Pileus Git - ~andy/linux/blob - 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(x) printk x
 103 #else
 104 #define ADBG(x)
 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         .dad_transmits          = 1,
 181         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 182         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 183         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 184 #ifdef CONFIG_IPV6_PRIVACY
 185         .use_tempaddr           = 0,
 186         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 187         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 188         .regen_max_retry        = REGEN_MAX_RETRY,
 189         .max_desync_factor      = MAX_DESYNC_FACTOR,
 190 #endif
 191         .max_addresses          = IPV6_MAX_ADDRESSES,
 192         .accept_ra_defrtr       = 1,
 193         .accept_ra_pinfo        = 1,
 194 #ifdef CONFIG_IPV6_ROUTER_PREF
 195         .accept_ra_rtr_pref     = 1,
 196         .rtr_probe_interval     = 60 * HZ,
 197 #ifdef CONFIG_IPV6_ROUTE_INFO
 198         .accept_ra_rt_info_max_plen = 0,
 199 #endif
 200 #endif
 201         .proxy_ndp              = 0,
 202         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 203         .disable_ipv6           = 0,
 204         .accept_dad             = 1,
 205 };
 206
 207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 208         .forwarding             = 0,
 209         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 210         .mtu6                   = IPV6_MIN_MTU,
 211         .accept_ra              = 1,
 212         .accept_redirects       = 1,
 213         .autoconf               = 1,
 214         .dad_transmits          = 1,
 215         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 216         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 217         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 218 #ifdef CONFIG_IPV6_PRIVACY
 219         .use_tempaddr           = 0,
 220         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 221         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 222         .regen_max_retry        = REGEN_MAX_RETRY,
 223         .max_desync_factor      = MAX_DESYNC_FACTOR,
 224 #endif
 225         .max_addresses          = IPV6_MAX_ADDRESSES,
 226         .accept_ra_defrtr       = 1,
 227         .accept_ra_pinfo        = 1,
 228 #ifdef CONFIG_IPV6_ROUTER_PREF
 229         .accept_ra_rtr_pref     = 1,
 230         .rtr_probe_interval     = 60 * HZ,
 231 #ifdef CONFIG_IPV6_ROUTE_INFO
 232         .accept_ra_rt_info_max_plen = 0,
 233 #endif
 234 #endif
 235         .proxy_ndp              = 0,
 236         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 237         .disable_ipv6           = 0,
 238         .accept_dad             = 1,
 239 };
 240
 241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 246 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
 247 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
 248 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
 249
 250 /* Check if a valid qdisc is available */
 251 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 252 {
 253         return !qdisc_tx_is_noop(dev);
 254 }
 255
 256 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 257 {
 258         if (del_timer(&idev->rs_timer))
 259                 __in6_dev_put(idev);
 260 }
 261
 262 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 263 {
 264         if (del_timer(&ifp->dad_timer))
 265                 __in6_ifa_put(ifp);
 266 }
 267
 268 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 269                                   unsigned long when)
 270 {
 271         if (!timer_pending(&idev->rs_timer))
 272                 in6_dev_hold(idev);
 273         mod_timer(&idev->rs_timer, jiffies + when);
 274 }
 275
 276 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 277                                    unsigned long when)
 278 {
 279         if (!timer_pending(&ifp->dad_timer))
 280                 in6_ifa_hold(ifp);
 281         mod_timer(&ifp->dad_timer, jiffies + when);
 282 }
 283
 284 static int snmp6_alloc_dev(struct inet6_dev *idev)
 285 {
 286         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 287                           sizeof(struct ipstats_mib),
 288                           __alignof__(struct ipstats_mib)) < 0)
 289                 goto err_ip;
 290         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 291                                         GFP_KERNEL);
 292         if (!idev->stats.icmpv6dev)
 293                 goto err_icmp;
 294         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 295                                            GFP_KERNEL);
 296         if (!idev->stats.icmpv6msgdev)
 297                 goto err_icmpmsg;
 298
 299         return 0;
 300
 301 err_icmpmsg:
 302         kfree(idev->stats.icmpv6dev);
 303 err_icmp:
 304         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 305 err_ip:
 306         return -ENOMEM;
 307 }
 308
 309 static void snmp6_free_dev(struct inet6_dev *idev)
 310 {
 311         kfree(idev->stats.icmpv6msgdev);
 312         kfree(idev->stats.icmpv6dev);
 313         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 314 }
 315
 316 /* Nobody refers to this device, we may destroy it. */
 317
 318 void in6_dev_finish_destroy(struct inet6_dev *idev)
 319 {
 320         struct net_device *dev = idev->dev;
 321
 322         WARN_ON(!list_empty(&idev->addr_list));
 323         WARN_ON(idev->mc_list != NULL);
 324         WARN_ON(timer_pending(&idev->rs_timer));
 325
 326 #ifdef NET_REFCNT_DEBUG
 327         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 328 #endif
 329         dev_put(dev);
 330         if (!idev->dead) {
 331                 pr_warn("Freeing alive inet6 device %p\n", idev);
 332                 return;
 333         }
 334         snmp6_free_dev(idev);
 335         kfree_rcu(idev, rcu);
 336 }
 337 EXPORT_SYMBOL(in6_dev_finish_destroy);
 338
 339 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 340 {
 341         struct inet6_dev *ndev;
 342
 343         ASSERT_RTNL();
 344
 345         if (dev->mtu < IPV6_MIN_MTU)
 346                 return NULL;
 347
 348         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 349
 350         if (ndev == NULL)
 351                 return NULL;
 352
 353         rwlock_init(&ndev->lock);
 354         ndev->dev = dev;
 355         INIT_LIST_HEAD(&ndev->addr_list);
 356         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 357                     (unsigned long)ndev);
 358         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 359         ndev->cnf.mtu6 = dev->mtu;
 360         ndev->cnf.sysctl = NULL;
 361         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 362         if (ndev->nd_parms == NULL) {
 363                 kfree(ndev);
 364                 return NULL;
 365         }
 366         if (ndev->cnf.forwarding)
 367                 dev_disable_lro(dev);
 368         /* We refer to the device */
 369         dev_hold(dev);
 370
 371         if (snmp6_alloc_dev(ndev) < 0) {
 372                 ADBG((KERN_WARNING
 373                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 374                         __func__, dev->name));
 375                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 376                 dev_put(dev);
 377                 kfree(ndev);
 378                 return NULL;
 379         }
 380
 381         if (snmp6_register_dev(ndev) < 0) {
 382                 ADBG((KERN_WARNING
 383                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 384                         __func__, dev->name));
 385                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 386                 ndev->dead = 1;
 387                 in6_dev_finish_destroy(ndev);
 388                 return NULL;
 389         }
 390
 391         /* One reference from device.  We must do this before
 392          * we invoke __ipv6_regen_rndid().
 393          */
 394         in6_dev_hold(ndev);
 395
 396         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 397                 ndev->cnf.accept_dad = -1;
 398
 399 #if IS_ENABLED(CONFIG_IPV6_SIT)
 400         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 401                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 402                 ndev->cnf.rtr_solicits = 0;
 403         }
 404 #endif
 405
 406 #ifdef CONFIG_IPV6_PRIVACY
 407         INIT_LIST_HEAD(&ndev->tempaddr_list);
 408         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 409         if ((dev->flags&IFF_LOOPBACK) ||
 410             dev->type == ARPHRD_TUNNEL ||
 411             dev->type == ARPHRD_TUNNEL6 ||
 412             dev->type == ARPHRD_SIT ||
 413             dev->type == ARPHRD_NONE) {
 414                 ndev->cnf.use_tempaddr = -1;
 415         } else {
 416                 in6_dev_hold(ndev);
 417                 ipv6_regen_rndid((unsigned long) ndev);
 418         }
 419 #endif
 420         ndev->token = in6addr_any;
 421
 422         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 423                 ndev->if_flags |= IF_READY;
 424
 425         ipv6_mc_init_dev(ndev);
 426         ndev->tstamp = jiffies;
 427         addrconf_sysctl_register(ndev);
 428         /* protected by rtnl_lock */
 429         rcu_assign_pointer(dev->ip6_ptr, ndev);
 430
 431         /* Join interface-local all-node multicast group */
 432         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         /* Join all-router multicast group if forwarding is set */
 438         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 439                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 static int inet6_netconf_msgsize_devconf(int type)
 463 {
 464         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 465                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 466
 467         /* type -1 is used for ALL */
 468         if (type == -1 || type == NETCONFA_FORWARDING)
 469                 size += nla_total_size(4);
 470 #ifdef CONFIG_IPV6_MROUTE
 471         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 472                 size += nla_total_size(4);
 473 #endif
 474
 475         return size;
 476 }
 477
 478 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 479                                       struct ipv6_devconf *devconf, u32 portid,
 480                                       u32 seq, int event, unsigned int flags,
 481                                       int type)
 482 {
 483         struct nlmsghdr  *nlh;
 484         struct netconfmsg *ncm;
 485
 486         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 487                         flags);
 488         if (nlh == NULL)
 489                 return -EMSGSIZE;
 490
 491         ncm = nlmsg_data(nlh);
 492         ncm->ncm_family = AF_INET6;
 493
 494         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 495                 goto nla_put_failure;
 496
 497         /* type -1 is used for ALL */
 498         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 499             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 500                 goto nla_put_failure;
 501 #ifdef CONFIG_IPV6_MROUTE
 502         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 503             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 504                         devconf->mc_forwarding) < 0)
 505                 goto nla_put_failure;
 506 #endif
 507         return nlmsg_end(skb, nlh);
 508
 509 nla_put_failure:
 510         nlmsg_cancel(skb, nlh);
 511         return -EMSGSIZE;
 512 }
 513
 514 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 515                                   struct ipv6_devconf *devconf)
 516 {
 517         struct sk_buff *skb;
 518         int err = -ENOBUFS;
 519
 520         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 521         if (skb == NULL)
 522                 goto errout;
 523
 524         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 525                                          RTM_NEWNETCONF, 0, type);
 526         if (err < 0) {
 527                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 528                 WARN_ON(err == -EMSGSIZE);
 529                 kfree_skb(skb);
 530                 goto errout;
 531         }
 532         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 533         return;
 534 errout:
 535         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 536 }
 537
 538 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 539         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 540         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 541 };
 542
 543 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 544                                      struct nlmsghdr *nlh)
 545 {
 546         struct net *net = sock_net(in_skb->sk);
 547         struct nlattr *tb[NETCONFA_MAX+1];
 548         struct netconfmsg *ncm;
 549         struct sk_buff *skb;
 550         struct ipv6_devconf *devconf;
 551         struct inet6_dev *in6_dev;
 552         struct net_device *dev;
 553         int ifindex;
 554         int err;
 555
 556         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 557                           devconf_ipv6_policy);
 558         if (err < 0)
 559                 goto errout;
 560
 561         err = EINVAL;
 562         if (!tb[NETCONFA_IFINDEX])
 563                 goto errout;
 564
 565         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 566         switch (ifindex) {
 567         case NETCONFA_IFINDEX_ALL:
 568                 devconf = net->ipv6.devconf_all;
 569                 break;
 570         case NETCONFA_IFINDEX_DEFAULT:
 571                 devconf = net->ipv6.devconf_dflt;
 572                 break;
 573         default:
 574                 dev = __dev_get_by_index(net, ifindex);
 575                 if (dev == NULL)
 576                         goto errout;
 577                 in6_dev = __in6_dev_get(dev);
 578                 if (in6_dev == NULL)
 579                         goto errout;
 580                 devconf = &in6_dev->cnf;
 581                 break;
 582         }
 583
 584         err = -ENOBUFS;
 585         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 586         if (skb == NULL)
 587                 goto errout;
 588
 589         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 590                                          NETLINK_CB(in_skb).portid,
 591                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 592                                          -1);
 593         if (err < 0) {
