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
  87 #ifdef CONFIG_IPV6_PRIVACY
  88 #include <linux/random.h>
  89 #endif
  90
  91 #include <linux/uaccess.h>
  92 #include <asm/unaligned.h>
  93
  94 #include <linux/proc_fs.h>
  95 #include <linux/seq_file.h>
  96 #include <linux/export.h>
  97
  98 /* Set to 3 to get tracing... */
  99 #define ACONF_DEBUG 2
 100
 101 #if ACONF_DEBUG >= 3
 102 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 103 #else
 104 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 105 #endif
 106
 107 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 108
 109 static inline u32 cstamp_delta(unsigned long cstamp)
 110 {
 111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112 }
 113
 114 #ifdef CONFIG_SYSCTL
 115 static void addrconf_sysctl_register(struct inet6_dev *idev);
 116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117 #else
 118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 119 {
 120 }
 121
 122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 123 {
 124 }
 125 #endif
 126
 127 #ifdef CONFIG_IPV6_PRIVACY
 128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130 static void ipv6_regen_rndid(unsigned long data);
 131 #endif
 132
 133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134 static int ipv6_count_addresses(struct inet6_dev *idev);
 135
 136 /*
 137  *      Configured unicast address hash table
 138  */
 139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 140 static DEFINE_SPINLOCK(addrconf_hash_lock);
 141
 142 static void addrconf_verify(unsigned long);
 143
 144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 145 static DEFINE_SPINLOCK(addrconf_verify_lock);
 146
 147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 149
 150 static void addrconf_type_change(struct net_device *dev,
 151                                  unsigned long event);
 152 static int addrconf_ifdown(struct net_device *dev, int how);
 153
 154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 155                                                   int plen,
 156                                                   const struct net_device *dev,
 157                                                   u32 flags, u32 noflags);
 158
 159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 160 static void addrconf_dad_timer(unsigned long data);
 161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 162 static void addrconf_dad_run(struct inet6_dev *idev);
 163 static void addrconf_rs_timer(unsigned long data);
 164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 166
 167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 168                                 struct prefix_info *pinfo);
 169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 170                                struct net_device *dev);
 171
 172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 173         .forwarding             = 0,
 174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 175         .mtu6                   = IPV6_MIN_MTU,
 176         .accept_ra              = 1,
 177         .accept_redirects       = 1,
 178         .autoconf               = 1,
 179         .force_mld_version      = 0,
 180         .mldv1_unsolicited_report_interval = 10 * HZ,
 181         .mldv2_unsolicited_report_interval = HZ,
 182         .dad_transmits          = 1,
 183         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 184         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 185         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 186 #ifdef CONFIG_IPV6_PRIVACY
 187         .use_tempaddr           = 0,
 188         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 189         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 190         .regen_max_retry        = REGEN_MAX_RETRY,
 191         .max_desync_factor      = MAX_DESYNC_FACTOR,
 192 #endif
 193         .max_addresses          = IPV6_MAX_ADDRESSES,
 194         .accept_ra_defrtr       = 1,
 195         .accept_ra_pinfo        = 1,
 196 #ifdef CONFIG_IPV6_ROUTER_PREF
 197         .accept_ra_rtr_pref     = 1,
 198         .rtr_probe_interval     = 60 * HZ,
 199 #ifdef CONFIG_IPV6_ROUTE_INFO
 200         .accept_ra_rt_info_max_plen = 0,
 201 #endif
 202 #endif
 203         .proxy_ndp              = 0,
 204         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 205         .disable_ipv6           = 0,
 206         .accept_dad             = 1,
 207         .suppress_frag_ndisc    = 1,
 208 };
 209
 210 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 211         .forwarding             = 0,
 212         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 213         .mtu6                   = IPV6_MIN_MTU,
 214         .accept_ra              = 1,
 215         .accept_redirects       = 1,
 216         .autoconf               = 1,
 217         .force_mld_version      = 0,
 218         .mldv1_unsolicited_report_interval = 10 * HZ,
 219         .mldv2_unsolicited_report_interval = HZ,
 220         .dad_transmits          = 1,
 221         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 222         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 223         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 224 #ifdef CONFIG_IPV6_PRIVACY
 225         .use_tempaddr           = 0,
 226         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 227         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 228         .regen_max_retry        = REGEN_MAX_RETRY,
 229         .max_desync_factor      = MAX_DESYNC_FACTOR,
 230 #endif
 231         .max_addresses          = IPV6_MAX_ADDRESSES,
 232         .accept_ra_defrtr       = 1,
 233         .accept_ra_pinfo        = 1,
 234 #ifdef CONFIG_IPV6_ROUTER_PREF
 235         .accept_ra_rtr_pref     = 1,
 236         .rtr_probe_interval     = 60 * HZ,
 237 #ifdef CONFIG_IPV6_ROUTE_INFO
 238         .accept_ra_rt_info_max_plen = 0,
 239 #endif
 240 #endif
 241         .proxy_ndp              = 0,
 242         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 243         .disable_ipv6           = 0,
 244         .accept_dad             = 1,
 245         .suppress_frag_ndisc    = 1,
 246 };
 247
 248 /* Check if a valid qdisc is available */
 249 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 250 {
 251         return !qdisc_tx_is_noop(dev);
 252 }
 253
 254 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 255 {
 256         if (del_timer(&idev->rs_timer))
 257                 __in6_dev_put(idev);
 258 }
 259
 260 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 261 {
 262         if (del_timer(&ifp->dad_timer))
 263                 __in6_ifa_put(ifp);
 264 }
 265
 266 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 267                                   unsigned long when)
 268 {
 269         if (!timer_pending(&idev->rs_timer))
 270                 in6_dev_hold(idev);
 271         mod_timer(&idev->rs_timer, jiffies + when);
 272 }
 273
 274 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 275                                    unsigned long when)
 276 {
 277         if (!timer_pending(&ifp->dad_timer))
 278                 in6_ifa_hold(ifp);
 279         mod_timer(&ifp->dad_timer, jiffies + when);
 280 }
 281
 282 static int snmp6_alloc_dev(struct inet6_dev *idev)
 283 {
 284         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 285                           sizeof(struct ipstats_mib),
 286                           __alignof__(struct ipstats_mib)) < 0)
 287                 goto err_ip;
 288         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 289                                         GFP_KERNEL);
 290         if (!idev->stats.icmpv6dev)
 291                 goto err_icmp;
 292         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 293                                            GFP_KERNEL);
 294         if (!idev->stats.icmpv6msgdev)
 295                 goto err_icmpmsg;
 296
 297         return 0;
 298
 299 err_icmpmsg:
 300         kfree(idev->stats.icmpv6dev);
 301 err_icmp:
 302         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 303 err_ip:
 304         return -ENOMEM;
 305 }
 306
 307 static void snmp6_free_dev(struct inet6_dev *idev)
 308 {
 309         kfree(idev->stats.icmpv6msgdev);
 310         kfree(idev->stats.icmpv6dev);
 311         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 312 }
 313
 314 /* Nobody refers to this device, we may destroy it. */
 315
 316 void in6_dev_finish_destroy(struct inet6_dev *idev)
 317 {
 318         struct net_device *dev = idev->dev;
 319
 320         WARN_ON(!list_empty(&idev->addr_list));
 321         WARN_ON(idev->mc_list != NULL);
 322         WARN_ON(timer_pending(&idev->rs_timer));
 323
 324 #ifdef NET_REFCNT_DEBUG
 325         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 326 #endif
 327         dev_put(dev);
 328         if (!idev->dead) {
 329                 pr_warn("Freeing alive inet6 device %p\n", idev);
 330                 return;
 331         }
 332         snmp6_free_dev(idev);
 333         kfree_rcu(idev, rcu);
 334 }
 335 EXPORT_SYMBOL(in6_dev_finish_destroy);
 336
 337 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 338 {
 339         struct inet6_dev *ndev;
 340
 341         ASSERT_RTNL();
 342
 343         if (dev->mtu < IPV6_MIN_MTU)
 344                 return NULL;
 345
 346         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 347
 348         if (ndev == NULL)
 349                 return NULL;
 350
 351         rwlock_init(&ndev->lock);
 352         ndev->dev = dev;
 353         INIT_LIST_HEAD(&ndev->addr_list);
 354         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 355                     (unsigned long)ndev);
 356         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 357         ndev->cnf.mtu6 = dev->mtu;
 358         ndev->cnf.sysctl = NULL;
 359         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 360         if (ndev->nd_parms == NULL) {
 361                 kfree(ndev);
 362                 return NULL;
 363         }
 364         if (ndev->cnf.forwarding)
 365                 dev_disable_lro(dev);
 366         /* We refer to the device */
 367         dev_hold(dev);
 368
 369         if (snmp6_alloc_dev(ndev) < 0) {
 370                 ADBG(KERN_WARNING
 371                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 372                         __func__, dev->name);
 373                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 374                 dev_put(dev);
 375                 kfree(ndev);
 376                 return NULL;
 377         }
 378
 379         if (snmp6_register_dev(ndev) < 0) {
 380                 ADBG(KERN_WARNING
 381                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 382                         __func__, dev->name);
 383                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 384                 ndev->dead = 1;
 385                 in6_dev_finish_destroy(ndev);
 386                 return NULL;
 387         }
 388
 389         /* One reference from device.  We must do this before
 390          * we invoke __ipv6_regen_rndid().
 391          */
 392         in6_dev_hold(ndev);
 393
 394         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 395                 ndev->cnf.accept_dad = -1;
 396
 397 #if IS_ENABLED(CONFIG_IPV6_SIT)
 398         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 399                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 400                 ndev->cnf.rtr_solicits = 0;
 401         }
 402 #endif
 403
 404 #ifdef CONFIG_IPV6_PRIVACY
 405         INIT_LIST_HEAD(&ndev->tempaddr_list);
 406         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 407         if ((dev->flags&IFF_LOOPBACK) ||
 408             dev->type == ARPHRD_TUNNEL ||
 409             dev->type == ARPHRD_TUNNEL6 ||
 410             dev->type == ARPHRD_SIT ||
 411             dev->type == ARPHRD_NONE) {
 412                 ndev->cnf.use_tempaddr = -1;
 413         } else {
 414                 in6_dev_hold(ndev);
 415                 ipv6_regen_rndid((unsigned long) ndev);
 416         }
 417 #endif
 418         ndev->token = in6addr_any;
 419
 420         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 421                 ndev->if_flags |= IF_READY;
 422
 423         ipv6_mc_init_dev(ndev);
 424         ndev->tstamp = jiffies;
 425         addrconf_sysctl_register(ndev);
 426         /* protected by rtnl_lock */
 427         rcu_assign_pointer(dev->ip6_ptr, ndev);
 428
 429         /* Join interface-local all-node multicast group */
 430         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 431
 432         /* Join all-node multicast group */
 433         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 434
 435         /* Join all-router multicast group if forwarding is set */
 436         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 437                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 438
 439         return ndev;
 440 }
 441
 442 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 443 {
 444         struct inet6_dev *idev;
 445
 446         ASSERT_RTNL();
 447
 448         idev = __in6_dev_get(dev);
 449         if (!idev) {
 450                 idev = ipv6_add_dev(dev);
 451                 if (!idev)
 452                         return NULL;
 453         }
 454
 455         if (dev->flags&IFF_UP)
 456                 ipv6_mc_up(idev);
 457         return idev;
 458 }
 459
 460 static int inet6_netconf_msgsize_devconf(int type)
 461 {
 462         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 463                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 464
 465         /* type -1 is used for ALL */
 466         if (type == -1 || type == NETCONFA_FORWARDING)
 467                 size += nla_total_size(4);
 468 #ifdef CONFIG_IPV6_MROUTE
 469         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 470                 size += nla_total_size(4);
 471 #endif
 472
 473         return size;
 474 }
 475
 476 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 477                                       struct ipv6_devconf *devconf, u32 portid,
 478                                       u32 seq, int event, unsigned int flags,
 479                                       int type)
 480 {
 481         struct nlmsghdr  *nlh;
 482         struct netconfmsg *ncm;
 483
 484         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 485                         flags);
 486         if (nlh == NULL)
 487                 return -EMSGSIZE;
 488
 489         ncm = nlmsg_data(nlh);
 490         ncm->ncm_family = AF_INET6;
 491
 492         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 493                 goto nla_put_failure;
 494
 495         /* type -1 is used for ALL */
 496         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 497             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 498                 goto nla_put_failure;
 499 #ifdef CONFIG_IPV6_MROUTE
 500         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 501             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 502                         devconf->mc_forwarding) < 0)
 503                 goto nla_put_failure;
 504 #endif
 505         return nlmsg_end(skb, nlh);
 506
 507 nla_put_failure:
 508         nlmsg_cancel(skb, nlh);
 509         return -EMSGSIZE;
 510 }
 511
 512 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 513                                   struct ipv6_devconf *devconf)
 514 {
 515         struct sk_buff *skb;
 516         int err = -ENOBUFS;
 517
 518         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 519         if (skb == NULL)
 520                 goto errout;
 521
 522         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 523                                          RTM_NEWNETCONF, 0, type);
 524         if (err < 0) {
 525                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 526                 WARN_ON(err == -EMSGSIZE);
 527                 kfree_skb(skb);
 528                 goto errout;
 529         }
 530         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 531         return;
 532 errout:
 533         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 534 }
 535
 536 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 537         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 538         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 539 };
 540
 541 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 542                                      struct nlmsghdr *nlh)
 543 {
 544         struct net *net = sock_net(in_skb->sk);
 545         struct nlattr *tb[NETCONFA_MAX+1];
 546         struct netconfmsg *ncm;
 547         struct sk_buff *skb;
 548         struct ipv6_devconf *devconf;
 549         struct inet6_dev *in6_dev;
 550         struct net_device *dev;
 551         int ifindex;
 552         int err;
 553
 554         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 555                           devconf_ipv6_policy);
 556         if (err < 0)
 557                 goto errout;
 558
 559         err = EINVAL;
 560         if (!tb[NETCONFA_IFINDEX])
 561                 goto errout;
 562
 563         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 564         switch (ifindex) {
 565         case NETCONFA_IFINDEX_ALL:
 566                 devconf = net->ipv6.devconf_all;
 567                 break;
 568         case NETCONFA_IFINDEX_DEFAULT:
 569                 devconf = net->ipv6.devconf_dflt;
 570                 break;
 571         default:
 572                 dev = __dev_get_by_index(net, ifindex);
 573                 if (dev == NULL)
 574                         goto errout;
 575                 in6_dev = __in6_dev_get(dev);
 576                 if (in6_dev == NULL)
 577                         goto errout;
 578                 devconf = &in6_dev->cnf;
 579                 break;
 580         }
 581
 582         err = -ENOBUFS;
 583         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 584         if (skb == NULL)
 585                 goto errout;
 586
 587         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 588                                          NETLINK_CB(in_skb).portid,
 589                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 590                                          -1);
 591         if (err < 0) {
 592                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 593                 WARN_ON(err == -EMSGSIZE);
 594                 kfree_skb(skb);
 595                 goto errout;
 596         }
 597         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 598 errout:
 599         return err;
 600 }
 601
 602 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 603                                       struct netlink_callback *cb)
 604 {
 605         struct net *net = sock_net(skb->sk);
 606         int h, s_h;
 607         int idx, s_idx;
 608         struct net_device *dev;
 609         struct inet6_dev *idev;
 610         struct hlist_head *head;
 611
 612         s_h = cb->args[0];
 613         s_idx = idx = cb->args[1];
 614
 615         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 616                 idx = 0;
 617                 head = &net->dev_index_head[h];
 618                 rcu_read_lock();
 619                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 620                           net->dev_base_seq;
 621                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 622                         if (idx < s_idx)
 623                                 goto cont;
 624                         idev = __in6_dev_get(dev);
 625                         if (!idev)
 626                                 goto cont;
 627
 628                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 629                                                        &idev->cnf,
 630                                                        NETLINK_CB(cb->skb).portid,
 631                                                        cb->nlh->nlmsg_seq,
 632                                                        RTM_NEWNETCONF,
 633                                                        NLM_F_MULTI,
 634                                                        -1) <= 0) {
 635                                 rcu_read_unlock();
 636                                 goto done;
 637                         }
 638                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 639 cont:
 640                         idx++;
 641                 }
 642                 rcu_read_unlock();
 643         }
 644         if (h == NETDEV_HASHENTRIES) {
 645                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 646                                                net->ipv6.devconf_all,
 647                                                NETLINK_CB(cb->skb).portid,
 648                                                cb->nlh->nlmsg_seq,
 649                                                RTM_NEWNETCONF, NLM_F_MULTI,
 650                                                -1) <= 0)
 651                         goto done;
 652                 else
 653                         h++;
 654         }
 655         if (h == NETDEV_HASHENTRIES + 1) {
 656                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 657                                                net->ipv6.devconf_dflt,
 658                                                NETLINK_CB(cb->skb).portid,
 659                                                cb->nlh->nlmsg_seq,
 660                                                RTM_NEWNETCONF, NLM_F_MULTI,
 661                                                -1) <= 0)
 662                         goto done;
 663                 else
 664                         h++;
 665         }
 666 done:
 667         cb->args[0] = h;
 668         cb->args[1] = idx;
 669
 670         return skb->len;
 671 }
 672
 673 #ifdef CONFIG_SYSCTL
 674 static void dev_forward_change(struct inet6_dev *idev)
 675 {
 676         struct net_device *dev;
 677         struct inet6_ifaddr *ifa;
 678
 679         if (!idev)
 680                 return;
 681         dev = idev->dev;
 682         if (idev->cnf.forwarding)
 683                 dev_disable_lro(dev);
 684         if (dev->flags & IFF_MULTICAST) {
 685                 if (idev->cnf.forwarding) {
 686                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 687                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 688                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 689                 } else {
 690                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 691                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 692                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 693                 }
 694         }
 695
 696         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 697                 if (ifa->flags&IFA_F_TENTATIVE)
 698                         continue;
 699                 if (idev->cnf.forwarding)
 700                         addrconf_join_anycast(ifa);
 701                 else
 702                         addrconf_leave_anycast(ifa);
 703         }
 704         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 705                                      dev->ifindex, &idev->cnf);
 706 }
 707
 708
 709 static void addrconf_forward_change(struct net *net, __s32 newf)
 710 {
 711         struct net_device *dev;
 712         struct inet6_dev *idev;
 713
 714         for_each_netdev(net, dev) {
 715                 idev = __in6_dev_get(dev);
 716                 if (idev) {
 717                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 718                         idev->cnf.forwarding = newf;
 719                         if (changed)
 720                                 dev_forward_change(idev);
 721                 }
 722         }
 723 }
 724
 725 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 726 {
 727         struct net *net;
 728         int old;
 729
 730         if (!rtnl_trylock())
 731                 return restart_syscall();
 732
 733         net = (struct net *)table->extra2;
 734         old = *p;
 735         *p = newf;
 736
 737         if (p == &net->ipv6.devconf_dflt->forwarding) {
 738                 if ((!newf) ^ (!old))
 739                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 740                                                      NETCONFA_IFINDEX_DEFAULT,
 741                                                      net->ipv6.devconf_dflt);
 742                 rtnl_unlock();
 743                 return 0;
 744         }
 745
 746         if (p == &net->ipv6.devconf_all->forwarding) {
 747                 net->ipv6.devconf_dflt->forwarding = newf;
 748                 addrconf_forward_change(net, newf);
 749                 if ((!newf) ^ (!old))
 750                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 751                                                      NETCONFA_IFINDEX_ALL,
 752                                                      net->ipv6.devconf_all);
 753         } else if ((!newf) ^ (!old))
 754                 dev_forward_change((struct inet6_dev *)table->extra1);
 755         rtnl_unlock();
 756
 757         if (newf)
 758                 rt6_purge_dflt_routers(net);
 759         return 1;
 760 }
 761 #endif
 762
 763 /* Nobody refers to this ifaddr, destroy it */
 764 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 765 {
 766         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 767
 768 #ifdef NET_REFCNT_DEBUG
 769         pr_debug("%s\n", __func__);
 770 #endif
 771
 772         in6_dev_put(ifp->idev);
 773
 774         if (del_timer(&ifp->dad_timer))
 775                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 776
 777         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 778                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 779                 return;
 780         }
 781         ip6_rt_put(ifp->rt);
 782
 783         kfree_rcu(ifp, rcu);
 784 }
 785
 786 static void
 787 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 788 {
 789         struct list_head *p;
 790         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 791
 792         /*
 793          * Each device address list is sorted in order of scope -
 794          * global before linklocal.
