enum addr_type_t type = MULTICAST_ADDR;
4202
4203         return inet6_dump_addr(skb, cb, type);
4204 }
4205
4206
4207 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4208 {
4209         enum addr_type_t type = ANYCAST_ADDR;
4210
4211         return inet6_dump_addr(skb, cb, type);
4212 }
4213
4214 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4215 {
4216         struct net *net = sock_net(in_skb->sk);
4217         struct ifaddrmsg *ifm;
4218         struct nlattr *tb[IFA_MAX+1];
4219         struct in6_addr *addr = NULL, *peer;
4220         struct net_device *dev = NULL;
4221         struct inet6_ifaddr *ifa;
4222         struct sk_buff *skb;
4223         int err;
4224
4225         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4226         if (err < 0)
4227                 goto errout;
4228
4229         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4230         if (addr == NULL) {
4231                 err = -EINVAL;
4232                 goto errout;
4233         }
4234
4235         ifm = nlmsg_data(nlh);
4236         if (ifm->ifa_index)
4237                 dev = __dev_get_by_index(net, ifm->ifa_index);
4238
4239         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4240         if (!ifa) {
4241                 err = -EADDRNOTAVAIL;
4242                 goto errout;
4243         }
4244
4245         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4246         if (!skb) {
4247                 err = -ENOBUFS;
4248                 goto errout_ifa;
4249         }
4250
4251         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4252                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4253         if (err < 0) {
4254                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4255                 WARN_ON(err == -EMSGSIZE);
4256                 kfree_skb(skb);
4257                 goto errout_ifa;
4258         }
4259         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4260 errout_ifa:
4261         in6_ifa_put(ifa);
4262 errout:
4263         return err;
4264 }
4265
4266 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4267 {
4268         struct sk_buff *skb;
4269         struct net *net = dev_net(ifa->idev->dev);
4270         int err = -ENOBUFS;
4271
4272         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4273         if (skb == NULL)
4274                 goto errout;
4275
4276         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4277         if (err < 0) {
4278                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4279                 WARN_ON(err == -EMSGSIZE);
4280                 kfree_skb(skb);
4281                 goto errout;
4282         }
4283         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4284         return;
4285 errout:
4286         if (err < 0)
4287                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4288 }
4289
4290 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4291                                 __s32 *array, int bytes)
4292 {
4293         BUG_ON(bytes < (DEVCONF_MAX * 4));
4294
4295         memset(array, 0, bytes);
4296         array[DEVCONF_FORWARDING] = cnf->forwarding;
4297         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4298         array[DEVCONF_MTU6] = cnf->mtu6;
4299         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4300         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4301         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4302         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4303         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4304         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4305                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4306         array[DEVCONF_RTR_SOLICIT_DELAY] =
4307                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4308         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4309         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4310                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4311         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4312                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4313         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4314         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4315         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4316         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4317         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4318         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4319         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4320         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4321 #ifdef CONFIG_IPV6_ROUTER_PREF
4322         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4323         array[DEVCONF_RTR_PROBE_INTERVAL] =
4324                 jiffies_to_msecs(cnf->rtr_probe_interval);
4325 #ifdef CONFIG_IPV6_ROUTE_INFO
4326         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4327 #endif
4328 #endif
4329         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4330         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4331 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4332         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4333 #endif
4334 #ifdef CONFIG_IPV6_MROUTE
4335         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4336 #endif
4337         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4338         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4339         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4340         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4341         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4342         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4343 }
4344
4345 static inline size_t inet6_ifla6_size(void)
4346 {
4347         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4348              + nla_total_size(sizeof(struct ifla_cacheinfo))
4349              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4350              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4351              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4352              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4353 }
4354
4355 static inline size_t inet6_if_nlmsg_size(void)
4356 {