 594                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 595                 WARN_ON(err == -EMSGSIZE);
 596                 kfree_skb(skb);
 597                 goto errout;
 598         }
 599         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 600 errout:
 601         return err;
 602 }
 603
 604 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 605                                       struct netlink_callback *cb)
 606 {
 607         struct net *net = sock_net(skb->sk);
 608         int h, s_h;
 609         int idx, s_idx;
 610         struct net_device *dev;
 611         struct inet6_dev *idev;
 612         struct hlist_head *head;
 613
 614         s_h = cb->args[0];
 615         s_idx = idx = cb->args[1];
 616
 617         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 618                 idx = 0;
 619                 head = &net->dev_index_head[h];
 620                 rcu_read_lock();
 621                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 622                           net->dev_base_seq;
 623                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 624                         if (idx < s_idx)
 625                                 goto cont;
 626                         idev = __in6_dev_get(dev);
 627                         if (!idev)
 628                                 goto cont;
 629
 630                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 631                                                        &idev->cnf,
 632                                                        NETLINK_CB(cb->skb).portid,
 633                                                        cb->nlh->nlmsg_seq,
 634                                                        RTM_NEWNETCONF,
 635                                                        NLM_F_MULTI,
 636                                                        -1) <= 0) {
 637                                 rcu_read_unlock();
 638                                 goto done;
 639                         }
 640                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 641 cont:
 642                         idx++;
 643                 }
 644                 rcu_read_unlock();
 645         }
 646         if (h == NETDEV_HASHENTRIES) {
 647                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 648                                                net->ipv6.devconf_all,
 649                                                NETLINK_CB(cb->skb).portid,
 650                                                cb->nlh->nlmsg_seq,
 651                                                RTM_NEWNETCONF, NLM_F_MULTI,
 652                                                -1) <= 0)
 653                         goto done;
 654                 else
 655                         h++;
 656         }
 657         if (h == NETDEV_HASHENTRIES + 1) {
 658                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 659                                                net->ipv6.devconf_dflt,
 660                                                NETLINK_CB(cb->skb).portid,
 661                                                cb->nlh->nlmsg_seq,
 662                                                RTM_NEWNETCONF, NLM_F_MULTI,
 663                                                -1) <= 0)
 664                         goto done;
 665                 else
 666                         h++;
 667         }
 668 done:
 669         cb->args[0] = h;
 670         cb->args[1] = idx;
 671
 672         return skb->len;
 673 }
 674
 675 #ifdef CONFIG_SYSCTL
 676 static void dev_forward_change(struct inet6_dev *idev)
 677 {
 678         struct net_device *dev;
 679         struct inet6_ifaddr *ifa;
 680
 681         if (!idev)
 682                 return;
 683         dev = idev->dev;
 684         if (idev->cnf.forwarding)
 685                 dev_disable_lro(dev);
 686         if (dev->flags & IFF_MULTICAST) {
 687                 if (idev->cnf.forwarding) {
 688                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 689                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 690                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 691                 } else {
 692                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 693                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 694                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 695                 }
 696         }
 697
 698         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 699                 if (ifa->flags&IFA_F_TENTATIVE)
 700                         continue;
 701                 if (idev->cnf.forwarding)
 702                         addrconf_join_anycast(ifa);
 703                 else
 704                         addrconf_leave_anycast(ifa);
 705         }
 706         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 707                                      dev->ifindex, &idev->cnf);
 708 }
 709
 710
 711 static void addrconf_forward_change(struct net *net, __s32 newf)
 712 {
 713         struct net_device *dev;
 714         struct inet6_dev *idev;
 715
 716         for_each_netdev(net, dev) {
 717                 idev = __in6_dev_get(dev);
 718                 if (idev) {
 719                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 720                         idev->cnf.forwarding = newf;
 721                         if (changed)
 722                                 dev_forward_change(idev);
 723                 }
 724         }
 725 }
 726
 727 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 728 {
 729         struct net *net;
 730         int old;
 731
 732         if (!rtnl_trylock())
 733                 return restart_syscall();
 734
 735         net = (struct net *)table->extra2;
 736         old = *p;
 737         *p = newf;
 738
 739         if (p == &net->ipv6.devconf_dflt->forwarding) {
 740                 if ((!newf) ^ (!old))
 741                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 742                                                      NETCONFA_IFINDEX_DEFAULT,
 743                                                      net->ipv6.devconf_dflt);
 744                 rtnl_unlock();
 745                 return 0;
 746         }
 747
 748         if (p == &net->ipv6.devconf_all->forwarding) {
 749                 net->ipv6.devconf_dflt->forwarding = newf;
 750                 addrconf_forward_change(net, newf);
 751                 if ((!newf) ^ (!old))
 752                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 753                                                      NETCONFA_IFINDEX_ALL,
 754                                                      net->ipv6.devconf_all);
 755         } else if ((!newf) ^ (!old))
 756                 dev_forward_change((struct inet6_dev *)table->extra1);
 757         rtnl_unlock();
 758
 759         if (newf)
 760                 rt6_purge_dflt_routers(net);
 761         return 1;
 762 }
 763 #endif
 764
 765 /* Nobody refers to this ifaddr, destroy it */
 766 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 767 {
 768         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 769
 770 #ifdef NET_REFCNT_DEBUG
 771         pr_debug("%s\n", __func__);
 772 #endif
 773
 774         in6_dev_put(ifp->idev);
 775
 776         if (del_timer(&ifp->dad_timer))
 777                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 778
 779         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 780                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 781                 return;
 782         }
 783         ip6_rt_put(ifp->rt);
 784
 785         kfree_rcu(ifp, rcu);
 786 }
 787
 788 static void
 789 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 790 {
 791         struct list_head *p;
 792         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 793
 794         /*
 795          * Each device address list is sorted in order of scope -
 796          * global before linklocal.
 797          */
 798         list_for_each(p, &idev->addr_list) {
 799                 struct inet6_ifaddr *ifa
 800                         = list_entry(p, struct inet6_ifaddr, if_list);
 801                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 802                         break;
 803         }
 804
 805         list_add_tail(&ifp->if_list, p);
 806 }
 807
 808 static u32 inet6_addr_hash(const struct in6_addr *addr)
 809 {
 810         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 811 }
 812
 813 /* On success it returns ifp with increased reference count */
 814
 815 static struct inet6_ifaddr *
 816 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 817               int scope, u32 flags)
 818 {
 819         struct inet6_ifaddr *ifa = NULL;
 820         struct rt6_info *rt;
 821         unsigned int hash;
 822         int err = 0;
 823         int addr_type = ipv6_addr_type(addr);
 824
 825         if (addr_type == IPV6_ADDR_ANY ||
 826             addr_type & IPV6_ADDR_MULTICAST ||
 827             (!(idev->dev->flags & IFF_LOOPBACK) &&
 828              addr_type & IPV6_ADDR_LOOPBACK))
 829                 return ERR_PTR(-EADDRNOTAVAIL);
 830
 831         rcu_read_lock_bh();
 832         if (idev->dead) {
 833                 err = -ENODEV;                  /*XXX*/
 834                 goto out2;
 835         }
 836
 837         if (idev->cnf.disable_ipv6) {
 838                 err = -EACCES;
 839                 goto out2;
 840         }
 841
 842         spin_lock(&addrconf_hash_lock);
 843
 844         /* Ignore adding duplicate addresses on an interface */
 845         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 846                 ADBG(("ipv6_add_addr: already assigned\n"));
 847                 err = -EEXIST;
 848                 goto out;
 849         }
 850
 851         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 852
 853         if (ifa == NULL) {
 854                 ADBG(("ipv6_add_addr: malloc failed\n"));
 855                 err = -ENOBUFS;
 856                 goto out;
 857         }
 858
 859         rt = addrconf_dst_alloc(idev, addr, false);
 860         if (IS_ERR(rt)) {
 861                 err = PTR_ERR(rt);
 862                 goto out;
 863         }
 864
 865         ifa->addr = *addr;
 866
 867         spin_lock_init(&ifa->lock);
 868         spin_lock_init(&ifa->state_lock);
 869         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 870                     (unsigned long)ifa);
 871         INIT_HLIST_NODE(&ifa->addr_lst);
 872         ifa->scope = scope;
 873         ifa->prefix_len = pfxlen;
 874         ifa->flags = flags | IFA_F_TENTATIVE;
 875         ifa->cstamp = ifa->tstamp = jiffies;
 876         ifa->tokenized = false;
 877
 878         ifa->rt = rt;
 879
 880         ifa->idev = idev;
 881         in6_dev_hold(idev);
 882         /* For caller */
 883         in6_ifa_hold(ifa);
 884
 885         /* Add to big hash table */
 886         hash = inet6_addr_hash(addr);
 887
 888         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 889         spin_unlock(&addrconf_hash_lock);
 890
 891         write_lock(&idev->lock);
 892         /* Add to inet6_dev unicast addr list. */
 893         ipv6_link_dev_addr(idev, ifa);
 894
 895 #ifdef CONFIG_IPV6_PRIVACY
 896         if (ifa->flags&IFA_F_TEMPORARY) {
 897                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 898                 in6_ifa_hold(ifa);
 899         }
 900 #endif
 901
 902         in6_ifa_hold(ifa);
 903         write_unlock(&idev->lock);
 904 out2:
 905         rcu_read_unlock_bh();
 906
 907         if (likely(err == 0))
 908                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 909         else {
 910                 kfree(ifa);
 911                 ifa = ERR_PTR(err);
 912         }
 913
 914         return ifa;
 915 out:
 916         spin_unlock(&addrconf_hash_lock);
 917         goto out2;
 918 }
 919
 920 /* This function wants to get referenced ifp and releases it before return */
 921
 922 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 923 {
 924         struct inet6_ifaddr *ifa, *ifn;
 925         struct inet6_dev *idev = ifp->idev;
 926         int state;
 927         int deleted = 0, onlink = 0;
 928         unsigned long expires = jiffies;
 929
 930         spin_lock_bh(&ifp->state_lock);
 931         state = ifp->state;
 932         ifp->state = INET6_IFADDR_STATE_DEAD;
 933         spin_unlock_bh(&ifp->state_lock);
 934
 935         if (state == INET6_IFADDR_STATE_DEAD)
 936                 goto out;
 937
 938         spin_lock_bh(&addrconf_hash_lock);
 939         hlist_del_init_rcu(&ifp->addr_lst);
 940         spin_unlock_bh(&addrconf_hash_lock);
 941
 942         write_lock_bh(&idev->lock);
 943 #ifdef CONFIG_IPV6_PRIVACY
 944         if (ifp->flags&IFA_F_TEMPORARY) {
 945                 list_del(&ifp->tmp_list);
 946                 if (ifp->ifpub) {
 947                         in6_ifa_put(ifp->ifpub);
 948                         ifp->ifpub = NULL;
 949                 }
 950                 __in6_ifa_put(ifp);
 951         }
 952 #endif
 953
 954         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 955                 if (ifa == ifp) {
 956                         list_del_init(&ifp->if_list);
 957                         __in6_ifa_put(ifp);
 958
 959                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 960                                 break;
 961                         deleted = 1;
 962                         continue;
 963                 } else if (ifp->flags & IFA_F_PERMANENT) {
 964                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 965                                               ifp->prefix_len)) {
 966                                 if (ifa->flags & IFA_F_PERMANENT) {
 967                                         onlink = 1;
 968                                         if (deleted)
 969                                                 break;
 970                                 } else {
 971                                         unsigned long lifetime;
 972
 973                                         if (!onlink)
 974                                                 onlink = -1;
 975
 976                                         spin_lock(&ifa->lock);
 977
 978                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 979                                         /*
 980                                          * Note: Because this address is
 981                                          * not permanent, lifetime <
 982                                          * LONG_MAX / HZ here.