 795          */
 796         list_for_each(p, &idev->addr_list) {
 797                 struct inet6_ifaddr *ifa
 798                         = list_entry(p, struct inet6_ifaddr, if_list);
 799                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 800                         break;
 801         }
 802
 803         list_add_tail(&ifp->if_list, p);
 804 }
 805
 806 static u32 inet6_addr_hash(const struct in6_addr *addr)
 807 {
 808         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 809 }
 810
 811 /* On success it returns ifp with increased reference count */
 812
 813 static struct inet6_ifaddr *
 814 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 815               const struct in6_addr *peer_addr, int pfxlen,
 816               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 817 {
 818         struct inet6_ifaddr *ifa = NULL;
 819         struct rt6_info *rt;
 820         unsigned int hash;
 821         int err = 0;
 822         int addr_type = ipv6_addr_type(addr);
 823
 824         if (addr_type == IPV6_ADDR_ANY ||
 825             addr_type & IPV6_ADDR_MULTICAST ||
 826             (!(idev->dev->flags & IFF_LOOPBACK) &&
 827              addr_type & IPV6_ADDR_LOOPBACK))
 828                 return ERR_PTR(-EADDRNOTAVAIL);
 829
 830         rcu_read_lock_bh();
 831         if (idev->dead) {
 832                 err = -ENODEV;                  /*XXX*/
 833                 goto out2;
 834         }
 835
 836         if (idev->cnf.disable_ipv6) {
 837                 err = -EACCES;
 838                 goto out2;
 839         }
 840
 841         spin_lock(&addrconf_hash_lock);
 842
 843         /* Ignore adding duplicate addresses on an interface */
 844         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 845                 ADBG("ipv6_add_addr: already assigned\n");
 846                 err = -EEXIST;
 847                 goto out;
 848         }
 849
 850         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 851
 852         if (ifa == NULL) {
 853                 ADBG("ipv6_add_addr: malloc failed\n");
 854                 err = -ENOBUFS;
 855                 goto out;
 856         }
 857
 858         rt = addrconf_dst_alloc(idev, addr, false);
 859         if (IS_ERR(rt)) {
 860                 err = PTR_ERR(rt);
 861                 goto out;
 862         }
 863
 864         ifa->addr = *addr;
 865         if (peer_addr)
 866                 ifa->peer_addr = *peer_addr;
 867
 868         spin_lock_init(&ifa->lock);
 869         spin_lock_init(&ifa->state_lock);
 870         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 871                     (unsigned long)ifa);
 872         INIT_HLIST_NODE(&ifa->addr_lst);
 873         ifa->scope = scope;
 874         ifa->prefix_len = pfxlen;
 875         ifa->flags = flags | IFA_F_TENTATIVE;
 876         ifa->valid_lft = valid_lft;
 877         ifa->prefered_lft = prefered_lft;
 878         ifa->cstamp = ifa->tstamp = jiffies;
 879         ifa->tokenized = false;
 880
 881         ifa->rt = rt;
 882
 883         ifa->idev = idev;
 884         in6_dev_hold(idev);
 885         /* For caller */
 886         in6_ifa_hold(ifa);
 887
 888         /* Add to big hash table */
 889         hash = inet6_addr_hash(addr);
 890
 891         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 892         spin_unlock(&addrconf_hash_lock);
 893
 894         write_lock(&idev->lock);
 895         /* Add to inet6_dev unicast addr list. */
 896         ipv6_link_dev_addr(idev, ifa);
 897
 898 #ifdef CONFIG_IPV6_PRIVACY
 899         if (ifa->flags&IFA_F_TEMPORARY) {
 900                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 901                 in6_ifa_hold(ifa);
 902         }
 903 #endif
 904
 905         in6_ifa_hold(ifa);
 906         write_unlock(&idev->lock);
 907 out2:
 908         rcu_read_unlock_bh();
 909
 910         if (likely(err == 0))
 911                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 912         else {
 913                 kfree(ifa);
 914                 ifa = ERR_PTR(err);
 915         }
 916
 917         return ifa;
 918 out:
 919         spin_unlock(&addrconf_hash_lock);
 920         goto out2;
 921 }
 922
 923 /* This function wants to get referenced ifp and releases it before return */
 924
 925 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 926 {
 927         struct inet6_ifaddr *ifa, *ifn;
 928         struct inet6_dev *idev = ifp->idev;
 929         int state;
 930         int deleted = 0, onlink = 0;
 931         unsigned long expires = jiffies;
 932
 933         spin_lock_bh(&ifp->state_lock);
 934         state = ifp->state;
 935         ifp->state = INET6_IFADDR_STATE_DEAD;
 936         spin_unlock_bh(&ifp->state_lock);
 937
 938         if (state == INET6_IFADDR_STATE_DEAD)
 939                 goto out;
 940
 941         spin_lock_bh(&addrconf_hash_lock);
 942         hlist_del_init_rcu(&ifp->addr_lst);
 943         spin_unlock_bh(&addrconf_hash_lock);
 944
 945         write_lock_bh(&idev->lock);
 946 #ifdef CONFIG_IPV6_PRIVACY
 947         if (ifp->flags&IFA_F_TEMPORARY) {
 948                 list_del(&ifp->tmp_list);
 949                 if (ifp->ifpub) {
 950                         in6_ifa_put(ifp->ifpub);
 951                         ifp->ifpub = NULL;
 952                 }
 953                 __in6_ifa_put(ifp);
 954         }
 955 #endif
 956
 957         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 958                 if (ifa == ifp) {
 959                         list_del_init(&ifp->if_list);
 960                         __in6_ifa_put(ifp);
 961
 962                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 963                                 break;
 964                         deleted = 1;
 965                         continue;
 966                 } else if (ifp->flags & IFA_F_PERMANENT) {
 967                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 968                                               ifp->prefix_len)) {
 969                                 if (ifa->flags & IFA_F_PERMANENT) {
 970                                         onlink = 1;
 971                                         if (deleted)
 972                                                 break;
 973                                 } else {
 974                                         unsigned long lifetime;
 975
 976                                         if (!onlink)
 977                                                 onlink = -1;
 978
 979                                         spin_lock(&ifa->lock);
 980
 981                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 982                                         /*
 983                                          * Note: Because this address is
 984                                          * not permanent, lifetime <
 985                                          * LONG_MAX / HZ here.
 986                                          */
 987                                         if (time_before(expires,
 988                                                         ifa->tstamp + lifetime * HZ))
 989                                                 expires = ifa->tstamp + lifetime * HZ;
 990                                         spin_unlock(&ifa->lock);
 991                                 }
 992                         }
 993                 }
 994         }
 995         write_unlock_bh(&idev->lock);
 996
 997         addrconf_del_dad_timer(ifp);
 998
 999         ipv6_ifa_notify(RTM_DELADDR, ifp);
1000
1001         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1002
1003         /*
1004          * Purge or update corresponding prefix
1005          *
1006          * 1) we don't purge prefix here if address was not permanent.
1007          *    prefix is managed by its own lifetime.
1008          * 2) if there're no addresses, delete prefix.
1009          * 3) if there're still other permanent address(es),
1010          *    corresponding prefix is still permanent.
1011          * 4) otherwise, update prefix lifetime to the
1012          *    longest valid lifetime among the corresponding
1013          *    addresses on the device.
1014          *    Note: subsequent RA will update lifetime.
1015          *
1016          * --yoshfuji
1017          */
1018         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1019                 struct in6_addr prefix;
1020                 struct rt6_info *rt;
1021
1022                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1023
1024                 rt = addrconf_get_prefix_route(&prefix,
1025                                                ifp->prefix_len,
1026                                                ifp->idev->dev,
1027                                                0, RTF_GATEWAY | RTF_DEFAULT);
1028
1029                 if (rt) {
1030                         if (onlink == 0) {
1031                                 ip6_del_rt(rt);
1032                                 rt = NULL;
1033                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1034                                 rt6_set_expires(rt, expires);
1035                         }
1036                 }
1037                 ip6_rt_put(rt);
1038         }
1039
1040         /* clean up prefsrc entries */
1041         rt6_remove_prefsrc(ifp);
1042 out:
1043         in6_ifa_put(ifp);
1044 }
1045
1046 #ifdef CONFIG_IPV6_PRIVACY
1047 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1048 {
1049         struct inet6_dev *idev = ifp->idev;
1050         struct in6_addr addr, *tmpaddr;
1051         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1052         unsigned long regen_advance;
1053         int tmp_plen;
1054         int ret = 0;
1055         int max_addresses;
1056         u32 addr_flags;
1057         unsigned long now = jiffies;
1058
1059         write_lock(&idev->lock);
1060         if (ift) {
1061                 spin_lock_bh(&ift->lock);
1062                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1063                 spin_unlock_bh(&ift->lock);
1064                 tmpaddr = &addr;
1065         } else {
1066                 tmpaddr = NULL;
1067         }
1068 retry:
1069         in6_dev_hold(idev);
1070         if (idev->cnf.use_tempaddr <= 0) {
1071                 write_unlock(&idev->lock);
1072                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1073                 in6_dev_put(idev);
1074                 ret = -1;
1075                 goto out;
1076         }
1077         spin_lock_bh(&ifp->lock);
1078         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1079                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1080                 spin_unlock_bh(&ifp->lock);
1081                 write_unlock(&idev->lock);
1082                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1083                         __func__);
1084                 in6_dev_put(idev);
1085                 ret = -1;
1086                 goto out;
1087         }
1088         in6_ifa_hold(ifp);
1089         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1090         __ipv6_try_regen_rndid(idev, tmpaddr);
1091         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1092         age = (now - ifp->tstamp) / HZ;
1093         tmp_valid_lft = min_t(__u32,
1094                               ifp->valid_lft,
1095                               idev->cnf.temp_valid_lft + age);
1096         tmp_prefered_lft = min_t(__u32,
1097                                  ifp->prefered_lft,
1098                                  idev->cnf.temp_prefered_lft + age -
1099                                  idev->cnf.max_desync_factor);
1100         tmp_plen = ifp->prefix_len;
1101         max_addresses = idev->cnf.max_addresses;
1102         tmp_tstamp = ifp->tstamp;
1103         spin_unlock_bh(&ifp->lock);
1104
1105         regen_advance = idev->cnf.regen_max_retry *
1106                         idev->cnf.dad_transmits *
1107                         idev->nd_parms->retrans_time / HZ;
1108         write_unlock(&idev->lock);
1109
1110         /* A temporary address is created only if this calculated Preferred
1111          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1112          * an implementation must not create a temporary address with a zero
1113          * Preferred Lifetime.
1114          */
1115         if (tmp_prefered_lft <= regen_advance) {
1116                 in6_ifa_put(ifp);
1117                 in6_dev_put(idev);
1118                 ret = -1;
1119                 goto out;
1120         }
1121
1122         addr_flags = IFA_F_TEMPORARY;
1123         /* set in addrconf_prefix_rcv() */
1124         if (ifp->flags & IFA_F_OPTIMISTIC)
1125                 addr_flags |= IFA_F_OPTIMISTIC;
1126
1127         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1128                             ipv6_addr_scope(&addr), addr_flags,
1129                             tmp_valid_lft, tmp_prefered_lft);
1130         if (IS_ERR(ift)) {
1131                 in6_ifa_put(ifp);
1132                 in6_dev_put(idev);
1133                 pr_info("%s: retry temporary address regeneration\n", __func__);
1134                 tmpaddr = &addr;
1135                 write_lock(&idev->lock);
1136                 goto retry;
1137         }
1138
1139         spin_lock_bh(&ift->lock);
1140         ift->ifpub = ifp;
1141         ift->cstamp = now;
1142         ift->tstamp = tmp_tstamp;
1143         spin_unlock_bh(&ift->lock);
1144
1145         addrconf_dad_start(ift);
1146         in6_ifa_put(ift);
1147         in6_dev_put(idev);
1148 out:
1149         return ret;
1150 }
1151 #endif
1152
1153 /*
1154  *      Choose an appropriate source address (RFC3484)
1155  */
1156 enum {
1157         IPV6_SADDR_RULE_INIT = 0,
1158         IPV6_SADDR_RULE_LOCAL,
1159         IPV6_SADDR_RULE_SCOPE,
1160         IPV6_SADDR_RULE_PREFERRED,
1161 #ifdef CONFIG_IPV6_MIP6
1162         IPV6_SADDR_RULE_HOA,
1163 #endif
1164         IPV6_SADDR_RULE_OIF,
1165         IPV6_SADDR_RULE_LABEL,
1166 #ifdef CONFIG_IPV6_PRIVACY
1167         IPV6_SADDR_RULE_PRIVACY,
1168 #endif
1169         IPV6_SADDR_RULE_ORCHID,
1170         IPV6_SADDR_RULE_PREFIX,
1171         IPV6_SADDR_RULE_MAX
1172 };
1173
1174 struct ipv6_saddr_score {
1175         int                     rule;
1176         int                     addr_type;
1177         struct inet6_ifaddr     *ifa;
1178         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1179         int                     scopedist;
1180         int                     matchlen;
1181 };
1182
1183 struct ipv6_saddr_dst {
1184         const struct in6_addr *addr;
1185         int ifindex;
1186         int scope;
1187         int label;
1188         unsigned int prefs;
1189 };
1190
1191 static inline int ipv6_saddr_preferred(int type)
1192 {
1193         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1194                 return 1;
1195         return 0;
1196 }
1197
1198 static int ipv6_get_saddr_eval(struct net *net,
1199                                struct ipv6_saddr_score *score,
1200                                struct ipv6_saddr_dst *dst,
1201                                int i)
1202 {
1203         int ret;
1204
1205         if (i <= score->rule) {
1206                 switch (i) {
1207                 case IPV6_SADDR_RULE_SCOPE:
1208                         ret = score->scopedist;
1209                         break;
1210                 case IPV6_SADDR_RULE_PREFIX:
1211                         ret = score->matchlen;
1212                         break;
1213                 default:
1214                         ret = !!test_bit(i, score->scorebits);
1215                 }
1216                 goto out;
1217         }
1218
1219         switch (i) {
1220         case IPV6_SADDR_RULE_INIT:
1221                 /* Rule 0: remember if hiscore is not ready yet */
1222                 ret = !!score->ifa;
1223                 break;
1224         case IPV6_SADDR_RULE_LOCAL:
1225                 /* Rule 1: Prefer same address */
1226                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1227                 break;
1228         case IPV6_SADDR_RULE_SCOPE:
1229                 /* Rule 2: Prefer appropriate scope
1230                  *
1231                  *      ret
1232                  *       ^
1233                  *    -1 |  d 15
1234                  *    ---+--+-+---> scope
1235                  *       |
1236                  *       |             d is scope of the destination.
1237                  *  B-d  |  \
1238                  *       |   \      <- smaller scope is better if
1239                  *  B-15 |    \        if scope is enough for destinaion.
1240                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1241                  * d-C-1 | /
1242                  *       |/         <- greater is better
1243                  *   -C  /             if scope is not enough for destination.
1244                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1245                  *
1246                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1247                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1248                  * Assume B = 0 and we get C > 29.