4357         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4358                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4359                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4360                + nla_total_size(4) /* IFLA_MTU */
4361                + nla_total_size(4) /* IFLA_LINK */
4362                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4363 }
4364
4365 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4366                                       int items, int bytes)
4367 {
4368         int i;
4369         int pad = bytes - sizeof(u64) * items;
4370         BUG_ON(pad < 0);
4371
4372         /* Use put_unaligned() because stats may not be aligned for u64. */
4373         put_unaligned(items, &stats[0]);
4374         for (i = 1; i < items; i++)
4375                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4376
4377         memset(&stats[items], 0, pad);
4378 }
4379
4380 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4381                                       int items, int bytes, size_t syncpoff)
4382 {
4383         int i;
4384         int pad = bytes - sizeof(u64) * items;
4385         BUG_ON(pad < 0);
4386
4387         /* Use put_unaligned() because stats may not be aligned for u64. */
4388         put_unaligned(items, &stats[0]);
4389         for (i = 1; i < items; i++)
4390                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4391
4392         memset(&stats[items], 0, pad);
4393 }
4394
4395 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4396                              int bytes)
4397 {
4398         switch (attrtype) {
4399         case IFLA_INET6_STATS:
4400                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4401                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4402                 break;
4403         case IFLA_INET6_ICMP6STATS:
4404                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4405                 break;
4406         }
4407 }
4408
4409 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4410 {
4411         struct nlattr *nla;
4412         struct ifla_cacheinfo ci;
4413
4414         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4415                 goto nla_put_failure;
4416         ci.max_reasm_len = IPV6_MAXPLEN;
4417         ci.tstamp = cstamp_delta(idev->tstamp);
4418         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4419         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4420         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4421                 goto nla_put_failure;
4422         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4423         if (nla == NULL)
4424                 goto nla_put_failure;
4425         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4426
4427         /* XXX - MC not implemented */
4428
4429         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4430         if (nla == NULL)
4431                 goto nla_put_failure;
4432         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4433
4434         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4435         if (nla == NULL)
4436                 goto nla_put_failure;
4437         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4438
4439         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4440         if (nla == NULL)
4441                 goto nla_put_failure;
4442
4443         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4444                 goto nla_put_failure;
4445
4446         read_lock_bh(&idev->lock);
4447         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4448         read_unlock_bh(&idev->lock);
4449
4450         return 0;
4451
4452 nla_put_failure:
4453         return -EMSGSIZE;
4454 }
4455
4456 static size_t inet6_get_link_af_size(const struct net_device *dev)
4457 {
4458         if (!__in6_dev_get(dev))
4459                 return 0;
4460
4461         return inet6_ifla6_size();
4462 }
4463
4464 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4465 {
4466         struct inet6_dev *idev = __in6_dev_get(dev);
4467
4468         if (!idev)
4469                 return -ENODATA;
4470
4471         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4472                 return -EMSGSIZE;
4473
4474         return 0;
4475 }
4476
4477 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4478 {
4479         struct inet6_ifaddr *ifp;
4480         struct net_device *dev = idev->dev;
4481         bool update_rs = false;
4482         struct in6_addr ll_addr;
4483
4484         ASSERT_RTNL();
4485
4486         if (token == NULL)
4487                 return -EINVAL;
4488         if (ipv6_addr_any(token))
4489                 return -EINVAL;
4490         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4491                 return -EINVAL;
4492         if (!ipv6_accept_ra(idev))
4493                 return -EINVAL;
4494         if (idev->cnf.rtr_solicits <= 0)
4495                 return -EINVAL;
4496
4497         write_lock_bh(&idev->lock);
4498
4499         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4500         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4501
4502         write_unlock_bh(&idev->lock);
4503
4504         if (!idev->dead && (idev->if_flags & IF_READY) &&
4505             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4506                              IFA_F_OPTIMISTIC)) {
4507
4508                 /* If we're not ready, then normal ifup will take care
4509                  * of this. Otherwise, we need to request our rs here.
4510                  */
4511                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4512                 update_rs = true;
4513         }
4514
4515         write_lock_bh(&idev->lock);
4516
4517         if (update_rs) {
4518                 idev->if_flags |= IF_RS_SENT;
4519                 idev->rs_probes = 1;
4520                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4521         }
4522
4523         /* Well, that's kinda nasty ... */
4524         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4525                 spin_lock(&ifp->lock);
4526                 if (ifp->tokenized) {
4527                         ifp->valid_lft = 0;
4528                         ifp->prefered_lft = 0;
4529                 }
4530                 spin_unlock(&ifp->lock);
4531         }
4532
4533         write_unlock_bh(&idev->lock);
4534         addrconf_verify_rtnl();
4535         return 0;
4536 }
4537
4538 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4539 {
4540         int err = -EINVAL;
4541         struct inet6_dev *idev = __in6_dev_get(dev);