 983                                          */
 984                                         if (time_before(expires,
 985                                                         ifa->tstamp + lifetime * HZ))
 986                                                 expires = ifa->tstamp + lifetime * HZ;
 987                                         spin_unlock(&ifa->lock);
 988                                 }
 989                         }
 990                 }
 991         }
 992         write_unlock_bh(&idev->lock);
 993
 994         addrconf_del_dad_timer(ifp);
 995
 996         ipv6_ifa_notify(RTM_DELADDR, ifp);
 997
 998         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
 999
1000         /*
1001          * Purge or update corresponding prefix
1002          *
1003          * 1) we don't purge prefix here if address was not permanent.
1004          *    prefix is managed by its own lifetime.
1005          * 2) if there're no addresses, delete prefix.
1006          * 3) if there're still other permanent address(es),
1007          *    corresponding prefix is still permanent.
1008          * 4) otherwise, update prefix lifetime to the
1009          *    longest valid lifetime among the corresponding
1010          *    addresses on the device.
1011          *    Note: subsequent RA will update lifetime.
1012          *
1013          * --yoshfuji
1014          */
1015         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1016                 struct in6_addr prefix;
1017                 struct rt6_info *rt;
1018
1019                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1020
1021                 rt = addrconf_get_prefix_route(&prefix,
1022                                                ifp->prefix_len,
1023                                                ifp->idev->dev,
1024                                                0, RTF_GATEWAY | RTF_DEFAULT);
1025
1026                 if (rt) {
1027                         if (onlink == 0) {
1028                                 ip6_del_rt(rt);
1029                                 rt = NULL;
1030                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1031                                 rt6_set_expires(rt, expires);
1032                         }
1033                 }
1034                 ip6_rt_put(rt);
1035         }
1036
1037         /* clean up prefsrc entries */
1038         rt6_remove_prefsrc(ifp);
1039 out:
1040         in6_ifa_put(ifp);
1041 }
1042
1043 #ifdef CONFIG_IPV6_PRIVACY
1044 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1045 {
1046         struct inet6_dev *idev = ifp->idev;
1047         struct in6_addr addr, *tmpaddr;
1048         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1049         unsigned long regen_advance;
1050         int tmp_plen;
1051         int ret = 0;
1052         int max_addresses;
1053         u32 addr_flags;
1054         unsigned long now = jiffies;
1055
1056         write_lock(&idev->lock);
1057         if (ift) {
1058                 spin_lock_bh(&ift->lock);
1059                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1060                 spin_unlock_bh(&ift->lock);
1061                 tmpaddr = &addr;
1062         } else {
1063                 tmpaddr = NULL;
1064         }
1065 retry:
1066         in6_dev_hold(idev);
1067         if (idev->cnf.use_tempaddr <= 0) {
1068                 write_unlock(&idev->lock);
1069                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1070                 in6_dev_put(idev);
1071                 ret = -1;
1072                 goto out;
1073         }
1074         spin_lock_bh(&ifp->lock);
1075         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1076                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1077                 spin_unlock_bh(&ifp->lock);
1078                 write_unlock(&idev->lock);
1079                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1080                         __func__);
1081                 in6_dev_put(idev);
1082                 ret = -1;
1083                 goto out;
1084         }
1085         in6_ifa_hold(ifp);
1086         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1087         __ipv6_try_regen_rndid(idev, tmpaddr);
1088         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1089         age = (now - ifp->tstamp) / HZ;
1090         tmp_valid_lft = min_t(__u32,
1091                               ifp->valid_lft,
1092                               idev->cnf.temp_valid_lft + age);
1093         tmp_prefered_lft = min_t(__u32,
1094                                  ifp->prefered_lft,
1095                                  idev->cnf.temp_prefered_lft + age -
1096                                  idev->cnf.max_desync_factor);
1097         tmp_plen = ifp->prefix_len;
1098         max_addresses = idev->cnf.max_addresses;
1099         tmp_tstamp = ifp->tstamp;
1100         spin_unlock_bh(&ifp->lock);
1101
1102         regen_advance = idev->cnf.regen_max_retry *
1103                         idev->cnf.dad_transmits *
1104                         idev->nd_parms->retrans_time / HZ;
1105         write_unlock(&idev->lock);
1106
1107         /* A temporary address is created only if this calculated Preferred
1108          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1109          * an implementation must not create a temporary address with a zero
1110          * Preferred Lifetime.
1111          */
1112         if (tmp_prefered_lft <= regen_advance) {
1113                 in6_ifa_put(ifp);
1114                 in6_dev_put(idev);
1115                 ret = -1;
1116                 goto out;
1117         }
1118
1119         addr_flags = IFA_F_TEMPORARY;
1120         /* set in addrconf_prefix_rcv() */
1121         if (ifp->flags & IFA_F_OPTIMISTIC)
1122                 addr_flags |= IFA_F_OPTIMISTIC;
1123
1124         ift = !max_addresses ||
1125               ipv6_count_addresses(idev) < max_addresses ?
1126                 ipv6_add_addr(idev, &addr, tmp_plen, ipv6_addr_scope(&addr),
1127                               addr_flags) : NULL;
1128         if (IS_ERR_OR_NULL(ift)) {
1129                 in6_ifa_put(ifp);
1130                 in6_dev_put(idev);
1131                 pr_info("%s: retry temporary address regeneration\n", __func__);
1132                 tmpaddr = &addr;
1133                 write_lock(&idev->lock);
1134                 goto retry;
1135         }
1136
1137         spin_lock_bh(&ift->lock);
1138         ift->ifpub = ifp;
1139         ift->valid_lft = tmp_valid_lft;
1140         ift->prefered_lft = tmp_prefered_lft;
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 static 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 ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1809 {
1810         switch (dev->type) {
1811         case ARPHRD_ETHER:
1812         case ARPHRD_FDDI:
1813                 return addrconf_ifid_eui48(eui, dev);
1814         case ARPHRD_ARCNET:
1815                 return addrconf_ifid_arcnet(eui, dev);
1816         case ARPHRD_INFINIBAND:
1817                 return addrconf_ifid_infiniband(eui, dev);
1818         case ARPHRD_SIT:
1819                 return addrconf_ifid_sit(eui, dev);
1820         case ARPHRD_IPGRE:
1821                 return addrconf_ifid_gre(eui, dev);
1822         case ARPHRD_IEEE802154:
1823                 return addrconf_ifid_eui64(eui, dev);
1824         case ARPHRD_IEEE1394:
1825                 return addrconf_ifid_ieee1394(eui, dev);
1826         }
1827         return -1;
1828 }
1829
1830 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1831 {
1832         int err = -1;
1833         struct inet6_ifaddr *ifp;
1834
1835         read_lock_bh(&idev->lock);
1836         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1837                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1838                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1839                         err = 0;
1840                         break;
1841                 }
1842         }
1843         read_unlock_bh(&idev->lock);
1844         return err;
1845 }
1846
1847 #ifdef CONFIG_IPV6_PRIVACY
1848 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1849 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1850 {
1851 regen:
1852         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1853         idev->rndid[0] &= ~0x02;
1854
1855         /*
1856          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1857          * check if generated address is not inappropriate
1858          *
1859          *  - Reserved subnet anycast (RFC 2526)
1860          *      11111101 11....11 1xxxxxxx
1861          *  - ISATAP (RFC4214) 6.1
1862          *      00-00-5E-FE-xx-xx-xx-xx
1863          *  - value 0
1864          *  - XXX: already assigned to an address on the device
1865          */
1866         if (idev->rndid[0] == 0xfd &&
1867             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1868             (idev->rndid[7]&0x80))
1869                 goto regen;
1870         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1871                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1872                         goto regen;
1873                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1874                         goto regen;
1875         }
1876 }
1877
1878 static void ipv6_regen_rndid(unsigned long data)
1879 {
1880         struct inet6_dev *idev = (struct inet6_dev *) data;
1881         unsigned long expires;
1882
1883         rcu_read_lock_bh();
1884         write_lock_bh(&idev->lock);
1885
1886         if (idev->dead)
1887                 goto out;
1888
1889         __ipv6_regen_rndid(idev);
1890
1891         expires = jiffies +
1892                 idev->cnf.temp_prefered_lft * HZ -
1893                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1894                 idev->cnf.max_desync_factor * HZ;
1895         if (time_before(expires, jiffies)) {
1896                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1897                         __func__, idev->dev->name);
1898                 goto out;
1899         }
1900
1901         if (!mod_timer(&idev->regen_timer, expires))
1902                 in6_dev_hold(idev);
1903
1904 out:
1905         write_unlock_bh(&idev->lock);
1906         rcu_read_unlock_bh();
1907         in6_dev_put(idev);
1908 }
1909
1910 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1911 {
1912         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1913                 __ipv6_regen_rndid(idev);
1914 }
1915 #endif
1916
1917 /*
1918  *      Add prefix route.