1249                  */
1250                 ret = __ipv6_addr_src_scope(score->addr_type);
1251                 if (ret >= dst->scope)
1252                         ret = -ret;
1253                 else
1254                         ret -= 128;     /* 30 is enough */
1255                 score->scopedist = ret;
1256                 break;
1257         case IPV6_SADDR_RULE_PREFERRED:
1258                 /* Rule 3: Avoid deprecated and optimistic addresses */
1259                 ret = ipv6_saddr_preferred(score->addr_type) ||
1260                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1261                 break;
1262 #ifdef CONFIG_IPV6_MIP6
1263         case IPV6_SADDR_RULE_HOA:
1264             {
1265                 /* Rule 4: Prefer home address */
1266                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1267                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1268                 break;
1269             }
1270 #endif
1271         case IPV6_SADDR_RULE_OIF:
1272                 /* Rule 5: Prefer outgoing interface */
1273                 ret = (!dst->ifindex ||
1274                        dst->ifindex == score->ifa->idev->dev->ifindex);
1275                 break;
1276         case IPV6_SADDR_RULE_LABEL:
1277                 /* Rule 6: Prefer matching label */
1278                 ret = ipv6_addr_label(net,
1279                                       &score->ifa->addr, score->addr_type,
1280                                       score->ifa->idev->dev->ifindex) == dst->label;
1281                 break;
1282 #ifdef CONFIG_IPV6_PRIVACY
1283         case IPV6_SADDR_RULE_PRIVACY:
1284             {
1285                 /* Rule 7: Prefer public address
1286                  * Note: prefer temporary address if use_tempaddr >= 2
1287                  */
1288                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1289                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1290                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1291                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1292                 break;
1293             }
1294 #endif
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                       unsigned char banned_flags)
1449 {
1450         struct inet6_ifaddr *ifp;
1451         int err = -EADDRNOTAVAIL;
1452
1453         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1454                 if (ifp->scope == IFA_LINK &&
1455                     !(ifp->flags & banned_flags)) {
1456                         *addr = ifp->addr;
1457                         err = 0;
1458                         break;
1459                 }
1460         }
1461         return err;
1462 }
1463
1464 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1465                     unsigned char banned_flags)
1466 {
1467         struct inet6_dev *idev;
1468         int err = -EADDRNOTAVAIL;
1469
1470         rcu_read_lock();
1471         idev = __in6_dev_get(dev);
1472         if (idev) {
1473                 read_lock_bh(&idev->lock);
1474                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1475                 read_unlock_bh(&idev->lock);
1476         }
1477         rcu_read_unlock();
1478         return err;
1479 }
1480
1481 static int ipv6_count_addresses(struct inet6_dev *idev)
1482 {
1483         int cnt = 0;
1484         struct inet6_ifaddr *ifp;
1485
1486         read_lock_bh(&idev->lock);
1487         list_for_each_entry(ifp, &idev->addr_list, if_list)
1488                 cnt++;
1489         read_unlock_bh(&idev->lock);
1490         return cnt;
1491 }
1492
1493 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1494                   const struct net_device *dev, int strict)
1495 {
1496         struct inet6_ifaddr *ifp;
1497         unsigned int hash = inet6_addr_hash(addr);
1498
1499         rcu_read_lock_bh();
1500         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1501                 if (!net_eq(dev_net(ifp->idev->dev), net))
1502                         continue;
1503                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1504                     !(ifp->flags&IFA_F_TENTATIVE) &&
1505                     (dev == NULL || ifp->idev->dev == dev ||
1506                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1507                         rcu_read_unlock_bh();
1508                         return 1;
1509                 }
1510         }
1511
1512         rcu_read_unlock_bh();
1513         return 0;
1514 }
1515 EXPORT_SYMBOL(ipv6_chk_addr);
1516
1517 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1518                                struct net_device *dev)
1519 {
1520         unsigned int hash = inet6_addr_hash(addr);
1521         struct inet6_ifaddr *ifp;
1522
1523         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1524                 if (!net_eq(dev_net(ifp->idev->dev), net))
1525                         continue;
1526                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1527                         if (dev == NULL || ifp->idev->dev == dev)
1528                                 return true;
1529                 }
1530         }
1531         return false;
1532 }
1533
1534 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1535 {
1536         struct inet6_dev *idev;
1537         struct inet6_ifaddr *ifa;
1538         int     onlink;
1539
1540         onlink = 0;
1541         rcu_read_lock();
1542         idev = __in6_dev_get(dev);
1543         if (idev) {
1544                 read_lock_bh(&idev->lock);
1545                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1546                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1547                                                    ifa->prefix_len);
1548                         if (onlink)
1549                                 break;
1550                 }
1551                 read_unlock_bh(&idev->lock);
1552         }
1553         rcu_read_unlock();
1554         return onlink;
1555 }
1556 EXPORT_SYMBOL(ipv6_chk_prefix);
1557
1558 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1559                                      struct net_device *dev, int strict)
1560 {
1561         struct inet6_ifaddr *ifp, *result = NULL;
1562         unsigned int hash = inet6_addr_hash(addr);
1563
1564         rcu_read_lock_bh();
1565         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1566                 if (!net_eq(dev_net(ifp->idev->dev), net))
1567                         continue;
1568                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1569                         if (dev == NULL || ifp->idev->dev == dev ||
1570                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1571                                 result = ifp;
1572                                 in6_ifa_hold(ifp);
1573                                 break;
1574                         }
1575                 }
1576         }
1577         rcu_read_unlock_bh();
1578
1579         return result;
1580 }
1581
1582 /* Gets referenced address, destroys ifaddr */
1583
1584 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1585 {
1586         if (ifp->flags&IFA_F_PERMANENT) {
1587                 spin_lock_bh(&ifp->lock);
1588                 addrconf_del_dad_timer(ifp);
1589                 ifp->flags |= IFA_F_TENTATIVE;
1590                 if (dad_failed)
1591                         ifp->flags |= IFA_F_DADFAILED;
1592                 spin_unlock_bh(&ifp->lock);
1593                 if (dad_failed)
1594                         ipv6_ifa_notify(0, ifp);
1595                 in6_ifa_put(ifp);
1596 #ifdef CONFIG_IPV6_PRIVACY
1597         } else if (ifp->flags&IFA_F_TEMPORARY) {
1598                 struct inet6_ifaddr *ifpub;
1599                 spin_lock_bh(&ifp->lock);
1600                 ifpub = ifp->ifpub;
1601                 if (ifpub) {
1602                         in6_ifa_hold(ifpub);
1603                         spin_unlock_bh(&ifp->lock);
1604                         ipv6_create_tempaddr(ifpub, ifp);
1605                         in6_ifa_put(ifpub);
1606                 } else {
1607                         spin_unlock_bh(&ifp->lock);
1608                 }
1609                 ipv6_del_addr(ifp);
1610 #endif
1611         } else
1612                 ipv6_del_addr(ifp);
1613 }
1614
1615 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1616 {
1617         int err = -ENOENT;
1618
1619         spin_lock(&ifp->state_lock);
1620         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1621                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1622                 err = 0;
1623         }
1624         spin_unlock(&ifp->state_lock);
1625
1626         return err;
1627 }
1628
1629 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1630 {
1631         struct inet6_dev *idev = ifp->idev;
1632
1633         if (addrconf_dad_end(ifp)) {
1634                 in6_ifa_put(ifp);
1635                 return;
1636         }
1637
1638         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1639                              ifp->idev->dev->name, &ifp->addr);
1640
1641         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1642                 struct in6_addr addr;
1643
1644                 addr.s6_addr32[0] = htonl(0xfe800000);
1645                 addr.s6_addr32[1] = 0;
1646
1647                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1648                     ipv6_addr_equal(&ifp->addr, &addr)) {
1649                         /* DAD failed for link-local based on MAC address */
1650                         idev->cnf.disable_ipv6 = 1;
1651
1652                         pr_info("%s: IPv6 being disabled!\n",
1653                                 ifp->idev->dev->name);
1654                 }
1655         }
1656
1657         addrconf_dad_stop(ifp, 1);
1658 }
1659
1660 /* Join to solicited addr multicast group. */
1661
1662 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1663 {
1664         struct in6_addr maddr;
1665
1666         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1667                 return;
1668
1669         addrconf_addr_solict_mult(addr, &maddr);
1670         ipv6_dev_mc_inc(dev, &maddr);
1671 }
1672
1673 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1674 {
1675         struct in6_addr maddr;
1676
1677         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1678                 return;
1679
1680         addrconf_addr_solict_mult(addr, &maddr);
1681         __ipv6_dev_mc_dec(idev, &maddr);
1682 }
1683
1684 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1685 {
1686         struct in6_addr addr;
1687         if (ifp->prefix_len == 127) /* RFC 6164 */
1688                 return;
1689         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1690         if (ipv6_addr_any(&addr))
1691                 return;
1692         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1693 }
1694
1695 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1696 {
1697         struct in6_addr addr;
1698         if (ifp->prefix_len == 127) /* RFC 6164 */
1699                 return;
1700         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1701         if (ipv6_addr_any(&addr))
1702                 return;
1703         __ipv6_dev_ac_dec(ifp->idev, &addr);
1704 }
1705
1706 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1707 {
1708         if (dev->addr_len != ETH_ALEN)
1709                 return -1;
1710         memcpy(eui, dev->dev_addr, 3);
1711         memcpy(eui + 5, dev->dev_addr + 3, 3);
1712
1713         /*
1714          * The zSeries OSA network cards can be shared among various
1715          * OS instances, but the OSA cards have only one MAC address.
1716          * This leads to duplicate address conflicts in conjunction
1717          * with IPv6 if more than one instance uses the same card.
1718          *
1719          * The driver for these cards can deliver a unique 16-bit
1720          * identifier for each instance sharing the same card.  It is
1721          * placed instead of 0xFFFE in the interface identifier.  The
1722          * "u" bit of the interface identifier is not inverted in this
1723          * case.  Hence the resulting interface identifier has local
1724          * scope according to RFC2373.
1725          */
1726         if (dev->dev_id) {
1727                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1728                 eui[4] = dev->dev_id & 0xFF;
1729         } else {
1730                 eui[3] = 0xFF;
1731                 eui[4] = 0xFE;
1732                 eui[0] ^= 2;
1733         }
1734         return 0;
1735 }
1736
1737 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1738 {
1739         if (dev->addr_len != IEEE802154_ADDR_LEN)
1740                 return -1;
1741         memcpy(eui, dev->dev_addr, 8);
1742         eui[0] ^= 2;
1743         return 0;
1744 }
1745
1746 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1747 {
1748         union fwnet_hwaddr *ha;
1749
1750         if (dev->addr_len != FWNET_ALEN)
1751                 return -1;
1752
1753         ha = (union fwnet_hwaddr *)dev->dev_addr;
1754
1755         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1756         eui[0] ^= 2;
1757         return 0;
1758 }
1759
1760 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1761 {
1762         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1763         if (dev->addr_len != ARCNET_ALEN)
1764                 return -1;
1765         memset(eui, 0, 7);
1766         eui[7] = *(u8 *)dev->dev_addr;
1767         return 0;
1768 }
1769
1770 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1771 {
1772         if (dev->addr_len != INFINIBAND_ALEN)
1773                 return -1;
1774         memcpy(eui, dev->dev_addr + 12, 8);
1775         eui[0] |= 2;
1776         return 0;
1777 }
1778
1779 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1780 {
1781         if (addr == 0)
1782                 return -1;
1783         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1784                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1785                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1786                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1787                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1788                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1789         eui[1] = 0;
1790         eui[2] = 0x5E;
1791         eui[3] = 0xFE;
1792         memcpy(eui + 4, &addr, 4);
1793         return 0;
1794 }
1795
1796 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1797 {
1798         if (dev->priv_flags & IFF_ISATAP)
1799                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1800         return -1;
1801 }
1802
1803 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1804 {
1805         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1806 }
1807
1808 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1809 {
1810         memcpy(eui, dev->perm_addr, 3);
1811         memcpy(eui + 5, dev->perm_addr + 3, 3);
1812         eui[3] = 0xFF;
1813         eui[4] = 0xFE;
1814         eui[0] ^= 2;
1815         return 0;
1816 }
1817
1818 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1819 {
1820         switch (dev->type) {
1821         case ARPHRD_ETHER:
1822         case ARPHRD_FDDI:
1823                 return addrconf_ifid_eui48(eui, dev);
1824         case ARPHRD_ARCNET:
1825                 return addrconf_ifid_arcnet(eui, dev);
1826         case ARPHRD_INFINIBAND:
1827                 return addrconf_ifid_infiniband(eui, dev);
1828         case ARPHRD_SIT:
1829                 return addrconf_ifid_sit(eui, dev);
1830         case ARPHRD_IPGRE:
1831                 return addrconf_ifid_gre(eui, dev);
1832         case ARPHRD_IEEE802154:
1833                 return addrconf_ifid_eui64(eui, dev);
1834         case ARPHRD_IEEE1394:
1835                 return addrconf_ifid_ieee1394(eui, dev);
1836         case ARPHRD_TUNNEL6:
1837                 return addrconf_ifid_ip6tnl(eui, dev);
1838         }
1839         return -1;
1840 }
1841
1842 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1843 {
1844         int err = -1;
1845         struct inet6_ifaddr *ifp;
1846
1847         read_lock_bh(&idev->lock);
1848         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1849                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1850                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1851                         err = 0;
1852                         break;
1853                 }
1854         }
1855         read_unlock_bh(&idev->lock);
1856         return err;
1857 }
1858
1859 #ifdef CONFIG_IPV6_PRIVACY
1860 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1861 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1862 {
1863 regen:
1864         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1865         idev->rndid[0] &= ~0x02;
1866
1867         /*
1868          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1869          * check if generated address is not inappropriate
1870          *
1871          *  - Reserved subnet anycast (RFC 2526)
1872          *      11111101 11....11 1xxxxxxx
1873          *  - ISATAP (RFC4214) 6.1
1874          *      00-00-5E-FE-xx-xx-xx-xx
1875          *  - value 0
1876          *  - XXX: already assigned to an address on the device
1877          */
1878         if (idev->rndid[0] == 0xfd &&
1879             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1880             (idev->rndid[7]&0x80))
1881                 goto regen;
1882         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1883                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1884                         goto regen;
1885                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1886                         goto regen;
1887         }
1888 }
1889
1890 static void ipv6_regen_rndid(unsigned long data)
1891 {
1892         struct inet6_dev *idev = (struct inet6_dev *) data;
1893         unsigned long expires;
1894
1895         rcu_read_lock_bh();
1896         write_lock_bh(&idev->lock);
1897
1898         if (idev->dead)
1899                 goto out;
1900
1901         __ipv6_regen_rndid(idev);
1902
1903         expires = jiffies +
1904                 idev->cnf.temp_prefered_lft * HZ -
1905                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1906                 idev->cnf.max_desync_factor * HZ;
1907         if (time_before(expires, jiffies)) {
1908                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1909                         __func__, idev->dev->name);
1910                 goto out;
1911         }
1912
1913         if (!mod_timer(&idev->regen_timer, expires))
1914                 in6_dev_hold(idev);
1915
1916 out:
1917         write_unlock_bh(&idev->lock);
1918         rcu_read_unlock_bh();
1919         in6_dev_put(idev);
1920 }
1921
1922 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1923 {
1924         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1925                 __ipv6_regen_rndid(idev);
1926 }
1927 #endif
1928
1929 /*
1930  *      Add prefix route.
1931  */
1932
1933 static void
1934 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1935                       unsigned long expires, u32 flags)
1936 {
1937         struct fib6_config cfg = {
1938                 .fc_table = RT6_TABLE_PREFIX,
1939                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1940                 .fc_ifindex = dev->ifindex,
1941                 .fc_expires = expires,
1942                 .fc_dst_len = plen,
1943                 .fc_flags = RTF_UP | flags,
1944                 .fc_nlinfo.nl_net = dev_net(dev),
1945                 .fc_protocol = RTPROT_KERNEL,
1946         };
1947
1948         cfg.fc_dst = *pfx;
1949
1950         /* Prevent useless cloning on PtP SIT.
1951            This thing is done here expecting that the whole
1952            class of non-broadcast devices need not cloning.
1953          */
1954 #if IS_ENABLED(CONFIG_IPV6_SIT)
1955         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1956                 cfg.fc_flags |= RTF_NONEXTHOP;
1957 #endif
1958
1959         ip6_route_add(&cfg);
1960 }
1961
1962
1963 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1964                                                   int plen,
1965                                                   const struct net_device *dev,
1966                                                   u32 flags, u32 noflags)
1967 {
1968         struct fib6_node *fn;
1969         struct rt6_info *rt = NULL;
1970         struct fib6_table *table;
1971
1972         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1973         if (table == NULL)
1974                 return NULL;
1975
1976         read_lock_bh(&table->tb6_lock);
1977         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1978         if (!fn)
1979                 goto out;
1980         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1981                 if (rt->dst.dev->ifindex != dev->ifindex)
1982                         continue;
1983                 if ((rt->rt6i_flags & flags) != flags)
1984                         continue;
1985                 if ((rt->rt6i_flags & noflags) != 0)
1986                         continue;
1987                 dst_hold(&rt->dst);
1988                 break;
1989         }
1990 out:
1991         read_unlock_bh(&table->tb6_lock);
1992         return rt;
1993 }
1994
1995
1996 /* Create "default" multicast route to the interface */
1997
1998 static void addrconf_add_mroute(struct net_device *dev)
1999 {
2000         struct fib6_config cfg = {
2001                 .fc_table = RT6_TABLE_LOCAL,
2002                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2003                 .fc_ifindex = dev->ifindex,
2004                 .fc_dst_len = 8,
2005                 .fc_flags = RTF_UP,
2006                 .fc_nlinfo.nl_net = dev_net(dev),
2007         };
2008
2009         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2010
2011         ip6_route_add(&cfg);
2012 }
2013
2014 #if IS_ENABLED(CONFIG_IPV6_SIT)
2015 static void sit_route_add(struct net_device *dev)
2016 {
2017         struct fib6_config cfg = {
2018                 .fc_table = RT6_TABLE_MAIN,
2019                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2020                 .fc_ifindex = dev->ifindex,
2021                 .fc_dst_len = 96,
2022                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2023                 .fc_nlinfo.nl_net = dev_net(dev),
2024         };
2025
2026         /* prefix length - 96 bits "::d.d.d.d" */
2027         ip6_route_add(&cfg);
2028 }
2029 #endif
2030
2031 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2032 {
2033         struct inet6_dev *idev;
2034
2035         ASSERT_RTNL();
2036
2037         idev = ipv6_find_idev(dev);
2038         if (!idev)
2039                 return ERR_PTR(-ENOBUFS);
2040
2041         if (idev->cnf.disable_ipv6)
2042                 return ERR_PTR(-EACCES);
2043
2044         /* Add default multicast route */
2045         if (!(dev->flags & IFF_LOOPBACK))
2046                 addrconf_add_mroute(dev);
2047
2048         return idev;
2049 }
2050
2051 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2052 {
2053         struct prefix_info *pinfo;
2054         __u32 valid_lft;
2055         __u32 prefered_lft;
2056         int addr_type;
2057         struct inet6_dev *in6_dev;
2058         struct net *net = dev_net(dev);
2059
2060         pinfo = (struct prefix_info *) opt;
2061
2062         if (len < sizeof(struct prefix_info)) {
2063                 ADBG("addrconf: prefix option too short\n");
2064                 return;
2065         }
2066
2067         /*
2068          *      Validation checks ([ADDRCONF], page 19)
2069          */
2070
2071         addr_type = ipv6_addr_type(&pinfo->prefix);
2072
2073         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2074                 return;
2075
2076         valid_lft = ntohl(pinfo->valid);
2077         prefered_lft = ntohl(pinfo->prefered);
2078
2079         if (prefered_lft > valid_lft) {
2080                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2081                 return;
2082         }
2083
2084         in6_dev = in6_dev_get(dev);
2085
2086         if (in6_dev == NULL) {
2087                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2088                                     dev->name);
2089                 return;
2090         }
2091
2092         /*
2093          *      Two things going on here:
2094          *      1) Add routes for on-link prefixes
2095          *      2) Configure prefixes with the auto flag set
2096          */
2097
2098         if (pinfo->onlink) {
2099                 struct rt6_info *rt;
2100                 unsigned long rt_expires;
2101
2102                 /* Avoid arithmetic overflow. Really, we could
2103                  * save rt_expires in seconds, likely valid_lft,
2104                  * but it would require division in fib gc, that it
2105                  * not good.