4542         struct nlattr *tb[IFLA_INET6_MAX + 1];
4543
4544         if (!idev)
4545                 return -EAFNOSUPPORT;
4546
4547         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4548                 BUG();
4549
4550         if (tb[IFLA_INET6_TOKEN]) {
4551                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4552                 if (err)
4553                         return err;
4554         }
4555
4556         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4557                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4558
4559                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4560                     mode != IN6_ADDR_GEN_MODE_NONE)
4561                         return -EINVAL;
4562                 idev->addr_gen_mode = mode;
4563                 err = 0;
4564         }
4565
4566         return err;
4567 }
4568
4569 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4570                              u32 portid, u32 seq, int event, unsigned int flags)
4571 {
4572         struct net_device *dev = idev->dev;
4573         struct ifinfomsg *hdr;
4574         struct nlmsghdr *nlh;
4575         void *protoinfo;
4576
4577         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4578         if (nlh == NULL)
4579                 return -EMSGSIZE;
4580
4581         hdr = nlmsg_data(nlh);
4582         hdr->ifi_family = AF_INET6;
4583         hdr->__ifi_pad = 0;
4584         hdr->ifi_type = dev->type;
4585         hdr->ifi_index = dev->ifindex;
4586         hdr->ifi_flags = dev_get_flags(dev);
4587         hdr->ifi_change = 0;
4588
4589         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4590             (dev->addr_len &&
4591              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4592             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4593             (dev->ifindex != dev->iflink &&
4594              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4595                 goto nla_put_failure;
4596         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4597         if (protoinfo == NULL)
4598                 goto nla_put_failure;
4599
4600         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4601                 goto nla_put_failure;
4602
4603         nla_nest_end(skb, protoinfo);
4604         return nlmsg_end(skb, nlh);
4605
4606 nla_put_failure:
4607         nlmsg_cancel(skb, nlh);
4608         return -EMSGSIZE;
4609 }
4610
4611 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4612 {
4613         struct net *net = sock_net(skb->sk);
4614         int h, s_h;
4615         int idx = 0, s_idx;
4616         struct net_device *dev;
4617         struct inet6_dev *idev;
4618         struct hlist_head *head;
4619
4620         s_h = cb->args[0];
4621         s_idx = cb->args[1];
4622
4623         rcu_read_lock();
4624         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4625                 idx = 0;
4626                 head = &net->dev_index_head[h];
4627                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4628                         if (idx < s_idx)
4629                                 goto cont;
4630                         idev = __in6_dev_get(dev);
4631                         if (!idev)
4632                                 goto cont;
4633                         if (inet6_fill_ifinfo(skb, idev,
4634                                               NETLINK_CB(cb->skb).portid,
4635                                               cb->nlh->nlmsg_seq,
4636                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4637                                 goto out;
4638 cont:
4639                         idx++;
4640                 }
4641         }
4642 out:
4643         rcu_read_unlock();
4644         cb->args[1] = idx;
4645         cb->args[0] = h;
4646
4647         return skb->len;
4648 }
4649
4650 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4651 {
4652         struct sk_buff *skb;
4653         struct net *net = dev_net(idev->dev);
4654         int err = -ENOBUFS;
4655
4656         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4657         if (skb == NULL)
4658                 goto errout;
4659
4660         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4661         if (err < 0) {
4662                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4663                 WARN_ON(err == -EMSGSIZE);
4664                 kfree_skb(skb);
4665                 goto errout;
4666         }
4667         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4668         return;
4669 errout:
4670         if (err < 0)
4671                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4672 }
4673
4674 static inline size_t inet6_prefix_nlmsg_size(void)
4675 {
4676         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4677                + nla_total_size(sizeof(struct in6_addr))
4678                + nla_total_size(sizeof(struct prefix_cacheinfo));
4679 }
4680
4681 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4682                              struct prefix_info *pinfo, u32 portid, u32 seq,
4683                              int event, unsigned int flags)
4684 {
4685         struct prefixmsg *pmsg;
4686         struct nlmsghdr *nlh;
4687         struct prefix_cacheinfo ci;
4688
4689         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4690         if (nlh == NULL)
4691                 return -EMSGSIZE;
4692
4693         pmsg = nlmsg_data(nlh);
4694         pmsg->prefix_family = AF_INET6;
4695         pmsg->prefix_pad1 = 0;
4696         pmsg->prefix_pad2 = 0;
4697         pmsg->prefix_ifindex = idev->dev->ifindex;
4698         pmsg->prefix_len = pinfo->prefix_len;
4699         pmsg->prefix_type = pinfo->type;
4700         pmsg->prefix_pad3 = 0;
4701         pmsg->prefix_flags = 0;
4702         if (pinfo->onlink)
4703                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4704         if (pinfo->autoconf)
4705                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4706
4707         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4708                 goto nla_put_failure;
4709         ci.preferred_time = ntohl(pinfo->prefered);
4710         ci.valid_time = ntohl(pinfo->valid);
4711         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4712                 goto nla_put_failure;
4713         return nlmsg_end(skb, nlh);
4714
4715 nla_put_failure:
4716         nlmsg_cancel(skb, nlh);
