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  * caller must hold RTNL */
1695 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1696 {
1697         struct in6_addr maddr;
1698
1699         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1700                 return;
1701
1702         addrconf_addr_solict_mult(addr, &maddr);
1703         ipv6_dev_mc_inc(dev, &maddr);
1704 }
1705
1706 /* caller must hold RTNL */
1707 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1708 {
1709         struct in6_addr maddr;
1710
1711         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1712                 return;
1713
1714         addrconf_addr_solict_mult(addr, &maddr);
1715         __ipv6_dev_mc_dec(idev, &maddr);
1716 }
1717
1718 /* caller must hold RTNL */
1719 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1720 {
1721         struct in6_addr addr;
1722
1723         if (ifp->prefix_len >= 127) /* RFC 6164 */
1724                 return;
1725         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1726         if (ipv6_addr_any(&addr))
1727                 return;
1728         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1729 }
1730
1731 /* caller must hold RTNL */
1732 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1733 {
1734         struct in6_addr addr;
1735
1736         if (ifp->prefix_len >= 127) /* RFC 6164 */
1737                 return;
1738         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1739         if (ipv6_addr_any(&addr))
1740                 return;
1741         __ipv6_dev_ac_dec(ifp->idev, &addr);
1742 }
1743
1744 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1745 {
1746         if (dev->addr_len != ETH_ALEN)
1747                 return -1;
1748         memcpy(eui, dev->dev_addr, 3);
1749         memcpy(eui + 5, dev->dev_addr + 3, 3);
1750
1751         /*
1752          * The zSeries OSA network cards can be shared among various
1753          * OS instances, but the OSA cards have only one MAC address.
1754          * This leads to duplicate address conflicts in conjunction
1755          * with IPv6 if more than one instance uses the same card.
1756          *
1757          * The driver for these cards can deliver a unique 16-bit
1758          * identifier for each instance sharing the same card.  It is
1759          * placed instead of 0xFFFE in the interface identifier.  The
1760          * "u" bit of the interface identifier is not inverted in this
1761          * case.  Hence the resulting interface identifier has local
1762          * scope according to RFC2373.
1763          */
1764         if (dev->dev_id) {
1765                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1766                 eui[4] = dev->dev_id & 0xFF;
1767         } else {
1768                 eui[3] = 0xFF;
1769                 eui[4] = 0xFE;
1770                 eui[0] ^= 2;
1771         }
1772         return 0;
1773 }
1774
1775 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1776 {
1777         if (dev->addr_len != IEEE802154_ADDR_LEN)
1778                 return -1;
1779         memcpy(eui, dev->dev_addr, 8);
1780         eui[0] ^= 2;
1781         return 0;
1782 }
1783
1784 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1785 {
1786         union fwnet_hwaddr *ha;
1787
1788         if (dev->addr_len != FWNET_ALEN)
1789                 return -1;
1790
1791         ha = (union fwnet_hwaddr *)dev->dev_addr;
1792
1793         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1794         eui[0] ^= 2;
1795         return 0;
1796 }
1797
1798 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1799 {
1800         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1801         if (dev->addr_len != ARCNET_ALEN)
1802                 return -1;
1803         memset(eui, 0, 7);
1804         eui[7] = *(u8 *)dev->dev_addr;
1805         return 0;
1806 }
1807
1808 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1809 {
1810         if (dev->addr_len != INFINIBAND_ALEN)
1811                 return -1;
1812         memcpy(eui, dev->dev_addr + 12, 8);
1813         eui[0] |= 2;
1814         return 0;
1815 }
1816
1817 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1818 {
1819         if (addr == 0)
1820                 return -1;
1821         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1822                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1823                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1824                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1825                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1826                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1827         eui[1] = 0;
1828         eui[2] = 0x5E;
1829         eui[3] = 0xFE;
1830         memcpy(eui + 4, &addr, 4);
1831         return 0;
1832 }
1833
1834 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1835 {
1836         if (dev->priv_flags & IFF_ISATAP)
1837                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1838         return -1;
1839 }
1840
1841 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1842 {
1843         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1844 }
1845
1846 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1847 {
1848         memcpy(eui, dev->perm_addr, 3);
1849         memcpy(eui + 5, dev->perm_addr + 3, 3);
1850         eui[3] = 0xFF;
1851         eui[4] = 0xFE;
1852         eui[0] ^= 2;
1853         return 0;
1854 }
1855
1856 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1857 {
1858         switch (dev->type) {
1859         case ARPHRD_ETHER:
1860         case ARPHRD_FDDI:
1861                 return addrconf_ifid_eui48(eui, dev);
1862         case ARPHRD_ARCNET:
1863                 return addrconf_ifid_arcnet(eui, dev);
1864         case ARPHRD_INFINIBAND:
1865                 return addrconf_ifid_infiniband(eui, dev);
1866         case ARPHRD_SIT:
1867                 return addrconf_ifid_sit(eui, dev);
1868         case ARPHRD_IPGRE:
1869                 return addrconf_ifid_gre(eui, dev);
1870         case ARPHRD_6LOWPAN:
1871         case ARPHRD_IEEE802154:
1872                 return addrconf_ifid_eui64(eui, dev);
1873         case ARPHRD_IEEE1394:
1874                 return addrconf_ifid_ieee1394(eui, dev);
1875         case ARPHRD_TUNNEL6:
1876                 return addrconf_ifid_ip6tnl(eui, dev);
1877         }
1878         return -1;
1879 }
1880
1881 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1882 {
1883         int err = -1;
1884         struct inet6_ifaddr *ifp;
1885
1886         read_lock_bh(&idev->lock);
1887         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1888                 if (ifp->scope > IFA_LINK)
1889                         break;
1890                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1891                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1892                         err = 0;
1893                         break;
1894                 }
1895         }
1896         read_unlock_bh(&idev->lock);
1897         return err;
1898 }
1899
1900 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1901 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1902 {
1903 regen:
1904         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1905         idev->rndid[0] &= ~0x02;
1906
1907         /*
1908          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1909          * check if generated address is not inappropriate
1910          *
1911          *  - Reserved subnet anycast (RFC 2526)
1912          *      11111101 11....11 1xxxxxxx
1913          *  - ISATAP (RFC4214) 6.1
1914          *      00-00-5E-FE-xx-xx-xx-xx
1915          *  - value 0
1916          *  - XXX: already assigned to an address on the device
1917          */
1918         if (idev->rndid[0] == 0xfd &&
1919             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1920             (idev->rndid[7]&0x80))
1921                 goto regen;
1922         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1923                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1924                         goto regen;
1925                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1926                         goto regen;
1927         }
1928 }
1929
1930 static void ipv6_regen_rndid(unsigned long data)
1931 {
1932         struct inet6_dev *idev = (struct inet6_dev *) data;
1933         unsigned long expires;
1934
1935         rcu_read_lock_bh();
1936         write_lock_bh(&idev->lock);
1937
1938         if (idev->dead)
1939                 goto out;
1940
1941         __ipv6_regen_rndid(idev);
1942
1943         expires = jiffies +
1944                 idev->cnf.temp_prefered_lft * HZ -
1945                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1946                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1947                 idev->cnf.max_desync_factor * HZ;
1948         if (time_before(expires, jiffies)) {
1949                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1950                         __func__, idev->dev->name);
1951                 goto out;
1952         }
1953
1954         if (!mod_timer(&idev->regen_timer, expires))
1955                 in6_dev_hold(idev);
1956
1957 out:
1958         write_unlock_bh(&idev->lock);
1959         rcu_read_unlock_bh();
1960         in6_dev_put(idev);
1961 }
1962
1963 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1964 {
1965         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1966                 __ipv6_regen_rndid(idev);
1967 }
1968
1969 /*
1970  *      Add prefix route.
1971  */
1972
1973 static void
1974 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1975                       unsigned long expires, u32 flags)
1976 {
1977         struct fib6_config cfg = {
1978                 .fc_table = RT6_TABLE_PREFIX,
1979                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1980                 .fc_ifindex = dev->ifindex,
1981                 .fc_expires = expires,
1982                 .fc_dst_len = plen,
1983                 .fc_flags = RTF_UP | flags,
1984                 .fc_nlinfo.nl_net = dev_net(dev),
1985                 .fc_protocol = RTPROT_KERNEL,
1986         };
1987
1988         cfg.fc_dst = *pfx;
1989
1990         /* Prevent useless cloning on PtP SIT.
1991            This thing is done here expecting that the whole
1992            class of non-broadcast devices need not cloning.
1993          */
1994 #if IS_ENABLED(CONFIG_IPV6_SIT)
1995         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1996                 cfg.fc_flags |= RTF_NONEXTHOP;
1997 #endif
1998
1999         ip6_route_add(&cfg);
2000 }
2001
2002
2003 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2004                                                   int plen,
2005                                                   const struct net_device *dev,
2006                                                   u32 flags, u32 noflags)
2007 {
2008         struct fib6_node *fn;
2009         struct rt6_info *rt = NULL;
2010         struct fib6_table *table;
2011
2012         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2013         if (table == NULL)
2014                 return NULL;
2015
2016         read_lock_bh(&table->tb6_lock);
2017         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2018         if (!fn)
2019                 goto out;
2020         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2021                 if (rt->dst.dev->ifindex != dev->ifindex)
2022                         continue;
2023                 if ((rt->rt6i_flags & flags) != flags)
2024                         continue;
2025                 if ((rt->rt6i_flags & noflags) != 0)
2026                         continue;
2027                 dst_hold(&rt->dst);
2028                 break;
2029         }
2030 out:
2031         read_unlock_bh(&table->tb6_lock);
2032         return rt;
2033 }
2034
2035
2036 /* Create "default" multicast route to the interface */
2037
2038 static void addrconf_add_mroute(struct net_device *dev)
2039 {
2040         struct fib6_config cfg = {
2041                 .fc_table = RT6_TABLE_LOCAL,
2042                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2043                 .fc_ifindex = dev->ifindex,
2044                 .fc_dst_len = 8,
2045                 .fc_flags = RTF_UP,
2046                 .fc_nlinfo.nl_net = dev_net(dev),
2047         };
2048
2049         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2050
2051         ip6_route_add(&cfg);
2052 }
2053
2054 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2055 {
2056         struct inet6_dev *idev;
2057
2058         ASSERT_RTNL();
2059
2060         idev = ipv6_find_idev(dev);
2061         if (!idev)
2062                 return ERR_PTR(-ENOBUFS);
2063
2064         if (idev->cnf.disable_ipv6)
2065                 return ERR_PTR(-EACCES);
2066
2067         /* Add default multicast route */
2068         if (!(dev->flags & IFF_LOOPBACK))
2069                 addrconf_add_mroute(dev);
2070
2071         return idev;
2072 }
2073
2074 static void manage_tempaddrs(struct inet6_dev *idev,
2075                              struct inet6_ifaddr *ifp,
2076                              __u32 valid_lft, __u32 prefered_lft,
2077                              bool create, unsigned long now)
2078 {
2079         u32 flags;
2080         struct inet6_ifaddr *ift;
2081
2082         read_lock_bh(&idev->lock);
2083         /* update all temporary addresses in the list */
2084         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2085                 int age, max_valid, max_prefered;
2086
2087                 if (ifp != ift->ifpub)
2088                         continue;
2089
2090                 /* RFC 4941 section 3.3:
2091                  * If a received option will extend the lifetime of a public
2092                  * address, the lifetimes of temporary addresses should
2093                  * be extended, subject to the overall constraint that no
2094                  * temporary addresses should ever remain "valid" or "preferred"
2095                  * for a time longer than (TEMP_VALID_LIFETIME) or
2096                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2097                  */
2098                 age = (now - ift->cstamp) / HZ;
2099                 max_valid = idev->cnf.temp_valid_lft - age;
2100                 if (max_valid < 0)
2101                         max_valid = 0;
2102
2103                 max_prefered = idev->cnf.temp_prefered_lft -
2104                                idev->cnf.max_desync_factor - age;
2105                 if (max_prefered < 0)
2106                         max_prefered = 0;
2107
2108                 if (valid_lft > max_valid)
2109                         valid_lft = max_valid;
2110
2111                 if (prefered_lft > max_prefered)
2112                         prefered_lft = max_prefered;
2113
2114                 spin_lock(&ift->lock);
2115                 flags = ift->flags;
2116                 ift->valid_lft = valid_lft;
2117                 ift->prefered_lft = prefered_lft;
2118                 ift->tstamp = now;
2119                 if (prefered_lft > 0)
2120                         ift->flags &= ~IFA_F_DEPRECATED;
2121
2122                 spin_unlock(&ift->lock);
2123                 if (!(flags&IFA_F_TENTATIVE))
2124                         ipv6_ifa_notify(0, ift);
2125         }
2126
2127         if ((create || list_empty(&idev->tempaddr_list)) &&
2128             idev->cnf.use_tempaddr > 0) {
2129                 /* When a new public address is created as described
2130                  * in [ADDRCONF], also create a new temporary address.
2131                  * Also create a temporary address if it's enabled but
2132                  * no temporary address currently exists.
2133                  */
2134                 read_unlock_bh(&idev->lock);
2135                 ipv6_create_tempaddr(ifp, NULL);
2136         } else {
2137                 read_unlock_bh(&idev->lock);
2138         }
2139 }
2140
2141 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2142 {
2143         struct prefix_info *pinfo;
2144         __u32 valid_lft;
2145         __u32 prefered_lft;
2146         int addr_type;
2147         struct inet6_dev *in6_dev;
2148         struct net *net = dev_net(dev);
2149
2150         pinfo = (struct prefix_info *) opt;
2151
2152         if (len < sizeof(struct prefix_info)) {
2153                 ADBG("addrconf: prefix option too short\n");
2154                 return;
2155         }
2156
2157         /*
2158          *      Validation checks ([ADDRCONF], page 19)
2159          */
2160
2161         addr_type = ipv6_addr_type(&pinfo->prefix);
2162
2163         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2164                 return;
2165
2166         valid_lft = ntohl(pinfo->valid);
2167         prefered_lft = ntohl(pinfo->prefered);
2168
2169         if (prefered_lft > valid_lft) {
2170                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2171                 return;
2172         }
2173
2174         in6_dev = in6_dev_get(dev);
2175
2176         if (in6_dev == NULL) {
2177                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2178                                     dev->name);
2179                 return;
2180         }
2181
2182         /*
2183          *      Two things going on here:
2184          *      1) Add routes for on-link prefixes
2185          *      2) Configure prefixes with the auto flag set
2186          */
2187
2188         if (pinfo->onlink) {
2189                 struct rt6_info *rt;
2190                 unsigned long rt_expires;
2191
2192                 /* Avoid arithmetic overflow. Really, we could
2193                  * save rt_expires in seconds, likely valid_lft,
2194                  * but it would require division in fib gc, that it
2195                  * not good.