2106                  */
2107                 if (HZ > USER_HZ)
2108                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2109                 else
2110                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2111
2112                 if (addrconf_finite_timeout(rt_expires))
2113                         rt_expires *= HZ;
2114
2115                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2116                                                pinfo->prefix_len,
2117                                                dev,
2118                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2119                                                RTF_GATEWAY | RTF_DEFAULT);
2120
2121                 if (rt) {
2122                         /* Autoconf prefix route */
2123                         if (valid_lft == 0) {
2124                                 ip6_del_rt(rt);
2125                                 rt = NULL;
2126                         } else if (addrconf_finite_timeout(rt_expires)) {
2127                                 /* not infinity */
2128                                 rt6_set_expires(rt, jiffies + rt_expires);
2129                         } else {
2130                                 rt6_clean_expires(rt);
2131                         }
2132                 } else if (valid_lft) {
2133                         clock_t expires = 0;
2134                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2135                         if (addrconf_finite_timeout(rt_expires)) {
2136                                 /* not infinity */
2137                                 flags |= RTF_EXPIRES;
2138                                 expires = jiffies_to_clock_t(rt_expires);
2139                         }
2140                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2141                                               dev, expires, flags);
2142                 }
2143                 ip6_rt_put(rt);
2144         }
2145
2146         /* Try to figure out our local address for this prefix */
2147
2148         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2149                 struct inet6_ifaddr *ifp;
2150                 struct in6_addr addr;
2151                 int create = 0, update_lft = 0;
2152                 bool tokenized = false;
2153
2154                 if (pinfo->prefix_len == 64) {
2155                         memcpy(&addr, &pinfo->prefix, 8);
2156
2157                         if (!ipv6_addr_any(&in6_dev->token)) {
2158                                 read_lock_bh(&in6_dev->lock);
2159                                 memcpy(addr.s6_addr + 8,
2160                                        in6_dev->token.s6_addr + 8, 8);
2161                                 read_unlock_bh(&in6_dev->lock);
2162                                 tokenized = true;
2163                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2164                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2165                                 in6_dev_put(in6_dev);
2166                                 return;
2167                         }
2168                         goto ok;
2169                 }
2170                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2171                                     pinfo->prefix_len);
2172                 in6_dev_put(in6_dev);
2173                 return;
2174
2175 ok:
2176
2177                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2178
2179                 if (ifp == NULL && valid_lft) {
2180                         int max_addresses = in6_dev->cnf.max_addresses;
2181                         u32 addr_flags = 0;
2182
2183 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2184                         if (in6_dev->cnf.optimistic_dad &&
2185                             !net->ipv6.devconf_all->forwarding && sllao)
2186                                 addr_flags = IFA_F_OPTIMISTIC;
2187 #endif
2188
2189                         /* Do not allow to create too much of autoconfigured
2190                          * addresses; this would be too easy way to crash kernel.
2191                          */
2192                         if (!max_addresses ||
2193                             ipv6_count_addresses(in6_dev) < max_addresses)
2194                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2195                                                     pinfo->prefix_len,
2196                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2197                                                     addr_flags, valid_lft,
2198                                                     prefered_lft);
2199
2200                         if (IS_ERR_OR_NULL(ifp)) {
2201                                 in6_dev_put(in6_dev);
2202                                 return;
2203                         }
2204
2205                         update_lft = 0;
2206                         create = 1;
2207                         ifp->cstamp = jiffies;
2208                         ifp->tokenized = tokenized;
2209                         addrconf_dad_start(ifp);
2210                 }
2211
2212                 if (ifp) {
2213                         int flags;
2214                         unsigned long now;
2215 #ifdef CONFIG_IPV6_PRIVACY
2216                         struct inet6_ifaddr *ift;
2217 #endif
2218                         u32 stored_lft;
2219
2220                         /* update lifetime (RFC2462 5.5.3 e) */
2221                         spin_lock(&ifp->lock);
2222                         now = jiffies;
2223                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2224                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2225                         else
2226                                 stored_lft = 0;
2227                         if (!update_lft && !create && stored_lft) {
2228                                 if (valid_lft > MIN_VALID_LIFETIME ||
2229                                     valid_lft > stored_lft)
2230                                         update_lft = 1;
2231                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2232                                         /* valid_lft <= stored_lft is always true */
2233                                         /*
2234                                          * RFC 4862 Section 5.5.3e:
2235                                          * "Note that the preferred lifetime of
2236                                          *  the corresponding address is always
2237                                          *  reset to the Preferred Lifetime in
2238                                          *  the received Prefix Information
2239                                          *  option, regardless of whether the
2240                                          *  valid lifetime is also reset or
2241                                          *  ignored."
2242                                          *
2243                                          *  So if the preferred lifetime in
2244                                          *  this advertisement is different
2245                                          *  than what we have stored, but the
2246                                          *  valid lifetime is invalid, just
2247                                          *  reset prefered_lft.
2248                                          *
2249                                          *  We must set the valid lifetime
2250                                          *  to the stored lifetime since we'll
2251                                          *  be updating the timestamp below,
2252                                          *  else we'll set it back to the
2253                                          *  minimum.
2254                                          */
2255                                         if (prefered_lft != ifp->prefered_lft) {
2256                                                 valid_lft = stored_lft;
2257                                                 update_lft = 1;
2258                                         }
2259                                 } else {
2260                                         valid_lft = MIN_VALID_LIFETIME;
2261                                         if (valid_lft < prefered_lft)
2262                                                 prefered_lft = valid_lft;
2263                                         update_lft = 1;
2264                                 }
2265                         }
2266
2267                         if (update_lft) {
2268                                 ifp->valid_lft = valid_lft;
2269                                 ifp->prefered_lft = prefered_lft;
2270                                 ifp->tstamp = now;
2271                                 flags = ifp->flags;
2272                                 ifp->flags &= ~IFA_F_DEPRECATED;
2273                                 spin_unlock(&ifp->lock);
2274
2275                                 if (!(flags&IFA_F_TENTATIVE))
2276                                         ipv6_ifa_notify(0, ifp);
2277                         } else
2278                                 spin_unlock(&ifp->lock);
2279
2280 #ifdef CONFIG_IPV6_PRIVACY
2281                         read_lock_bh(&in6_dev->lock);
2282                         /* update all temporary addresses in the list */
2283                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2284                                             tmp_list) {
2285                                 int age, max_valid, max_prefered;
2286
2287                                 if (ifp != ift->ifpub)
2288                                         continue;
2289
2290                                 /*
2291                                  * RFC 4941 section 3.3:
2292                                  * If a received option will extend the lifetime
2293                                  * of a public address, the lifetimes of
2294                                  * temporary addresses should be extended,
2295                                  * subject to the overall constraint that no
2296                                  * temporary addresses should ever remain
2297                                  * "valid" or "preferred" for a time longer than
2298                                  * (TEMP_VALID_LIFETIME) or
2299                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2300                                  * respectively.
2301                                  */
2302                                 age = (now - ift->cstamp) / HZ;
2303                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2304                                 if (max_valid < 0)
2305                                         max_valid = 0;
2306
2307                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2308                                                in6_dev->cnf.max_desync_factor -
2309                                                age;
2310                                 if (max_prefered < 0)
2311                                         max_prefered = 0;
2312
2313                                 if (valid_lft > max_valid)
2314                                         valid_lft = max_valid;
2315
2316                                 if (prefered_lft > max_prefered)
2317                                         prefered_lft = max_prefered;
2318
2319                                 spin_lock(&ift->lock);
2320                                 flags = ift->flags;
2321                                 ift->valid_lft = valid_lft;
2322                                 ift->prefered_lft = prefered_lft;
2323                                 ift->tstamp = now;
2324                                 if (prefered_lft > 0)
2325                                         ift->flags &= ~IFA_F_DEPRECATED;
2326
2327                                 spin_unlock(&ift->lock);
2328                                 if (!(flags&IFA_F_TENTATIVE))
2329                                         ipv6_ifa_notify(0, ift);
2330                         }
2331
2332                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2333                                 /*
2334                                  * When a new public address is created as
2335                                  * described in [ADDRCONF], also create a new
2336                                  * temporary address. Also create a temporary
2337                                  * address if it's enabled but no temporary
2338                                  * address currently exists.
2339                                  */
2340                                 read_unlock_bh(&in6_dev->lock);
2341                                 ipv6_create_tempaddr(ifp, NULL);
2342                         } else {
2343                                 read_unlock_bh(&in6_dev->lock);
2344                         }
2345 #endif
2346                         in6_ifa_put(ifp);
2347                         addrconf_verify(0);
2348                 }
2349         }
2350         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2351         in6_dev_put(in6_dev);
2352 }
2353
2354 /*
2355  *      Set destination address.
2356  *      Special case for SIT interfaces where we create a new "virtual"
2357  *      device.
2358  */
2359 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2360 {
2361         struct in6_ifreq ireq;
2362         struct net_device *dev;
2363         int err = -EINVAL;
2364
2365         rtnl_lock();
2366
2367         err = -EFAULT;
2368         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2369                 goto err_exit;
2370
2371         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2372
2373         err = -ENODEV;
2374         if (dev == NULL)
2375                 goto err_exit;
2376
2377 #if IS_ENABLED(CONFIG_IPV6_SIT)
2378         if (dev->type == ARPHRD_SIT) {
2379                 const struct net_device_ops *ops = dev->netdev_ops;
2380                 struct ifreq ifr;
2381                 struct ip_tunnel_parm p;
2382
2383                 err = -EADDRNOTAVAIL;
2384                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2385                         goto err_exit;
2386
2387                 memset(&p, 0, sizeof(p));
2388                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2389                 p.iph.saddr = 0;
2390                 p.iph.version = 4;
2391                 p.iph.ihl = 5;
2392                 p.iph.protocol = IPPROTO_IPV6;
2393                 p.iph.ttl = 64;
2394                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2395
2396                 if (ops->ndo_do_ioctl) {
2397                         mm_segment_t oldfs = get_fs();
2398
2399                         set_fs(KERNEL_DS);
2400                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2401                         set_fs(oldfs);
2402                 } else
2403                         err = -EOPNOTSUPP;
2404
2405                 if (err == 0) {
2406                         err = -ENOBUFS;
2407                         dev = __dev_get_by_name(net, p.name);
2408                         if (!dev)
2409                                 goto err_exit;
2410                         err = dev_open(dev);
2411                 }
2412         }
2413 #endif
2414
2415 err_exit:
2416         rtnl_unlock();
2417         return err;
2418 }
2419
2420 /*
2421  *      Manual configuration of address on an interface
2422  */
2423 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2424                           const struct in6_addr *peer_pfx,
2425                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2426                           __u32 valid_lft)
2427 {
2428         struct inet6_ifaddr *ifp;
2429         struct inet6_dev *idev;
2430         struct net_device *dev;
2431         int scope;
2432         u32 flags;
2433         clock_t expires;
2434         unsigned long timeout;
2435
2436         ASSERT_RTNL();
2437
2438         if (plen > 128)
2439                 return -EINVAL;
2440
2441         /* check the lifetime */
2442         if (!valid_lft || prefered_lft > valid_lft)
2443                 return -EINVAL;
2444
2445         dev = __dev_get_by_index(net, ifindex);
2446         if (!dev)
2447                 return -ENODEV;
2448
2449         idev = addrconf_add_dev(dev);
2450         if (IS_ERR(idev))
2451                 return PTR_ERR(idev);
2452
2453         scope = ipv6_addr_scope(pfx);
2454
2455         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2456         if (addrconf_finite_timeout(timeout)) {
2457                 expires = jiffies_to_clock_t(timeout * HZ);
2458                 valid_lft = timeout;
2459                 flags = RTF_EXPIRES;
2460         } else {
2461                 expires = 0;
2462                 flags = 0;
2463                 ifa_flags |= IFA_F_PERMANENT;
2464         }
2465
2466         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2467         if (addrconf_finite_timeout(timeout)) {
2468                 if (timeout == 0)
2469                         ifa_flags |= IFA_F_DEPRECATED;
2470                 prefered_lft = timeout;
2471         }
2472
2473         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2474                             valid_lft, prefered_lft);
2475
2476         if (!IS_ERR(ifp)) {
2477                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2478                                       expires, flags);
2479                 /*
2480                  * Note that section 3.1 of RFC 4429 indicates
2481                  * that the Optimistic flag should not be set for
2482                  * manually configured addresses
2483                  */
2484                 addrconf_dad_start(ifp);
2485                 in6_ifa_put(ifp);
2486                 addrconf_verify(0);
2487                 return 0;
2488         }
2489
2490         return PTR_ERR(ifp);
2491 }
2492
2493 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2494                           unsigned int plen)
2495 {
2496         struct inet6_ifaddr *ifp;
2497         struct inet6_dev *idev;
2498         struct net_device *dev;
2499
2500         if (plen > 128)
2501                 return -EINVAL;
2502
2503         dev = __dev_get_by_index(net, ifindex);
2504         if (!dev)
2505                 return -ENODEV;
2506
2507         if ((idev = __in6_dev_get(dev)) == NULL)
2508                 return -ENXIO;
2509
2510         read_lock_bh(&idev->lock);
2511         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2512                 if (ifp->prefix_len == plen &&
2513                     ipv6_addr_equal(pfx, &ifp->addr)) {
2514                         in6_ifa_hold(ifp);
2515                         read_unlock_bh(&idev->lock);
2516
2517                         ipv6_del_addr(ifp);
2518                         return 0;
2519                 }
2520         }
2521         read_unlock_bh(&idev->lock);
2522         return -EADDRNOTAVAIL;
2523 }
2524
2525
2526 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2527 {
2528         struct in6_ifreq ireq;
2529         int err;
2530
2531         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2532                 return -EPERM;
2533
2534         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2535                 return -EFAULT;
2536
2537         rtnl_lock();
2538         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2539                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2540                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2541         rtnl_unlock();
2542         return err;
2543 }
2544
2545 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2546 {
2547         struct in6_ifreq ireq;
2548         int err;
2549
2550         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2551                 return -EPERM;
2552
2553         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2554                 return -EFAULT;
2555
2556         rtnl_lock();
2557         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2558                              ireq.ifr6_prefixlen);
2559         rtnl_unlock();
2560         return err;
2561 }
2562
2563 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2564                      int plen, int scope)
2565 {
2566         struct inet6_ifaddr *ifp;
2567
2568         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2569                             scope, IFA_F_PERMANENT, 0, 0);
2570         if (!IS_ERR(ifp)) {
2571                 spin_lock_bh(&ifp->lock);
2572                 ifp->flags &= ~IFA_F_TENTATIVE;
2573                 spin_unlock_bh(&ifp->lock);
2574                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2575                 in6_ifa_put(ifp);
2576         }
2577 }
2578
2579 #if IS_ENABLED(CONFIG_IPV6_SIT)
2580 static void sit_add_v4_addrs(struct inet6_dev *idev)
2581 {
2582         struct in6_addr addr;
2583         struct net_device *dev;
2584         struct net *net = dev_net(idev->dev);
2585         int scope;
2586
2587         ASSERT_RTNL();
2588
2589         memset(&addr, 0, sizeof(struct in6_addr));
2590         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2591
2592         if (idev->dev->flags&IFF_POINTOPOINT) {
2593                 addr.s6_addr32[0] = htonl(0xfe800000);
2594                 scope = IFA_LINK;
2595         } else {
2596                 scope = IPV6_ADDR_COMPATv4;
2597         }
2598
2599         if (addr.s6_addr32[3]) {
2600                 add_addr(idev, &addr, 128, scope);
2601                 return;
2602         }
2603
2604         for_each_netdev(net, dev) {
2605                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2606                 if (in_dev && (dev->flags & IFF_UP)) {
2607                         struct in_ifaddr *ifa;
2608
2609                         int flag = scope;
2610
2611                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2612                                 int plen;
2613
2614                                 addr.s6_addr32[3] = ifa->ifa_local;
2615
2616                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2617                                         continue;
2618                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2619                                         if (idev->dev->flags&IFF_POINTOPOINT)
2620                                                 continue;
2621                                         flag |= IFA_HOST;
2622                                 }
2623                                 if (idev->dev->flags&IFF_POINTOPOINT)
2624                                         plen = 64;
2625                                 else
2626                                         plen = 96;
2627
2628                                 add_addr(idev, &addr, plen, flag);
2629                         }
2630                 }
2631         }
2632 }
2633 #endif
2634
2635 static void init_loopback(struct net_device *dev)
2636 {
2637         struct inet6_dev  *idev;
2638         struct net_device *sp_dev;
2639         struct inet6_ifaddr *sp_ifa;
2640         struct rt6_info *sp_rt;
2641
2642         /* ::1 */
2643
2644         ASSERT_RTNL();
2645
2646         if ((idev = ipv6_find_idev(dev)) == NULL) {
2647                 pr_debug("%s: add_dev failed\n", __func__);
2648                 return;
2649         }
2650
2651         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2652
2653         /* Add routes to other interface's IPv6 addresses */
2654         for_each_netdev(dev_net(dev), sp_dev) {
2655                 if (!strcmp(sp_dev->name, dev->name))
2656                         continue;
2657
2658                 idev = __in6_dev_get(sp_dev);
2659                 if (!idev)
2660                         continue;
2661
2662                 read_lock_bh(&idev->lock);
2663                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2664
2665                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2666                                 continue;
2667
2668                         if (sp_ifa->rt)
2669                                 continue;
2670
2671                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2672
2673                         /* Failure cases are ignored */
2674                         if (!IS_ERR(sp_rt)) {
2675                                 sp_ifa->rt = sp_rt;
2676                                 ip6_ins_rt(sp_rt);
2677                         }
2678                 }
2679                 read_unlock_bh(&idev->lock);
2680         }
2681 }
2682
2683 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2684 {
2685         struct inet6_ifaddr *ifp;
2686         u32 addr_flags = IFA_F_PERMANENT;
2687
2688 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2689         if (idev->cnf.optimistic_dad &&
2690             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2691                 addr_flags |= IFA_F_OPTIMISTIC;
2692 #endif
2693
2694
2695         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2696         if (!IS_ERR(ifp)) {
2697                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2698                 addrconf_dad_start(ifp);
2699                 in6_ifa_put(ifp);
2700         }
2701 }
2702
2703 static void addrconf_dev_config(struct net_device *dev)
2704 {
2705         struct in6_addr addr;
2706         struct inet6_dev *idev;
2707
2708         ASSERT_RTNL();
2709
2710         if ((dev->type != ARPHRD_ETHER) &&
2711             (dev->type != ARPHRD_FDDI) &&
2712             (dev->type != ARPHRD_ARCNET) &&
2713             (dev->type != ARPHRD_INFINIBAND) &&
2714             (dev->type != ARPHRD_IEEE802154) &&
2715             (dev->type != ARPHRD_IEEE1394) &&
2716             (dev->type != ARPHRD_TUNNEL6)) {
2717                 /* Alas, we support only Ethernet autoconfiguration. */
2718                 return;
2719         }
2720
2721         idev = addrconf_add_dev(dev);
2722         if (IS_ERR(idev))
2723                 return;
2724
2725         memset(&addr, 0, sizeof(struct in6_addr));
2726         addr.s6_addr32[0] = htonl(0xFE800000);
2727
2728         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2729                 addrconf_add_linklocal(idev, &addr);
2730 }
2731
2732 #if IS_ENABLED(CONFIG_IPV6_SIT)
2733 static void addrconf_sit_config(struct net_device *dev)
2734 {
2735         struct inet6_dev *idev;
2736
2737         ASSERT_RTNL();
2738
2739         /*
2740          * Configure the tunnel with one of our IPv4
2741          * addresses... we should configure all of
2742          * our v4 addrs in the tunnel
2743          */
2744
2745         if ((idev = ipv6_find_idev(dev)) == NULL) {
2746                 pr_debug("%s: add_dev failed\n", __func__);
2747                 return;
2748         }
2749
2750         if (dev->priv_flags & IFF_ISATAP) {
2751                 struct in6_addr addr;
2752
2753                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2754                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2755                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2756                         addrconf_add_linklocal(idev, &addr);
2757                 return;
2758         }
2759
2760         sit_add_v4_addrs(idev);
2761
2762         if (dev->flags&IFF_POINTOPOINT)
2763                 addrconf_add_mroute(dev);
2764         else
2765                 sit_route_add(dev);
2766 }
2767 #endif
2768
2769 #if IS_ENABLED(CONFIG_NET_IPGRE)
2770 static void addrconf_gre_config(struct net_device *dev)
2771 {
2772         struct inet6_dev *idev;
2773         struct in6_addr addr;
2774
2775         ASSERT_RTNL();
2776
2777         if ((idev = ipv6_find_idev(dev)) == NULL) {
2778                 pr_debug("%s: add_dev failed\n", __func__);
2779                 return;
2780         }
2781
2782         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2783         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2784
2785         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2786                 addrconf_add_linklocal(idev, &addr);
2787 }
2788 #endif
2789
2790 static inline int
2791 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2792 {
2793         struct in6_addr lladdr;
2794
2795         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2796                 addrconf_add_linklocal(idev, &lladdr);
2797                 return 0;
2798         }
2799         return -1;
2800 }
2801
2802 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2803                            void *ptr)
2804 {
2805         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2806         struct inet6_dev *idev = __in6_dev_get(dev);
2807         int run_pending = 0;
2808         int err;
2809
2810         switch (event) {
2811         case NETDEV_REGISTER:
2812                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2813                         idev = ipv6_add_dev(dev);
2814                         if (!idev)
2815                                 return notifier_from_errno(-ENOMEM);
2816                 }
2817                 break;
2818
2819         case NETDEV_UP:
2820         case NETDEV_CHANGE:
2821                 if (dev->flags & IFF_SLAVE)
2822                         break;
2823
2824                 if (event == NETDEV_UP) {
2825                         if (!addrconf_qdisc_ok(dev)) {
2826                                 /* device is not ready yet. */
2827                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2828                                         dev->name);
2829                                 break;
2830                         }
2831
2832                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2833                                 idev = ipv6_add_dev(dev);
2834
2835                         if (idev) {
2836                                 idev->if_flags |= IF_READY;
2837                                 run_pending = 1;
2838                         }
2839                 } else {
2840                         if (!addrconf_qdisc_ok(dev)) {
2841                                 /* device is still not ready. */
2842                                 break;
2843                         }
2844
2845                         if (idev) {
2846                                 if (idev->if_flags & IF_READY)
2847                                         /* device is already configured. */
2848                                         break;
2849                                 idev->if_flags |= IF_READY;
2850                         }
2851
2852                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2853                                 dev->name);
2854
2855                         run_pending = 1;
2856                 }
2857
2858                 switch (dev->type) {
2859 #if IS_ENABLED(CONFIG_IPV6_SIT)
2860                 case ARPHRD_SIT:
2861                         addrconf_sit_config(dev);
2862                         break;
2863 #endif
2864 #if IS_ENABLED(CONFIG_NET_IPGRE)
2865                 case ARPHRD_IPGRE:
2866                         addrconf_gre_config(dev);
2867                         break;
2868 #endif
2869                 case ARPHRD_LOOPBACK:
2870                         init_loopback(dev);
2871                         break;
2872
2873                 default:
2874                         addrconf_dev_config(dev);
2875                         break;
2876                 }
2877
2878                 if (idev) {
2879                         if (run_pending)
2880                                 addrconf_dad_run(idev);
2881
2882                         /*
2883                          * If the MTU changed during the interface down,
2884                          * when the interface up, the changed MTU must be
2885                          * reflected in the idev as well as routers.