4717         return -EMSGSIZE;
4718 }
4719
4720 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4721                          struct prefix_info *pinfo)
4722 {
4723         struct sk_buff *skb;
4724         struct net *net = dev_net(idev->dev);
4725         int err = -ENOBUFS;
4726
4727         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4728         if (skb == NULL)
4729                 goto errout;
4730
4731         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4732         if (err < 0) {
4733                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4734                 WARN_ON(err == -EMSGSIZE);
4735                 kfree_skb(skb);
4736                 goto errout;
4737         }
4738         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4739         return;
4740 errout:
4741         if (err < 0)
4742                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4743 }
4744
4745 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4746 {
4747         struct net *net = dev_net(ifp->idev->dev);
4748
4749         if (event)
4750                 ASSERT_RTNL();
4751
4752         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4753
4754         switch (event) {
4755         case RTM_NEWADDR:
4756                 /*
4757                  * If the address was optimistic
4758                  * we inserted the route at the start of
4759                  * our DAD process, so we don't need
4760                  * to do it again
4761                  */
4762                 if (!(ifp->rt->rt6i_node))
4763                         ip6_ins_rt(ifp->rt);
4764                 if (ifp->idev->cnf.forwarding)
4765                         addrconf_join_anycast(ifp);
4766                 if (!ipv6_addr_any(&ifp->peer_addr))
4767                         addrconf_prefix_route(&ifp->peer_addr, 128,
4768                                               ifp->idev->dev, 0, 0);
4769                 break;
4770         case RTM_DELADDR:
4771                 if (ifp->idev->cnf.forwarding)
4772                         addrconf_leave_anycast(ifp);
4773                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4774                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4775                         struct rt6_info *rt;
4776                         struct net_device *dev = ifp->idev->dev;
4777
4778                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4779                                         dev->ifindex, 1);
4780                         if (rt) {
4781                                 dst_hold(&rt->dst);
4782                                 if (ip6_del_rt(rt))
4783                                         dst_free(&rt->dst);
4784                         }
4785                 }
4786                 dst_hold(&ifp->rt->dst);
4787
4788                 if (ip6_del_rt(ifp->rt))
4789                         dst_free(&ifp->rt->dst);
4790                 break;
4791         }
4792         atomic_inc(&net->ipv6.dev_addr_genid);
4793         rt_genid_bump_ipv6(net);
4794 }
4795
4796 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4797 {
4798         rcu_read_lock_bh();
4799         if (likely(ifp->idev->dead == 0))
4800                 __ipv6_ifa_notify(event, ifp);
4801         rcu_read_unlock_bh();
4802 }
4803
4804 #ifdef CONFIG_SYSCTL
4805
4806 static
4807 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4808                            void __user *buffer, size_t *lenp, loff_t *ppos)
4809 {
4810         int *valp = ctl->data;
4811         int val = *valp;
4812         loff_t pos = *ppos;
4813         struct ctl_table lctl;
4814         int ret;
4815
4816         /*
4817          * ctl->data points to idev->cnf.forwarding, we should
4818          * not modify it until we get the rtnl lock.
4819          */
4820         lctl = *ctl;
4821         lctl.data = &val;
4822
4823         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4824
4825         if (write)
4826                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4827         if (ret)
4828                 *ppos = pos;
4829         return ret;
4830 }
4831
4832 static void dev_disable_change(struct inet6_dev *idev)
4833 {
4834         struct netdev_notifier_info info;
4835
4836         if (!idev || !idev->dev)
4837                 return;
4838
4839         netdev_notifier_info_init(&info, idev->dev);
4840         if (idev->cnf.disable_ipv6)
4841                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4842         else
4843                 addrconf_notify(NULL, NETDEV_UP, &info);
4844 }
4845
4846 static void addrconf_disable_change(struct net *net, __s32 newf)
4847 {
4848         struct net_device *dev;
4849         struct inet6_dev *idev;
4850
4851         rcu_read_lock();
4852         for_each_netdev_rcu(net, dev) {
4853                 idev = __in6_dev_get(dev);
4854                 if (idev) {
4855                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4856                         idev->cnf.disable_ipv6 = newf;
4857                         if (changed)
4858                                 dev_disable_change(idev);
4859                 }
4860         }
4861         rcu_read_unlock();
4862 }
4863
4864 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4865 {
4866         struct net *net;
4867         int old;
4868
4869         if (!rtnl_trylock())
4870                 return restart_syscall();
4871
4872         net = (struct net *)table->extra2;
4873         old = *p;
4874         *p = newf;
4875
4876         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4877                 rtnl_unlock();
4878                 return 0;
4879         }
4880
4881         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4882                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4883                 addrconf_disable_change(net, newf);
4884         } else if ((!newf) ^ (!old))
4885                 dev_disable_change((struct inet6_dev *)table->extra1);
4886
4887         rtnl_unlock();
4888         return 0;
4889 }
4890
4891 static
4892 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4893                             void __user *buffer, size_t *lenp, loff_t *ppos)
4894 {
4895         int *valp = ctl->data;
4896         int val = *valp;
4897         loff_t pos = *ppos;
4898         struct ctl_table lctl;
4899         int ret;
4900
4901         /*
4902          * ctl->data points to idev->cnf.disable_ipv6, we should
4903          * not modify it until we get the rtnl lock.