2196                  */
2197                 if (HZ > USER_HZ)
2198                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2199                 else
2200                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2201
2202                 if (addrconf_finite_timeout(rt_expires))
2203                         rt_expires *= HZ;
2204
2205                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2206                                                pinfo->prefix_len,
2207                                                dev,
2208                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2209                                                RTF_GATEWAY | RTF_DEFAULT);
2210
2211                 if (rt) {
2212                         /* Autoconf prefix route */
2213                         if (valid_lft == 0) {
2214                                 ip6_del_rt(rt);
2215                                 rt = NULL;
2216                         } else if (addrconf_finite_timeout(rt_expires)) {
2217                                 /* not infinity */
2218                                 rt6_set_expires(rt, jiffies + rt_expires);
2219                         } else {
2220                                 rt6_clean_expires(rt);
2221                         }
2222                 } else if (valid_lft) {
2223                         clock_t expires = 0;
2224                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2225                         if (addrconf_finite_timeout(rt_expires)) {
2226                                 /* not infinity */
2227                                 flags |= RTF_EXPIRES;
2228                                 expires = jiffies_to_clock_t(rt_expires);
2229                         }
2230                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2231                                               dev, expires, flags);
2232                 }
2233                 ip6_rt_put(rt);
2234         }
2235
2236         /* Try to figure out our local address for this prefix */
2237
2238         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2239                 struct inet6_ifaddr *ifp;
2240                 struct in6_addr addr;
2241                 int create = 0, update_lft = 0;
2242                 bool tokenized = false;
2243
2244                 if (pinfo->prefix_len == 64) {
2245                         memcpy(&addr, &pinfo->prefix, 8);
2246
2247                         if (!ipv6_addr_any(&in6_dev->token)) {
2248                                 read_lock_bh(&in6_dev->lock);
2249                                 memcpy(addr.s6_addr + 8,
2250                                        in6_dev->token.s6_addr + 8, 8);
2251                                 read_unlock_bh(&in6_dev->lock);
2252                                 tokenized = true;
2253                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2254                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2255                                 in6_dev_put(in6_dev);
2256                                 return;
2257                         }
2258                         goto ok;
2259                 }
2260                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2261                                     pinfo->prefix_len);
2262                 in6_dev_put(in6_dev);
2263                 return;
2264
2265 ok:
2266
2267                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2268
2269                 if (ifp == NULL && valid_lft) {
2270                         int max_addresses = in6_dev->cnf.max_addresses;
2271                         u32 addr_flags = 0;
2272
2273 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2274                         if (in6_dev->cnf.optimistic_dad &&
2275                             !net->ipv6.devconf_all->forwarding && sllao)
2276                                 addr_flags = IFA_F_OPTIMISTIC;
2277 #endif
2278
2279                         /* Do not allow to create too much of autoconfigured
2280                          * addresses; this would be too easy way to crash kernel.
2281                          */
2282                         if (!max_addresses ||
2283                             ipv6_count_addresses(in6_dev) < max_addresses)
2284                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2285                                                     pinfo->prefix_len,
2286                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2287                                                     addr_flags, valid_lft,
2288                                                     prefered_lft);
2289
2290                         if (IS_ERR_OR_NULL(ifp)) {
2291                                 in6_dev_put(in6_dev);
2292                                 return;
2293                         }
2294
2295                         update_lft = 0;
2296                         create = 1;
2297                         spin_lock_bh(&ifp->lock);
2298                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2299                         ifp->cstamp = jiffies;
2300                         ifp->tokenized = tokenized;
2301                         spin_unlock_bh(&ifp->lock);
2302                         addrconf_dad_start(ifp);
2303                 }
2304
2305                 if (ifp) {
2306                         u32 flags;
2307                         unsigned long now;
2308                         u32 stored_lft;
2309
2310                         /* update lifetime (RFC2462 5.5.3 e) */
2311                         spin_lock(&ifp->lock);
2312                         now = jiffies;
2313                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2314                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2315                         else
2316                                 stored_lft = 0;
2317                         if (!update_lft && !create && stored_lft) {
2318                                 const u32 minimum_lft = min(
2319                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2320                                 valid_lft = max(valid_lft, minimum_lft);
2321
2322                                 /* RFC4862 Section 5.5.3e:
2323                                  * "Note that the preferred lifetime of the
2324                                  *  corresponding address is always reset to
2325                                  *  the Preferred Lifetime in the received
2326                                  *  Prefix Information option, regardless of
2327                                  *  whether the valid lifetime is also reset or
2328                                  *  ignored."
2329                                  *
2330                                  * So we should always update prefered_lft here.
2331                                  */
2332                                 update_lft = 1;
2333                         }
2334
2335                         if (update_lft) {
2336                                 ifp->valid_lft = valid_lft;
2337                                 ifp->prefered_lft = prefered_lft;
2338                                 ifp->tstamp = now;
2339                                 flags = ifp->flags;
2340                                 ifp->flags &= ~IFA_F_DEPRECATED;
2341                                 spin_unlock(&ifp->lock);
2342
2343                                 if (!(flags&IFA_F_TENTATIVE))
2344                                         ipv6_ifa_notify(0, ifp);
2345                         } else
2346                                 spin_unlock(&ifp->lock);
2347
2348                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2349                                          create, now);
2350
2351                         in6_ifa_put(ifp);
2352                         addrconf_verify();
2353                 }
2354         }
2355         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2356         in6_dev_put(in6_dev);
2357 }
2358
2359 /*
2360  *      Set destination address.
2361  *      Special case for SIT interfaces where we create a new "virtual"
2362  *      device.
2363  */
2364 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2365 {
2366         struct in6_ifreq ireq;
2367         struct net_device *dev;
2368         int err = -EINVAL;
2369
2370         rtnl_lock();
2371
2372         err = -EFAULT;
2373         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2374                 goto err_exit;
2375
2376         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2377
2378         err = -ENODEV;
2379         if (dev == NULL)
2380                 goto err_exit;
2381
2382 #if IS_ENABLED(CONFIG_IPV6_SIT)
2383         if (dev->type == ARPHRD_SIT) {
2384                 const struct net_device_ops *ops = dev->netdev_ops;
2385                 struct ifreq ifr;
2386                 struct ip_tunnel_parm p;
2387
2388                 err = -EADDRNOTAVAIL;
2389                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2390                         goto err_exit;
2391
2392                 memset(&p, 0, sizeof(p));
2393                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2394                 p.iph.saddr = 0;
2395                 p.iph.version = 4;
2396                 p.iph.ihl = 5;
2397                 p.iph.protocol = IPPROTO_IPV6;
2398                 p.iph.ttl = 64;
2399                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2400
2401                 if (ops->ndo_do_ioctl) {
2402                         mm_segment_t oldfs = get_fs();
2403
2404                         set_fs(KERNEL_DS);
2405                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2406                         set_fs(oldfs);
2407                 } else
2408                         err = -EOPNOTSUPP;
2409
2410                 if (err == 0) {
2411                         err = -ENOBUFS;
2412                         dev = __dev_get_by_name(net, p.name);
2413                         if (!dev)
2414                                 goto err_exit;
2415                         err = dev_open(dev);
2416                 }
2417         }
2418 #endif
2419
2420 err_exit:
2421         rtnl_unlock();
2422         return err;
2423 }
2424
2425 /*
2426  *      Manual configuration of address on an interface
2427  */
2428 static int inet6_addr_add(struct net *net, int ifindex,
2429                           const struct in6_addr *pfx,
2430                           const struct in6_addr *peer_pfx,
2431                           unsigned int plen, __u32 ifa_flags,
2432                           __u32 prefered_lft, __u32 valid_lft)
2433 {
2434         struct inet6_ifaddr *ifp;
2435         struct inet6_dev *idev;
2436         struct net_device *dev;
2437         int scope;
2438         u32 flags;
2439         clock_t expires;
2440         unsigned long timeout;
2441
2442         ASSERT_RTNL();
2443
2444         if (plen > 128)
2445                 return -EINVAL;
2446
2447         /* check the lifetime */
2448         if (!valid_lft || prefered_lft > valid_lft)
2449                 return -EINVAL;
2450
2451         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2452                 return -EINVAL;
2453
2454         dev = __dev_get_by_index(net, ifindex);
2455         if (!dev)
2456                 return -ENODEV;
2457
2458         idev = addrconf_add_dev(dev);
2459         if (IS_ERR(idev))
2460                 return PTR_ERR(idev);
2461
2462         scope = ipv6_addr_scope(pfx);
2463
2464         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2465         if (addrconf_finite_timeout(timeout)) {
2466                 expires = jiffies_to_clock_t(timeout * HZ);
2467                 valid_lft = timeout;
2468                 flags = RTF_EXPIRES;
2469         } else {
2470                 expires = 0;
2471                 flags = 0;
2472                 ifa_flags |= IFA_F_PERMANENT;
2473         }
2474
2475         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2476         if (addrconf_finite_timeout(timeout)) {
2477                 if (timeout == 0)
2478                         ifa_flags |= IFA_F_DEPRECATED;
2479                 prefered_lft = timeout;
2480         }
2481
2482         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2483                             valid_lft, prefered_lft);
2484
2485         if (!IS_ERR(ifp)) {
2486                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2487                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2488                                               expires, flags);
2489                 }
2490
2491                 /*
2492                  * Note that section 3.1 of RFC 4429 indicates
2493                  * that the Optimistic flag should not be set for
2494                  * manually configured addresses
2495                  */
2496                 addrconf_dad_start(ifp);
2497                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2498                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2499                                          true, jiffies);
2500                 in6_ifa_put(ifp);
2501                 addrconf_verify_rtnl();
2502                 return 0;
2503         }
2504
2505         return PTR_ERR(ifp);
2506 }
2507
2508 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2509                           const struct in6_addr *pfx, unsigned int plen)
2510 {
2511         struct inet6_ifaddr *ifp;
2512         struct inet6_dev *idev;
2513         struct net_device *dev;
2514
2515         if (plen > 128)
2516                 return -EINVAL;
2517
2518         dev = __dev_get_by_index(net, ifindex);
2519         if (!dev)
2520                 return -ENODEV;
2521
2522         if ((idev = __in6_dev_get(dev)) == NULL)
2523                 return -ENXIO;
2524
2525         read_lock_bh(&idev->lock);
2526         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2527                 if (ifp->prefix_len == plen &&
2528                     ipv6_addr_equal(pfx, &ifp->addr)) {
2529                         in6_ifa_hold(ifp);
2530                         read_unlock_bh(&idev->lock);
2531
2532                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2533                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2534                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2535                                                  jiffies);
2536                         ipv6_del_addr(ifp);
2537                         addrconf_verify_rtnl();
2538                         return 0;
2539                 }
2540         }
2541         read_unlock_bh(&idev->lock);
2542         return -EADDRNOTAVAIL;
2543 }
2544
2545
2546 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2547 {
2548         struct in6_ifreq ireq;
2549         int err;
2550
2551         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2552                 return -EPERM;
2553
2554         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2555                 return -EFAULT;
2556
2557         rtnl_lock();
2558         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2559                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2560                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2561         rtnl_unlock();
2562         return err;
2563 }
2564
2565 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2566 {
2567         struct in6_ifreq ireq;
2568         int err;
2569
2570         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2571                 return -EPERM;
2572
2573         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2574                 return -EFAULT;
2575
2576         rtnl_lock();
2577         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2578                              ireq.ifr6_prefixlen);
2579         rtnl_unlock();
2580         return err;
2581 }
2582
2583 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2584                      int plen, int scope)
2585 {
2586         struct inet6_ifaddr *ifp;
2587
2588         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2589                             scope, IFA_F_PERMANENT,
2590                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2591         if (!IS_ERR(ifp)) {
2592                 spin_lock_bh(&ifp->lock);
2593                 ifp->flags &= ~IFA_F_TENTATIVE;
2594                 spin_unlock_bh(&ifp->lock);
2595                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2596                 in6_ifa_put(ifp);
2597         }
2598 }
2599
2600 #if IS_ENABLED(CONFIG_IPV6_SIT)
2601 static void sit_add_v4_addrs(struct inet6_dev *idev)
2602 {
2603         struct in6_addr addr;
2604         struct net_device *dev;
2605         struct net *net = dev_net(idev->dev);
2606         int scope, plen;
2607         u32 pflags = 0;
2608
2609         ASSERT_RTNL();
2610
2611         memset(&addr, 0, sizeof(struct in6_addr));
2612         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2613
2614         if (idev->dev->flags&IFF_POINTOPOINT) {
2615                 addr.s6_addr32[0] = htonl(0xfe800000);
2616                 scope = IFA_LINK;
2617                 plen = 64;
2618         } else {
2619                 scope = IPV6_ADDR_COMPATv4;
2620                 plen = 96;
2621                 pflags |= RTF_NONEXTHOP;
2622         }
2623
2624         if (addr.s6_addr32[3]) {
2625                 add_addr(idev, &addr, plen, scope);
2626                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2627                 return;
2628         }
2629
2630         for_each_netdev(net, dev) {
2631                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2632                 if (in_dev && (dev->flags & IFF_UP)) {
2633                         struct in_ifaddr *ifa;
2634
2635                         int flag = scope;
2636
2637                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2638
2639                                 addr.s6_addr32[3] = ifa->ifa_local;
2640
2641                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2642                                         continue;
2643                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2644                                         if (idev->dev->flags&IFF_POINTOPOINT)
2645                                                 continue;
2646                                         flag |= IFA_HOST;
2647                                 }
2648
2649                                 add_addr(idev, &addr, plen, flag);
2650                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2651                                                       pflags);
2652                         }
2653                 }
2654         }
2655 }
2656 #endif
2657
2658 static void init_loopback(struct net_device *dev)
2659 {
2660         struct inet6_dev  *idev;
2661         struct net_device *sp_dev;
2662         struct inet6_ifaddr *sp_ifa;
2663         struct rt6_info *sp_rt;
2664
2665         /* ::1 */
2666
2667         ASSERT_RTNL();
2668
2669         if ((idev = ipv6_find_idev(dev)) == NULL) {
2670                 pr_debug("%s: add_dev failed\n", __func__);
2671                 return;
2672         }
2673
2674         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2675
2676         /* Add routes to other interface's IPv6 addresses */
2677         for_each_netdev(dev_net(dev), sp_dev) {
2678                 if (!strcmp(sp_dev->name, dev->name))
2679                         continue;
2680
2681                 idev = __in6_dev_get(sp_dev);
2682                 if (!idev)
2683                         continue;
2684
2685                 read_lock_bh(&idev->lock);
2686                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2687
2688                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2689                                 continue;
2690
2691                         if (sp_ifa->rt) {
2692                                 /* This dst has been added to garbage list when
2693                                  * lo device down, release this obsolete dst and
2694                                  * reallocate a new router for ifa.