2886                          */
2887                         if (idev->cnf.mtu6 != dev->mtu &&
2888                             dev->mtu >= IPV6_MIN_MTU) {
2889                                 rt6_mtu_change(dev, dev->mtu);
2890                                 idev->cnf.mtu6 = dev->mtu;
2891                         }
2892                         idev->tstamp = jiffies;
2893                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2894
2895                         /*
2896                          * If the changed mtu during down is lower than
2897                          * IPV6_MIN_MTU stop IPv6 on this interface.
2898                          */
2899                         if (dev->mtu < IPV6_MIN_MTU)
2900                                 addrconf_ifdown(dev, 1);
2901                 }
2902                 break;
2903
2904         case NETDEV_CHANGEMTU:
2905                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2906                         rt6_mtu_change(dev, dev->mtu);
2907                         idev->cnf.mtu6 = dev->mtu;
2908                         break;
2909                 }
2910
2911                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2912                         idev = ipv6_add_dev(dev);
2913                         if (idev)
2914                                 break;
2915                 }
2916
2917                 /*
2918                  * MTU falled under IPV6_MIN_MTU.
2919                  * Stop IPv6 on this interface.
2920                  */
2921
2922         case NETDEV_DOWN:
2923         case NETDEV_UNREGISTER:
2924                 /*
2925                  *      Remove all addresses from this interface.
2926                  */
2927                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2928                 break;
2929
2930         case NETDEV_CHANGENAME:
2931                 if (idev) {
2932                         snmp6_unregister_dev(idev);
2933                         addrconf_sysctl_unregister(idev);
2934                         addrconf_sysctl_register(idev);
2935                         err = snmp6_register_dev(idev);
2936                         if (err)
2937                                 return notifier_from_errno(err);
2938                 }
2939                 break;
2940
2941         case NETDEV_PRE_TYPE_CHANGE:
2942         case NETDEV_POST_TYPE_CHANGE:
2943                 addrconf_type_change(dev, event);
2944                 break;
2945         }
2946
2947         return NOTIFY_OK;
2948 }
2949
2950 /*
2951  *      addrconf module should be notified of a device going up
2952  */
2953 static struct notifier_block ipv6_dev_notf = {
2954         .notifier_call = addrconf_notify,
2955 };
2956
2957 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2958 {
2959         struct inet6_dev *idev;
2960         ASSERT_RTNL();
2961
2962         idev = __in6_dev_get(dev);
2963
2964         if (event == NETDEV_POST_TYPE_CHANGE)
2965                 ipv6_mc_remap(idev);
2966         else if (event == NETDEV_PRE_TYPE_CHANGE)
2967                 ipv6_mc_unmap(idev);
2968 }
2969
2970 static int addrconf_ifdown(struct net_device *dev, int how)
2971 {
2972         struct net *net = dev_net(dev);
2973         struct inet6_dev *idev;
2974         struct inet6_ifaddr *ifa;
2975         int state, i;
2976
2977         ASSERT_RTNL();
2978
2979         rt6_ifdown(net, dev);
2980         neigh_ifdown(&nd_tbl, dev);
2981
2982         idev = __in6_dev_get(dev);
2983         if (idev == NULL)
2984                 return -ENODEV;
2985
2986         /*
2987          * Step 1: remove reference to ipv6 device from parent device.
2988          *         Do not dev_put!
2989          */
2990         if (how) {
2991                 idev->dead = 1;
2992
2993                 /* protected by rtnl_lock */
2994                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2995
2996                 /* Step 1.5: remove snmp6 entry */
2997                 snmp6_unregister_dev(idev);
2998
2999         }
3000
3001         /* Step 2: clear hash table */
3002         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3003                 struct hlist_head *h = &inet6_addr_lst[i];
3004
3005                 spin_lock_bh(&addrconf_hash_lock);
3006         restart:
3007                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3008                         if (ifa->idev == idev) {
3009                                 hlist_del_init_rcu(&ifa->addr_lst);
3010                                 addrconf_del_dad_timer(ifa);
3011                                 goto restart;
3012                         }
3013                 }
3014                 spin_unlock_bh(&addrconf_hash_lock);
3015         }
3016
3017         write_lock_bh(&idev->lock);
3018
3019         addrconf_del_rs_timer(idev);
3020
3021         /* Step 2: clear flags for stateless addrconf */
3022         if (!how)
3023                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3024
3025 #ifdef CONFIG_IPV6_PRIVACY
3026         if (how && del_timer(&idev->regen_timer))
3027                 in6_dev_put(idev);
3028
3029         /* Step 3: clear tempaddr list */
3030         while (!list_empty(&idev->tempaddr_list)) {
3031                 ifa = list_first_entry(&idev->tempaddr_list,
3032                                        struct inet6_ifaddr, tmp_list);
3033                 list_del(&ifa->tmp_list);
3034                 write_unlock_bh(&idev->lock);
3035                 spin_lock_bh(&ifa->lock);
3036
3037                 if (ifa->ifpub) {
3038                         in6_ifa_put(ifa->ifpub);
3039                         ifa->ifpub = NULL;
3040                 }
3041                 spin_unlock_bh(&ifa->lock);
3042                 in6_ifa_put(ifa);
3043                 write_lock_bh(&idev->lock);
3044         }
3045 #endif
3046
3047         while (!list_empty(&idev->addr_list)) {
3048                 ifa = list_first_entry(&idev->addr_list,
3049                                        struct inet6_ifaddr, if_list);
3050                 addrconf_del_dad_timer(ifa);
3051
3052                 list_del(&ifa->if_list);
3053
3054                 write_unlock_bh(&idev->lock);
3055
3056                 spin_lock_bh(&ifa->state_lock);
3057                 state = ifa->state;
3058                 ifa->state = INET6_IFADDR_STATE_DEAD;
3059                 spin_unlock_bh(&ifa->state_lock);
3060
3061                 if (state != INET6_IFADDR_STATE_DEAD) {
3062                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3063                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3064                 }
3065                 in6_ifa_put(ifa);
3066
3067                 write_lock_bh(&idev->lock);
3068         }
3069
3070         write_unlock_bh(&idev->lock);
3071
3072         /* Step 5: Discard multicast list */
3073         if (how)
3074                 ipv6_mc_destroy_dev(idev);
3075         else
3076                 ipv6_mc_down(idev);
3077
3078         idev->tstamp = jiffies;
3079
3080         /* Last: Shot the device (if unregistered) */
3081         if (how) {
3082                 addrconf_sysctl_unregister(idev);
3083                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3084                 neigh_ifdown(&nd_tbl, dev);
3085                 in6_dev_put(idev);
3086         }
3087         return 0;
3088 }
3089
3090 static void addrconf_rs_timer(unsigned long data)
3091 {
3092         struct inet6_dev *idev = (struct inet6_dev *)data;
3093         struct net_device *dev = idev->dev;
3094         struct in6_addr lladdr;
3095
3096         write_lock(&idev->lock);
3097         if (idev->dead || !(idev->if_flags & IF_READY))
3098                 goto out;
3099
3100         if (!ipv6_accept_ra(idev))
3101                 goto out;
3102
3103         /* Announcement received after solicitation was sent */
3104         if (idev->if_flags & IF_RA_RCVD)
3105                 goto out;
3106
3107         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3108                 write_unlock(&idev->lock);
3109                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3110                         ndisc_send_rs(dev, &lladdr,
3111                                       &in6addr_linklocal_allrouters);
3112                 else
3113                         goto put;
3114
3115                 write_lock(&idev->lock);
3116                 /* The wait after the last probe can be shorter */
3117                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3118                                              idev->cnf.rtr_solicits) ?
3119                                       idev->cnf.rtr_solicit_delay :
3120                                       idev->cnf.rtr_solicit_interval);
3121         } else {
3122                 /*
3123                  * Note: we do not support deprecated "all on-link"
3124                  * assumption any longer.
3125                  */
3126                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3127         }
3128
3129 out:
3130         write_unlock(&idev->lock);
3131 put:
3132         in6_dev_put(idev);
3133 }
3134
3135 /*
3136  *      Duplicate Address Detection
3137  */
3138 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3139 {
3140         unsigned long rand_num;
3141         struct inet6_dev *idev = ifp->idev;
3142
3143         if (ifp->flags & IFA_F_OPTIMISTIC)
3144                 rand_num = 0;
3145         else
3146                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3147
3148         ifp->dad_probes = idev->cnf.dad_transmits;
3149         addrconf_mod_dad_timer(ifp, rand_num);
3150 }
3151
3152 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3153 {
3154         struct inet6_dev *idev = ifp->idev;
3155         struct net_device *dev = idev->dev;
3156
3157         addrconf_join_solict(dev, &ifp->addr);
3158
3159         net_srandom(ifp->addr.s6_addr32[3]);
3160
3161         read_lock_bh(&idev->lock);
3162         spin_lock(&ifp->lock);
3163         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3164                 goto out;
3165
3166         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3167             idev->cnf.accept_dad < 1 ||
3168             !(ifp->flags&IFA_F_TENTATIVE) ||
3169             ifp->flags & IFA_F_NODAD) {
3170                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3171                 spin_unlock(&ifp->lock);
3172                 read_unlock_bh(&idev->lock);
3173
3174                 addrconf_dad_completed(ifp);
3175                 return;
3176         }
3177
3178         if (!(idev->if_flags & IF_READY)) {
3179                 spin_unlock(&ifp->lock);
3180                 read_unlock_bh(&idev->lock);
3181                 /*
3182                  * If the device is not ready:
3183                  * - keep it tentative if it is a permanent address.
3184                  * - otherwise, kill it.
3185                  */
3186                 in6_ifa_hold(ifp);
3187                 addrconf_dad_stop(ifp, 0);
3188                 return;
3189         }
3190
3191         /*
3192          * Optimistic nodes can start receiving
3193          * Frames right away
3194          */
3195         if (ifp->flags & IFA_F_OPTIMISTIC)
3196                 ip6_ins_rt(ifp->rt);
3197
3198         addrconf_dad_kick(ifp);
3199 out:
3200         spin_unlock(&ifp->lock);
3201         read_unlock_bh(&idev->lock);
3202 }
3203
3204 static void addrconf_dad_timer(unsigned long data)
3205 {
3206         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3207         struct inet6_dev *idev = ifp->idev;
3208         struct in6_addr mcaddr;
3209
3210         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3211                 goto out;
3212
3213         write_lock(&idev->lock);
3214         if (idev->dead || !(idev->if_flags & IF_READY)) {
3215                 write_unlock(&idev->lock);
3216                 goto out;
3217         }
3218
3219         spin_lock(&ifp->lock);
3220         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3221                 spin_unlock(&ifp->lock);
3222                 write_unlock(&idev->lock);
3223                 goto out;
3224         }
3225
3226         if (ifp->dad_probes == 0) {
3227                 /*
3228                  * DAD was successful
3229                  */
3230
3231                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3232                 spin_unlock(&ifp->lock);
3233                 write_unlock(&idev->lock);
3234
3235                 addrconf_dad_completed(ifp);
3236
3237                 goto out;
3238         }
3239
3240         ifp->dad_probes--;
3241         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3242         spin_unlock(&ifp->lock);
3243         write_unlock(&idev->lock);
3244
3245         /* send a neighbour solicitation for our addr */
3246         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3247         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3248 out:
3249         in6_ifa_put(ifp);
3250 }
3251
3252 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3253 {
3254         struct net_device *dev = ifp->idev->dev;
3255         struct in6_addr lladdr;
3256         bool send_rs, send_mld;
3257
3258         addrconf_del_dad_timer(ifp);
3259
3260         /*
3261          *      Configure the address for reception. Now it is valid.
3262          */
3263
3264         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3265
3266         /* If added prefix is link local and we are prepared to process
3267            router advertisements, start sending router solicitations.
3268          */
3269
3270         read_lock_bh(&ifp->idev->lock);
3271         spin_lock(&ifp->lock);
3272         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3273                    ifp->idev->valid_ll_addr_cnt == 1;
3274         send_rs = send_mld &&
3275                   ipv6_accept_ra(ifp->idev) &&
3276                   ifp->idev->cnf.rtr_solicits > 0 &&
3277                   (dev->flags&IFF_LOOPBACK) == 0;
3278         spin_unlock(&ifp->lock);
3279         read_unlock_bh(&ifp->idev->lock);
3280
3281         /* While dad is in progress mld report's source address is in6_addrany.
3282          * Resend with proper ll now.