4904          */
4905         lctl = *ctl;
4906         lctl.data = &val;
4907
4908         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4909
4910         if (write)
4911                 ret = addrconf_disable_ipv6(ctl, valp, val);
4912         if (ret)
4913                 *ppos = pos;
4914         return ret;
4915 }
4916
4917 static
4918 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4919                               void __user *buffer, size_t *lenp, loff_t *ppos)
4920 {
4921         int *valp = ctl->data;
4922         int ret;
4923         int old, new;
4924
4925         old = *valp;
4926         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4927         new = *valp;
4928
4929         if (write && old != new) {
4930                 struct net *net = ctl->extra2;
4931
4932                 if (!rtnl_trylock())
4933                         return restart_syscall();
4934
4935                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4936                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4937                                                      NETCONFA_IFINDEX_DEFAULT,
4938                                                      net->ipv6.devconf_dflt);
4939                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4940                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4941                                                      NETCONFA_IFINDEX_ALL,
4942                                                      net->ipv6.devconf_all);
4943                 else {
4944                         struct inet6_dev *idev = ctl->extra1;
4945
4946                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4947                                                      idev->dev->ifindex,
4948                                                      &idev->cnf);
4949                 }
4950                 rtnl_unlock();
4951         }
4952
4953         return ret;
4954 }
4955
4956
4957 static struct addrconf_sysctl_table
4958 {
4959         struct ctl_table_header *sysctl_header;
4960         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4961 } addrconf_sysctl __read_mostly = {
4962         .sysctl_header = NULL,
4963         .addrconf_vars = {
4964                 {
4965                         .procname       = "forwarding",
4966                         .data           = &ipv6_devconf.forwarding,
4967                         .maxlen         = sizeof(int),
4968                         .mode           = 0644,
4969                         .proc_handler   = addrconf_sysctl_forward,
4970                 },
4971                 {
4972                         .procname       = "hop_limit",
4973                         .data           = &ipv6_devconf.hop_limit,
4974                         .maxlen         = sizeof(int),
4975                         .mode           = 0644,
4976                         .proc_handler   = proc_dointvec,
4977                 },
4978                 {
4979                         .procname       = "mtu",
4980                         .data           = &ipv6_devconf.mtu6,
4981                         .maxlen         = sizeof(int),
4982                         .mode           = 0644,
4983                         .proc_handler   = proc_dointvec,
4984                 },
4985                 {
4986                         .procname       = "accept_ra",
4987                         .data           = &ipv6_devconf.accept_ra,
4988                         .maxlen         = sizeof(int),
4989                         .mode           = 0644,
4990                         .proc_handler   = proc_dointvec,
4991                 },
4992                 {
4993                         .procname       = "accept_redirects",
4994                         .data           = &ipv6_devconf.accept_redirects,
4995                         .maxlen         = sizeof(int),
4996                         .mode           = 0644,
4997                         .proc_handler   = proc_dointvec,
4998                 },
4999                 {
5000                         .procname       = "autoconf",
5001                         .data           = &ipv6_devconf.autoconf,
5002                         .maxlen         = sizeof(int),
5003                         .mode           = 0644,
5004                         .proc_handler   = proc_dointvec,
5005                 },
5006                 {
5007                         .procname       = "dad_transmits",
5008                         .data           = &ipv6_devconf.dad_transmits,
5009                         .maxlen         = sizeof(int),
5010                         .mode           = 0644,
5011                         .proc_handler   = proc_dointvec,
5012                 },
5013                 {
5014                         .procname       = "router_solicitations",
5015                         .data           = &ipv6_devconf.rtr_solicits,
5016                         .maxlen         = sizeof(int),
5017                         .mode           = 0644,
5018                         .proc_handler   = proc_dointvec,
5019                 },
5020                 {
5021                         .procname       = "router_solicitation_interval",
5022                         .data           = &ipv6_devconf.rtr_solicit_interval,
5023                         .maxlen         = sizeof(int),
5024                         .mode           = 0644,
5025                         .proc_handler   = proc_dointvec_jiffies,
5026                 },
5027                 {
5028                         .procname       = "router_solicitation_delay",
5029                         .data           = &ipv6_devconf.rtr_solicit_delay,
5030                         .maxlen         = sizeof(int),
5031                         .mode           = 0644,
5032                         .proc_handler   = proc_dointvec_jiffies,
5033                 },
5034                 {
5035                         .procname       = "force_mld_version",
5036                         .data           = &ipv6_devconf.force_mld_version,