2695                                  */
2696                                 if (sp_ifa->rt->dst.obsolete > 0) {
2697                                         ip6_rt_put(sp_ifa->rt);
2698                                         sp_ifa->rt = NULL;
2699                                 } else {
2700                                         continue;
2701                                 }
2702                         }
2703
2704                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2705
2706                         /* Failure cases are ignored */
2707                         if (!IS_ERR(sp_rt)) {
2708                                 sp_ifa->rt = sp_rt;
2709                                 ip6_ins_rt(sp_rt);
2710                         }
2711                 }
2712                 read_unlock_bh(&idev->lock);
2713         }
2714 }
2715
2716 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2717 {
2718         struct inet6_ifaddr *ifp;
2719         u32 addr_flags = IFA_F_PERMANENT;
2720
2721 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2722         if (idev->cnf.optimistic_dad &&
2723             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2724                 addr_flags |= IFA_F_OPTIMISTIC;
2725 #endif
2726
2727
2728         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2729                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2730         if (!IS_ERR(ifp)) {
2731                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2732                 addrconf_dad_start(ifp);
2733                 in6_ifa_put(ifp);
2734         }
2735 }
2736
2737 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2738 {
2739         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2740                 struct in6_addr addr;
2741
2742                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2743                 /* addrconf_add_linklocal also adds a prefix_route and we
2744                  * only need to care about prefix routes if ipv6_generate_eui64
2745                  * couldn't generate one.
2746                  */
2747                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2748                         addrconf_add_linklocal(idev, &addr);
2749                 else if (prefix_route)
2750                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2751         }
2752 }
2753
2754 static void addrconf_dev_config(struct net_device *dev)
2755 {
2756         struct inet6_dev *idev;
2757
2758         ASSERT_RTNL();
2759
2760         if ((dev->type != ARPHRD_ETHER) &&
2761             (dev->type != ARPHRD_FDDI) &&
2762             (dev->type != ARPHRD_ARCNET) &&
2763             (dev->type != ARPHRD_INFINIBAND) &&
2764             (dev->type != ARPHRD_IEEE802154) &&
2765             (dev->type != ARPHRD_IEEE1394) &&
2766             (dev->type != ARPHRD_TUNNEL6) &&
2767             (dev->type != ARPHRD_6LOWPAN)) {
2768                 /* Alas, we support only Ethernet autoconfiguration. */
2769                 return;
2770         }
2771
2772         idev = addrconf_add_dev(dev);
2773         if (IS_ERR(idev))
2774                 return;
2775
2776         addrconf_addr_gen(idev, false);
2777 }
2778
2779 #if IS_ENABLED(CONFIG_IPV6_SIT)
2780 static void addrconf_sit_config(struct net_device *dev)
2781 {
2782         struct inet6_dev *idev;
2783
2784         ASSERT_RTNL();
2785
2786         /*
2787          * Configure the tunnel with one of our IPv4
2788          * addresses... we should configure all of
2789          * our v4 addrs in the tunnel
2790          */
2791
2792         if ((idev = ipv6_find_idev(dev)) == NULL) {
2793                 pr_debug("%s: add_dev failed\n", __func__);
2794                 return;
2795         }
2796
2797         if (dev->priv_flags & IFF_ISATAP) {
2798                 addrconf_addr_gen(idev, false);
2799                 return;
2800         }
2801
2802         sit_add_v4_addrs(idev);
2803
2804         if (dev->flags&IFF_POINTOPOINT)
2805                 addrconf_add_mroute(dev);
2806 }
2807 #endif
2808
2809 #if IS_ENABLED(CONFIG_NET_IPGRE)
2810 static void addrconf_gre_config(struct net_device *dev)
2811 {
2812         struct inet6_dev *idev;
2813
2814         ASSERT_RTNL();
2815
2816         if ((idev = ipv6_find_idev(dev)) == NULL) {
2817                 pr_debug("%s: add_dev failed\n", __func__);
2818                 return;
2819         }
2820
2821         addrconf_addr_gen(idev, true);
2822 }
2823 #endif
2824
2825 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2826                            void *ptr)
2827 {
2828         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2829         struct inet6_dev *idev = __in6_dev_get(dev);
2830         int run_pending = 0;
2831         int err;
2832
2833         switch (event) {
2834         case NETDEV_REGISTER:
2835                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2836                         idev = ipv6_add_dev(dev);
2837                         if (IS_ERR(idev))
2838                                 return notifier_from_errno(PTR_ERR(idev));
2839                 }
2840                 break;
2841
2842         case NETDEV_UP:
2843         case NETDEV_CHANGE:
2844                 if (dev->flags & IFF_SLAVE)
2845                         break;
2846
2847                 if (event == NETDEV_UP) {
2848                         if (!addrconf_qdisc_ok(dev)) {
2849                                 /* device is not ready yet. */
2850                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2851                                         dev->name);
2852                                 break;
2853                         }
2854
2855                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2856                                 idev = ipv6_add_dev(dev);
2857
2858                         if (!IS_ERR_OR_NULL(idev)) {
2859                                 idev->if_flags |= IF_READY;
2860                                 run_pending = 1;
2861                         }
2862                 } else {
2863                         if (!addrconf_qdisc_ok(dev)) {
2864                                 /* device is still not ready. */
2865                                 break;
2866                         }
2867
2868                         if (idev) {
2869                                 if (idev->if_flags & IF_READY)
2870                                         /* device is already configured. */
2871                                         break;
2872                                 idev->if_flags |= IF_READY;
2873                         }
2874
2875                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2876                                 dev->name);
2877
2878                         run_pending = 1;
2879                 }
2880
2881                 switch (dev->type) {
2882 #if IS_ENABLED(CONFIG_IPV6_SIT)
2883                 case ARPHRD_SIT:
2884                         addrconf_sit_config(dev);
2885                         break;
2886 #endif
2887 #if IS_ENABLED(CONFIG_NET_IPGRE)
2888                 case ARPHRD_IPGRE:
2889                         addrconf_gre_config(dev);
2890                         break;
2891 #endif
2892                 case ARPHRD_LOOPBACK:
2893                         init_loopback(dev);
2894                         break;
2895
2896                 default:
2897                         addrconf_dev_config(dev);
2898                         break;
2899                 }
2900
2901                 if (!IS_ERR_OR_NULL(idev)) {
2902                         if (run_pending)
2903                                 addrconf_dad_run(idev);
2904
2905                         /*
2906                          * If the MTU changed during the interface down,
2907                          * when the interface up, the changed MTU must be
2908                          * reflected in the idev as well as routers.
2909                          */
2910                         if (idev->cnf.mtu6 != dev->mtu &&
2911                             dev->mtu >= IPV6_MIN_MTU) {
2912                                 rt6_mtu_change(dev, dev->mtu);
2913                                 idev->cnf.mtu6 = dev->mtu;
2914                         }
2915                         idev->tstamp = jiffies;
2916                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2917
2918                         /*
2919                          * If the changed mtu during down is lower than
2920                          * IPV6_MIN_MTU stop IPv6 on this interface.
2921                          */
2922                         if (dev->mtu < IPV6_MIN_MTU)
2923                                 addrconf_ifdown(dev, 1);
2924                 }
2925                 break;
2926
2927         case NETDEV_CHANGEMTU:
2928                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2929                         rt6_mtu_change(dev, dev->mtu);
2930                         idev->cnf.mtu6 = dev->mtu;
2931                         break;
2932                 }
2933
2934                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2935                         idev = ipv6_add_dev(dev);
2936                         if (!IS_ERR(idev))
2937                                 break;
2938                 }
2939
2940                 /*
2941                  * if MTU under IPV6_MIN_MTU.
2942                  * Stop IPv6 on this interface.
2943                  */
2944
2945         case NETDEV_DOWN:
2946         case NETDEV_UNREGISTER:
2947                 /*
2948                  *      Remove all addresses from this interface.
2949                  */
2950                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2951                 break;
2952
2953         case NETDEV_CHANGENAME:
2954                 if (idev) {
2955                         snmp6_unregister_dev(idev);
2956                         addrconf_sysctl_unregister(idev);
2957                         err = addrconf_sysctl_register(idev);
2958                         if (err)
2959                                 return notifier_from_errno(err);
2960                         err = snmp6_register_dev(idev);
2961                         if (err) {
2962                                 addrconf_sysctl_unregister(idev);
2963                                 return notifier_from_errno(err);
2964                         }
2965                 }
2966                 break;
2967
2968         case NETDEV_PRE_TYPE_CHANGE:
2969         case NETDEV_POST_TYPE_CHANGE:
2970                 addrconf_type_change(dev, event);
2971                 break;
2972         }
2973
2974         return NOTIFY_OK;
2975 }
2976
2977 /*
2978  *      addrconf module should be notified of a device going up
2979  */
2980 static struct notifier_block ipv6_dev_notf = {
2981         .notifier_call = addrconf_notify,
2982 };
2983
2984 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2985 {
2986         struct inet6_dev *idev;
2987         ASSERT_RTNL();
2988
2989         idev = __in6_dev_get(dev);
2990
2991         if (event == NETDEV_POST_TYPE_CHANGE)
2992                 ipv6_mc_remap(idev);
2993         else if (event == NETDEV_PRE_TYPE_CHANGE)
2994                 ipv6_mc_unmap(idev);
2995 }
2996
2997 static int addrconf_ifdown(struct net_device *dev, int how)
2998 {
2999         struct net *net = dev_net(dev);
3000         struct inet6_dev *idev;
3001         struct inet6_ifaddr *ifa;
3002         int state, i;
3003
3004         ASSERT_RTNL();
3005
3006         rt6_ifdown(net, dev);
3007         neigh_ifdown(&nd_tbl, dev);
3008
3009         idev = __in6_dev_get(dev);
3010         if (idev == NULL)
3011                 return -ENODEV;
3012
3013         /*
3014          * Step 1: remove reference to ipv6 device from parent device.
3015          *         Do not dev_put!
3016          */
3017         if (how) {
3018                 idev->dead = 1;
3019
3020                 /* protected by rtnl_lock */
3021                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3022
3023                 /* Step 1.5: remove snmp6 entry */
3024                 snmp6_unregister_dev(idev);
3025
3026         }
3027
3028         /* Step 2: clear hash table */
3029         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3030                 struct hlist_head *h = &inet6_addr_lst[i];
3031
3032                 spin_lock_bh(&addrconf_hash_lock);
3033         restart:
3034                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3035                         if (ifa->idev == idev) {
3036                                 hlist_del_init_rcu(&ifa->addr_lst);
3037                                 addrconf_del_dad_work(ifa);
3038                                 goto restart;
3039                         }
3040                 }
3041                 spin_unlock_bh(&addrconf_hash_lock);
3042         }
3043
3044         write_lock_bh(&idev->lock);
3045
3046         addrconf_del_rs_timer(idev);
3047
3048         /* Step 2: clear flags for stateless addrconf */
3049         if (!how)
3050                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3051
3052         if (how && del_timer(&idev->regen_timer))
3053                 in6_dev_put(idev);
3054
3055         /* Step 3: clear tempaddr list */
3056         while (!list_empty(&idev->tempaddr_list)) {
3057                 ifa = list_first_entry(&idev->tempaddr_list,
3058                                        struct inet6_ifaddr, tmp_list);
3059                 list_del(&ifa->tmp_list);
3060                 write_unlock_bh(&idev->lock);
3061                 spin_lock_bh(&ifa->lock);
3062
3063                 if (ifa->ifpub) {
3064                         in6_ifa_put(ifa->ifpub);
3065                         ifa->ifpub = NULL;
3066                 }
3067                 spin_unlock_bh(&ifa->lock);
3068                 in6_ifa_put(ifa);
3069                 write_lock_bh(&idev->lock);
3070         }
3071
3072         while (!list_empty(&idev->addr_list)) {
3073                 ifa = list_first_entry(&idev->addr_list,
3074                                        struct inet6_ifaddr, if_list);
3075                 addrconf_del_dad_work(ifa);
3076
3077                 list_del(&ifa->if_list);
3078
3079                 write_unlock_bh(&idev->lock);
3080
3081                 spin_lock_bh(&ifa->state_lock);
3082                 state = ifa->state;
3083                 ifa->state = INET6_IFADDR_STATE_DEAD;
3084                 spin_unlock_bh(&ifa->state_lock);
3085
3086                 if (state != INET6_IFADDR_STATE_DEAD) {
3087                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3088                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3089                 }
3090                 in6_ifa_put(ifa);
3091
3092                 write_lock_bh(&idev->lock);
3093         }
3094
3095         write_unlock_bh(&idev->lock);
3096
3097         /* Step 5: Discard multicast list */
3098         if (how)
3099                 ipv6_mc_destroy_dev(idev);
3100         else
3101                 ipv6_mc_down(idev);
3102
3103         idev->tstamp = jiffies;
3104
3105         /* Last: Shot the device (if unregistered) */
3106         if (how) {
3107                 addrconf_sysctl_unregister(idev);
3108                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3109                 neigh_ifdown(&nd_tbl, dev);
3110                 in6_dev_put(idev);
3111         }
3112         return 0;
3113 }
3114
3115 static void addrconf_rs_timer(unsigned long data)
3116 {
3117         struct inet6_dev *idev = (struct inet6_dev *)data;
3118         struct net_device *dev = idev->dev;
3119         struct in6_addr lladdr;
3120
3121         write_lock(&idev->lock);
3122         if (idev->dead || !(idev->if_flags & IF_READY))
3123                 goto out;
3124
3125         if (!ipv6_accept_ra(idev))
3126                 goto out;
3127
3128         /* Announcement received after solicitation was sent */
3129         if (idev->if_flags & IF_RA_RCVD)
3130                 goto out;
3131
3132         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3133                 write_unlock(&idev->lock);
3134                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3135                         ndisc_send_rs(dev, &lladdr,
3136                                       &in6addr_linklocal_allrouters);
3137                 else
3138                         goto put;
3139
3140                 write_lock(&idev->lock);
3141                 /* The wait after the last probe can be shorter */
3142                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3143                                              idev->cnf.rtr_solicits) ?