3283          */
3284         if (send_mld)
3285                 ipv6_mc_dad_complete(ifp->idev);
3286
3287         if (send_rs) {
3288                 /*
3289                  *      If a host as already performed a random delay
3290                  *      [...] as part of DAD [...] there is no need
3291                  *      to delay again before sending the first RS
3292                  */
3293                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3294                         return;
3295                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3296
3297                 write_lock_bh(&ifp->idev->lock);
3298                 spin_lock(&ifp->lock);
3299                 ifp->idev->rs_probes = 1;
3300                 ifp->idev->if_flags |= IF_RS_SENT;
3301                 addrconf_mod_rs_timer(ifp->idev,
3302                                       ifp->idev->cnf.rtr_solicit_interval);
3303                 spin_unlock(&ifp->lock);
3304                 write_unlock_bh(&ifp->idev->lock);
3305         }
3306 }
3307
3308 static void addrconf_dad_run(struct inet6_dev *idev)
3309 {
3310         struct inet6_ifaddr *ifp;
3311
3312         read_lock_bh(&idev->lock);
3313         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3314                 spin_lock(&ifp->lock);
3315                 if (ifp->flags & IFA_F_TENTATIVE &&
3316                     ifp->state == INET6_IFADDR_STATE_DAD)
3317                         addrconf_dad_kick(ifp);
3318                 spin_unlock(&ifp->lock);
3319         }
3320         read_unlock_bh(&idev->lock);
3321 }
3322
3323 #ifdef CONFIG_PROC_FS
3324 struct if6_iter_state {
3325         struct seq_net_private p;
3326         int bucket;
3327         int offset;
3328 };
3329
3330 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3331 {
3332         struct inet6_ifaddr *ifa = NULL;
3333         struct if6_iter_state *state = seq->private;
3334         struct net *net = seq_file_net(seq);
3335         int p = 0;
3336
3337         /* initial bucket if pos is 0 */
3338         if (pos == 0) {
3339                 state->bucket = 0;
3340                 state->offset = 0;
3341         }
3342
3343         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3344                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3345                                          addr_lst) {
3346                         if (!net_eq(dev_net(ifa->idev->dev), net))
3347                                 continue;
3348                         /* sync with offset */
3349                         if (p < state->offset) {
3350                                 p++;
3351                                 continue;
3352                         }
3353                         state->offset++;
3354                         return ifa;
3355                 }
3356
3357                 /* prepare for next bucket */
3358                 state->offset = 0;
3359                 p = 0;
3360         }
3361         return NULL;
3362 }
3363
3364 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3365                                          struct inet6_ifaddr *ifa)
3366 {
3367         struct if6_iter_state *state = seq->private;
3368         struct net *net = seq_file_net(seq);
3369
3370         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3371                 if (!net_eq(dev_net(ifa->idev->dev), net))
3372                         continue;
3373                 state->offset++;
3374                 return ifa;
3375         }
3376
3377         while (++state->bucket < IN6_ADDR_HSIZE) {
3378                 state->offset = 0;
3379                 hlist_for_each_entry_rcu_bh(ifa,
3380                                      &inet6_addr_lst[state->bucket], addr_lst) {
3381                         if (!net_eq(dev_net(ifa->idev->dev), net))
3382                                 continue;
3383                         state->offset++;
3384                         return ifa;
3385                 }
3386         }
3387
3388         return NULL;
3389 }
3390
3391 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3392         __acquires(rcu_bh)
3393 {
3394         rcu_read_lock_bh();
3395         return if6_get_first(seq, *pos);
3396 }
3397
3398 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3399 {
3400         struct inet6_ifaddr *ifa;
3401
3402         ifa = if6_get_next(seq, v);
3403         ++*pos;
3404         return ifa;
3405 }
3406
3407 static void if6_seq_stop(struct seq_file *seq, void *v)
3408         __releases(rcu_bh)
3409 {
3410         rcu_read_unlock_bh();
3411 }
3412
3413 static int if6_seq_show(struct seq_file *seq, void *v)
3414 {
3415         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3416         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3417                    &ifp->addr,
3418                    ifp->idev->dev->ifindex,
3419                    ifp->prefix_len,
3420                    ifp->scope,
3421                    ifp->flags,
3422                    ifp->idev->dev->name);
3423         return 0;
3424 }
3425
3426 static const struct seq_operations if6_seq_ops = {
3427         .start  = if6_seq_start,
3428         .next   = if6_seq_next,
3429         .show   = if6_seq_show,
3430         .stop   = if6_seq_stop,
3431 };
3432
3433 static int if6_seq_open(struct inode *inode, struct file *file)
3434 {
3435         return seq_open_net(inode, file, &if6_seq_ops,
3436                             sizeof(struct if6_iter_state));
3437 }
3438
3439 static const struct file_operations if6_fops = {
3440         .owner          = THIS_MODULE,
3441         .open           = if6_seq_open,
3442         .read           = seq_read,
3443         .llseek         = seq_lseek,
3444         .release        = seq_release_net,
3445 };
3446
3447 static int __net_init if6_proc_net_init(struct net *net)
3448 {
3449         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3450                 return -ENOMEM;
3451         return 0;
3452 }
3453
3454 static void __net_exit if6_proc_net_exit(struct net *net)
3455 {
3456         remove_proc_entry("if_inet6", net->proc_net);
3457 }
3458
3459 static struct pernet_operations if6_proc_net_ops = {
3460        .init = if6_proc_net_init,
3461        .exit = if6_proc_net_exit,
3462 };
3463
3464 int __init if6_proc_init(void)
3465 {
3466         return register_pernet_subsys(&if6_proc_net_ops);
3467 }
3468
3469 void if6_proc_exit(void)
3470 {
3471         unregister_pernet_subsys(&if6_proc_net_ops);
3472 }
3473 #endif  /* CONFIG_PROC_FS */
3474
3475 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3476 /* Check if address is a home address configured on any interface. */
3477 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3478 {
3479         int ret = 0;
3480         struct inet6_ifaddr *ifp = NULL;
3481         unsigned int hash = inet6_addr_hash(addr);
3482
3483         rcu_read_lock_bh();
3484         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3485                 if (!net_eq(dev_net(ifp->idev->dev), net))
3486                         continue;
3487                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3488                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3489                         ret = 1;
3490                         break;
3491                 }
3492         }
3493         rcu_read_unlock_bh();
3494         return ret;
3495 }
3496 #endif
3497
3498 /*
3499  *      Periodic address status verification
3500  */
3501
3502 static void addrconf_verify(unsigned long foo)
3503 {
3504         unsigned long now, next, next_sec, next_sched;
3505         struct inet6_ifaddr *ifp;
3506         int i;
3507
3508         rcu_read_lock_bh();
3509         spin_lock(&addrconf_verify_lock);
3510         now = jiffies;
3511         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3512
3513         del_timer(&addr_chk_timer);
3514
3515         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3516 restart:
3517                 hlist_for_each_entry_rcu_bh(ifp,
3518                                          &inet6_addr_lst[i], addr_lst) {
3519                         unsigned long age;
3520
3521                         if (ifp->flags & IFA_F_PERMANENT)
3522                                 continue;
3523
3524                         spin_lock(&ifp->lock);
3525                         /* We try to batch several events at once. */
3526                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3527
3528                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3529                             age >= ifp->valid_lft) {
3530                                 spin_unlock(&ifp->lock);
3531                                 in6_ifa_hold(ifp);
3532                                 ipv6_del_addr(ifp);
3533                                 goto restart;
3534                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3535                                 spin_unlock(&ifp->lock);
3536                                 continue;
3537                         } else if (age >= ifp->prefered_lft) {
3538                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3539                                 int deprecate = 0;
3540
3541                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3542                                         deprecate = 1;
3543                                         ifp->flags |= IFA_F_DEPRECATED;
3544                                 }
3545
3546                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3547                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3548
3549                                 spin_unlock(&ifp->lock);
3550
3551                                 if (deprecate) {
3552                                         in6_ifa_hold(ifp);
3553
3554                                         ipv6_ifa_notify(0, ifp);
3555                                         in6_ifa_put(ifp);
3556                                         goto restart;
3557                                 }
3558 #ifdef CONFIG_IPV6_PRIVACY
3559                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3560                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3561                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3562                                         ifp->idev->cnf.dad_transmits *
3563                                         ifp->idev->nd_parms->retrans_time / HZ;
3564
3565                                 if (age >= ifp->prefered_lft - regen_advance) {
3566                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3567                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3568                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3569                                         if (!ifp->regen_count && ifpub) {
3570                                                 ifp->regen_count++;
3571                                                 in6_ifa_hold(ifp);
3572                                                 in6_ifa_hold(ifpub);
3573                                                 spin_unlock(&ifp->lock);
3574
3575                                                 spin_lock(&ifpub->lock);
3576                                                 ifpub->regen_count = 0;
3577                                                 spin_unlock(&ifpub->lock);
3578                                                 ipv6_create_tempaddr(ifpub, ifp);
3579                                                 in6_ifa_put(ifpub);
3580                                                 in6_ifa_put(ifp);
3581                                                 goto restart;
3582                                         }
3583                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3584                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3585                                 spin_unlock(&ifp->lock);
3586 #endif
3587                         } else {
3588                                 /* ifp->prefered_lft <= ifp->valid_lft */
3589                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3590                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3591                                 spin_unlock(&ifp->lock);
3592                         }
3593                 }
3594         }
3595
3596         next_sec = round_jiffies_up(next);
3597         next_sched = next;
3598
3599         /* If rounded timeout is accurate enough, accept it. */
3600         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3601                 next_sched = next_sec;
3602
3603         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3604         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3605                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3606
3607         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3608               now, next, next_sec, next_sched);
3609
3610         addr_chk_timer.expires = next_sched;
3611         add_timer(&addr_chk_timer);
3612         spin_unlock(&addrconf_verify_lock);
3613         rcu_read_unlock_bh();
3614 }
3615
3616 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3617                                      struct in6_addr **peer_pfx)
3618 {
3619         struct in6_addr *pfx = NULL;
3620
3621         *peer_pfx = NULL;
3622
3623         if (addr)
3624                 pfx = nla_data(addr);
3625
3626         if (local) {
3627                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3628                         *peer_pfx = pfx;
3629                 pfx = nla_data(local);
3630         }
3631
3632         return pfx;
3633 }
3634
3635 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3636         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3637         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3638         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3639 };
3640
3641 static int
3642 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3643 {
3644         struct net *net = sock_net(skb->sk);
3645         struct ifaddrmsg *ifm;
3646         struct nlattr *tb[IFA_MAX+1];
3647         struct in6_addr *pfx, *peer_pfx;
3648         int err;
3649
3650         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3651         if (err < 0)
3652                 return err;
3653
3654         ifm = nlmsg_data(nlh);
3655         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3656         if (pfx == NULL)
3657                 return -EINVAL;
3658
3659         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3660 }
3661
3662 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3663                              u32 prefered_lft, u32 valid_lft)
3664 {
3665         u32 flags;
3666         clock_t expires;
3667         unsigned long timeout;
3668
3669         if (!valid_lft || (prefered_lft > valid_lft))
3670                 return -EINVAL;
3671
3672         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3673         if (addrconf_finite_timeout(timeout)) {
3674                 expires = jiffies_to_clock_t(timeout * HZ);
3675                 valid_lft = timeout;
3676                 flags = RTF_EXPIRES;
3677         } else {
3678                 expires = 0;
3679                 flags = 0;
3680                 ifa_flags |= IFA_F_PERMANENT;
3681         }
3682
3683         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3684         if (addrconf_finite_timeout(timeout)) {
3685                 if (timeout == 0)
3686                         ifa_flags |= IFA_F_DEPRECATED;
3687                 prefered_lft = timeout;
3688         }
3689
3690         spin_lock_bh(&ifp->lock);
3691         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3692         ifp->tstamp = jiffies;
3693         ifp->valid_lft = valid_lft;
3694         ifp->prefered_lft = prefered_lft;
3695
3696         spin_unlock_bh(&ifp->lock);
3697         if (!(ifp->flags&IFA_F_TENTATIVE))
3698                 ipv6_ifa_notify(0, ifp);
3699
3700         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3701                               expires, flags);
3702         addrconf_verify(0);
3703
3704         return 0;
3705 }
3706
3707 static int
3708 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3709 {
3710         struct net *net = sock_net(skb->sk);
3711         struct ifaddrmsg *ifm;
3712         struct nlattr *tb[IFA_MAX+1];
3713         struct in6_addr *pfx, *peer_pfx;
3714         struct inet6_ifaddr *ifa;
3715         struct net_device *dev;
3716         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3717         u8 ifa_flags;
3718         int err;
3719
3720         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3721         if (err < 0)
3722                 return err;
3723
3724         ifm = nlmsg_data(nlh);
3725         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3726         if (pfx == NULL)
3727                 return -EINVAL;
3728
3729         if (tb[IFA_CACHEINFO]) {
3730                 struct ifa_cacheinfo *ci;
3731
3732                 ci = nla_data(tb[IFA_CACHEINFO]);
3733                 valid_lft = ci->ifa_valid;
3734                 preferred_lft = ci->ifa_prefered;
3735         } else {
3736                 preferred_lft = INFINITY_LIFE_TIME;
3737                 valid_lft = INFINITY_LIFE_TIME;
3738         }
3739
3740         dev =  __dev_get_by_index(net, ifm->ifa_index);
3741         if (dev == NULL)
3742                 return -ENODEV;
3743
3744         /* We ignore other flags so far. */
3745         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3746
3747         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3748         if (ifa == NULL) {
3749                 /*
3750                  * It would be best to check for !NLM_F_CREATE here but
3751                  * userspace alreay relies on not having to provide this.
3752                  */
3753                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3754                                       ifm->ifa_prefixlen, ifa_flags,
3755                                       preferred_lft, valid_lft);
3756         }
3757
3758         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3759             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3760                 err = -EEXIST;
3761         else
3762                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3763
3764         in6_ifa_put(ifa);
3765
3766         return err;
3767 }
3768
3769 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3770                           u8 scope, int ifindex)
3771 {
3772         struct ifaddrmsg *ifm;
3773
3774         ifm = nlmsg_data(nlh);
3775         ifm->ifa_family = AF_INET6;
3776         ifm->ifa_prefixlen = prefixlen;
3777         ifm->ifa_flags = flags;
3778         ifm->ifa_scope = scope;
3779         ifm->ifa_index = ifindex;
3780 }
3781
3782 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3783                          unsigned long tstamp, u32 preferred, u32 valid)
3784 {
3785         struct ifa_cacheinfo ci;
3786
3787         ci.cstamp = cstamp_delta(cstamp);
3788         ci.tstamp = cstamp_delta(tstamp);
3789         ci.ifa_prefered = preferred;
3790         ci.ifa_valid = valid;
3791
3792         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3793 }
3794
3795 static inline int rt_scope(int ifa_scope)
3796 {
3797         if (ifa_scope & IFA_HOST)
3798                 return RT_SCOPE_HOST;
3799         else if (ifa_scope & IFA_LINK)
3800                 return RT_SCOPE_LINK;
3801         else if (ifa_scope & IFA_SITE)
3802                 return RT_SCOPE_SITE;
3803         else
3804                 return RT_SCOPE_UNIVERSE;
3805 }
3806
3807 static inline int inet6_ifaddr_msgsize(void)
3808 {
3809         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3810                + nla_total_size(16) /* IFA_LOCAL */
3811                + nla_total_size(16) /* IFA_ADDRESS */
3812                + nla_total_size(sizeof(struct ifa_cacheinfo));
3813 }
3814
3815 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3816                              u32 portid, u32 seq, int event, unsigned int flags)
3817 {
3818         struct nlmsghdr  *nlh;
3819         u32 preferred, valid;
3820
3821         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3822         if (nlh == NULL)
3823                 return -EMSGSIZE;
3824
3825         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3826                       ifa->idev->dev->ifindex);
3827
3828         if (!(ifa->flags&IFA_F_PERMANENT)) {
3829                 preferred = ifa->prefered_lft;
3830                 valid = ifa->valid_lft;
3831                 if (preferred != INFINITY_LIFE_TIME) {
3832                         long tval = (jiffies - ifa->tstamp)/HZ;
3833                         if (preferred > tval)
3834                                 preferred -= tval;
3835                         else
3836                                 preferred = 0;
3837                         if (valid != INFINITY_LIFE_TIME) {
3838                                 if (valid > tval)
3839                                         valid -= tval;
3840                                 else
3841                                         valid = 0;
3842                         }
3843                 }
3844         } else {
3845                 preferred = INFINITY_LIFE_TIME;
3846                 valid = INFINITY_LIFE_TIME;
3847         }
3848
3849         if (!ipv6_addr_any(&ifa->peer_addr)) {
3850                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3851                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3852                         goto error;
3853         } else
3854                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3855                         goto error;
3856
3857         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3858                 goto error;
3859
3860         return nlmsg_end(skb, nlh);
3861
3862 error:
3863         nlmsg_cancel(skb, nlh);
3864         return -EMSGSIZE;
3865 }
3866
3867 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3868                                 u32 portid, u32 seq, int event, u16 flags)
3869 {
3870         struct nlmsghdr  *nlh;
3871         u8 scope = RT_SCOPE_UNIVERSE;
3872         int ifindex = ifmca->idev->dev->ifindex;
3873
3874         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3875                 scope = RT_SCOPE_SITE;
3876
3877         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3878         if (nlh == NULL)
3879                 return -EMSGSIZE;
3880
3881         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3882         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3883             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3884                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3885                 nlmsg_cancel(skb, nlh);
3886                 return -EMSGSIZE;
3887         }
3888
3889         return nlmsg_end(skb, nlh);
3890 }
3891
3892 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3893                                 u32 portid, u32 seq, int event, unsigned int flags)
3894 {
3895         struct nlmsghdr  *nlh;
3896         u8 scope = RT_SCOPE_UNIVERSE;
3897         int ifindex = ifaca->aca_idev->dev->ifindex;
3898
3899         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3900                 scope = RT_SCOPE_SITE;
3901
3902         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3903         if (nlh == NULL)
3904                 return -EMSGSIZE;
3905
3906         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3907         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3908             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3909                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3910                 nlmsg_cancel(skb, nlh);
3911                 return -EMSGSIZE;
3912         }
3913
3914         return nlmsg_end(skb, nlh);
3915 }
3916
3917 enum addr_type_t {
3918         UNICAST_ADDR,
3919         MULTICAST_ADDR,
3920         ANYCAST_ADDR,
3921 };
3922
3923 /* called with rcu_read_lock() */
3924 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3925                           struct netlink_callback *cb, enum addr_type_t type,
3926                           int s_ip_idx, int *p_ip_idx)
3927 {
3928         struct ifmcaddr6 *ifmca;
3929         struct ifacaddr6 *ifaca;
3930         int err = 1;
3931         int ip_idx = *p_ip_idx;
3932
3933         read_lock_bh(&idev->lock);
3934         switch (type) {
3935         case UNICAST_ADDR: {
3936                 struct inet6_ifaddr *ifa;
3937
3938                 /* unicast address incl. temp addr */
3939                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3940                         if (++ip_idx < s_ip_idx)
3941                                 continue;
3942                         err = inet6_fill_ifaddr(skb, ifa,
3943                                                 NETLINK_CB(cb->skb).portid,
3944                                                 cb->nlh->nlmsg_seq,
3945                                                 RTM_NEWADDR,
3946                                                 NLM_F_MULTI);
3947                         if (err <= 0)
3948                                 break;
3949                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3950                 }
3951                 break;
3952         }
3953         case MULTICAST_ADDR:
3954                 /* multicast address */
3955                 for (ifmca = idev->mc_list; ifmca;
3956                      ifmca = ifmca->next, ip_idx++) {
3957                         if (ip_idx < s_ip_idx)
3958                                 continue;
3959                         err = inet6_fill_ifmcaddr(skb, ifmca,
3960                                                   NETLINK_CB(cb->skb).portid,
3961                                                   cb->nlh->nlmsg_seq,
3962                                                   RTM_GETMULTICAST,
3963                                                   NLM_F_MULTI);
3964                         if (err <= 0)
3965                                 break;
3966                 }