1919  */
1920
1921 static void
1922 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1923                       unsigned long expires, u32 flags)
1924 {
1925         struct fib6_config cfg = {
1926                 .fc_table = RT6_TABLE_PREFIX,
1927                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1928                 .fc_ifindex = dev->ifindex,
1929                 .fc_expires = expires,
1930                 .fc_dst_len = plen,
1931                 .fc_flags = RTF_UP | flags,
1932                 .fc_nlinfo.nl_net = dev_net(dev),
1933                 .fc_protocol = RTPROT_KERNEL,
1934         };
1935
1936         cfg.fc_dst = *pfx;
1937
1938         /* Prevent useless cloning on PtP SIT.
1939            This thing is done here expecting that the whole
1940            class of non-broadcast devices need not cloning.
1941          */
1942 #if IS_ENABLED(CONFIG_IPV6_SIT)
1943         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1944                 cfg.fc_flags |= RTF_NONEXTHOP;
1945 #endif
1946
1947         ip6_route_add(&cfg);
1948 }
1949
1950
1951 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1952                                                   int plen,
1953                                                   const struct net_device *dev,
1954                                                   u32 flags, u32 noflags)
1955 {
1956         struct fib6_node *fn;
1957         struct rt6_info *rt = NULL;
1958         struct fib6_table *table;
1959
1960         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1961         if (table == NULL)
1962                 return NULL;
1963
1964         read_lock_bh(&table->tb6_lock);
1965         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1966         if (!fn)
1967                 goto out;
1968         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1969                 if (rt->dst.dev->ifindex != dev->ifindex)
1970                         continue;
1971                 if ((rt->rt6i_flags & flags) != flags)
1972                         continue;
1973                 if ((rt->rt6i_flags & noflags) != 0)
1974                         continue;
1975                 dst_hold(&rt->dst);
1976                 break;
1977         }
1978 out:
1979         read_unlock_bh(&table->tb6_lock);
1980         return rt;
1981 }
1982
1983
1984 /* Create "default" multicast route to the interface */
1985
1986 static void addrconf_add_mroute(struct net_device *dev)
1987 {
1988         struct fib6_config cfg = {
1989                 .fc_table = RT6_TABLE_LOCAL,
1990                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1991                 .fc_ifindex = dev->ifindex,
1992                 .fc_dst_len = 8,
1993                 .fc_flags = RTF_UP,
1994                 .fc_nlinfo.nl_net = dev_net(dev),
1995         };
1996
1997         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1998
1999         ip6_route_add(&cfg);
2000 }
2001
2002 #if IS_ENABLED(CONFIG_IPV6_SIT)
2003 static void sit_route_add(struct net_device *dev)
2004 {
2005         struct fib6_config cfg = {
2006                 .fc_table = RT6_TABLE_MAIN,
2007                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2008                 .fc_ifindex = dev->ifindex,
2009                 .fc_dst_len = 96,
2010                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2011                 .fc_nlinfo.nl_net = dev_net(dev),
2012         };
2013
2014         /* prefix length - 96 bits "::d.d.d.d" */
2015         ip6_route_add(&cfg);
2016 }
2017 #endif
2018
2019 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2020 {
2021         struct inet6_dev *idev;
2022
2023         ASSERT_RTNL();
2024
2025         idev = ipv6_find_idev(dev);
2026         if (!idev)
2027                 return ERR_PTR(-ENOBUFS);
2028
2029         if (idev->cnf.disable_ipv6)
2030                 return ERR_PTR(-EACCES);
2031
2032         /* Add default multicast route */
2033         if (!(dev->flags & IFF_LOOPBACK))
2034                 addrconf_add_mroute(dev);
2035
2036         return idev;
2037 }
2038
2039 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2040 {
2041         struct prefix_info *pinfo;
2042         __u32 valid_lft;
2043         __u32 prefered_lft;
2044         int addr_type;
2045         struct inet6_dev *in6_dev;
2046         struct net *net = dev_net(dev);
2047
2048         pinfo = (struct prefix_info *) opt;
2049
2050         if (len < sizeof(struct prefix_info)) {
2051                 ADBG(("addrconf: prefix option too short\n"));
2052                 return;
2053         }
2054
2055         /*
2056          *      Validation checks ([ADDRCONF], page 19)
2057          */
2058
2059         addr_type = ipv6_addr_type(&pinfo->prefix);
2060
2061         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2062                 return;
2063
2064         valid_lft = ntohl(pinfo->valid);
2065         prefered_lft = ntohl(pinfo->prefered);
2066
2067         if (prefered_lft > valid_lft) {
2068                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2069                 return;
2070         }
2071
2072         in6_dev = in6_dev_get(dev);
2073
2074         if (in6_dev == NULL) {
2075                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2076                                     dev->name);
2077                 return;
2078         }
2079
2080         /*
2081          *      Two things going on here:
2082          *      1) Add routes for on-link prefixes
2083          *      2) Configure prefixes with the auto flag set
2084          */
2085
2086         if (pinfo->onlink) {
2087                 struct rt6_info *rt;
2088                 unsigned long rt_expires;
2089
2090                 /* Avoid arithmetic overflow. Really, we could
2091                  * save rt_expires in seconds, likely valid_lft,
2092                  * but it would require division in fib gc, that it
2093                  * not good.
2094                  */
2095                 if (HZ > USER_HZ)
2096                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2097                 else
2098                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2099
2100                 if (addrconf_finite_timeout(rt_expires))
2101                         rt_expires *= HZ;
2102
2103                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2104                                                pinfo->prefix_len,
2105                                                dev,
2106                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2107                                                RTF_GATEWAY | RTF_DEFAULT);
2108
2109                 if (rt) {
2110                         /* Autoconf prefix route */
2111                         if (valid_lft == 0) {
2112                                 ip6_del_rt(rt);
2113                                 rt = NULL;
2114                         } else if (addrconf_finite_timeout(rt_expires)) {
2115                                 /* not infinity */
2116                                 rt6_set_expires(rt, jiffies + rt_expires);
2117                         } else {
2118                                 rt6_clean_expires(rt);
2119                         }
2120                 } else if (valid_lft) {
2121                         clock_t expires = 0;
2122                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2123                         if (addrconf_finite_timeout(rt_expires)) {
2124                                 /* not infinity */
2125                                 flags |= RTF_EXPIRES;
2126                                 expires = jiffies_to_clock_t(rt_expires);
2127                         }
2128                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2129                                               dev, expires, flags);
2130                 }
2131                 ip6_rt_put(rt);
2132         }
2133
2134         /* Try to figure out our local address for this prefix */
2135
2136         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2137                 struct inet6_ifaddr *ifp;
2138                 struct in6_addr addr;
2139                 int create = 0, update_lft = 0;
2140                 bool tokenized = false;
2141
2142                 if (pinfo->prefix_len == 64) {
2143                         memcpy(&addr, &pinfo->prefix, 8);
2144
2145                         if (!ipv6_addr_any(&in6_dev->token)) {
2146                                 read_lock_bh(&in6_dev->lock);
2147                                 memcpy(addr.s6_addr + 8,
2148                                        in6_dev->token.s6_addr + 8, 8);
2149                                 read_unlock_bh(&in6_dev->lock);
2150                                 tokenized = true;
2151                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2152                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2153                                 in6_dev_put(in6_dev);
2154                                 return;
2155                         }
2156                         goto ok;
2157                 }
2158                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2159                                     pinfo->prefix_len);
2160                 in6_dev_put(in6_dev);
2161                 return;
2162
2163 ok:
2164
2165                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2166
2167                 if (ifp == NULL && valid_lft) {
2168                         int max_addresses = in6_dev->cnf.max_addresses;
2169                         u32 addr_flags = 0;
2170
2171 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2172                         if (in6_dev->cnf.optimistic_dad &&
2173                             !net->ipv6.devconf_all->forwarding && sllao)
2174                                 addr_flags = IFA_F_OPTIMISTIC;
2175 #endif
2176
2177                         /* Do not allow to create too much of autoconfigured
2178                          * addresses; this would be too easy way to crash kernel.
2179                          */
2180                         if (!max_addresses ||
2181                             ipv6_count_addresses(in6_dev) < max_addresses)
2182                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2183                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2184                                                     addr_flags);
2185
2186                         if (IS_ERR_OR_NULL(ifp)) {
2187                                 in6_dev_put(in6_dev);
2188                                 return;
2189                         }
2190
2191                         update_lft = create = 1;
2192                         ifp->cstamp = jiffies;
2193                         ifp->tokenized = tokenized;
2194                         addrconf_dad_start(ifp);
2195                 }
2196
2197                 if (ifp) {
2198                         int flags;
2199                         unsigned long now;
2200 #ifdef CONFIG_IPV6_PRIVACY
2201                         struct inet6_ifaddr *ift;
2202 #endif
2203                         u32 stored_lft;
2204
2205                         /* update lifetime (RFC2462 5.5.3 e) */
2206                         spin_lock(&ifp->lock);
2207                         now = jiffies;
2208                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2209                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2210                         else
2211                                 stored_lft = 0;
2212                         if (!update_lft && stored_lft) {
2213                                 if (valid_lft > MIN_VALID_LIFETIME ||
2214                                     valid_lft > stored_lft)
2215                                         update_lft = 1;
2216                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2217                                         /* valid_lft <= stored_lft is always true */
2218                                         /*
2219                                          * RFC 4862 Section 5.5.3e:
2220                                          * "Note that the preferred lifetime of
2221                                          *  the corresponding address is always
2222                                          *  reset to the Preferred Lifetime in
2223                                          *  the received Prefix Information
2224                                          *  option, regardless of whether the
2225                                          *  valid lifetime is also reset or
2226                                          *  ignored."