3144                                       idev->cnf.rtr_solicit_delay :
3145                                       idev->cnf.rtr_solicit_interval);
3146         } else {
3147                 /*
3148                  * Note: we do not support deprecated "all on-link"
3149                  * assumption any longer.
3150                  */
3151                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3152         }
3153
3154 out:
3155         write_unlock(&idev->lock);
3156 put:
3157         in6_dev_put(idev);
3158 }
3159
3160 /*
3161  *      Duplicate Address Detection
3162  */
3163 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3164 {
3165         unsigned long rand_num;
3166         struct inet6_dev *idev = ifp->idev;
3167
3168         if (ifp->flags & IFA_F_OPTIMISTIC)
3169                 rand_num = 0;
3170         else
3171                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3172
3173         ifp->dad_probes = idev->cnf.dad_transmits;
3174         addrconf_mod_dad_work(ifp, rand_num);
3175 }
3176
3177 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3178 {
3179         struct inet6_dev *idev = ifp->idev;
3180         struct net_device *dev = idev->dev;
3181
3182         addrconf_join_solict(dev, &ifp->addr);
3183
3184         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3185
3186         read_lock_bh(&idev->lock);
3187         spin_lock(&ifp->lock);
3188         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3189                 goto out;
3190
3191         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3192             idev->cnf.accept_dad < 1 ||
3193             !(ifp->flags&IFA_F_TENTATIVE) ||
3194             ifp->flags & IFA_F_NODAD) {
3195                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3196                 spin_unlock(&ifp->lock);
3197                 read_unlock_bh(&idev->lock);
3198
3199                 addrconf_dad_completed(ifp);
3200                 return;
3201         }
3202
3203         if (!(idev->if_flags & IF_READY)) {
3204                 spin_unlock(&ifp->lock);
3205                 read_unlock_bh(&idev->lock);
3206                 /*
3207                  * If the device is not ready:
3208                  * - keep it tentative if it is a permanent address.
3209                  * - otherwise, kill it.
3210                  */
3211                 in6_ifa_hold(ifp);
3212                 addrconf_dad_stop(ifp, 0);
3213                 return;
3214         }
3215
3216         /*
3217          * Optimistic nodes can start receiving
3218          * Frames right away
3219          */
3220         if (ifp->flags & IFA_F_OPTIMISTIC)
3221                 ip6_ins_rt(ifp->rt);
3222
3223         addrconf_dad_kick(ifp);
3224 out:
3225         spin_unlock(&ifp->lock);
3226         read_unlock_bh(&idev->lock);
3227 }
3228
3229 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3230 {
3231         bool begin_dad = false;
3232
3233         spin_lock_bh(&ifp->state_lock);
3234         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3235                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3236                 begin_dad = true;
3237         }
3238         spin_unlock_bh(&ifp->state_lock);
3239
3240         if (begin_dad)
3241                 addrconf_mod_dad_work(ifp, 0);
3242 }
3243
3244 static void addrconf_dad_work(struct work_struct *w)
3245 {
3246         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3247                                                 struct inet6_ifaddr,
3248                                                 dad_work);
3249         struct inet6_dev *idev = ifp->idev;
3250         struct in6_addr mcaddr;
3251
3252         enum {
3253                 DAD_PROCESS,
3254                 DAD_BEGIN,
3255                 DAD_ABORT,
3256         } action = DAD_PROCESS;
3257
3258         rtnl_lock();
3259
3260         spin_lock_bh(&ifp->state_lock);
3261         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3262                 action = DAD_BEGIN;
3263                 ifp->state = INET6_IFADDR_STATE_DAD;
3264         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3265                 action = DAD_ABORT;
3266                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3267         }
3268         spin_unlock_bh(&ifp->state_lock);
3269
3270         if (action == DAD_BEGIN) {
3271                 addrconf_dad_begin(ifp);
3272                 goto out;
3273         } else if (action == DAD_ABORT) {
3274                 addrconf_dad_stop(ifp, 1);
3275                 goto out;
3276         }
3277
3278         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3279                 goto out;
3280
3281         write_lock_bh(&idev->lock);
3282         if (idev->dead || !(idev->if_flags & IF_READY)) {
3283                 write_unlock_bh(&idev->lock);
3284                 goto out;
3285         }
3286
3287         spin_lock(&ifp->lock);
3288         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3289                 spin_unlock(&ifp->lock);
3290                 write_unlock_bh(&idev->lock);
3291                 goto out;
3292         }
3293
3294         if (ifp->dad_probes == 0) {
3295                 /*
3296                  * DAD was successful
3297                  */
3298
3299                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3300                 spin_unlock(&ifp->lock);
3301                 write_unlock_bh(&idev->lock);
3302
3303                 addrconf_dad_completed(ifp);
3304
3305                 goto out;
3306         }
3307
3308         ifp->dad_probes--;
3309         addrconf_mod_dad_work(ifp,
3310                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3311         spin_unlock(&ifp->lock);
3312         write_unlock_bh(&idev->lock);
3313
3314         /* send a neighbour solicitation for our addr */
3315         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3316         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3317 out:
3318         in6_ifa_put(ifp);
3319         rtnl_unlock();
3320 }
3321
3322 /* ifp->idev must be at least read locked */
3323 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3324 {
3325         struct inet6_ifaddr *ifpiter;
3326         struct inet6_dev *idev = ifp->idev;
3327
3328         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3329                 if (ifpiter->scope > IFA_LINK)
3330                         break;
3331                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3332                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3333                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3334                     IFA_F_PERMANENT)
3335                         return false;
3336         }
3337         return true;
3338 }
3339
3340 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3341 {
3342         struct net_device *dev = ifp->idev->dev;
3343         struct in6_addr lladdr;
3344         bool send_rs, send_mld;
3345
3346         addrconf_del_dad_work(ifp);
3347
3348         /*
3349          *      Configure the address for reception. Now it is valid.
3350          */
3351
3352         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3353
3354         /* If added prefix is link local and we are prepared to process
3355            router advertisements, start sending router solicitations.
3356          */
3357
3358         read_lock_bh(&ifp->idev->lock);
3359         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3360         send_rs = send_mld &&
3361                   ipv6_accept_ra(ifp->idev) &&
3362                   ifp->idev->cnf.rtr_solicits > 0 &&
3363                   (dev->flags&IFF_LOOPBACK) == 0;
3364         read_unlock_bh(&ifp->idev->lock);
3365
3366         /* While dad is in progress mld report's source address is in6_addrany.
3367          * Resend with proper ll now.
3368          */
3369         if (send_mld)
3370                 ipv6_mc_dad_complete(ifp->idev);
3371
3372         if (send_rs) {
3373                 /*
3374                  *      If a host as already performed a random delay
3375                  *      [...] as part of DAD [...] there is no need
3376                  *      to delay again before sending the first RS
3377                  */
3378                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3379                         return;
3380                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3381
3382                 write_lock_bh(&ifp->idev->lock);
3383                 spin_lock(&ifp->lock);
3384                 ifp->idev->rs_probes = 1;
3385                 ifp->idev->if_flags |= IF_RS_SENT;
3386                 addrconf_mod_rs_timer(ifp->idev,
3387                                       ifp->idev->cnf.rtr_solicit_interval);
3388                 spin_unlock(&ifp->lock);
3389                 write_unlock_bh(&ifp->idev->lock);
3390         }
3391 }
3392
3393 static void addrconf_dad_run(struct inet6_dev *idev)
3394 {
3395         struct inet6_ifaddr *ifp;
3396
3397         read_lock_bh(&idev->lock);
3398         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3399                 spin_lock(&ifp->lock);
3400                 if (ifp->flags & IFA_F_TENTATIVE &&
3401                     ifp->state == INET6_IFADDR_STATE_DAD)
3402                         addrconf_dad_kick(ifp);
3403                 spin_unlock(&ifp->lock);
3404         }
3405         read_unlock_bh(&idev->lock);
3406 }
3407
3408 #ifdef CONFIG_PROC_FS
3409 struct if6_iter_state {
3410         struct seq_net_private p;
3411         int bucket;
3412         int offset;
3413 };
3414
3415 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3416 {
3417         struct inet6_ifaddr *ifa = NULL;
3418         struct if6_iter_state *state = seq->private;
3419         struct net *net = seq_file_net(seq);
3420         int p = 0;
3421
3422         /* initial bucket if pos is 0 */
3423         if (pos == 0) {
3424                 state->bucket = 0;
3425                 state->offset = 0;
3426         }
3427
3428         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3429                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3430                                          addr_lst) {
3431                         if (!net_eq(dev_net(ifa->idev->dev), net))
3432                                 continue;
3433                         /* sync with offset */
3434                         if (p < state->offset) {
3435                                 p++;
3436                                 continue;
3437                         }
3438                         state->offset++;
3439                         return ifa;
3440                 }
3441
3442                 /* prepare for next bucket */
3443                 state->offset = 0;
3444                 p = 0;
3445         }
3446         return NULL;
3447 }
3448
3449 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3450                                          struct inet6_ifaddr *ifa)
3451 {
3452         struct if6_iter_state *state = seq->private;
3453         struct net *net = seq_file_net(seq);
3454
3455         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3456                 if (!net_eq(dev_net(ifa->idev->dev), net))
3457                         continue;
3458                 state->offset++;
3459                 return ifa;
3460         }
3461
3462         while (++state->bucket < IN6_ADDR_HSIZE) {
3463                 state->offset = 0;
3464                 hlist_for_each_entry_rcu_bh(ifa,
3465                                      &inet6_addr_lst[state->bucket], addr_lst) {
3466                         if (!net_eq(dev_net(ifa->idev->dev), net))
3467                                 continue;
3468                         state->offset++;
3469                         return ifa;
3470                 }
3471         }
3472
3473         return NULL;
3474 }
3475
3476 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3477         __acquires(rcu_bh)
3478 {
3479         rcu_read_lock_bh();
3480         return if6_get_first(seq, *pos);
3481 }
3482
3483 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3484 {
3485         struct inet6_ifaddr *ifa;
3486
3487         ifa = if6_get_next(seq, v);
3488         ++*pos;
3489         return ifa;
3490 }
3491
3492 static void if6_seq_stop(struct seq_file *seq, void *v)
3493         __releases(rcu_bh)
3494 {
3495         rcu_read_unlock_bh();
3496 }
3497
3498 static int if6_seq_show(struct seq_file *seq, void *v)
3499 {
3500         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3501         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3502                    &ifp->addr,
3503                    ifp->idev->dev->ifindex,
3504                    ifp->prefix_len,
3505                    ifp->scope,
3506                    (u8) ifp->flags,
3507                    ifp->idev->dev->name);
3508         return 0;
3509 }
3510
3511 static const struct seq_operations if6_seq_ops = {
3512         .start  = if6_seq_start,
3513         .next   = if6_seq_next,
3514         .show   = if6_seq_show,
3515         .stop   = if6_seq_stop,
3516 };
3517
3518 static int if6_seq_open(struct inode *inode, struct file *file)
3519 {
3520         return seq_open_net(inode, file, &if6_seq_ops,
3521                             sizeof(struct if6_iter_state));
3522 }
3523
3524 static const struct file_operations if6_fops = {
3525         .owner          = THIS_MODULE,
3526         .open           = if6_seq_open,
3527         .read           = seq_read,
3528         .llseek         = seq_lseek,
3529         .release        = seq_release_net,
3530 };
3531
3532 static int __net_init if6_proc_net_init(struct net *net)
3533 {
3534         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3535                 return -ENOMEM;
3536         return 0;
3537 }
3538
3539 static void __net_exit if6_proc_net_exit(struct net *net)
3540 {
3541         remove_proc_entry("if_inet6", net->proc_net);
3542 }
3543
3544 static struct pernet_operations if6_proc_net_ops = {
3545        .init = if6_proc_net_init,
3546        .exit = if6_proc_net_exit,
3547 };
3548
3549 int __init if6_proc_init(void)
3550 {
3551         return register_pernet_subsys(&if6_proc_net_ops);
3552 }
3553
3554 void if6_proc_exit(void)
3555 {
3556         unregister_pernet_subsys(&if6_proc_net_ops);
3557 }
3558 #endif  /* CONFIG_PROC_FS */
3559
3560 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3561 /* Check if address is a home address configured on any interface. */
3562 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3563 {
3564         int ret = 0;
3565         struct inet6_ifaddr *ifp = NULL;
3566         unsigned int hash = inet6_addr_hash(addr);
3567
3568         rcu_read_lock_bh();
3569         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3570                 if (!net_eq(dev_net(ifp->idev->dev), net))
3571                         continue;
3572                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3573                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3574                         ret = 1;
3575                         break;
3576                 }
3577         }
3578         rcu_read_unlock_bh();
3579         return ret;
3580 }
3581 #endif
3582
3583 /*
3584  *      Periodic address status verification
3585  */
3586
3587 static void addrconf_verify_rtnl(void)
3588 {
3589         unsigned long now, next, next_sec, next_sched;
3590         struct inet6_ifaddr *ifp;
3591         int i;
3592
3593         ASSERT_RTNL();
3594
3595         rcu_read_lock_bh();
3596         now = jiffies;
3597         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3598
3599         cancel_delayed_work(&addr_chk_work);
3600
3601         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3602 restart:
3603                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3604                         unsigned long age;
3605
3606                         /* When setting preferred_lft to a value not zero or
3607                          * infinity, while valid_lft is infinity
3608                          * IFA_F_PERMANENT has a non-infinity life time.