3967                 break;
3968         case ANYCAST_ADDR:
3969                 /* anycast address */
3970                 for (ifaca = idev->ac_list; ifaca;
3971                      ifaca = ifaca->aca_next, ip_idx++) {
3972                         if (ip_idx < s_ip_idx)
3973                                 continue;
3974                         err = inet6_fill_ifacaddr(skb, ifaca,
3975                                                   NETLINK_CB(cb->skb).portid,
3976                                                   cb->nlh->nlmsg_seq,
3977                                                   RTM_GETANYCAST,
3978                                                   NLM_F_MULTI);
3979                         if (err <= 0)
3980                                 break;
3981                 }
3982                 break;
3983         default:
3984                 break;
3985         }
3986         read_unlock_bh(&idev->lock);
3987         *p_ip_idx = ip_idx;
3988         return err;
3989 }
3990
3991 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3992                            enum addr_type_t type)
3993 {
3994         struct net *net = sock_net(skb->sk);
3995         int h, s_h;
3996         int idx, ip_idx;
3997         int s_idx, s_ip_idx;
3998         struct net_device *dev;
3999         struct inet6_dev *idev;
4000         struct hlist_head *head;
4001
4002         s_h = cb->args[0];
4003         s_idx = idx = cb->args[1];
4004         s_ip_idx = ip_idx = cb->args[2];
4005
4006         rcu_read_lock();
4007         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4008         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4009                 idx = 0;
4010                 head = &net->dev_index_head[h];
4011                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4012                         if (idx < s_idx)
4013                                 goto cont;
4014                         if (h > s_h || idx > s_idx)
4015                                 s_ip_idx = 0;
4016                         ip_idx = 0;
4017                         idev = __in6_dev_get(dev);
4018                         if (!idev)
4019                                 goto cont;
4020
4021                         if (in6_dump_addrs(idev, skb, cb, type,
4022                                            s_ip_idx, &ip_idx) <= 0)
4023                                 goto done;
4024 cont:
4025                         idx++;
4026                 }
4027         }
4028 done:
4029         rcu_read_unlock();
4030         cb->args[0] = h;
4031         cb->args[1] = idx;
4032         cb->args[2] = ip_idx;
4033
4034         return skb->len;
4035 }
4036
4037 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4038 {
4039         enum addr_type_t type = UNICAST_ADDR;
4040
4041         return inet6_dump_addr(skb, cb, type);
4042 }
4043
4044 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4045 {
4046         enum addr_type_t type = MULTICAST_ADDR;
4047
4048         return inet6_dump_addr(skb, cb, type);
4049 }
4050
4051
4052 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4053 {
4054         enum addr_type_t type = ANYCAST_ADDR;
4055
4056         return inet6_dump_addr(skb, cb, type);
4057 }
4058
4059 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4060 {
4061         struct net *net = sock_net(in_skb->sk);
4062         struct ifaddrmsg *ifm;
4063         struct nlattr *tb[IFA_MAX+1];
4064         struct in6_addr *addr = NULL, *peer;
4065         struct net_device *dev = NULL;
4066         struct inet6_ifaddr *ifa;
4067         struct sk_buff *skb;
4068         int err;
4069
4070         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4071         if (err < 0)
4072                 goto errout;
4073
4074         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4075         if (addr == NULL) {
4076                 err = -EINVAL;
4077                 goto errout;
4078         }
4079
4080         ifm = nlmsg_data(nlh);
4081         if (ifm->ifa_index)
4082                 dev = __dev_get_by_index(net, ifm->ifa_index);
4083
4084         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4085         if (!ifa) {
4086                 err = -EADDRNOTAVAIL;
4087                 goto errout;
4088         }
4089
4090         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4091         if (!skb) {
4092                 err = -ENOBUFS;
4093                 goto errout_ifa;
4094         }
4095
4096         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4097                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4098         if (err < 0) {
4099                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4100                 WARN_ON(err == -EMSGSIZE);
4101                 kfree_skb(skb);
4102                 goto errout_ifa;
4103         }
4104         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4105 errout_ifa:
4106         in6_ifa_put(ifa);
4107 errout:
4108         return err;
4109 }
4110
4111 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4112 {
4113         struct sk_buff *skb;
4114         struct net *net = dev_net(ifa->idev->dev);
4115         int err = -ENOBUFS;
4116
4117         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4118         if (skb == NULL)
4119                 goto errout;
4120
4121         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4122         if (err < 0) {
4123                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4124                 WARN_ON(err == -EMSGSIZE);
4125                 kfree_skb(skb);
4126                 goto errout;
4127         }
4128         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4129         return;
4130 errout:
4131         if (err < 0)
4132                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4133 }
4134
4135 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4136                                 __s32 *array, int bytes)
4137 {
4138         BUG_ON(bytes < (DEVCONF_MAX * 4));
4139
4140         memset(array, 0, bytes);
4141         array[DEVCONF_FORWARDING] = cnf->forwarding;
4142         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4143         array[DEVCONF_MTU6] = cnf->mtu6;
4144         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4145         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4146         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4147         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4148         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4149         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4150                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4151         array[DEVCONF_RTR_SOLICIT_DELAY] =
4152                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4153         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4154         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4155                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4156         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4157                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4158 #ifdef CONFIG_IPV6_PRIVACY
4159         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4160         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4161         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4162         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4163         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4164 #endif
4165         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4166         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4167         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4168 #ifdef CONFIG_IPV6_ROUTER_PREF
4169         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4170         array[DEVCONF_RTR_PROBE_INTERVAL] =
4171                 jiffies_to_msecs(cnf->rtr_probe_interval);
4172 #ifdef CONFIG_IPV6_ROUTE_INFO
4173         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4174 #endif
4175 #endif
4176         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4177         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4178 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4179         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4180 #endif
4181 #ifdef CONFIG_IPV6_MROUTE
4182         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4183 #endif
4184         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4185         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4186         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4187         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4188         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4189 }
4190
4191 static inline size_t inet6_ifla6_size(void)
4192 {
4193         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4194              + nla_total_size(sizeof(struct ifla_cacheinfo))
4195              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4196              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4197              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4198              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4199 }
4200
4201 static inline size_t inet6_if_nlmsg_size(void)
4202 {
4203         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4204                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4205                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4206                + nla_total_size(4) /* IFLA_MTU */
4207                + nla_total_size(4) /* IFLA_LINK */
4208                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4209 }
4210
4211 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4212                                       int items, int bytes)
4213 {
4214         int i;
4215         int pad = bytes - sizeof(u64) * items;
4216         BUG_ON(pad < 0);
4217
4218         /* Use put_unaligned() because stats may not be aligned for u64. */
4219         put_unaligned(items, &stats[0]);
4220         for (i = 1; i < items; i++)
4221                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4222
4223         memset(&stats[items], 0, pad);
4224 }
4225
4226 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4227                                       int items, int bytes, size_t syncpoff)
4228 {
4229         int i;
4230         int pad = bytes - sizeof(u64) * items;
4231         BUG_ON(pad < 0);
4232
4233         /* Use put_unaligned() because stats may not be aligned for u64. */
4234         put_unaligned(items, &stats[0]);
4235         for (i = 1; i < items; i++)
4236                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4237
4238         memset(&stats[items], 0, pad);
4239 }
4240
4241 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4242                              int bytes)
4243 {
4244         switch (attrtype) {
4245         case IFLA_INET6_STATS:
4246                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4247                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4248                 break;
4249         case IFLA_INET6_ICMP6STATS:
4250                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4251                 break;
4252         }
4253 }
4254
4255 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4256 {
4257         struct nlattr *nla;
4258         struct ifla_cacheinfo ci;
4259
4260         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4261                 goto nla_put_failure;
4262         ci.max_reasm_len = IPV6_MAXPLEN;
4263         ci.tstamp = cstamp_delta(idev->tstamp);
4264         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4265         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4266         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4267                 goto nla_put_failure;
4268         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4269         if (nla == NULL)
4270                 goto nla_put_failure;
4271         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4272
4273         /* XXX - MC not implemented */
4274
4275         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4276         if (nla == NULL)
4277                 goto nla_put_failure;
4278         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4279
4280         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4281         if (nla == NULL)
4282                 goto nla_put_failure;
4283         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4284
4285         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4286         if (nla == NULL)
4287                 goto nla_put_failure;
4288         read_lock_bh(&idev->lock);
4289         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4290         read_unlock_bh(&idev->lock);
4291
4292         return 0;
4293
4294 nla_put_failure:
4295         return -EMSGSIZE;
4296 }
4297
4298 static size_t inet6_get_link_af_size(const struct net_device *dev)
4299 {
4300         if (!__in6_dev_get(dev))
4301                 return 0;
4302
4303         return inet6_ifla6_size();
4304 }
4305
4306 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4307 {
4308         struct inet6_dev *idev = __in6_dev_get(dev);
4309
4310         if (!idev)
4311                 return -ENODATA;
4312
4313         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4314                 return -EMSGSIZE;
4315
4316         return 0;
4317 }
4318
4319 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4320 {
4321         struct inet6_ifaddr *ifp;
4322         struct net_device *dev = idev->dev;
4323         bool update_rs = false;
4324         struct in6_addr ll_addr;
4325
4326         if (token == NULL)
4327                 return -EINVAL;
4328         if (ipv6_addr_any(token))
4329                 return -EINVAL;
4330         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4331                 return -EINVAL;
4332         if (!ipv6_accept_ra(idev))
4333                 return -EINVAL;
4334         if (idev->cnf.rtr_solicits <= 0)
4335                 return -EINVAL;
4336
4337         write_lock_bh(&idev->lock);
4338
4339         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4340         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4341
4342         write_unlock_bh(&idev->lock);
4343
4344         if (!idev->dead && (idev->if_flags & IF_READY) &&
4345             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4346                              IFA_F_OPTIMISTIC)) {
4347
4348                 /* If we're not ready, then normal ifup will take care
4349                  * of this. Otherwise, we need to request our rs here.
4350                  */
4351                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4352                 update_rs = true;
4353         }
4354
4355         write_lock_bh(&idev->lock);
4356
4357         if (update_rs) {
4358                 idev->if_flags |= IF_RS_SENT;
4359                 idev->rs_probes = 1;
4360                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4361         }
4362
4363         /* Well, that's kinda nasty ... */
4364         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4365                 spin_lock(&ifp->lock);
4366                 if (ifp->tokenized) {
4367                         ifp->valid_lft = 0;
4368                         ifp->prefered_lft = 0;
4369                 }
4370                 spin_unlock(&ifp->lock);
4371         }
4372
4373         write_unlock_bh(&idev->lock);
4374         addrconf_verify(0);
4375         return 0;
4376 }
4377
4378 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4379 {
4380         int err = -EINVAL;
4381         struct inet6_dev *idev = __in6_dev_get(dev);
4382         struct nlattr *tb[IFLA_INET6_MAX + 1];
4383
4384         if (!idev)
4385                 return -EAFNOSUPPORT;
4386
4387         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4388                 BUG();
4389
4390         if (tb[IFLA_INET6_TOKEN])
4391                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4392
4393         return err;
4394 }
4395
4396 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4397                              u32 portid, u32 seq, int event, unsigned int flags)
4398 {
4399         struct net_device *dev = idev->dev;
4400         struct ifinfomsg *hdr;
4401         struct nlmsghdr *nlh;
4402         void *protoinfo;
4403
4404         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4405         if (nlh == NULL)
4406                 return -EMSGSIZE;
4407
4408         hdr = nlmsg_data(nlh);
4409         hdr->ifi_family = AF_INET6;
4410         hdr->__ifi_pad = 0;
4411         hdr->ifi_type = dev->type;
4412         hdr->ifi_index = dev->ifindex;
4413         hdr->ifi_flags = dev_get_flags(dev);
4414         hdr->ifi_change = 0;
4415
4416         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4417             (dev->addr_len &&
4418              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4419             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4420             (dev->ifindex != dev->iflink &&
4421              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4422                 goto nla_put_failure;
4423         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4424         if (protoinfo == NULL)
4425                 goto nla_put_failure;
4426
4427         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4428                 goto nla_put_failure;
4429
4430         nla_nest_end(skb, protoinfo);
4431         return nlmsg_end(skb, nlh);
4432
4433 nla_put_failure:
4434         nlmsg_cancel(skb, nlh);
4435         return -EMSGSIZE;
4436 }
4437
4438 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4439 {
4440         struct net *net = sock_net(skb->sk);
4441         int h, s_h;
4442         int idx = 0, s_idx;
4443         struct net_device *dev;
4444         struct inet6_dev *idev;
4445         struct hlist_head *head;
4446
4447         s_h = cb->args[0];
4448         s_idx = cb->args[1];
4449
4450         rcu_read_lock();
4451         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4452                 idx = 0;
4453                 head = &net->dev_index_head[h];
4454                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4455                         if (idx < s_idx)
4456                                 goto cont;
4457                         idev = __in6_dev_get(dev);
4458                         if (!idev)
4459                                 goto cont;
4460                         if (inet6_fill_ifinfo(skb, idev,
4461                                               NETLINK_CB(cb->skb).portid,
4462                                               cb->nlh->nlmsg_seq,
4463                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4464                                 goto out;
4465 cont:
4466                         idx++;
4467                 }
4468         }
4469 out:
4470         rcu_read_unlock();
4471         cb->args[1] = idx;
4472         cb->args[0] = h;
4473
4474         return skb->len;
4475 }
4476
4477 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4478 {
4479         struct sk_buff *skb;
4480         struct net *net = dev_net(idev->dev);
4481         int err = -ENOBUFS;
4482
4483         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4484         if (skb == NULL)
4485                 goto errout;
4486
4487         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4488         if (err < 0) {
4489                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4490                 WARN_ON(err == -EMSGSIZE);
4491                 kfree_skb(skb);
4492                 goto errout;
4493         }
4494         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4495         return;
4496 errout:
4497         if (err < 0)
4498                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4499 }
4500
4501 static inline size_t inet6_prefix_nlmsg_size(void)
4502 {
4503         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4504                + nla_total_size(sizeof(struct in6_addr))
4505                + nla_total_size(sizeof(struct prefix_cacheinfo));
4506 }
4507
4508 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4509                              struct prefix_info *pinfo, u32 portid, u32 seq,
4510                              int event, unsigned int flags)
4511 {
4512         struct prefixmsg *pmsg;
4513         struct nlmsghdr *nlh;
4514         struct prefix_cacheinfo ci;
4515
4516         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4517         if (nlh == NULL)
4518                 return -EMSGSIZE;
4519
4520         pmsg = nlmsg_data(nlh);
4521         pmsg->prefix_family = AF_INET6;
4522         pmsg->prefix_pad1 = 0;
4523         pmsg->prefix_pad2 = 0;
4524         pmsg->prefix_ifindex = idev->dev->ifindex;
4525         pmsg->prefix_len = pinfo->prefix_len;
4526         pmsg->prefix_type = pinfo->type;
4527         pmsg->prefix_pad3 = 0;
4528         pmsg->prefix_flags = 0;
4529         if (pinfo->onlink)
4530                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4531         if (pinfo->autoconf)
4532                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4533
4534         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4535                 goto nla_put_failure;
4536         ci.preferred_time = ntohl(pinfo->prefered);
4537         ci.valid_time = ntohl(pinfo->valid);
4538         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4539                 goto nla_put_failure;
4540         return nlmsg_end(skb, nlh);
4541
4542 nla_put_failure:
4543         nlmsg_cancel(skb, nlh);
4544         return -EMSGSIZE;
4545 }
4546
4547 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4548                          struct prefix_info *pinfo)
4549 {
4550         struct sk_buff *skb;
4551         struct net *net = dev_net(idev->dev);
4552         int err = -ENOBUFS;
4553
4554         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4555         if (skb == NULL)
4556                 goto errout;
4557
4558         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4559         if (err < 0) {
4560                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4561                 WARN_ON(err == -EMSGSIZE);
4562                 kfree_skb(skb);
4563                 goto errout;
4564         }
4565         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4566         return;
4567 errout:
4568         if (err < 0)
4569                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4570 }
4571
4572 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4573 {
4574         write_lock_bh(&ifp->idev->lock);
4575         spin_lock(&ifp->lock);
4576         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4577                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4578             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4579                 ifp->idev->valid_ll_addr_cnt += count;
4580         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4581         spin_unlock(&ifp->lock);
4582         write_unlock_bh(&ifp->idev->lock);
4583 }
4584
4585 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4586 {
4587         struct net *net = dev_net(ifp->idev->dev);
4588
4589         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4590
4591         switch (event) {
4592         case RTM_NEWADDR:
4593                 update_valid_ll_addr_cnt(ifp, 1);
4594
4595                 /*
4596                  * If the address was optimistic
4597                  * we inserted the route at the start of
4598                  * our DAD process, so we don't need
4599                  * to do it again
4600                  */
4601                 if (!(ifp->rt->rt6i_node))
4602                         ip6_ins_rt(ifp->rt);
4603                 if (ifp->idev->cnf.forwarding)
4604                         addrconf_join_anycast(ifp);
4605                 if (!ipv6_addr_any(&ifp->peer_addr))
4606                         addrconf_prefix_route(&ifp->peer_addr, 128,
4607                                               ifp->idev->dev, 0, 0);
4608                 break;
4609         case RTM_DELADDR:
4610                 update_valid_ll_addr_cnt(ifp, -1);
4611
4612                 if (ifp->idev->cnf.forwarding)
4613                         addrconf_leave_anycast(ifp);
4614                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4615                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4616                         struct rt6_info *rt;
4617                         struct net_device *dev = ifp->idev->dev;
4618
4619                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4620                                         dev->ifindex, 1);
4621                         if (rt) {
4622                                 dst_hold(&rt->dst);
4623                                 if (ip6_del_rt(rt))
4624                                         dst_free(&rt->dst);
4625                         }
4626                 }
4627                 dst_hold(&ifp->rt->dst);
4628
4629                 if (ip6_del_rt(ifp->rt))
4630                         dst_free(&ifp->rt->dst);
4631                 break;
4632         }
4633         atomic_inc(&net->ipv6.dev_addr_genid);
4634         rt_genid_bump_ipv6(net);
4635 }
4636
4637 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4638 {
4639         rcu_read_lock_bh();
4640         if (likely(ifp->idev->dead == 0))
4641                 __ipv6_ifa_notify(event, ifp);
4642         rcu_read_unlock_bh();
4643 }
4644
4645 #ifdef CONFIG_SYSCTL
4646
4647 static
4648 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4649                            void __user *buffer, size_t *lenp, loff_t *ppos)
4650 {
4651         int *valp = ctl->data;
4652         int val = *valp;
4653         loff_t pos = *ppos;
4654         struct ctl_table lctl;
4655         int ret;
4656
4657         /*
4658          * ctl->data points to idev->cnf.forwarding, we should
4659          * not modify it until we get the rtnl lock.