2227                                          *
2228                                          *  So if the preferred lifetime in
2229                                          *  this advertisement is different
2230                                          *  than what we have stored, but the
2231                                          *  valid lifetime is invalid, just
2232                                          *  reset prefered_lft.
2233                                          *
2234                                          *  We must set the valid lifetime
2235                                          *  to the stored lifetime since we'll
2236                                          *  be updating the timestamp below,
2237                                          *  else we'll set it back to the
2238                                          *  minimum.
2239                                          */
2240                                         if (prefered_lft != ifp->prefered_lft) {
2241                                                 valid_lft = stored_lft;
2242                                                 update_lft = 1;
2243                                         }
2244                                 } else {
2245                                         valid_lft = MIN_VALID_LIFETIME;
2246                                         if (valid_lft < prefered_lft)
2247                                                 prefered_lft = valid_lft;
2248                                         update_lft = 1;
2249                                 }
2250                         }
2251
2252                         if (update_lft) {
2253                                 ifp->valid_lft = valid_lft;
2254                                 ifp->prefered_lft = prefered_lft;
2255                                 ifp->tstamp = now;
2256                                 flags = ifp->flags;
2257                                 ifp->flags &= ~IFA_F_DEPRECATED;
2258                                 spin_unlock(&ifp->lock);
2259
2260                                 if (!(flags&IFA_F_TENTATIVE))
2261                                         ipv6_ifa_notify(0, ifp);
2262                         } else
2263                                 spin_unlock(&ifp->lock);
2264
2265 #ifdef CONFIG_IPV6_PRIVACY
2266                         read_lock_bh(&in6_dev->lock);
2267                         /* update all temporary addresses in the list */
2268                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2269                                             tmp_list) {
2270                                 int age, max_valid, max_prefered;
2271
2272                                 if (ifp != ift->ifpub)
2273                                         continue;
2274
2275                                 /*
2276                                  * RFC 4941 section 3.3:
2277                                  * If a received option will extend the lifetime
2278                                  * of a public address, the lifetimes of
2279                                  * temporary addresses should be extended,
2280                                  * subject to the overall constraint that no
2281                                  * temporary addresses should ever remain
2282                                  * "valid" or "preferred" for a time longer than
2283                                  * (TEMP_VALID_LIFETIME) or
2284                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2285                                  * respectively.
2286                                  */
2287                                 age = (now - ift->cstamp) / HZ;
2288                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2289                                 if (max_valid < 0)
2290                                         max_valid = 0;
2291
2292                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2293                                                in6_dev->cnf.max_desync_factor -
2294                                                age;
2295                                 if (max_prefered < 0)
2296                                         max_prefered = 0;
2297
2298                                 if (valid_lft > max_valid)
2299                                         valid_lft = max_valid;
2300
2301                                 if (prefered_lft > max_prefered)
2302                                         prefered_lft = max_prefered;
2303
2304                                 spin_lock(&ift->lock);
2305                                 flags = ift->flags;
2306                                 ift->valid_lft = valid_lft;
2307                                 ift->prefered_lft = prefered_lft;
2308                                 ift->tstamp = now;
2309                                 if (prefered_lft > 0)
2310                                         ift->flags &= ~IFA_F_DEPRECATED;
2311
2312                                 spin_unlock(&ift->lock);
2313                                 if (!(flags&IFA_F_TENTATIVE))
2314                                         ipv6_ifa_notify(0, ift);
2315                         }
2316
2317                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2318                                 /*
2319                                  * When a new public address is created as
2320                                  * described in [ADDRCONF], also create a new
2321                                  * temporary address. Also create a temporary
2322                                  * address if it's enabled but no temporary
2323                                  * address currently exists.
2324                                  */
2325                                 read_unlock_bh(&in6_dev->lock);
2326                                 ipv6_create_tempaddr(ifp, NULL);
2327                         } else {
2328                                 read_unlock_bh(&in6_dev->lock);
2329                         }
2330 #endif
2331                         in6_ifa_put(ifp);
2332                         addrconf_verify(0);
2333                 }
2334         }
2335         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2336         in6_dev_put(in6_dev);
2337 }
2338
2339 /*
2340  *      Set destination address.
2341  *      Special case for SIT interfaces where we create a new "virtual"
2342  *      device.
2343  */
2344 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2345 {
2346         struct in6_ifreq ireq;
2347         struct net_device *dev;
2348         int err = -EINVAL;
2349
2350         rtnl_lock();
2351
2352         err = -EFAULT;
2353         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2354                 goto err_exit;
2355
2356         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2357
2358         err = -ENODEV;
2359         if (dev == NULL)
2360                 goto err_exit;
2361
2362 #if IS_ENABLED(CONFIG_IPV6_SIT)
2363         if (dev->type == ARPHRD_SIT) {
2364                 const struct net_device_ops *ops = dev->netdev_ops;
2365                 struct ifreq ifr;
2366                 struct ip_tunnel_parm p;
2367
2368                 err = -EADDRNOTAVAIL;
2369                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2370                         goto err_exit;
2371
2372                 memset(&p, 0, sizeof(p));
2373                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2374                 p.iph.saddr = 0;
2375                 p.iph.version = 4;
2376                 p.iph.ihl = 5;
2377                 p.iph.protocol = IPPROTO_IPV6;
2378                 p.iph.ttl = 64;
2379                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2380
2381                 if (ops->ndo_do_ioctl) {
2382                         mm_segment_t oldfs = get_fs();
2383
2384                         set_fs(KERNEL_DS);
2385                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2386                         set_fs(oldfs);
2387                 } else
2388                         err = -EOPNOTSUPP;
2389
2390                 if (err == 0) {
2391                         err = -ENOBUFS;
2392                         dev = __dev_get_by_name(net, p.name);
2393                         if (!dev)
2394                                 goto err_exit;
2395                         err = dev_open(dev);
2396                 }
2397         }
2398 #endif
2399
2400 err_exit:
2401         rtnl_unlock();
2402         return err;
2403 }
2404
2405 /*
2406  *      Manual configuration of address on an interface
2407  */
2408 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2409                           const struct in6_addr *peer_pfx,
2410                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2411                           __u32 valid_lft)
2412 {
2413         struct inet6_ifaddr *ifp;
2414         struct inet6_dev *idev;
2415         struct net_device *dev;
2416         int scope;
2417         u32 flags;
2418         clock_t expires;
2419         unsigned long timeout;
2420
2421         ASSERT_RTNL();
2422
2423         if (plen > 128)
2424                 return -EINVAL;
2425
2426         /* check the lifetime */
2427         if (!valid_lft || prefered_lft > valid_lft)
2428                 return -EINVAL;
2429
2430         dev = __dev_get_by_index(net, ifindex);
2431         if (!dev)
2432                 return -ENODEV;
2433
2434         idev = addrconf_add_dev(dev);
2435         if (IS_ERR(idev))
2436                 return PTR_ERR(idev);
2437
2438         scope = ipv6_addr_scope(pfx);
2439
2440         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2441         if (addrconf_finite_timeout(timeout)) {
2442                 expires = jiffies_to_clock_t(timeout * HZ);
2443                 valid_lft = timeout;
2444                 flags = RTF_EXPIRES;
2445         } else {
2446                 expires = 0;
2447                 flags = 0;
2448                 ifa_flags |= IFA_F_PERMANENT;
2449         }
2450
2451         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2452         if (addrconf_finite_timeout(timeout)) {
2453                 if (timeout == 0)
2454                         ifa_flags |= IFA_F_DEPRECATED;
2455                 prefered_lft = timeout;
2456         }
2457
2458         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2459
2460         if (!IS_ERR(ifp)) {
2461                 spin_lock_bh(&ifp->lock);
2462                 ifp->valid_lft = valid_lft;
2463                 ifp->prefered_lft = prefered_lft;
2464                 ifp->tstamp = jiffies;
2465                 if (peer_pfx)
2466                         ifp->peer_addr = *peer_pfx;
2467                 spin_unlock_bh(&ifp->lock);
2468
2469                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2470                                       expires, flags);
2471                 /*
2472                  * Note that section 3.1 of RFC 4429 indicates
2473                  * that the Optimistic flag should not be set for
2474                  * manually configured addresses
2475                  */
2476                 addrconf_dad_start(ifp);
2477                 in6_ifa_put(ifp);
2478                 addrconf_verify(0);
2479                 return 0;
2480         }
2481
2482         return PTR_ERR(ifp);
2483 }
2484
2485 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2486                           unsigned int plen)
2487 {
2488         struct inet6_ifaddr *ifp;
2489         struct inet6_dev *idev;
2490         struct net_device *dev;
2491
2492         if (plen > 128)
2493                 return -EINVAL;
2494
2495         dev = __dev_get_by_index(net, ifindex);
2496         if (!dev)
2497                 return -ENODEV;
2498
2499         if ((idev = __in6_dev_get(dev)) == NULL)
2500                 return -ENXIO;
2501
2502         read_lock_bh(&idev->lock);
2503         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2504                 if (ifp->prefix_len == plen &&
2505                     ipv6_addr_equal(pfx, &ifp->addr)) {
2506                         in6_ifa_hold(ifp);
2507                         read_unlock_bh(&idev->lock);
2508
2509                         ipv6_del_addr(ifp);
2510                         return 0;
2511                 }
2512         }
2513         read_unlock_bh(&idev->lock);
2514         return -EADDRNOTAVAIL;
2515 }
2516
2517
2518 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2519 {
2520         struct in6_ifreq ireq;
2521         int err;
2522
2523         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2524                 return -EPERM;
2525
2526         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2527                 return -EFAULT;
2528
2529         rtnl_lock();
2530         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2531                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2532                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2533         rtnl_unlock();
2534         return err;
2535 }
2536
2537 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2538 {
2539         struct in6_ifreq ireq;
2540         int err;
2541
2542         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2543                 return -EPERM;