3609                          */
3610                         if ((ifp->flags & IFA_F_PERMANENT) &&
3611                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3612                                 continue;
3613
3614                         spin_lock(&ifp->lock);
3615                         /* We try to batch several events at once. */
3616                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3617
3618                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3619                             age >= ifp->valid_lft) {
3620                                 spin_unlock(&ifp->lock);
3621                                 in6_ifa_hold(ifp);
3622                                 ipv6_del_addr(ifp);
3623                                 goto restart;
3624                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3625                                 spin_unlock(&ifp->lock);
3626                                 continue;
3627                         } else if (age >= ifp->prefered_lft) {
3628                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3629                                 int deprecate = 0;
3630
3631                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3632                                         deprecate = 1;
3633                                         ifp->flags |= IFA_F_DEPRECATED;
3634                                 }
3635
3636                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3637                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3638                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3639
3640                                 spin_unlock(&ifp->lock);
3641
3642                                 if (deprecate) {
3643                                         in6_ifa_hold(ifp);
3644
3645                                         ipv6_ifa_notify(0, ifp);
3646                                         in6_ifa_put(ifp);
3647                                         goto restart;
3648                                 }
3649                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3650                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3651                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3652                                         ifp->idev->cnf.dad_transmits *
3653                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3654
3655                                 if (age >= ifp->prefered_lft - regen_advance) {
3656                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3657                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3658                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3659                                         if (!ifp->regen_count && ifpub) {
3660                                                 ifp->regen_count++;
3661                                                 in6_ifa_hold(ifp);
3662                                                 in6_ifa_hold(ifpub);
3663                                                 spin_unlock(&ifp->lock);
3664
3665                                                 spin_lock(&ifpub->lock);
3666                                                 ifpub->regen_count = 0;
3667                                                 spin_unlock(&ifpub->lock);
3668                                                 ipv6_create_tempaddr(ifpub, ifp);
3669                                                 in6_ifa_put(ifpub);
3670                                                 in6_ifa_put(ifp);
3671                                                 goto restart;
3672                                         }
3673                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3674                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3675                                 spin_unlock(&ifp->lock);
3676                         } else {
3677                                 /* ifp->prefered_lft <= ifp->valid_lft */
3678                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3679                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3680                                 spin_unlock(&ifp->lock);
3681                         }
3682                 }
3683         }
3684
3685         next_sec = round_jiffies_up(next);
3686         next_sched = next;
3687
3688         /* If rounded timeout is accurate enough, accept it. */
3689         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3690                 next_sched = next_sec;
3691
3692         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3693         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3694                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3695
3696         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3697               now, next, next_sec, next_sched);
3698         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3699         rcu_read_unlock_bh();
3700 }
3701
3702 static void addrconf_verify_work(struct work_struct *w)
3703 {
3704         rtnl_lock();
3705         addrconf_verify_rtnl();
3706         rtnl_unlock();
3707 }
3708
3709 static void addrconf_verify(void)
3710 {
3711         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3712 }
3713
3714 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3715                                      struct in6_addr **peer_pfx)
3716 {
3717         struct in6_addr *pfx = NULL;
3718
3719         *peer_pfx = NULL;
3720
3721         if (addr)
3722                 pfx = nla_data(addr);
3723
3724         if (local) {
3725                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3726                         *peer_pfx = pfx;
3727                 pfx = nla_data(local);
3728         }
3729
3730         return pfx;
3731 }
3732
3733 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3734         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3735         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3736         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3737         [IFA_FLAGS]             = { .len = sizeof(u32) },
3738 };
3739
3740 static int
3741 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3742 {
3743         struct net *net = sock_net(skb->sk);
3744         struct ifaddrmsg *ifm;
3745         struct nlattr *tb[IFA_MAX+1];
3746         struct in6_addr *pfx, *peer_pfx;
3747         u32 ifa_flags;
3748         int err;
3749
3750         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3751         if (err < 0)
3752                 return err;
3753
3754         ifm = nlmsg_data(nlh);
3755         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3756         if (pfx == NULL)
3757                 return -EINVAL;
3758
3759         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3760
3761         /* We ignore other flags so far. */
3762         ifa_flags &= IFA_F_MANAGETEMPADDR;
3763
3764         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3765                               ifm->ifa_prefixlen);
3766 }
3767
3768 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3769                              u32 prefered_lft, u32 valid_lft)
3770 {
3771         u32 flags;
3772         clock_t expires;
3773         unsigned long timeout;
3774         bool was_managetempaddr;
3775         bool had_prefixroute;
3776
3777         ASSERT_RTNL();
3778
3779         if (!valid_lft || (prefered_lft > valid_lft))
3780                 return -EINVAL;
3781
3782         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3783             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3784                 return -EINVAL;
3785
3786         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3787         if (addrconf_finite_timeout(timeout)) {
3788                 expires = jiffies_to_clock_t(timeout * HZ);
3789                 valid_lft = timeout;
3790                 flags = RTF_EXPIRES;
3791         } else {
3792                 expires = 0;
3793                 flags = 0;
3794                 ifa_flags |= IFA_F_PERMANENT;
3795         }
3796
3797         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3798         if (addrconf_finite_timeout(timeout)) {
3799                 if (timeout == 0)
3800                         ifa_flags |= IFA_F_DEPRECATED;
3801                 prefered_lft = timeout;
3802         }
3803
3804         spin_lock_bh(&ifp->lock);
3805         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3806         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3807                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
3808         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3809                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3810                         IFA_F_NOPREFIXROUTE);
3811         ifp->flags |= ifa_flags;
3812         ifp->tstamp = jiffies;
3813         ifp->valid_lft = valid_lft;
3814         ifp->prefered_lft = prefered_lft;
3815
3816         spin_unlock_bh(&ifp->lock);
3817         if (!(ifp->flags&IFA_F_TENTATIVE))
3818                 ipv6_ifa_notify(0, ifp);
3819
3820         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3821                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3822                                       expires, flags);
3823         } else if (had_prefixroute) {
3824                 enum cleanup_prefix_rt_t action;
3825                 unsigned long rt_expires;
3826
3827                 write_lock_bh(&ifp->idev->lock);
3828                 action = check_cleanup_prefix_route(ifp, &rt_expires);
3829                 write_unlock_bh(&ifp->idev->lock);
3830
3831                 if (action != CLEANUP_PREFIX_RT_NOP) {
3832                         cleanup_prefix_route(ifp, rt_expires,
3833                                 action == CLEANUP_PREFIX_RT_DEL);
3834                 }
3835         }
3836
3837         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3838                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3839                         valid_lft = prefered_lft = 0;
3840                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3841                                  !was_managetempaddr, jiffies);
3842         }
3843
3844         addrconf_verify_rtnl();
3845
3846         return 0;
3847 }
3848
3849 static int
3850 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3851 {
3852         struct net *net = sock_net(skb->sk);
3853         struct ifaddrmsg *ifm;
3854         struct nlattr *tb[IFA_MAX+1];
3855         struct in6_addr *pfx, *peer_pfx;
3856         struct inet6_ifaddr *ifa;
3857         struct net_device *dev;
3858         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3859         u32 ifa_flags;
3860         int err;
3861
3862         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3863         if (err < 0)
3864                 return err;
3865
3866         ifm = nlmsg_data(nlh);
3867         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3868         if (pfx == NULL)
3869                 return -EINVAL;
3870
3871         if (tb[IFA_CACHEINFO]) {
3872                 struct ifa_cacheinfo *ci;
3873
3874                 ci = nla_data(tb[IFA_CACHEINFO]);
3875                 valid_lft = ci->ifa_valid;
3876                 preferred_lft = ci->ifa_prefered;
3877         } else {
3878                 preferred_lft = INFINITY_LIFE_TIME;
3879                 valid_lft = INFINITY_LIFE_TIME;
3880         }
3881
3882         dev =  __dev_get_by_index(net, ifm->ifa_index);
3883         if (dev == NULL)
3884                 return -ENODEV;
3885
3886         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3887
3888         /* We ignore other flags so far. */
3889         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3890                      IFA_F_NOPREFIXROUTE;
3891
3892         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3893         if (ifa == NULL) {
3894                 /*
3895                  * It would be best to check for !NLM_F_CREATE here but
3896                  * userspace already relies on not having to provide this.
3897                  */
3898                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3899                                       ifm->ifa_prefixlen, ifa_flags,
3900                                       preferred_lft, valid_lft);
3901         }
3902
3903         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3904             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3905                 err = -EEXIST;
3906         else
3907                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3908
3909         in6_ifa_put(ifa);
3910
3911         return err;
3912 }
3913
3914 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3915                           u8 scope, int ifindex)
3916 {
3917         struct ifaddrmsg *ifm;
3918
3919         ifm = nlmsg_data(nlh);
3920         ifm->ifa_family = AF_INET6;
3921         ifm->ifa_prefixlen = prefixlen;
3922         ifm->ifa_flags = flags;
3923         ifm->ifa_scope = scope;
3924         ifm->ifa_index = ifindex;
3925 }
3926
3927 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3928                          unsigned long tstamp, u32 preferred, u32 valid)
3929 {
3930         struct ifa_cacheinfo ci;
3931
3932         ci.cstamp = cstamp_delta(cstamp);
3933         ci.tstamp = cstamp_delta(tstamp);
3934         ci.ifa_prefered = preferred;
3935         ci.ifa_valid = valid;
3936
3937         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3938 }
3939
3940 static inline int rt_scope(int ifa_scope)
3941 {
3942         if (ifa_scope & IFA_HOST)
3943                 return RT_SCOPE_HOST;
3944         else if (ifa_scope & IFA_LINK)
3945                 return RT_SCOPE_LINK;
3946         else if (ifa_scope & IFA_SITE)
3947                 return RT_SCOPE_SITE;
3948         else
3949                 return RT_SCOPE_UNIVERSE;
3950 }
3951
3952 static inline int inet6_ifaddr_msgsize(void)
3953 {
3954         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3955                + nla_total_size(16) /* IFA_LOCAL */
3956                + nla_total_size(16) /* IFA_ADDRESS */
3957                + nla_total_size(sizeof(struct ifa_cacheinfo))
3958                + nla_total_size(4)  /* IFA_FLAGS */;
3959 }
3960
3961 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3962                              u32 portid, u32 seq, int event, unsigned int flags)
3963 {
3964         struct nlmsghdr  *nlh;
3965         u32 preferred, valid;
3966
3967         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3968         if (nlh == NULL)
3969                 return -EMSGSIZE;
3970
3971         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3972                       ifa->idev->dev->ifindex);
3973
3974         if (!((ifa->flags&IFA_F_PERMANENT) &&
3975               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3976                 preferred = ifa->prefered_lft;
3977                 valid = ifa->valid_lft;
3978                 if (preferred != INFINITY_LIFE_TIME) {
3979                         long tval = (jiffies - ifa->tstamp)/HZ;
3980                         if (preferred > tval)
3981                                 preferred -= tval;
3982                         else
3983                                 preferred = 0;
3984                         if (valid != INFINITY_LIFE_TIME) {
3985                                 if (valid > tval)
3986                                         valid -= tval;
3987                                 else
3988                                         valid = 0;
3989                         }
3990                 }
3991         } else {
3992                 preferred = INFINITY_LIFE_TIME;
3993                 valid = INFINITY_LIFE_TIME;
3994         }
3995
3996         if (!ipv6_addr_any(&ifa->peer_addr)) {
3997                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3998                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3999                         goto error;
4000         } else
4001                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4002                         goto error;
4003
4004         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4005                 goto error;
4006
4007         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4008                 goto error;
4009
4010         return nlmsg_end(skb, nlh);
4011
4012 error:
4013         nlmsg_cancel(skb, nlh);
4014         return -EMSGSIZE;
4015 }
4016
4017 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4018                                 u32 portid, u32 seq, int event, u16 flags)
4019 {
4020         struct nlmsghdr  *nlh;
4021         u8 scope = RT_SCOPE_UNIVERSE;
4022         int ifindex = ifmca->idev->dev->ifindex;
4023
4024         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4025                 scope = RT_SCOPE_SITE;
4026
4027         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4028         if (nlh == NULL)
4029                 return -EMSGSIZE;
4030
4031         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4032         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4033             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4034                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4035                 nlmsg_cancel(skb, nlh);
4036                 return -EMSGSIZE;
4037         }
4038
4039         return nlmsg_end(skb, nlh);
4040 }
4041
4042 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4043                                 u32 portid, u32 seq, int event, unsigned int flags)
4044 {
4045         struct nlmsghdr  *nlh;
4046         u8 scope = RT_SCOPE_UNIVERSE;
4047         int ifindex = ifaca->aca_idev->dev->ifindex;