5037                         .maxlen         = sizeof(int),
5038                         .mode           = 0644,
5039                         .proc_handler   = proc_dointvec,
5040                 },
5041                 {
5042                         .procname       = "mldv1_unsolicited_report_interval",
5043                         .data           =
5044                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5045                         .maxlen         = sizeof(int),
5046                         .mode           = 0644,
5047                         .proc_handler   = proc_dointvec_ms_jiffies,
5048                 },
5049                 {
5050                         .procname       = "mldv2_unsolicited_report_interval",
5051                         .data           =
5052                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5053                         .maxlen         = sizeof(int),
5054                         .mode           = 0644,
5055                         .proc_handler   = proc_dointvec_ms_jiffies,
5056                 },
5057                 {
5058                         .procname       = "use_tempaddr",
5059                         .data           = &ipv6_devconf.use_tempaddr,
5060                         .maxlen         = sizeof(int),
5061                         .mode           = 0644,
5062                         .proc_handler   = proc_dointvec,
5063                 },
5064                 {
5065                         .procname       = "temp_valid_lft",
5066                         .data           = &ipv6_devconf.temp_valid_lft,
5067                         .maxlen         = sizeof(int),
5068                         .mode           = 0644,
5069                         .proc_handler   = proc_dointvec,
5070                 },
5071                 {
5072                         .procname       = "temp_prefered_lft",
5073                         .data           = &ipv6_devconf.temp_prefered_lft,
5074                         .maxlen         = sizeof(int),
5075                         .mode           = 0644,
5076                         .proc_handler   = proc_dointvec,
5077                 },
5078                 {
5079                         .procname       = "regen_max_retry",
5080                         .data           = &ipv6_devconf.regen_max_retry,
5081                         .maxlen         = sizeof(int),
5082                         .mode           = 0644,
5083                         .proc_handler   = proc_dointvec,
5084                 },
5085                 {
5086                         .procname       = "max_desync_factor",
5087                         .data           = &ipv6_devconf.max_desync_factor,
5088                         .maxlen         = sizeof(int),
5089                         .mode           = 0644,
5090                         .proc_handler   = proc_dointvec,
5091                 },
5092                 {
5093                         .procname       = "max_addresses",
5094                         .data           = &ipv6_devconf.max_addresses,
5095                         .maxlen         = sizeof(int),
5096                         .mode           = 0644,
5097                         .proc_handler   = proc_dointvec,
5098                 },
5099                 {
5100                         .procname       = "accept_ra_defrtr",
5101                         .data           = &ipv6_devconf.accept_ra_defrtr,
5102                         .maxlen         = sizeof(int),
5103                         .mode           = 0644,
5104                         .proc_handler   = proc_dointvec,
5105                 },
5106                 {
5107                         .procname       = "accept_ra_pinfo",
5108                         .data           = &ipv6_devconf.accept_ra_pinfo,
5109                         .maxlen         = sizeof(int),
5110                         .mode           = 0644,
5111                         .proc_handler   = proc_dointvec,
5112                 },
5113 #ifdef CONFIG_IPV6_ROUTER_PREF
5114                 {
5115                         .procname       = "accept_ra_rtr_pref",
5116                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5117                         .maxlen         = sizeof(int),
5118                         .mode           = 0644,
5119                         .proc_handler   = proc_dointvec,
5120                 },
5121                 {
5122                         .procname       = "router_probe_interval",
5123                         .data           = &ipv6_devconf.rtr_probe_interval,
5124                         .maxlen         = sizeof(int),
5125                         .mode           = 0644,
5126                         .proc_handler   = proc_dointvec_jiffies,
5127                 },
5128 #ifdef CONFIG_IPV6_ROUTE_INFO
5129                 {
5130                         .procname       = "accept_ra_rt_info_max_plen",
5131                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5132                         .maxlen         = sizeof(int),
5133                         .mode           = 0644,
5134                         .proc_handler   = proc_dointvec,
5135                 },
5136 #endif
5137 #endif
5138                 {
5139                         .procname       = "proxy_ndp",
5140                         .data           = &ipv6_devconf.proxy_ndp,
5141                         .maxlen         = sizeof(int),
5142                         .mode           = 0644,
5143                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5144                 },
5145                 {
5146                         .procname       = "accept_source_route",
5147                         .data           = &ipv6_devconf.accept_source_route,
5148                         .maxlen         = sizeof(int),
5149                         .mode           = 0644,
5150                         .proc_handler   = proc_dointvec,
5151                 },
5152 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5153                 {