2544
2545         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2546                 return -EFAULT;
2547
2548         rtnl_lock();
2549         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2550                              ireq.ifr6_prefixlen);
2551         rtnl_unlock();
2552         return err;
2553 }
2554
2555 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2556                      int plen, int scope)
2557 {
2558         struct inet6_ifaddr *ifp;
2559
2560         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2561         if (!IS_ERR(ifp)) {
2562                 spin_lock_bh(&ifp->lock);
2563                 ifp->flags &= ~IFA_F_TENTATIVE;
2564                 spin_unlock_bh(&ifp->lock);
2565                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2566                 in6_ifa_put(ifp);
2567         }
2568 }
2569
2570 #if IS_ENABLED(CONFIG_IPV6_SIT)
2571 static void sit_add_v4_addrs(struct inet6_dev *idev)
2572 {
2573         struct in6_addr addr;
2574         struct net_device *dev;
2575         struct net *net = dev_net(idev->dev);
2576         int scope;
2577
2578         ASSERT_RTNL();
2579
2580         memset(&addr, 0, sizeof(struct in6_addr));
2581         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2582
2583         if (idev->dev->flags&IFF_POINTOPOINT) {
2584                 addr.s6_addr32[0] = htonl(0xfe800000);
2585                 scope = IFA_LINK;
2586         } else {
2587                 scope = IPV6_ADDR_COMPATv4;
2588         }
2589
2590         if (addr.s6_addr32[3]) {
2591                 add_addr(idev, &addr, 128, scope);
2592                 return;
2593         }
2594
2595         for_each_netdev(net, dev) {
2596                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2597                 if (in_dev && (dev->flags & IFF_UP)) {
2598                         struct in_ifaddr *ifa;
2599
2600                         int flag = scope;
2601
2602                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2603                                 int plen;
2604
2605                                 addr.s6_addr32[3] = ifa->ifa_local;
2606
2607                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2608                                         continue;
2609                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2610                                         if (idev->dev->flags&IFF_POINTOPOINT)
2611                                                 continue;
2612                                         flag |= IFA_HOST;
2613                                 }
2614                                 if (idev->dev->flags&IFF_POINTOPOINT)
2615                                         plen = 64;
2616                                 else
2617                                         plen = 96;
2618
2619                                 add_addr(idev, &addr, plen, flag);
2620                         }
2621                 }
2622         }
2623 }
2624 #endif
2625
2626 static void init_loopback(struct net_device *dev)
2627 {
2628         struct inet6_dev  *idev;
2629         struct net_device *sp_dev;
2630         struct inet6_ifaddr *sp_ifa;
2631         struct rt6_info *sp_rt;
2632
2633         /* ::1 */
2634
2635         ASSERT_RTNL();
2636
2637         if ((idev = ipv6_find_idev(dev)) == NULL) {
2638                 pr_debug("%s: add_dev failed\n", __func__);
2639                 return;
2640         }
2641
2642         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2643
2644         /* Add routes to other interface's IPv6 addresses */
2645         for_each_netdev(dev_net(dev), sp_dev) {
2646                 if (!strcmp(sp_dev->name, dev->name))
2647                         continue;
2648
2649                 idev = __in6_dev_get(sp_dev);
2650                 if (!idev)
2651                         continue;
2652
2653                 read_lock_bh(&idev->lock);
2654                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2655
2656                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2657                                 continue;
2658
2659                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2660
2661                         /* Failure cases are ignored */
2662                         if (!IS_ERR(sp_rt)) {
2663                                 sp_ifa->rt = sp_rt;
2664                                 ip6_ins_rt(sp_rt);
2665                         }
2666                 }
2667                 read_unlock_bh(&idev->lock);
2668         }
2669 }
2670
2671 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2672 {
2673         struct inet6_ifaddr *ifp;
2674         u32 addr_flags = IFA_F_PERMANENT;
2675
2676 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2677         if (idev->cnf.optimistic_dad &&
2678             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2679                 addr_flags |= IFA_F_OPTIMISTIC;
2680 #endif
2681
2682
2683         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2684         if (!IS_ERR(ifp)) {
2685                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2686                 addrconf_dad_start(ifp);
2687                 in6_ifa_put(ifp);
2688         }
2689 }
2690
2691 static void addrconf_dev_config(struct net_device *dev)
2692 {
2693         struct in6_addr addr;
2694         struct inet6_dev *idev;
2695
2696         ASSERT_RTNL();
2697
2698         if ((dev->type != ARPHRD_ETHER) &&
2699             (dev->type != ARPHRD_FDDI) &&
2700             (dev->type != ARPHRD_ARCNET) &&
2701             (dev->type != ARPHRD_INFINIBAND) &&
2702             (dev->type != ARPHRD_IEEE802154) &&
2703             (dev->type != ARPHRD_IEEE1394)) {
2704                 /* Alas, we support only Ethernet autoconfiguration. */
2705                 return;
2706         }
2707
2708         idev = addrconf_add_dev(dev);
2709         if (IS_ERR(idev))
2710                 return;
2711
2712         memset(&addr, 0, sizeof(struct in6_addr));
2713         addr.s6_addr32[0] = htonl(0xFE800000);
2714
2715         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2716                 addrconf_add_linklocal(idev, &addr);
2717 }
2718
2719 #if IS_ENABLED(CONFIG_IPV6_SIT)
2720 static void addrconf_sit_config(struct net_device *dev)
2721 {
2722         struct inet6_dev *idev;
2723
2724         ASSERT_RTNL();
2725
2726         /*
2727          * Configure the tunnel with one of our IPv4
2728          * addresses... we should configure all of
2729          * our v4 addrs in the tunnel
2730          */
2731
2732         if ((idev = ipv6_find_idev(dev)) == NULL) {
2733                 pr_debug("%s: add_dev failed\n", __func__);
2734                 return;
2735         }
2736
2737         if (dev->priv_flags & IFF_ISATAP) {
2738                 struct in6_addr addr;
2739
2740                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2741                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2742                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2743                         addrconf_add_linklocal(idev, &addr);
2744                 return;
2745         }
2746
2747         sit_add_v4_addrs(idev);
2748
2749         if (dev->flags&IFF_POINTOPOINT)
2750                 addrconf_add_mroute(dev);
2751         else
2752                 sit_route_add(dev);
2753 }
2754 #endif
2755
2756 #if IS_ENABLED(CONFIG_NET_IPGRE)
2757 static void addrconf_gre_config(struct net_device *dev)
2758 {
2759         struct inet6_dev *idev;
2760         struct in6_addr addr;
2761
2762         ASSERT_RTNL();
2763
2764         if ((idev = ipv6_find_idev(dev)) == NULL) {
2765                 pr_debug("%s: add_dev failed\n", __func__);
2766                 return;
2767         }
2768
2769         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2770         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2771
2772         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2773                 addrconf_add_linklocal(idev, &addr);
2774 }
2775 #endif
2776
2777 static inline int
2778 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2779 {
2780         struct in6_addr lladdr;
2781
2782         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2783                 addrconf_add_linklocal(idev, &lladdr);
2784                 return 0;
2785         }
2786         return -1;
2787 }
2788
2789 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2790 {
2791         struct net_device *link_dev;
2792         struct net *net = dev_net(idev->dev);
2793
2794         /* first try to inherit the link-local address from the link device */
2795         if (idev->dev->iflink &&
2796             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2797                 if (!ipv6_inherit_linklocal(idev, link_dev))
2798                         return;
2799         }
2800         /* then try to inherit it from any device */
2801         for_each_netdev(net, link_dev) {
2802                 if (!ipv6_inherit_linklocal(idev, link_dev))
2803                         return;
2804         }
2805         pr_debug("init ip6-ip6: add_linklocal failed\n");
2806 }
2807
2808 /*
2809  * Autoconfigure tunnel with a link-local address so routing protocols,
2810  * DHCPv6, MLD etc. can be run over the virtual link
2811  */
2812
2813 static void addrconf_ip6_tnl_config(struct net_device *dev)
2814 {
2815         struct inet6_dev *idev;
2816
2817         ASSERT_RTNL();
2818
2819         idev = addrconf_add_dev(dev);
2820         if (IS_ERR(idev)) {
2821                 pr_debug("init ip6-ip6: add_dev failed\n");
2822                 return;
2823         }
2824         ip6_tnl_add_linklocal(idev);
2825 }
2826
2827 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2828                            void *ptr)
2829 {
2830         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2831         struct inet6_dev *idev = __in6_dev_get(dev);
2832         int run_pending = 0;
2833         int err;
2834
2835         switch (event) {
2836         case NETDEV_REGISTER:
2837                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2838                         idev = ipv6_add_dev(dev);
2839                         if (!idev)
2840                                 return notifier_from_errno(-ENOMEM);
2841                 }
2842                 break;
2843
2844         case NETDEV_UP:
2845         case NETDEV_CHANGE:
2846                 if (dev->flags & IFF_SLAVE)
2847                         break;
2848
2849                 if (event == NETDEV_UP) {
2850                         if (!addrconf_qdisc_ok(dev)) {
2851                                 /* device is not ready yet. */
2852                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2853                                         dev->name);
2854                                 break;
2855                         }
2856
2857                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2858                                 idev = ipv6_add_dev(dev);
2859
2860                         if (idev) {
2861                                 idev->if_flags |= IF_READY;
2862                                 run_pending = 1;
2863                         }
2864                 } else {
2865                         if (!addrconf_qdisc_ok(dev)) {
2866                                 /* device is still not ready. */
2867                                 break;
2868                         }
2869
2870                         if (idev) {
2871                                 if (idev->if_flags & IF_READY)
2872                                         /* device is already configured. */
2873                                         break;
2874                                 idev->if_flags |= IF_READY;
2875                         }
2876
2877                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2878                                 dev->name);
2879
2880                         run_pending = 1;
2881                 }
2882
2883                 switch (dev->type) {
2884 #if IS_ENABLED(CONFIG_IPV6_SIT)
2885                 case ARPHRD_SIT:
2886                         addrconf_sit_config(dev);
2887                         break;
2888 #endif
2889 #if IS_ENABLED(CONFIG_NET_IPGRE)
2890                 case ARPHRD_IPGRE:
2891                         addrconf_gre_config(dev);
2892                         break;
2893 #endif
2894                 case ARPHRD_TUNNEL6:
2895                         addrconf_ip6_tnl_config(dev);
2896                         break;
2897                 case ARPHRD_LOOPBACK:
2898                         init_loopback(dev);
2899                         break;
2900
2901                 default:
2902                         addrconf_dev_config(dev);
2903                         break;
2904                 }
2905
2906                 if (idev) {
2907                         if (run_pending)
2908                                 addrconf_dad_run(idev);
2909
2910                         /*
2911                          * If the MTU changed during the interface down,
2912                          * when the interface up, the changed MTU must be
2913                          * reflected in the idev as well as routers.