4048
4049         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4050                 scope = RT_SCOPE_SITE;
4051
4052         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4053         if (nlh == NULL)
4054                 return -EMSGSIZE;
4055
4056         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4057         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4058             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4059                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4060                 nlmsg_cancel(skb, nlh);
4061                 return -EMSGSIZE;
4062         }
4063
4064         return nlmsg_end(skb, nlh);
4065 }
4066
4067 enum addr_type_t {
4068         UNICAST_ADDR,
4069         MULTICAST_ADDR,
4070         ANYCAST_ADDR,
4071 };
4072
4073 /* called with rcu_read_lock() */
4074 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4075                           struct netlink_callback *cb, enum addr_type_t type,
4076                           int s_ip_idx, int *p_ip_idx)
4077 {
4078         struct ifmcaddr6 *ifmca;
4079         struct ifacaddr6 *ifaca;
4080         int err = 1;
4081         int ip_idx = *p_ip_idx;
4082
4083         read_lock_bh(&idev->lock);
4084         switch (type) {
4085         case UNICAST_ADDR: {
4086                 struct inet6_ifaddr *ifa;
4087
4088                 /* unicast address incl. temp addr */
4089                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4090                         if (++ip_idx < s_ip_idx)
4091                                 continue;
4092                         err = inet6_fill_ifaddr(skb, ifa,
4093                                                 NETLINK_CB(cb->skb).portid,
4094                                                 cb->nlh->nlmsg_seq,
4095                                                 RTM_NEWADDR,
4096                                                 NLM_F_MULTI);
4097                         if (err <= 0)
4098                                 break;
4099                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4100                 }
4101                 break;
4102         }
4103         case MULTICAST_ADDR:
4104                 /* multicast address */
4105                 for (ifmca = idev->mc_list; ifmca;
4106                      ifmca = ifmca->next, ip_idx++) {
4107                         if (ip_idx < s_ip_idx)
4108                                 continue;
4109                         err = inet6_fill_ifmcaddr(skb, ifmca,
4110                                                   NETLINK_CB(cb->skb).portid,
4111                                                   cb->nlh->nlmsg_seq,
4112                                                   RTM_GETMULTICAST,
4113                                                   NLM_F_MULTI);
4114                         if (err <= 0)
4115                                 break;
4116                 }
4117                 break;
4118         case ANYCAST_ADDR:
4119                 /* anycast address */
4120                 for (ifaca = idev->ac_list; ifaca;
4121                      ifaca = ifaca->aca_next, ip_idx++) {
4122                         if (ip_idx < s_ip_idx)
4123                                 continue;
4124                         err = inet6_fill_ifacaddr(skb, ifaca,
4125                                                   NETLINK_CB(cb->skb).portid,
4126                                                   cb->nlh->nlmsg_seq,
4127                                                   RTM_GETANYCAST,
4128                                                   NLM_F_MULTI);
4129                         if (err <= 0)
4130                                 break;
4131                 }
4132                 break;
4133         default:
4134                 break;
4135         }
4136         read_unlock_bh(&idev->lock);
4137         *p_ip_idx = ip_idx;
4138         return err;
4139 }
4140
4141 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4142                            enum addr_type_t type)
4143 {
4144         struct net *net = sock_net(skb->sk);
4145         int h, s_h;
4146         int idx, ip_idx;
4147         int s_idx, s_ip_idx;
4148         struct net_device *dev;
4149         struct inet6_dev *idev;
4150         struct hlist_head *head;
4151
4152         s_h = cb->args[0];
4153         s_idx = idx = cb->args[1];
4154         s_ip_idx = ip_idx = cb->args[2];
4155
4156         rcu_read_lock();
4157         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4158         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4159                 idx = 0;
4160                 head = &net->dev_index_head[h];
4161                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4162                         if (idx < s_idx)
4163                                 goto cont;
4164                         if (h > s_h || idx > s_idx)
4165                                 s_ip_idx = 0;
4166                         ip_idx = 0;
4167                         idev = __in6_dev_get(dev);
4168                         if (!idev)
4169                                 goto cont;
4170
4171                         if (in6_dump_addrs(idev, skb, cb, type,
4172                                            s_ip_idx, &ip_idx) <= 0)
4173                                 goto done;
4174 cont:
4175                         idx++;
4176                 }
4177         }
4178 done:
4179         rcu_read_unlock();
4180         cb->args[0] = h;
4181         cb->args[1] = idx;
4182         cb->args[2] = ip_idx;
4183
4184         return skb->len;
4185 }
4186
4187 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4188 {
4189         enum addr_type_t type = UNICAST_ADDR;
4190
4191         return inet6_dump_addr(skb, cb, type);
4192 }
4193
4194 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4195 {
4196         enum addr_type_t type = MULTICAST_ADDR;
4197
4198         return inet6_dump_addr(skb, cb, type);
4199 }
4200
4201
4202 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4203 {
4204         enum addr_type_t type = ANYCAST_ADDR;
4205
4206         return inet6_dump_addr(skb, cb, type);
4207 }
4208
4209 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4210 {
4211         struct net *net = sock_net(in_skb->sk);
4212         struct ifaddrmsg *ifm;
4213         struct nlattr *tb[IFA_MAX+1];
4214         struct in6_addr *addr = NULL, *peer;
4215         struct net_device *dev = NULL;
4216         struct inet6_ifaddr *ifa;
4217         struct sk_buff *skb;
4218         int err;
4219
4220         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4221         if (err < 0)
4222                 goto errout;
4223
4224         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4225         if (addr == NULL) {
4226                 err = -EINVAL;
4227                 goto errout;
4228         }
4229
4230         ifm = nlmsg_data(nlh);
4231         if (ifm->ifa_index)
4232                 dev = __dev_get_by_index(net, ifm->ifa_index);
4233
4234         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4235         if (!ifa) {
4236                 err = -EADDRNOTAVAIL;
4237                 goto errout;
4238         }
4239
4240         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4241         if (!skb) {
4242                 err = -ENOBUFS;
4243                 goto errout_ifa;
4244         }
4245
4246         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4247                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4248         if (err < 0) {
4249                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4250                 WARN_ON(err == -EMSGSIZE);
4251                 kfree_skb(skb);
4252                 goto errout_ifa;
4253         }
4254         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4255 errout_ifa:
4256         in6_ifa_put(ifa);
4257 errout:
4258         return err;
4259 }
4260
4261 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4262 {
4263         struct sk_buff *skb;
4264         struct net *net = dev_net(ifa->idev->dev);
4265         int err = -ENOBUFS;
4266
4267         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4268         if (skb == NULL)
4269                 goto errout;
4270
4271         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4272         if (err < 0) {
4273                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4274                 WARN_ON(err == -EMSGSIZE);
4275                 kfree_skb(skb);
4276                 goto errout;
4277         }
4278         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4279         return;
4280 errout:
4281         if (err < 0)
4282                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4283 }
4284
4285 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4286                                 __s32 *array, int bytes)
4287 {
4288         BUG_ON(bytes < (DEVCONF_MAX * 4));
4289
4290         memset(array, 0, bytes);
4291         array[DEVCONF_FORWARDING] = cnf->forwarding;
4292         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4293         array[DEVCONF_MTU6] = cnf->mtu6;
4294         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4295         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4296         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4297         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4298         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4299         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4300                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4301         array[DEVCONF_RTR_SOLICIT_DELAY] =
4302                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4303         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4304         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4305                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4306         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4307                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4308         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4309         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4310         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4311         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4312         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4313         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4314         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4315         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4316 #ifdef CONFIG_IPV6_ROUTER_PREF
4317         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4318         array[DEVCONF_RTR_PROBE_INTERVAL] =
4319                 jiffies_to_msecs(cnf->rtr_probe_interval);
4320 #ifdef CONFIG_IPV6_ROUTE_INFO
4321         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4322 #endif
4323 #endif
4324         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4325         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4326 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4327         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4328 #endif
4329 #ifdef CONFIG_IPV6_MROUTE
4330         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4331 #endif
4332         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4333         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4334         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4335         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4336         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4337         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4338 }
4339
4340 static inline size_t inet6_ifla6_size(void)
4341 {
4342         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4343              + nla_total_size(sizeof(struct ifla_cacheinfo))
4344              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4345              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4346              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4347              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4348 }
4349
4350 static inline size_t inet6_if_nlmsg_size(void)
4351 {
4352         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4353                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4354                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4355                + nla_total_size(4) /* IFLA_MTU */
4356                + nla_total_size(4) /* IFLA_LINK */
4357                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4358 }
4359
4360 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4361                                       int items, int bytes)
4362 {
4363         int i;
4364         int pad = bytes - sizeof(u64) * items;
4365         BUG_ON(pad < 0);
4366
4367         /* Use put_unaligned() because stats may not be aligned for u64. */
4368         put_unaligned(items, &stats[0]);
4369         for (i = 1; i < items; i++)
4370                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4371
4372         memset(&stats[items], 0, pad);
4373 }
4374
4375 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4376                                       int items, int bytes, size_t syncpoff)
4377 {
4378         int i;
4379         int pad = bytes - sizeof(u64) * items;
4380         BUG_ON(pad < 0);
4381
4382         /* Use put_unaligned() because stats may not be aligned for u64. */
4383         put_unaligned(items, &stats[0]);
4384         for (i = 1; i < items; i++)
4385                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4386
4387         memset(&stats[items], 0, pad);
4388 }
4389
4390 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4391                              int bytes)
4392 {
4393         switch (attrtype) {
4394         case IFLA_INET6_STATS:
4395                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4396                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4397                 break;
4398         case IFLA_INET6_ICMP6STATS:
4399                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4400                 break;
4401         }
4402 }
4403
4404 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4405 {
4406         struct nlattr *nla;
4407         struct ifla_cacheinfo ci;
4408
4409         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4410                 goto nla_put_failure;
4411         ci.max_reasm_len = IPV6_MAXPLEN;
4412         ci.tstamp = cstamp_delta(idev->tstamp);
4413         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4414         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4415         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4416                 goto nla_put_failure;
4417         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4418         if (nla == NULL)
4419                 goto nla_put_failure;
4420         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4421
4422         /* XXX - MC not implemented */
4423
4424         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4425         if (nla == NULL)
4426                 goto nla_put_failure;
4427         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4428
4429         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4430         if (nla == NULL)
4431                 goto nla_put_failure;
4432         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4433
4434         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4435         if (nla == NULL)
4436                 goto nla_put_failure;
4437
4438         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4439                 goto nla_put_failure;
4440
4441         read_lock_bh(&idev->lock);
4442         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4443         read_unlock_bh(&idev->lock);
4444
4445         return 0;
4446
4447 nla_put_failure:
4448         return -EMSGSIZE;
4449 }
4450
4451 static size_t inet6_get_link_af_size(const struct net_device *dev)
4452 {
4453         if (!__in6_dev_get(dev))
4454                 return 0;
4455
4456         return inet6_ifla6_size();
4457 }
4458
4459 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4460 {
4461         struct inet6_dev *idev = __in6_dev_get(dev);
4462
4463         if (!idev)
4464                 return -ENODATA;
4465
4466         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4467                 return -EMSGSIZE;
4468
4469         return 0;
4470 }
4471
4472 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4473 {
4474         struct inet6_ifaddr *ifp;
4475         struct net_device *dev = idev->dev;
4476         bool update_rs = false;
4477         struct in6_addr ll_addr;
4478
4479         ASSERT_RTNL();
4480
4481         if (token == NULL)
4482                 return -EINVAL;
4483         if (ipv6_addr_any(token))
4484                 return -EINVAL;
4485         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4486                 return -EINVAL;
4487         if (!ipv6_accept_ra(idev))
4488                 return -EINVAL;
4489         if (idev->cnf.rtr_solicits <= 0)
4490                 return -EINVAL;
4491
4492         write_lock_bh(&idev->lock);
4493
4494         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4495         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4496
4497         write_unlock_bh(&idev->lock);
4498
4499         if (!idev->dead && (idev->if_flags & IF_READY) &&
4500             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4501                              IFA_F_OPTIMISTIC)) {
4502
4503                 /* If we're not ready, then normal ifup will take care
4504                  * of this. Otherwise, we need to request our rs here.