4660          */
4661         lctl = *ctl;
4662         lctl.data = &val;
4663
4664         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4665
4666         if (write)
4667                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4668         if (ret)
4669                 *ppos = pos;
4670         return ret;
4671 }
4672
4673 static void dev_disable_change(struct inet6_dev *idev)
4674 {
4675         struct netdev_notifier_info info;
4676
4677         if (!idev || !idev->dev)
4678                 return;
4679
4680         netdev_notifier_info_init(&info, idev->dev);
4681         if (idev->cnf.disable_ipv6)
4682                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4683         else
4684                 addrconf_notify(NULL, NETDEV_UP, &info);
4685 }
4686
4687 static void addrconf_disable_change(struct net *net, __s32 newf)
4688 {
4689         struct net_device *dev;
4690         struct inet6_dev *idev;
4691
4692         rcu_read_lock();
4693         for_each_netdev_rcu(net, dev) {
4694                 idev = __in6_dev_get(dev);
4695                 if (idev) {
4696                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4697                         idev->cnf.disable_ipv6 = newf;
4698                         if (changed)
4699                                 dev_disable_change(idev);
4700                 }
4701         }
4702         rcu_read_unlock();
4703 }
4704
4705 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4706 {
4707         struct net *net;
4708         int old;
4709
4710         if (!rtnl_trylock())
4711                 return restart_syscall();
4712
4713         net = (struct net *)table->extra2;
4714         old = *p;
4715         *p = newf;
4716
4717         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4718                 rtnl_unlock();
4719                 return 0;
4720         }
4721
4722         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4723                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4724                 addrconf_disable_change(net, newf);
4725         } else if ((!newf) ^ (!old))
4726                 dev_disable_change((struct inet6_dev *)table->extra1);
4727
4728         rtnl_unlock();
4729         return 0;
4730 }
4731
4732 static
4733 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4734                             void __user *buffer, size_t *lenp, loff_t *ppos)
4735 {
4736         int *valp = ctl->data;
4737         int val = *valp;
4738         loff_t pos = *ppos;
4739         struct ctl_table lctl;
4740         int ret;
4741
4742         /*
4743          * ctl->data points to idev->cnf.disable_ipv6, we should
4744          * not modify it until we get the rtnl lock.
4745          */
4746         lctl = *ctl;
4747         lctl.data = &val;
4748
4749         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4750
4751         if (write)
4752                 ret = addrconf_disable_ipv6(ctl, valp, val);
4753         if (ret)
4754                 *ppos = pos;
4755         return ret;
4756 }
4757
4758 static struct addrconf_sysctl_table
4759 {
4760         struct ctl_table_header *sysctl_header;
4761         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4762 } addrconf_sysctl __read_mostly = {
4763         .sysctl_header = NULL,
4764         .addrconf_vars = {
4765                 {
4766                         .procname       = "forwarding",
4767                         .data           = &ipv6_devconf.forwarding,
4768                         .maxlen         = sizeof(int),
4769                         .mode           = 0644,
4770                         .proc_handler   = addrconf_sysctl_forward,
4771                 },
4772                 {
4773                         .procname       = "hop_limit",
4774                         .data           = &ipv6_devconf.hop_limit,
4775                         .maxlen         = sizeof(int),
4776                         .mode           = 0644,
4777                         .proc_handler   = proc_dointvec,
4778                 },
4779                 {
4780                         .procname       = "mtu",
4781                         .data           = &ipv6_devconf.mtu6,
4782                         .maxlen         = sizeof(int),
4783                         .mode           = 0644,
4784                         .proc_handler   = proc_dointvec,
4785                 },
4786                 {
4787                         .procname       = "accept_ra",
4788                         .data           = &ipv6_devconf.accept_ra,
4789                         .maxlen         = sizeof(int),
4790                         .mode           = 0644,
4791                         .proc_handler   = proc_dointvec,
4792                 },
4793                 {
4794                         .procname       = "accept_redirects",
4795                         .data           = &ipv6_devconf.accept_redirects,
4796                         .maxlen         = sizeof(int),
4797                         .mode           = 0644,
4798                         .proc_handler   = proc_dointvec,
4799                 },
4800                 {
4801                         .procname       = "autoconf",
4802                         .data           = &ipv6_devconf.autoconf,
4803                         .maxlen         = sizeof(int),
4804                         .mode           = 0644,
4805                         .proc_handler   = proc_dointvec,
4806                 },
4807                 {
4808                         .procname       = "dad_transmits",
4809                         .data           = &ipv6_devconf.dad_transmits,
4810                         .maxlen         = sizeof(int),
4811                         .mode           = 0644,
4812                         .proc_handler   = proc_dointvec,
4813                 },
4814                 {
4815                         .procname       = "router_solicitations",
4816                         .data           = &ipv6_devconf.rtr_solicits,
4817                         .maxlen         = sizeof(int),
4818                         .mode           = 0644,
4819                         .proc_handler   = proc_dointvec,
4820                 },
4821                 {
4822                         .procname       = "router_solicitation_interval",
4823                         .data           = &ipv6_devconf.rtr_solicit_interval,
4824                         .maxlen         = sizeof(int),
4825                         .mode           = 0644,
4826                         .proc_handler   = proc_dointvec_jiffies,
4827                 },
4828                 {
4829                         .procname       = "router_solicitation_delay",
4830                         .data           = &ipv6_devconf.rtr_solicit_delay,
4831                         .maxlen         = sizeof(int),
4832                         .mode           = 0644,
4833                         .proc_handler   = proc_dointvec_jiffies,
4834                 },
4835                 {
4836                         .procname       = "force_mld_version",
4837                         .data           = &ipv6_devconf.force_mld_version,
4838                         .maxlen         = sizeof(int),
4839                         .mode           = 0644,
4840                         .proc_handler   = proc_dointvec,
4841                 },
4842                 {
4843                         .procname       = "mldv1_unsolicited_report_interval",
4844                         .data           =
4845                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4846                         .maxlen         = sizeof(int),
4847                         .mode           = 0644,
4848                         .proc_handler   = proc_dointvec_ms_jiffies,
4849                 },
4850                 {
4851                         .procname       = "mldv2_unsolicited_report_interval",
4852                         .data           =
4853                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4854                         .maxlen         = sizeof(int),
4855                         .mode           = 0644,
4856                         .proc_handler   = proc_dointvec_ms_jiffies,
4857                 },
4858 #ifdef CONFIG_IPV6_PRIVACY
4859                 {
4860                         .procname       = "use_tempaddr",
4861                         .data           = &ipv6_devconf.use_tempaddr,
4862                         .maxlen         = sizeof(int),
4863                         .mode           = 0644,
4864                         .proc_handler   = proc_dointvec,
4865                 },
4866                 {
4867                         .procname       = "temp_valid_lft",
4868                         .data           = &ipv6_devconf.temp_valid_lft,
4869                         .maxlen         = sizeof(int),
4870                         .mode           = 0644,
4871                         .proc_handler   = proc_dointvec,
4872                 },
4873                 {
4874                         .procname       = "temp_prefered_lft",
4875                         .data           = &ipv6_devconf.temp_prefered_lft,
4876                         .maxlen         = sizeof(int),
4877                         .mode           = 0644,
4878                         .proc_handler   = proc_dointvec,
4879                 },
4880                 {
4881                         .procname       = "regen_max_retry",
4882                         .data           = &ipv6_devconf.regen_max_retry,
4883                         .maxlen         = sizeof(int),
4884                         .mode           = 0644,
4885                         .proc_handler   = proc_dointvec,
4886                 },
4887                 {
4888                         .procname       = "max_desync_factor",
4889                         .data           = &ipv6_devconf.max_desync_factor,
4890                         .maxlen         = sizeof(int),
4891                         .mode           = 0644,
4892                         .proc_handler   = proc_dointvec,
4893                 },
4894 #endif
4895                 {
4896                         .procname       = "max_addresses",
4897                         .data           = &ipv6_devconf.max_addresses,
4898                         .maxlen         = sizeof(int),
4899                         .mode           = 0644,
4900                         .proc_handler   = proc_dointvec,
4901                 },
4902                 {
4903                         .procname       = "accept_ra_defrtr",
4904                         .data           = &ipv6_devconf.accept_ra_defrtr,
4905                         .maxlen         = sizeof(int),
4906                         .mode           = 0644,
4907                         .proc_handler   = proc_dointvec,
4908                 },
4909                 {
4910                         .procname       = "accept_ra_pinfo",
4911                         .data           = &ipv6_devconf.accept_ra_pinfo,
4912                         .maxlen         = sizeof(int),
4913                         .mode           = 0644,
4914                         .proc_handler   = proc_dointvec,
4915                 },
4916 #ifdef CONFIG_IPV6_ROUTER_PREF
4917                 {
4918                         .procname       = "accept_ra_rtr_pref",
4919                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4920                         .maxlen         = sizeof(int),
4921                         .mode           = 0644,
4922                         .proc_handler   = proc_dointvec,
4923                 },
4924                 {
4925                         .procname       = "router_probe_interval",
4926                         .data           = &ipv6_devconf.rtr_probe_interval,
4927                         .maxlen         = sizeof(int),
4928                         .mode           = 0644,
4929                         .proc_handler   = proc_dointvec_jiffies,
4930                 },
4931 #ifdef CONFIG_IPV6_ROUTE_INFO
4932                 {
4933                         .procname       = "accept_ra_rt_info_max_plen",
4934                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4935                         .maxlen         = sizeof(int),
4936                         .mode           = 0644,
4937                         .proc_handler   = proc_dointvec,
4938                 },
4939 #endif
4940 #endif
4941                 {
4942                         .procname       = "proxy_ndp",
4943                         .data           = &ipv6_devconf.proxy_ndp,
4944                         .maxlen         = sizeof(int),
4945                         .mode           = 0644,
4946                         .proc_handler   = proc_dointvec,
4947                 },
4948                 {
4949                         .procname       = "accept_source_route",
4950                         .data           = &ipv6_devconf.accept_source_route,
4951                         .maxlen         = sizeof(int),
4952                         .mode           = 0644,
4953                         .proc_handler   = proc_dointvec,
4954                 },
4955 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4956                 {
4957                         .procname       = "optimistic_dad",
4958                         .data           = &ipv6_devconf.optimistic_dad,
4959                         .maxlen         = sizeof(int),
4960                         .mode           = 0644,
4961                         .proc_handler   = proc_dointvec,
4962
4963                 },
4964 #endif
4965 #ifdef CONFIG_IPV6_MROUTE
4966                 {
4967                         .procname       = "mc_forwarding",
4968                         .data           = &ipv6_devconf.mc_forwarding,
4969                         .maxlen         = sizeof(int),
4970                         .mode           = 0444,
4971                         .proc_handler   = proc_dointvec,
4972                 },
4973 #endif
4974                 {
4975                         .procname       = "disable_ipv6",
4976                         .data           = &ipv6_devconf.disable_ipv6,
4977                         .maxlen         = sizeof(int),
4978                         .mode           = 0644,
4979                         .proc_handler   = addrconf_sysctl_disable,
4980                 },
4981                 {
4982                         .procname       = "accept_dad",
4983                         .data           = &ipv6_devconf.accept_dad,
4984                         .maxlen         = sizeof(int),
4985                         .mode           = 0644,
4986                         .proc_handler   = proc_dointvec,
4987                 },
4988                 {
4989                         .procname       = "force_tllao",
4990                         .data           = &ipv6_devconf.force_tllao,
4991                         .maxlen         = sizeof(int),
4992                         .mode           = 0644,
4993                         .proc_handler   = proc_dointvec
4994                 },
4995                 {
4996                         .procname       = "ndisc_notify",
4997                         .data           = &ipv6_devconf.ndisc_notify,
4998                         .maxlen         = sizeof(int),
4999                         .mode           = 0644,
5000                         .proc_handler   = proc_dointvec
5001                 },
5002                 {
5003                         .procname       = "suppress_frag_ndisc",
5004                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5005                         .maxlen         = sizeof(int),
5006                         .mode           = 0644,
5007                         .proc_handler   = proc_dointvec
5008                 },
5009                 {
5010                         /* sentinel */
5011                 }
5012         },
5013 };
5014
5015 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5016                 struct inet6_dev *idev, struct ipv6_devconf *p)
5017 {
5018         int i;
5019         struct addrconf_sysctl_table *t;
5020         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5021
5022         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5023         if (t == NULL)
5024                 goto out;
5025
5026         for (i = 0; t->addrconf_vars[i].data; i++) {
5027                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5028                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5029                 t->addrconf_vars[i].extra2 = net;
5030         }
5031
5032         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5033
5034         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5035         if (t->sysctl_header == NULL)
5036                 goto free;
5037
5038         p->sysctl = t;
5039         return 0;
5040
5041 free:
5042         kfree(t);
5043 out:
5044         return -ENOBUFS;
5045 }
5046
5047 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5048 {
5049         struct addrconf_sysctl_table *t;
5050
5051         if (p->sysctl == NULL)
5052                 return;
5053
5054         t = p->sysctl;
5055         p->sysctl = NULL;
5056         unregister_net_sysctl_table(t->sysctl_header);
5057         kfree(t);
5058 }
5059
5060 static void addrconf_sysctl_register(struct inet6_dev *idev)
5061 {
5062         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5063                               &ndisc_ifinfo_sysctl_change);
5064         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5065                                         idev, &idev->cnf);
5066 }
5067
5068 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5069 {
5070         __addrconf_sysctl_unregister(&idev->cnf);
5071         neigh_sysctl_unregister(idev->nd_parms);
5072 }
5073
5074
5075 #endif
5076
5077 static int __net_init addrconf_init_net(struct net *net)
5078 {
5079         int err = -ENOMEM;
5080         struct ipv6_devconf *all, *dflt;
5081
5082         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5083         if (all == NULL)
5084                 goto err_alloc_all;
5085
5086         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5087         if (dflt == NULL)
5088                 goto err_alloc_dflt;
5089
5090         /* these will be inherited by all namespaces */
5091         dflt->autoconf = ipv6_defaults.autoconf;
5092         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5093
5094         net->ipv6.devconf_all = all;
5095         net->ipv6.devconf_dflt = dflt;
5096
5097 #ifdef CONFIG_SYSCTL
5098         err = __addrconf_sysctl_register(net, "all", NULL, all);
5099         if (err < 0)
5100                 goto err_reg_all;
5101
5102         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5103         if (err < 0)
5104                 goto err_reg_dflt;
5105 #endif
5106         return 0;
5107
5108 #ifdef CONFIG_SYSCTL
5109 err_reg_dflt:
5110         __addrconf_sysctl_unregister(all);
5111 err_reg_all:
5112         kfree(dflt);
5113 #endif
5114 err_alloc_dflt:
5115         kfree(all);
5116 err_alloc_all:
5117         return err;
5118 }
5119
5120 static void __net_exit addrconf_exit_net(struct net *net)
5121 {
5122 #ifdef CONFIG_SYSCTL
5123         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5124         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5125 #endif
5126         if (!net_eq(net, &init_net)) {
5127                 kfree(net->ipv6.devconf_dflt);
5128                 kfree(net->ipv6.devconf_all);
5129         }
5130 }
5131
5132 static struct pernet_operations addrconf_ops = {
5133         .init = addrconf_init_net,
5134         .exit = addrconf_exit_net,
5135 };
5136
5137 static struct rtnl_af_ops inet6_ops = {
5138         .family           = AF_INET6,
5139         .fill_link_af     = inet6_fill_link_af,
5140         .get_link_af_size = inet6_get_link_af_size,
5141         .set_link_af      = inet6_set_link_af,
5142 };
5143
5144 /*
5145  *      Init / cleanup code
5146  */
5147
5148 int __init addrconf_init(void)
5149 {
5150         int i, err;
5151
5152         err = ipv6_addr_label_init();
5153         if (err < 0) {
5154                 pr_crit("%s: cannot initialize default policy table: %d\n",
5155                         __func__, err);
5156                 goto out;
5157         }
5158
5159         err = register_pernet_subsys(&addrconf_ops);
5160         if (err < 0)
5161                 goto out_addrlabel;
5162
5163         /* The addrconf netdev notifier requires that loopback_dev
5164          * has it's ipv6 private information allocated and setup
5165          * before it can bring up and give link-local addresses
5166          * to other devices which are up.
5167          *
5168          * Unfortunately, loopback_dev is not necessarily the first
5169          * entry in the global dev_base list of net devices.  In fact,
5170          * it is likely to be the very last entry on that list.
5171          * So this causes the notifier registry below to try and
5172          * give link-local addresses to all devices besides loopback_dev
5173          * first, then loopback_dev, which cases all the non-loopback_dev
5174          * devices to fail to get a link-local address.
5175          *
5176          * So, as a temporary fix, allocate the ipv6 structure for
5177          * loopback_dev first by hand.
5178          * Longer term, all of the dependencies ipv6 has upon the loopback
5179          * device and it being up should be removed.
5180          */
5181         rtnl_lock();
5182         if (!ipv6_add_dev(init_net.loopback_dev))
5183                 err = -ENOMEM;
5184         rtnl_unlock();
5185         if (err)
5186                 goto errlo;
5187
5188         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5189                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5190
5191         register_netdevice_notifier(&ipv6_dev_notf);
5192
5193         addrconf_verify(0);
5194
5195         err = rtnl_af_register(&inet6_ops);
5196         if (err < 0)
5197                 goto errout_af;
5198
5199         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5200                               NULL);
5201         if (err < 0)
5202                 goto errout;
5203
5204         /* Only the first call to __rtnl_register can fail */
5205         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5206         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5207         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5208                         inet6_dump_ifaddr, NULL);
5209         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5210                         inet6_dump_ifmcaddr, NULL);
5211         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5212                         inet6_dump_ifacaddr, NULL);
5213         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5214                         inet6_netconf_dump_devconf, NULL);
5215
5216         ipv6_addr_label_rtnl_register();
5217
5218         return 0;
5219 errout:
5220         rtnl_af_unregister(&inet6_ops);
5221 errout_af:
5222         unregister_netdevice_notifier(&ipv6_dev_notf);
5223 errlo:
5224         unregister_pernet_subsys(&addrconf_ops);
5225 out_addrlabel:
5226         ipv6_addr_label_cleanup();
5227 out:
5228         return err;
5229 }
5230
5231 void addrconf_cleanup(void)
5232 {
5233         struct net_device *dev;
5234         int i;
5235
5236         unregister_netdevice_notifier(&ipv6_dev_notf);
5237         unregister_pernet_subsys(&addrconf_ops);
5238         ipv6_addr_label_cleanup();
5239
5240         rtnl_lock();
5241
5242         __rtnl_af_unregister(&inet6_ops);
5243
5244         /* clean dev list */
5245         for_each_netdev(&init_net, dev) {
5246                 if (__in6_dev_get(dev) == NULL)
5247                         continue;
5248                 addrconf_ifdown(dev, 1);
5249         }
5250         addrconf_ifdown(init_net.loopback_dev, 2);
5251
5252         /*
5253          *      Check hash table.
5254          */
5255         spin_lock_bh(&addrconf_hash_lock);
5256         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5257                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5258         spin_unlock_bh(&addrconf_hash_lock);
5259
5260         del_timer(&addr_chk_timer);
5261         rtnl_unlock();
5262 }