2914                          */
2915                         if (idev->cnf.mtu6 != dev->mtu &&
2916                             dev->mtu >= IPV6_MIN_MTU) {
2917                                 rt6_mtu_change(dev, dev->mtu);
2918                                 idev->cnf.mtu6 = dev->mtu;
2919                         }
2920                         idev->tstamp = jiffies;
2921                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2922
2923                         /*
2924                          * If the changed mtu during down is lower than
2925                          * IPV6_MIN_MTU stop IPv6 on this interface.
2926                          */
2927                         if (dev->mtu < IPV6_MIN_MTU)
2928                                 addrconf_ifdown(dev, 1);
2929                 }
2930                 break;
2931
2932         case NETDEV_CHANGEMTU:
2933                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2934                         rt6_mtu_change(dev, dev->mtu);
2935                         idev->cnf.mtu6 = dev->mtu;
2936                         break;
2937                 }
2938
2939                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2940                         idev = ipv6_add_dev(dev);
2941                         if (idev)
2942                                 break;
2943                 }
2944
2945                 /*
2946                  * MTU falled under IPV6_MIN_MTU.
2947                  * Stop IPv6 on this interface.
2948                  */
2949
2950         case NETDEV_DOWN:
2951         case NETDEV_UNREGISTER:
2952                 /*
2953                  *      Remove all addresses from this interface.
2954                  */
2955                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2956                 break;
2957
2958         case NETDEV_CHANGENAME:
2959                 if (idev) {
2960                         snmp6_unregister_dev(idev);
2961                         addrconf_sysctl_unregister(idev);
2962                         addrconf_sysctl_register(idev);
2963                         err = snmp6_register_dev(idev);
2964                         if (err)
2965                                 return notifier_from_errno(err);
2966                 }
2967                 break;
2968
2969         case NETDEV_PRE_TYPE_CHANGE:
2970         case NETDEV_POST_TYPE_CHANGE:
2971                 addrconf_type_change(dev, event);
2972                 break;
2973         }
2974
2975         return NOTIFY_OK;
2976 }
2977
2978 /*
2979  *      addrconf module should be notified of a device going up
2980  */
2981 static struct notifier_block ipv6_dev_notf = {
2982         .notifier_call = addrconf_notify,
2983 };
2984
2985 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2986 {
2987         struct inet6_dev *idev;
2988         ASSERT_RTNL();
2989
2990         idev = __in6_dev_get(dev);
2991
2992         if (event == NETDEV_POST_TYPE_CHANGE)
2993                 ipv6_mc_remap(idev);
2994         else if (event == NETDEV_PRE_TYPE_CHANGE)
2995                 ipv6_mc_unmap(idev);
2996 }
2997
2998 static int addrconf_ifdown(struct net_device *dev, int how)
2999 {
3000         struct net *net = dev_net(dev);
3001         struct inet6_dev *idev;
3002         struct inet6_ifaddr *ifa;
3003         int state, i;
3004
3005         ASSERT_RTNL();
3006
3007         rt6_ifdown(net, dev);
3008         neigh_ifdown(&nd_tbl, dev);
3009
3010         idev = __in6_dev_get(dev);
3011         if (idev == NULL)
3012                 return -ENODEV;
3013
3014         /*
3015          * Step 1: remove reference to ipv6 device from parent device.
3016          *         Do not dev_put!
3017          */
3018         if (how) {
3019                 idev->dead = 1;
3020
3021                 /* protected by rtnl_lock */
3022                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3023
3024                 /* Step 1.5: remove snmp6 entry */
3025                 snmp6_unregister_dev(idev);
3026
3027         }
3028
3029         /* Step 2: clear hash table */
3030         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3031                 struct hlist_head *h = &inet6_addr_lst[i];
3032
3033                 spin_lock_bh(&addrconf_hash_lock);
3034         restart:
3035                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3036                         if (ifa->idev == idev) {
3037                                 hlist_del_init_rcu(&ifa->addr_lst);
3038                                 addrconf_del_dad_timer(ifa);
3039                                 goto restart;
3040                         }
3041                 }
3042                 spin_unlock_bh(&addrconf_hash_lock);
3043         }
3044
3045         write_lock_bh(&idev->lock);
3046
3047         addrconf_del_rs_timer(idev);
3048
3049         /* Step 2: clear flags for stateless addrconf */
3050         if (!how)
3051                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3052
3053 #ifdef CONFIG_IPV6_PRIVACY
3054         if (how && del_timer(&idev->regen_timer))
3055                 in6_dev_put(idev);
3056
3057         /* Step 3: clear tempaddr list */
3058         while (!list_empty(&idev->tempaddr_list)) {
3059                 ifa = list_first_entry(&idev->tempaddr_list,
3060                                        struct inet6_ifaddr, tmp_list);
3061                 list_del(&ifa->tmp_list);
3062                 write_unlock_bh(&idev->lock);
3063                 spin_lock_bh(&ifa->lock);
3064
3065                 if (ifa->ifpub) {
3066                         in6_ifa_put(ifa->ifpub);
3067                         ifa->ifpub = NULL;
3068                 }
3069                 spin_unlock_bh(&ifa->lock);
3070                 in6_ifa_put(ifa);
3071                 write_lock_bh(&idev->lock);
3072         }
3073 #endif
3074
3075         while (!list_empty(&idev->addr_list)) {
3076                 ifa = list_first_entry(&idev->addr_list,
3077                                        struct inet6_ifaddr, if_list);
3078                 addrconf_del_dad_timer(ifa);
3079
3080                 list_del(&ifa->if_list);
3081
3082                 write_unlock_bh(&idev->lock);
3083
3084                 spin_lock_bh(&ifa->state_lock);
3085                 state = ifa->state;
3086                 ifa->state = INET6_IFADDR_STATE_DEAD;
3087                 spin_unlock_bh(&ifa->state_lock);
3088
3089                 if (state != INET6_IFADDR_STATE_DEAD) {
3090                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3091                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3092                 }
3093                 in6_ifa_put(ifa);
3094
3095                 write_lock_bh(&idev->lock);
3096         }
3097
3098         write_unlock_bh(&idev->lock);
3099
3100         /* Step 5: Discard multicast list */
3101         if (how)
3102                 ipv6_mc_destroy_dev(idev);
3103         else
3104                 ipv6_mc_down(idev);
3105
3106         idev->tstamp = jiffies;
3107
3108         /* Last: Shot the device (if unregistered) */
3109         if (how) {
3110                 addrconf_sysctl_unregister(idev);
3111                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3112                 neigh_ifdown(&nd_tbl, dev);
3113                 in6_dev_put(idev);
3114         }
3115         return 0;
3116 }
3117
3118 static void addrconf_rs_timer(unsigned long data)
3119 {
3120         struct inet6_dev *idev = (struct inet6_dev *)data;
3121         struct in6_addr lladdr;
3122
3123         write_lock(&idev->lock);
3124         if (idev->dead || !(idev->if_flags & IF_READY))
3125                 goto out;
3126
3127         if (!ipv6_accept_ra(idev))
3128                 goto out;
3129
3130         /* Announcement received after solicitation was sent */
3131         if (idev->if_flags & IF_RA_RCVD)
3132                 goto out;
3133
3134         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3135                 if (!__ipv6_get_lladdr(idev, &lladdr, IFA_F_TENTATIVE))
3136                         ndisc_send_rs(idev->dev, &lladdr,
3137                                       &in6addr_linklocal_allrouters);
3138                 else
3139                         goto out;
3140
3141                 /* The wait after the last probe can be shorter */
3142                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3143                                              idev->cnf.rtr_solicits) ?
3144                                       idev->cnf.rtr_solicit_delay :
3145                                       idev->cnf.rtr_solicit_interval);
3146         } else {
3147                 /*
3148                  * Note: we do not support deprecated "all on-link"
3149                  * assumption any longer.
3150                  */
3151                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3152         }
3153
3154 out:
3155         write_unlock(&idev->lock);
3156         in6_dev_put(idev);
3157 }
3158
3159 /*
3160  *      Duplicate Address Detection
3161  */
3162 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3163 {
3164         unsigned long rand_num;
3165         struct inet6_dev *idev = ifp->idev;
3166
3167         if (ifp->flags & IFA_F_OPTIMISTIC)
3168                 rand_num = 0;
3169         else
3170                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3171
3172         ifp->dad_probes = idev->cnf.dad_transmits;
3173         addrconf_mod_dad_timer(ifp, rand_num);
3174 }
3175
3176 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3177 {
3178         struct inet6_dev *idev = ifp->idev;
3179         struct net_device *dev = idev->dev;
3180
3181         addrconf_join_solict(dev, &ifp->addr);
3182
3183         net_srandom(ifp->addr.s6_addr32[3]);
3184
3185         read_lock_bh(&idev->lock);
3186         spin_lock(&ifp->lock);
3187         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3188                 goto out;
3189
3190         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3191             idev->cnf.accept_dad < 1 ||
3192             !(ifp->flags&IFA_F_TENTATIVE) ||
3193             ifp->flags & IFA_F_NODAD) {
3194                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3195                 spin_unlock(&ifp->lock);
3196                 read_unlock_bh(&idev->lock);
3197
3198                 addrconf_dad_completed(ifp);
3199                 return;
3200         }
3201
3202         if (!(idev->if_flags & IF_READY)) {
3203                 spin_unlock(&ifp->lock);
3204                 read_unlock_bh(&idev->lock);
3205                 /*
3206                  * If the device is not ready:
3207                  * - keep it tentative if it is a permanent address.
3208                  * - otherwise, kill it.