4505                  */
4506                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4507                 update_rs = true;
4508         }
4509
4510         write_lock_bh(&idev->lock);
4511
4512         if (update_rs) {
4513                 idev->if_flags |= IF_RS_SENT;
4514                 idev->rs_probes = 1;
4515                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4516         }
4517
4518         /* Well, that's kinda nasty ... */
4519         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4520                 spin_lock(&ifp->lock);
4521                 if (ifp->tokenized) {
4522                         ifp->valid_lft = 0;
4523                         ifp->prefered_lft = 0;
4524                 }
4525                 spin_unlock(&ifp->lock);
4526         }
4527
4528         write_unlock_bh(&idev->lock);
4529         addrconf_verify_rtnl();
4530         return 0;
4531 }
4532
4533 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4534 {
4535         int err = -EINVAL;
4536         struct inet6_dev *idev = __in6_dev_get(dev);
4537         struct nlattr *tb[IFLA_INET6_MAX + 1];
4538
4539         if (!idev)
4540                 return -EAFNOSUPPORT;
4541
4542         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4543                 BUG();
4544
4545         if (tb[IFLA_INET6_TOKEN]) {
4546                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4547                 if (err)
4548                         return err;
4549         }
4550
4551         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4552                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4553
4554                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4555                     mode != IN6_ADDR_GEN_MODE_NONE)
4556                         return -EINVAL;
4557                 idev->addr_gen_mode = mode;
4558                 err = 0;
4559         }
4560
4561         return err;
4562 }
4563
4564 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4565                              u32 portid, u32 seq, int event, unsigned int flags)
4566 {
4567         struct net_device *dev = idev->dev;
4568         struct ifinfomsg *hdr;
4569         struct nlmsghdr *nlh;
4570         void *protoinfo;
4571
4572         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4573         if (nlh == NULL)
4574                 return -EMSGSIZE;
4575
4576         hdr = nlmsg_data(nlh);
4577         hdr->ifi_family = AF_INET6;
4578         hdr->__ifi_pad = 0;
4579         hdr->ifi_type = dev->type;
4580         hdr->ifi_index = dev->ifindex;
4581         hdr->ifi_flags = dev_get_flags(dev);
4582         hdr->ifi_change = 0;
4583
4584         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4585             (dev->addr_len &&
4586              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4587             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4588             (dev->ifindex != dev->iflink &&
4589              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4590                 goto nla_put_failure;
4591         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4592         if (protoinfo == NULL)
4593                 goto nla_put_failure;
4594
4595         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4596                 goto nla_put_failure;
4597
4598         nla_nest_end(skb, protoinfo);
4599         return nlmsg_end(skb, nlh);
4600
4601 nla_put_failure:
4602         nlmsg_cancel(skb, nlh);
4603         return -EMSGSIZE;
4604 }
4605
4606 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4607 {
4608         struct net *net = sock_net(skb->sk);
4609         int h, s_h;
4610         int idx = 0, s_idx;
4611         struct net_device *dev;
4612         struct inet6_dev *idev;
4613         struct hlist_head *head;
4614
4615         s_h = cb->args[0];
4616         s_idx = cb->args[1];
4617
4618         rcu_read_lock();
4619         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4620                 idx = 0;
4621                 head = &net->dev_index_head[h];
4622                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4623                         if (idx < s_idx)
4624                                 goto cont;
4625                         idev = __in6_dev_get(dev);
4626                         if (!idev)
4627                                 goto cont;
4628                         if (inet6_fill_ifinfo(skb, idev,
4629                                               NETLINK_CB(cb->skb).portid,
4630                                               cb->nlh->nlmsg_seq,
4631                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4632                                 goto out;
4633 cont:
4634                         idx++;
4635                 }
4636         }
4637 out:
4638         rcu_read_unlock();
4639         cb->args[1] = idx;
4640         cb->args[0] = h;
4641
4642         return skb->len;
4643 }
4644
4645 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4646 {
4647         struct sk_buff *skb;
4648         struct net *net = dev_net(idev->dev);
4649         int err = -ENOBUFS;
4650
4651         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4652         if (skb == NULL)
4653                 goto errout;
4654
4655         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4656         if (err < 0) {
4657                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4658                 WARN_ON(err == -EMSGSIZE);
4659                 kfree_skb(skb);
4660                 goto errout;
4661         }
4662         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4663         return;
4664 errout:
4665         if (err < 0)
4666                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4667 }
4668
4669 static inline size_t inet6_prefix_nlmsg_size(void)
4670 {
4671         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4672                + nla_total_size(sizeof(struct in6_addr))
4673                + nla_total_size(sizeof(struct prefix_cacheinfo));
4674 }
4675
4676 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4677                              struct prefix_info *pinfo, u32 portid, u32 seq,
4678                              int event, unsigned int flags)
4679 {
4680         struct prefixmsg *pmsg;
4681         struct nlmsghdr *nlh;
4682         struct prefix_cacheinfo ci;
4683
4684         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4685         if (nlh == NULL)
4686                 return -EMSGSIZE;
4687
4688         pmsg = nlmsg_data(nlh);
4689         pmsg->prefix_family = AF_INET6;
4690         pmsg->prefix_pad1 = 0;
4691         pmsg->prefix_pad2 = 0;
4692         pmsg->prefix_ifindex = idev->dev->ifindex;
4693         pmsg->prefix_len = pinfo->prefix_len;
4694         pmsg->prefix_type = pinfo->type;
4695         pmsg->prefix_pad3 = 0;
4696         pmsg->prefix_flags = 0;
4697         if (pinfo->onlink)
4698                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4699         if (pinfo->autoconf)
4700                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4701
4702         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4703                 goto nla_put_failure;
4704         ci.preferred_time = ntohl(pinfo->prefered);
4705         ci.valid_time = ntohl(pinfo->valid);
4706         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4707                 goto nla_put_failure;
4708         return nlmsg_end(skb, nlh);
4709
4710 nla_put_failure:
4711         nlmsg_cancel(skb, nlh);
4712         return -EMSGSIZE;
4713 }
4714
4715 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4716                          struct prefix_info *pinfo)
4717 {
4718         struct sk_buff *skb;
4719         struct net *net = dev_net(idev->dev);
4720         int err = -ENOBUFS;
4721
4722         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4723         if (skb == NULL)
4724                 goto errout;
4725
4726         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4727         if (err < 0) {
4728                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4729                 WARN_ON(err == -EMSGSIZE);
4730                 kfree_skb(skb);
4731                 goto errout;
4732         }
4733         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4734         return;
4735 errout:
4736         if (err < 0)
4737                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4738 }
4739
4740 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4741 {
4742         struct net *net = dev_net(ifp->idev->dev);
4743
4744         if (event)
4745                 ASSERT_RTNL();
4746
4747         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4748
4749         switch (event) {
4750         case RTM_NEWADDR:
4751                 /*
4752                  * If the address was optimistic
4753                  * we inserted the route at the start of
4754                  * our DAD process, so we don't need
4755                  * to do it again
4756                  */
4757                 if (!(ifp->rt->rt6i_node))
4758                         ip6_ins_rt(ifp->rt);
4759                 if (ifp->idev->cnf.forwarding)
4760                         addrconf_join_anycast(ifp);
4761                 if (!ipv6_addr_any(&ifp->peer_addr))
4762                         addrconf_prefix_route(&ifp->peer_addr, 128,
4763                                               ifp->idev->dev, 0, 0);
4764                 break;
4765         case RTM_DELADDR:
4766                 if (ifp->idev->cnf.forwarding)
4767                         addrconf_leave_anycast(ifp);
4768                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4769                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4770                         struct rt6_info *rt;
4771                         struct net_device *dev = ifp->idev->dev;
4772
4773                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4774                                         dev->ifindex, 1);
4775                         if (rt) {
4776                                 dst_hold(&rt->dst);
4777                                 if (ip6_del_rt(rt))
4778                                         dst_free(&rt->dst);
4779                         }
4780                 }
4781                 dst_hold(&ifp->rt->dst);
4782
4783                 if (ip6_del_rt(ifp->rt))
4784                         dst_free(&ifp->rt->dst);
4785                 break;
4786         }
4787         atomic_inc(&net->ipv6.dev_addr_genid);
4788         rt_genid_bump_ipv6(net);
4789 }
4790
4791 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4792 {
4793         rcu_read_lock_bh();
4794         if (likely(ifp->idev->dead == 0))
4795                 __ipv6_ifa_notify(event, ifp);
4796         rcu_read_unlock_bh();
4797 }
4798
4799 #ifdef CONFIG_SYSCTL
4800
4801 static
4802 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4803                            void __user *buffer, size_t *lenp, loff_t *ppos)
4804 {
4805         int *valp = ctl->data;
4806         int val = *valp;
4807         loff_t pos = *ppos;
4808         struct ctl_table lctl;
4809         int ret;
4810
4811         /*
4812          * ctl->data points to idev->cnf.forwarding, we should
4813          * not modify it until we get the rtnl lock.
4814          */
4815         lctl = *ctl;
4816         lctl.data = &val;
4817
4818         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4819
4820         if (write)
4821                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4822         if (ret)
4823                 *ppos = pos;
4824         return ret;
4825 }
4826
4827 static void dev_disable_change(struct inet6_dev *idev)
4828 {
4829         struct netdev_notifier_info info;
4830
4831         if (!idev || !idev->dev)
4832                 return;
4833
4834         netdev_notifier_info_init(&info, idev->dev);
4835         if (idev->cnf.disable_ipv6)
4836                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4837         else
4838                 addrconf_notify(NULL, NETDEV_UP, &info);
4839 }
4840
4841 static void addrconf_disable_change(struct net *net, __s32 newf)
4842 {
4843         struct net_device *dev;
4844         struct inet6_dev *idev;
4845
4846         rcu_read_lock();
4847         for_each_netdev_rcu(net, dev) {
4848                 idev = __in6_dev_get(dev);
4849                 if (idev) {
4850                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4851                         idev->cnf.disable_ipv6 = newf;
4852                         if (changed)
4853                                 dev_disable_change(idev);
4854                 }
4855         }
4856         rcu_read_unlock();
4857 }
4858
4859 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4860 {
4861         struct net *net;
4862         int old;
4863
4864         if (!rtnl_trylock())
4865                 return restart_syscall();
4866
4867         net = (struct net *)table->extra2;
4868         old = *p;
4869         *p = newf;
4870
4871         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4872                 rtnl_unlock();
4873                 return 0;
4874         }
4875
4876         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4877                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4878                 addrconf_disable_change(net, newf);
4879         } else if ((!newf) ^ (!old))
4880                 dev_disable_change((struct inet6_dev *)table->extra1);
4881
4882         rtnl_unlock();
4883         return 0;
4884 }
4885
4886 static
4887 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4888                             void __user *buffer, size_t *lenp, loff_t *ppos)
4889 {
4890         int *valp = ctl->data;
4891         int val = *valp;
4892         loff_t pos = *ppos;
4893         struct ctl_table lctl;
4894         int ret;
4895
4896         /*
4897          * ctl->data points to idev->cnf.disable_ipv6, we should
4898          * not modify it until we get the rtnl lock.
4899          */
4900         lctl = *ctl;
4901         lctl.data = &val;
4902
4903         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4904
4905         if (write)
4906                 ret = addrconf_disable_ipv6(ctl, valp, val);
4907         if (ret)
4908                 *ppos = pos;
4909         return ret;
4910 }
4911
4912 static
4913 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4914                               void __user *buffer, size_t *lenp, loff_t *ppos)
4915 {
4916         int *valp = ctl->data;
4917         int ret;
4918         int old, new;
4919
4920         old = *valp;
4921         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4922         new = *valp;
4923
4924         if (write && old != new) {
4925                 struct net *net = ctl->extra2;
4926
4927                 if (!rtnl_trylock())
4928                         return restart_syscall();
4929
4930                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4931                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4932                                                      NETCONFA_IFINDEX_DEFAULT,
4933                                                      net->ipv6.devconf_dflt);
4934                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4935                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4936                                                      NETCONFA_IFINDEX_ALL,
4937                                                      net->ipv6.devconf_all);
4938                 else {
4939                         struct inet6_dev *idev = ctl->extra1;
4940
4941                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4942                                                      idev->dev->ifindex,
4943                                                      &idev->cnf);
4944                 }
4945                 rtnl_unlock();
4946         }
4947
4948         return ret;
4949 }
4950
4951
4952 static struct addrconf_sysctl_table
4953 {
4954         struct ctl_table_header *sysctl_header;
4955         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4956 } addrconf_sysctl __read_mostly = {
4957         .sysctl_header = NULL,
4958         .addrconf_vars = {
4959                 {
4960                         .procname       = "forwarding",
4961                         .data           = &ipv6_devconf.forwarding,
4962                         .maxlen         = sizeof(int),
4963                         .mode           = 0644,
4964                         .proc_handler   = addrconf_sysctl_forward,
4965                 },
4966                 {
4967                         .procname       = "hop_limit",
4968                         .data           = &ipv6_devconf.hop_limit,
4969                         .maxlen         = sizeof(int),
4970                         .mode           = 0644,
4971                         .proc_handler   = proc_dointvec,
4972                 },
4973                 {
4974                         .procname       = "mtu",
4975                         .data           = &ipv6_devconf.mtu6,
4976                         .maxlen         = sizeof(int),
4977                         .mode           = 0644,
4978                         .proc_handler   = proc_dointvec,
4979                 },
4980                 {
4981                         .procname       = "accept_ra",
4982                         .data           = &ipv6_devconf.accept_ra,
4983                         .maxlen         = sizeof(int),
4984                         .mode           = 0644,
4985                         .proc_handler   = proc_dointvec,
4986                 },
4987                 {
4988                         .procname       = "accept_redirects",
4989                         .data           = &ipv6_devconf.accept_redirects,
4990                         .maxlen         = sizeof(int),
4991                         .mode           = 0644,
4992                         .proc_handler   = proc_dointvec,
4993                 },
4994                 {
4995                         .procname       = "autoconf",
4996                         .data           = &ipv6_devconf.autoconf,
4997                         .maxlen         = sizeof(int),
4998                         .mode           = 0644,
4999                         .proc_handler   = proc_dointvec,
5000                 },
5001                 {
5002                         .procname       = "dad_transmits",
5003                         .data           = &ipv6_devconf.dad_transmits,
5004                         .maxlen         = sizeof(int),
5005                         .mode           = 0644,
5006                         .proc_handler   = proc_dointvec,
5007                 },
5008                 {
5009                         .procname       = "router_solicitations",
5010                         .data           = &ipv6_devconf.rtr_solicits,
5011                         .maxlen         = sizeof(int),
5012                         .mode           = 0644,