5154                         .procname       = "optimistic_dad",
5155                         .data           = &ipv6_devconf.optimistic_dad,
5156                         .maxlen         = sizeof(int),
5157                         .mode           = 0644,
5158                         .proc_handler   = proc_dointvec,
5159
5160                 },
5161 #endif
5162 #ifdef CONFIG_IPV6_MROUTE
5163                 {
5164                         .procname       = "mc_forwarding",
5165                         .data           = &ipv6_devconf.mc_forwarding,
5166                         .maxlen         = sizeof(int),
5167                         .mode           = 0444,
5168                         .proc_handler   = proc_dointvec,
5169                 },
5170 #endif
5171                 {
5172                         .procname       = "disable_ipv6",
5173                         .data           = &ipv6_devconf.disable_ipv6,
5174                         .maxlen         = sizeof(int),
5175                         .mode           = 0644,
5176                         .proc_handler   = addrconf_sysctl_disable,
5177                 },
5178                 {
5179                         .procname       = "accept_dad",
5180                         .data           = &ipv6_devconf.accept_dad,
5181                         .maxlen         = sizeof(int),
5182                         .mode           = 0644,
5183                         .proc_handler   = proc_dointvec,
5184                 },
5185                 {
5186                         .procname       = "force_tllao",
5187                         .data           = &ipv6_devconf.force_tllao,
5188                         .maxlen         = sizeof(int),
5189                         .mode           = 0644,
5190                         .proc_handler   = proc_dointvec
5191                 },
5192                 {
5193                         .procname       = "ndisc_notify",
5194                         .data           = &ipv6_devconf.ndisc_notify,
5195                         .maxlen         = sizeof(int),
5196                         .mode           = 0644,
5197                         .proc_handler   = proc_dointvec
5198                 },
5199                 {
5200                         .procname       = "suppress_frag_ndisc",
5201                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5202                         .maxlen         = sizeof(int),
5203                         .mode           = 0644,
5204                         .proc_handler   = proc_dointvec
5205                 },
5206                 {
5207                         .procname       = "accept_ra_from_local",
5208                         .data           = &ipv6_devconf.accept_ra_from_local,
5209                         .maxlen         = sizeof(int),
5210                         .mode           = 0644,
5211                         .proc_handler   = proc_dointvec,
5212                 },
5213                 {
5214                         /* sentinel */
5215                 }
5216         },
5217 };
5218
5219 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5220                 struct inet6_dev *idev, struct ipv6_devconf *p)
5221 {
5222         int i;
5223         struct addrconf_sysctl_table *t;
5224         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5225
5226         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5227         if (t == NULL)
5228                 goto out;
5229
5230         for (i = 0; t->addrconf_vars[i].data; i++) {
5231                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5232                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5233                 t->addrconf_vars[i].extra2 = net;
5234         }
5235
5236         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5237
5238         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5239         if (t->sysctl_header == NULL)
5240                 goto free;
5241
5242         p->sysctl = t;
5243         return 0;
5244
5245 free:
5246         kfree(t);
5247 out:
5248         return -ENOBUFS;
5249 }
5250
5251 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5252 {
5253         struct addrconf_sysctl_table *t;
5254
5255         if (p->sysctl == NULL)
5256                 return;
5257
5258         t = p->sysctl;
5259         p->sysctl = NULL;
5260         unregister_net_sysctl_table(t->sysctl_header);
5261         kfree(t);
5262 }
5263
5264 static int addrconf_sysctl_register(struct inet6_dev *idev)
5265 {
5266         int err;
5267
5268         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5269                 return -EINVAL;
5270
5271         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5272                                     &ndisc_ifinfo_sysctl_change);
5273         if (err)
5274                 return err;
5275         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5276                                          idev, &idev->cnf);
5277         if (err)
5278                 neigh_sysctl_unregister(idev->nd_parms);
5279
5280         return err;
5281 }
5282
5283 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5284 {
5285         __addrconf_sysctl_unregister(&idev->cnf);
5286         neigh_sysctl_unregister(idev->nd_parms);
5287 }
5288
5289
5290 #endif
5291
5292 static int __net_init addrconf_init_net(struct net *net)
5293 {
5294         int err = -ENOMEM;
5295         struct ipv6_devconf *all, *dflt;
5296
5297         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5298         if (all == NULL)
5299                 goto err_alloc_all;
5300
5301         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5302         if (dflt == NULL)
5303                 goto err_alloc_dflt;
5304
5305         /* these will be inherited by all namespaces */
5306         dflt->autoconf = ipv6_defaults.autoconf;
5307         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5308
5309         net->ipv6.devconf_all = all;
5310         net->ipv6.devconf_dflt = dflt;
5311
5312 #ifdef CONFIG_SYSCTL
5313         err = __addrconf_sysctl_register(net, "all", NULL, all);