5013                         .proc_handler   = proc_dointvec,
5014                 },
5015                 {
5016                         .procname       = "router_solicitation_interval",
5017                         .data           = &ipv6_devconf.rtr_solicit_interval,
5018                         .maxlen         = sizeof(int),
5019                         .mode           = 0644,
5020                         .proc_handler   = proc_dointvec_jiffies,
5021                 },
5022                 {
5023                         .procname       = "router_solicitation_delay",
5024                         .data           = &ipv6_devconf.rtr_solicit_delay,
5025                         .maxlen         = sizeof(int),
5026                         .mode           = 0644,
5027                         .proc_handler   = proc_dointvec_jiffies,
5028                 },
5029                 {
5030                         .procname       = "force_mld_version",
5031                         .data           = &ipv6_devconf.force_mld_version,
5032                         .maxlen         = sizeof(int),
5033                         .mode           = 0644,
5034                         .proc_handler   = proc_dointvec,
5035                 },
5036                 {
5037                         .procname       = "mldv1_unsolicited_report_interval",
5038                         .data           =
5039                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5040                         .maxlen         = sizeof(int),
5041                         .mode           = 0644,
5042                         .proc_handler   = proc_dointvec_ms_jiffies,
5043                 },
5044                 {
5045                         .procname       = "mldv2_unsolicited_report_interval",
5046                         .data           =
5047                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5048                         .maxlen         = sizeof(int),
5049                         .mode           = 0644,
5050                         .proc_handler   = proc_dointvec_ms_jiffies,
5051                 },
5052                 {
5053                         .procname       = "use_tempaddr",
5054                         .data           = &ipv6_devconf.use_tempaddr,
5055                         .maxlen         = sizeof(int),
5056                         .mode           = 0644,
5057                         .proc_handler   = proc_dointvec,
5058                 },
5059                 {
5060                         .procname       = "temp_valid_lft",
5061                         .data           = &ipv6_devconf.temp_valid_lft,
5062                         .maxlen         = sizeof(int),
5063                         .mode           = 0644,
5064                         .proc_handler   = proc_dointvec,
5065                 },
5066                 {
5067                         .procname       = "temp_prefered_lft",
5068                         .data           = &ipv6_devconf.temp_prefered_lft,
5069                         .maxlen         = sizeof(int),
5070                         .mode           = 0644,
5071                         .proc_handler   = proc_dointvec,
5072                 },
5073                 {
5074                         .procname       = "regen_max_retry",
5075                         .data           = &ipv6_devconf.regen_max_retry,
5076                         .maxlen         = sizeof(int),
5077                         .mode           = 0644,
5078                         .proc_handler   = proc_dointvec,
5079                 },
5080                 {
5081                         .procname       = "max_desync_factor",
5082                         .data           = &ipv6_devconf.max_desync_factor,
5083                         .maxlen         = sizeof(int),
5084                         .mode           = 0644,
5085                         .proc_handler   = proc_dointvec,
5086                 },
5087                 {
5088                         .procname       = "max_addresses",
5089                         .data           = &ipv6_devconf.max_addresses,
5090                         .maxlen         = sizeof(int),
5091                         .mode           = 0644,
5092                         .proc_handler   = proc_dointvec,
5093                 },
5094                 {
5095                         .procname       = "accept_ra_defrtr",
5096                         .data           = &ipv6_devconf.accept_ra_defrtr,
5097                         .maxlen         = sizeof(int),
5098                         .mode           = 0644,
5099                         .proc_handler   = proc_dointvec,
5100                 },
5101                 {
5102                         .procname       = "accept_ra_pinfo",
5103                         .data           = &ipv6_devconf.accept_ra_pinfo,
5104                         .maxlen         = sizeof(int),
5105                         .mode           = 0644,
5106                         .proc_handler   = proc_dointvec,
5107                 },
5108 #ifdef CONFIG_IPV6_ROUTER_PREF
5109                 {
5110                         .procname       = "accept_ra_rtr_pref",
5111                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5112                         .maxlen         = sizeof(int),
5113                         .mode           = 0644,
5114                         .proc_handler   = proc_dointvec,
5115                 },
5116                 {
5117                         .procname       = "router_probe_interval",
5118                         .data           = &ipv6_devconf.rtr_probe_interval,
5119                         .maxlen         = sizeof(int),
5120                         .mode           = 0644,
5121                         .proc_handler   = proc_dointvec_jiffies,
5122                 },
5123 #ifdef CONFIG_IPV6_ROUTE_INFO
5124                 {
5125                         .procname       = "accept_ra_rt_info_max_plen",
5126                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5127                         .maxlen         = sizeof(int),
5128                         .mode           = 0644,
5129                         .proc_handler   = proc_dointvec,
5130                 },
5131 #endif
5132 #endif
5133                 {
5134                         .procname       = "proxy_ndp",
5135                         .data           = &ipv6_devconf.proxy_ndp,
5136                         .maxlen         = sizeof(int),
5137                         .mode           = 0644,
5138                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5139                 },
5140                 {
5141                         .procname       = "accept_source_route",
5142                         .data           = &ipv6_devconf.accept_source_route,
5143                         .maxlen         = sizeof(int),
5144                         .mode           = 0644,
5145                         .proc_handler   = proc_dointvec,
5146                 },
5147 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5148                 {
5149                         .procname       = "optimistic_dad",
5150                         .data           = &ipv6_devconf.optimistic_dad,
5151                         .maxlen         = sizeof(int),
5152                         .mode           = 0644,
5153                         .proc_handler   = proc_dointvec,
5154
5155                 },
5156 #endif
5157 #ifdef CONFIG_IPV6_MROUTE
5158                 {
5159                         .procname       = "mc_forwarding",
5160                         .data           = &ipv6_devconf.mc_forwarding,
5161                         .maxlen         = sizeof(int),
5162                         .mode           = 0444,
5163                         .proc_handler   = proc_dointvec,
5164                 },
5165 #endif
5166                 {
5167                         .procname       = "disable_ipv6",
5168                         .data           = &ipv6_devconf.disable_ipv6,
5169                         .maxlen         = sizeof(int),
5170                         .mode           = 0644,
5171                         .proc_handler   = addrconf_sysctl_disable,
5172                 },
5173                 {
5174                         .procname       = "accept_dad",
5175                         .data           = &ipv6_devconf.accept_dad,
5176                         .maxlen         = sizeof(int),
5177                         .mode           = 0644,
5178                         .proc_handler   = proc_dointvec,
5179                 },
5180                 {
5181                         .procname       = "force_tllao",
5182                         .data           = &ipv6_devconf.force_tllao,
5183                         .maxlen         = sizeof(int),
5184                         .mode           = 0644,
5185                         .proc_handler   = proc_dointvec
5186                 },
5187                 {
5188                         .procname       = "ndisc_notify",
5189                         .data           = &ipv6_devconf.ndisc_notify,
5190                         .maxlen         = sizeof(int),
5191                         .mode           = 0644,
5192                         .proc_handler   = proc_dointvec
5193                 },
5194                 {
5195                         .procname       = "suppress_frag_ndisc",
5196                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5197                         .maxlen         = sizeof(int),
5198                         .mode           = 0644,
5199                         .proc_handler   = proc_dointvec
5200                 },
5201                 {
5202                         .procname       = "accept_ra_from_local",
5203                         .data           = &ipv6_devconf.accept_ra_from_local,
5204                         .maxlen         = sizeof(int),
5205                         .mode           = 0644,
5206                         .proc_handler   = proc_dointvec,
5207                 },
5208                 {
5209                         /* sentinel */
5210                 }
5211         },
5212 };
5213
5214 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5215                 struct inet6_dev *idev, struct ipv6_devconf *p)
5216 {
5217         int i;
5218         struct addrconf_sysctl_table *t;
5219         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5220
5221         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5222         if (t == NULL)
5223                 goto out;
5224
5225         for (i = 0; t->addrconf_vars[i].data; i++) {
5226                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5227                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5228                 t->addrconf_vars[i].extra2 = net;
5229         }
5230
5231         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5232
5233         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5234         if (t->sysctl_header == NULL)
5235                 goto free;
5236
5237         p->sysctl = t;
5238         return 0;
5239
5240 free:
5241         kfree(t);
5242 out:
5243         return -ENOBUFS;
5244 }
5245
5246 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5247 {
5248         struct addrconf_sysctl_table *t;
5249
5250         if (p->sysctl == NULL)
5251                 return;
5252
5253         t = p->sysctl;
5254         p->sysctl = NULL;
5255         unregister_net_sysctl_table(t->sysctl_header);
5256         kfree(t);
5257 }
5258
5259 static int addrconf_sysctl_register(struct inet6_dev *idev)
5260 {
5261         int err;
5262
5263         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5264                 return -EINVAL;
5265
5266         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5267                                     &ndisc_ifinfo_sysctl_change);
5268         if (err)
5269                 return err;
5270         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5271                                          idev, &idev->cnf);
5272         if (err)
5273                 neigh_sysctl_unregister(idev->nd_parms);
5274
5275         return err;
5276 }
5277
5278 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5279 {
5280         __addrconf_sysctl_unregister(&idev->cnf);
5281         neigh_sysctl_unregister(idev->nd_parms);
5282 }
5283
5284
5285 #endif
5286
5287 static int __net_init addrconf_init_net(struct net *net)
5288 {
5289         int err = -ENOMEM;
5290         struct ipv6_devconf *all, *dflt;
5291
5292         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5293         if (all == NULL)
5294                 goto err_alloc_all;
5295
5296         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5297         if (dflt == NULL)
5298                 goto err_alloc_dflt;
5299
5300         /* these will be inherited by all namespaces */
5301         dflt->autoconf = ipv6_defaults.autoconf;
5302         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5303
5304         net->ipv6.devconf_all = all;
5305         net->ipv6.devconf_dflt = dflt;
5306
5307 #ifdef CONFIG_SYSCTL
5308         err = __addrconf_sysctl_register(net, "all", NULL, all);
5309         if (err < 0)
5310                 goto err_reg_all;
5311
5312         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5313         if (err < 0)
5314                 goto err_reg_dflt;
5315 #endif
5316         return 0;
5317
5318 #ifdef CONFIG_SYSCTL
5319 err_reg_dflt:
5320         __addrconf_sysctl_unregister(all);
5321 err_reg_all:
5322         kfree(dflt);
5323 #endif
5324 err_alloc_dflt:
5325         kfree(all);
5326 err_alloc_all:
5327         return err;
5328 }
5329
5330 static void __net_exit addrconf_exit_net(struct net *net)
5331 {
5332 #ifdef CONFIG_SYSCTL
5333         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5334         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5335 #endif
5336         if (!net_eq(net, &init_net)) {
5337                 kfree(net->ipv6.devconf_dflt);
5338                 kfree(net->ipv6.devconf_all);
5339         }
5340 }
5341
5342 static struct pernet_operations addrconf_ops = {
5343         .init = addrconf_init_net,
5344         .exit = addrconf_exit_net,
5345 };
5346
5347 static struct rtnl_af_ops inet6_ops = {
5348         .family           = AF_INET6,
5349         .fill_link_af     = inet6_fill_link_af,
5350         .get_link_af_size = inet6_get_link_af_size,
5351         .set_link_af      = inet6_set_link_af,
5352 };
5353
5354 /*
5355  *      Init / cleanup code
5356  */
5357
5358 int __init addrconf_init(void)
5359 {
5360         struct inet6_dev *idev;
5361         int i, err;
5362
5363         err = ipv6_addr_label_init();
5364         if (err < 0) {
5365                 pr_crit("%s: cannot initialize default policy table: %d\n",
5366                         __func__, err);
5367                 goto out;
5368         }
5369
5370         err = register_pernet_subsys(&addrconf_ops);
5371         if (err < 0)
5372                 goto out_addrlabel;
5373
5374         addrconf_wq = create_workqueue("ipv6_addrconf");
5375         if (!addrconf_wq) {
5376                 err = -ENOMEM;
5377                 goto out_nowq;
5378         }
5379
5380         /* The addrconf netdev notifier requires that loopback_dev
5381          * has it's ipv6 private information allocated and setup
5382          * before it can bring up and give link-local addresses
5383          * to other devices which are up.
5384          *
5385          * Unfortunately, loopback_dev is not necessarily the first
5386          * entry in the global dev_base list of net devices.  In fact,
5387          * it is likely to be the very last entry on that list.
5388          * So this causes the notifier registry below to try and
5389          * give link-local addresses to all devices besides loopback_dev
5390          * first, then loopback_dev, which cases all the non-loopback_dev
5391          * devices to fail to get a link-local address.
5392          *
5393          * So, as a temporary fix, allocate the ipv6 structure for
5394          * loopback_dev first by hand.
5395          * Longer term, all of the dependencies ipv6 has upon the loopback
5396          * device and it being up should be removed.
5397          */
5398         rtnl_lock();
5399         idev = ipv6_add_dev(init_net.loopback_dev);
5400         rtnl_unlock();
5401         if (IS_ERR(idev)) {
5402                 err = PTR_ERR(idev);
5403                 goto errlo;
5404         }
5405
5406         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5407                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5408
5409         register_netdevice_notifier(&ipv6_dev_notf);
5410
5411         addrconf_verify();
5412
5413         rtnl_af_register(&inet6_ops);
5414
5415         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5416                               NULL);
5417         if (err < 0)
5418                 goto errout;
5419
5420         /* Only the first call to __rtnl_register can fail */
5421         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5422         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5423         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5424                         inet6_dump_ifaddr, NULL);
5425         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5426                         inet6_dump_ifmcaddr, NULL);
5427         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5428                         inet6_dump_ifacaddr, NULL);
5429         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5430                         inet6_netconf_dump_devconf, NULL);
5431
5432         ipv6_addr_label_rtnl_register();
5433
5434         return 0;
5435 errout:
5436         rtnl_af_unregister(&inet6_ops);
5437         unregister_netdevice_notifier(&ipv6_dev_notf);
5438 errlo:
5439         destroy_workqueue(addrconf_wq);
5440 out_nowq:
5441         unregister_pernet_subsys(&addrconf_ops);
5442 out_addrlabel:
5443         ipv6_addr_label_cleanup();
5444 out:
5445         return err;
5446 }
5447
5448 void addrconf_cleanup(void)
5449 {
5450         struct net_device *dev;
5451         int i;
5452
5453         unregister_netdevice_notifier(&ipv6_dev_notf);
5454         unregister_pernet_subsys(&addrconf_ops);
5455         ipv6_addr_label_cleanup();
5456
5457         rtnl_lock();
5458
5459         __rtnl_af_unregister(&inet6_ops);
5460
5461         /* clean dev list */
5462         for_each_netdev(&init_net, dev) {
5463                 if (__in6_dev_get(dev) == NULL)
5464                         continue;
5465                 addrconf_ifdown(dev, 1);
5466         }
5467         addrconf_ifdown(init_net.loopback_dev, 2);
5468
5469         /*
5470          *      Check hash table.
5471          */
5472         spin_lock_bh(&addrconf_hash_lock);
5473         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5474                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5475         spin_unlock_bh(&addrconf_hash_lock);
5476         cancel_delayed_work(&addr_chk_work);
5477         rtnl_unlock();
5478
5479         destroy_workqueue(addrconf_wq);
5480 }