5314         if (err < 0)
5315                 goto err_reg_all;
5316
5317         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5318         if (err < 0)
5319                 goto err_reg_dflt;
5320 #endif
5321         return 0;
5322
5323 #ifdef CONFIG_SYSCTL
5324 err_reg_dflt:
5325         __addrconf_sysctl_unregister(all);
5326 err_reg_all:
5327         kfree(dflt);
5328 #endif
5329 err_alloc_dflt:
5330         kfree(all);
5331 err_alloc_all:
5332         return err;
5333 }
5334
5335 static void __net_exit addrconf_exit_net(struct net *net)
5336 {
5337 #ifdef CONFIG_SYSCTL
5338         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5339         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5340 #endif
5341         if (!net_eq(net, &init_net)) {
5342                 kfree(net->ipv6.devconf_dflt);
5343                 kfree(net->ipv6.devconf_all);
5344         }
5345 }
5346
5347 static struct pernet_operations addrconf_ops = {
5348         .init = addrconf_init_net,
5349         .exit = addrconf_exit_net,
5350 };
5351
5352 static struct rtnl_af_ops inet6_ops = {
5353         .family           = AF_INET6,
5354         .fill_link_af     = inet6_fill_link_af,
5355         .get_link_af_size = inet6_get_link_af_size,
5356         .set_link_af      = inet6_set_link_af,
5357 };
5358
5359 /*
5360  *      Init / cleanup code
5361  */
5362
5363 int __init addrconf_init(void)
5364 {
5365         struct inet6_dev *idev;
5366         int i, err;
5367
5368         err = ipv6_addr_label_init();
5369         if (err < 0) {
5370                 pr_crit("%s: cannot initialize default policy table: %d\n",
5371                         __func__, err);
5372                 goto out;
5373         }
5374
5375         err = register_pernet_subsys(&addrconf_ops);
5376         if (err < 0)
5377                 goto out_addrlabel;
5378
5379         addrconf_wq = create_workqueue("ipv6_addrconf");
5380         if (!addrconf_wq) {
5381                 err = -ENOMEM;
5382                 goto out_nowq;
5383         }
5384
5385         /* The addrconf netdev notifier requires that loopback_dev
5386          * has it's ipv6 private information allocated and setup
5387          * before it can bring up and give link-local addresses
5388          * to other devices which are up.
5389          *
5390          * Unfortunately, loopback_dev is not necessarily the first
5391          * entry in the global dev_base list of net devices.  In fact,
5392          * it is likely to be the very last entry on that list.
5393          * So this causes the notifier registry below to try and
5394          * give link-local addresses to all devices besides loopback_dev
5395          * first, then loopback_dev, which cases all the non-loopback_dev
5396          * devices to fail to get a link-local address.
5397          *
5398          * So, as a temporary fix, allocate the ipv6 structure for
5399          * loopback_dev first by hand.
5400          * Longer term, all of the dependencies ipv6 has upon the loopback
5401          * device and it being up should be removed.
5402          */
5403         rtnl_lock();
5404         idev = ipv6_add_dev(init_net.loopback_dev);
5405         rtnl_unlock();
5406         if (IS_ERR(idev)) {
5407                 err = PTR_ERR(idev);
5408                 goto errlo;
5409         }
5410
5411         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5412                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5413
5414         register_netdevice_notifier(&ipv6_dev_notf);
5415
5416         addrconf_verify();
5417
5418         rtnl_af_register(&inet6_ops);
5419
5420         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5421                               NULL);
5422         if (err < 0)
5423                 goto errout;
5424
5425         /* Only the first call to __rtnl_register can fail */
5426         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5427         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5428         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5429                         inet6_dump_ifaddr, NULL);
5430         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5431                         inet6_dump_ifmcaddr, NULL);
5432         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5433                         inet6_dump_ifacaddr, NULL);
5434         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5435                         inet6_netconf_dump_devconf, NULL);
5436
5437         ipv6_addr_label_rtnl_register();
5438
5439         return 0;
5440 errout:
5441         rtnl_af_unregister(&inet6_ops);
5442         unregister_netdevice_notifier(&ipv6_dev_notf);
5443 errlo:
5444         destroy_workqueue(addrconf_wq);
5445 out_nowq:
5446         unregister_pernet_subsys(&addrconf_ops);
5447 out_addrlabel:
5448         ipv6_addr_label_cleanup();
5449 out:
5450         return err;
5451 }
5452
5453 void addrconf_cleanup(void)
5454 {
5455         struct net_device *dev;
5456         int i;
5457
5458         unregister_netdevice_notifier(&ipv6_dev_notf);
5459         unregister_pernet_subsys(&addrconf_ops);
5460         ipv6_addr_label_cleanup();
5461
5462         rtnl_lock();
5463
5464         __rtnl_af_unregister(&inet6_ops);
5465
5466         /* clean dev list */
5467         for_each_netdev(&init_net, dev) {
5468                 if (__in6_dev_get(dev) == NULL)
5469                         continue;
5470                 addrconf_ifdown(dev, 1);
5471         }
5472         addrconf_ifdown(init_net.loopback_dev, 2);
5473
5474         /*
5475          *      Check hash table.
5476          */
5477         spin_lock_bh(&addrconf_hash_lock);
5478         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5479                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5480         spin_unlock_bh(&addrconf_hash_lock);
5481         cancel_delayed_work(&addr_chk_work);
5482         rtnl_unlock();
5483
5484         destroy_workqueue(addrconf_wq);
5485 }