3209                  */
3210                 in6_ifa_hold(ifp);
3211                 addrconf_dad_stop(ifp, 0);
3212                 return;
3213         }
3214
3215         /*
3216          * Optimistic nodes can start receiving
3217          * Frames right away
3218          */
3219         if (ifp->flags & IFA_F_OPTIMISTIC)
3220                 ip6_ins_rt(ifp->rt);
3221
3222         addrconf_dad_kick(ifp);
3223 out:
3224         spin_unlock(&ifp->lock);
3225         read_unlock_bh(&idev->lock);
3226 }
3227
3228 static void addrconf_dad_timer(unsigned long data)
3229 {
3230         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3231         struct inet6_dev *idev = ifp->idev;
3232         struct in6_addr mcaddr;
3233
3234         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3235                 goto out;
3236
3237         write_lock(&idev->lock);
3238         if (idev->dead || !(idev->if_flags & IF_READY)) {
3239                 write_unlock(&idev->lock);
3240                 goto out;
3241         }
3242
3243         spin_lock(&ifp->lock);
3244         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3245                 spin_unlock(&ifp->lock);
3246                 write_unlock(&idev->lock);
3247                 goto out;
3248         }
3249
3250         if (ifp->dad_probes == 0) {
3251                 /*
3252                  * DAD was successful
3253                  */
3254
3255                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3256                 spin_unlock(&ifp->lock);
3257                 write_unlock(&idev->lock);
3258
3259                 addrconf_dad_completed(ifp);
3260
3261                 goto out;
3262         }
3263
3264         ifp->dad_probes--;
3265         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3266         spin_unlock(&ifp->lock);
3267         write_unlock(&idev->lock);
3268
3269         /* send a neighbour solicitation for our addr */
3270         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3271         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3272 out:
3273         in6_ifa_put(ifp);
3274 }
3275
3276 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3277 {
3278         struct net_device *dev = ifp->idev->dev;
3279         struct in6_addr lladdr;
3280
3281         addrconf_del_dad_timer(ifp);
3282
3283         /*
3284          *      Configure the address for reception. Now it is valid.
3285          */
3286
3287         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3288
3289         /* If added prefix is link local and we are prepared to process
3290            router advertisements, start sending router solicitations.
3291          */
3292
3293         if (ipv6_accept_ra(ifp->idev) &&
3294             ifp->idev->cnf.rtr_solicits > 0 &&
3295             (dev->flags&IFF_LOOPBACK) == 0 &&
3296             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3297                 /*
3298                  *      If a host as already performed a random delay
3299                  *      [...] as part of DAD [...] there is no need
3300                  *      to delay again before sending the first RS
3301                  */
3302                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3303                         ndisc_send_rs(dev, &lladdr,
3304                                       &in6addr_linklocal_allrouters);
3305                 else
3306                         return;
3307
3308                 write_lock_bh(&ifp->idev->lock);
3309                 spin_lock(&ifp->lock);
3310                 ifp->idev->rs_probes = 1;
3311                 ifp->idev->if_flags |= IF_RS_SENT;
3312                 addrconf_mod_rs_timer(ifp->idev,
3313                                       ifp->idev->cnf.rtr_solicit_interval);
3314                 spin_unlock(&ifp->lock);
3315                 write_unlock_bh(&ifp->idev->lock);
3316         }
3317 }
3318
3319 static void addrconf_dad_run(struct inet6_dev *idev)
3320 {
3321         struct inet6_ifaddr *ifp;
3322
3323         read_lock_bh(&idev->lock);
3324         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3325                 spin_lock(&ifp->lock);
3326                 if (ifp->flags & IFA_F_TENTATIVE &&
3327                     ifp->state == INET6_IFADDR_STATE_DAD)
3328                         addrconf_dad_kick(ifp);
3329                 spin_unlock(&ifp->lock);
3330         }
3331         read_unlock_bh(&idev->lock);
3332 }
3333
3334 #ifdef CONFIG_PROC_FS
3335 struct if6_iter_state {
3336         struct seq_net_private p;
3337         int bucket;
3338         int offset;
3339 };
3340
3341 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3342 {
3343         struct inet6_ifaddr *ifa = NULL;
3344         struct if6_iter_state *state = seq->private;
3345         struct net *net = seq_file_net(seq);
3346         int p = 0;
3347
3348         /* initial bucket if pos is 0 */
3349         if (pos == 0) {
3350                 state->bucket = 0;
3351                 state->offset = 0;
3352         }
3353
3354         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3355                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3356                                          addr_lst) {
3357                         if (!net_eq(dev_net(ifa->idev->dev), net))
3358                                 continue;
3359                         /* sync with offset */
3360                         if (p < state->offset) {
3361                                 p++;
3362                                 continue;
3363                         }
3364                         state->offset++;
3365                         return ifa;
3366                 }
3367
3368                 /* prepare for next bucket */
3369                 state->offset = 0;
3370                 p = 0;
3371         }
3372         return NULL;
3373 }
3374
3375 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3376                                          struct inet6_ifaddr *ifa)
3377 {
3378         struct if6_iter_state *state = seq->private;
3379         struct net *net = seq_file_net(seq);
3380
3381         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3382                 if (!net_eq(dev_net(ifa->idev->dev), net))
3383                         continue;
3384                 state->offset++;
3385                 return ifa;
3386         }
3387
3388         while (++state->bucket < IN6_ADDR_HSIZE) {
3389                 state->offset = 0;
3390                 hlist_for_each_entry_rcu_bh(ifa,
3391                                      &inet6_addr_lst[state->bucket], addr_lst) {
3392                         if (!net_eq(dev_net(ifa->idev->dev), net))
3393                                 continue;
3394                         state->offset++;
3395                         return ifa;
3396                 }
3397         }
3398
3399         return NULL;
3400 }
3401
3402 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3403         __acquires(rcu_bh)
3404 {
3405         rcu_read_lock_bh();
3406         return if6_get_first(seq, *pos);
3407 }
3408
3409 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3410 {
3411         struct inet6_ifaddr *ifa;
3412
3413         ifa = if6_get_next(seq, v);
3414         ++*pos;
3415         return ifa;
3416 }
3417
3418 static void if6_seq_stop(struct seq_file *seq, void *v)
3419         __releases(rcu_bh)
3420 {
3421         rcu_read_unlock_bh();
3422 }
3423
3424 static int if6_seq_show(struct seq_file *seq, void *v)
3425 {
3426         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3427         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3428                    &ifp->addr,
3429                    ifp->idev->dev->ifindex,
3430                    ifp->prefix_len,
3431                    ifp->scope,
3432                    ifp->flags,
3433                    ifp->idev->dev->name);
3434         return 0;
3435 }
3436
3437 static const struct seq_operations if6_seq_ops = {
3438         .start  = if6_seq_start,
3439         .next   = if6_seq_next,
3440         .show   = if6_seq_show,
3441         .stop   = if6_seq_stop,
3442 };
3443
3444 static int if6_seq_open(struct inode *inode, struct file *file)
3445 {
3446         return seq_open_net(inode, file, &if6_seq_ops,
3447                             sizeof(struct if6_iter_state));
3448 }
3449
3450 static const struct file_operations if6_fops = {
3451         .owner          = THIS_MODULE,
3452         .open           = if6_seq_open,
3453         .read           = seq_read,
3454         .llseek         = seq_lseek,
3455         .release        = seq_release_net,
3456 };
3457
3458 static int __net_init if6_proc_net_init(struct net *net)
3459 {
3460         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3461                 return -ENOMEM;
3462         return 0;
3463 }
3464
3465 static void __net_exit if6_proc_net_exit(struct net *net)
3466 {
3467         remove_proc_entry("if_inet6", net->proc_net);
3468 }
3469
3470 static struct pernet_operations if6_proc_net_ops = {
3471        .init = if6_proc_net_init,
3472        .exit = if6_proc_net_exit,
3473 };
3474
3475 int __init if6_proc_init(void)
3476 {
3477         return register_pernet_subsys(&if6_proc_net_ops);
3478 }
3479
3480 void if6_proc_exit(void)
3481 {
3482         unregister_pernet_subsys(&if6_proc_net_ops);
3483 }
3484 #endif  /* CONFIG_PROC_FS */
3485
3486 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3487 /* Check if address is a home address configured on any interface. */
3488 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3489 {
3490         int ret = 0;
3491         struct inet6_ifaddr *ifp = NULL;
3492         unsigned int hash = inet6_addr_hash(addr);
3493
3494         rcu_read_lock_bh();
3495         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3496                 if (!net_eq(dev_net(ifp->idev->dev), net))
3497                         continue;
3498                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3499                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3500                         ret = 1;
3501                         break;
3502                 }
3503         }
3504         rcu_read_unlock_bh();
3505         return ret;
3506 }
3507 #endif
3508
3509 /*
3510  *      Periodic address status verification
3511  */
3512
3513 static void addrconf_verify(unsigned long foo)
3514 {
3515         unsigned long now, next, next_sec, next_sched;
3516         struct inet6_ifaddr *ifp;
3517         int i;
3518
3519         rcu_read_lock_bh();
3520         spin_lock(&addrconf_verify_lock);
3521         now = jiffies;
3522         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3523
3524         del_timer(&addr_chk_timer);
3525
3526         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3527 restart:
3528                 hlist_for_each_entry_rcu_bh(ifp,
3529                                          &inet6_addr_lst[i], addr_lst) {
3530                         unsigned long age;
3531
3532                         if (ifp->flags & IFA_F_PERMANENT)
3533                                 continue;
3534
3535                         spin_lock(&ifp->lock);
3536                         /* We try to batch several events at once. */
3537                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3538
3539                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3540                             age >= ifp->valid_lft) {
3541                                 spin_unlock(&ifp->lock);
3542                                 in6_ifa_hold(ifp);
3543                                 ipv6_del_addr(ifp);
3544                                 goto restart;
3545                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3546                                 spin_unlock(&ifp->lock);
3547                                 continue;
3548                         } else if (age >= ifp->prefered_lft) {
3549                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3550                                 int deprecate = 0;
3551
3552                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3553                                         deprecate = 1;
3554                                         ifp->flags |= IFA_F_DEPRECATED;
3555                                 }
3556
3557                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3558                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3559
3560                                 spin_unlock(&ifp->lock);
3561
3562                                 if (deprecate) {
3563                                         in6_ifa_hold(ifp);
3564
3565                                         ipv6_ifa_notify(0, ifp);
3566                                         in6_ifa_put(ifp);
3567                                         goto restart;
3568                                 }
3569 #ifdef CONFIG_IPV6_PRIVACY
3570                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3571                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3572                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3573                                         ifp->idev->cnf.dad_transmits *
3574                                         ifp->idev->nd_parms->retrans_time / HZ;
3575
3576                                 if (age >= ifp->prefered_lft - regen_advance) {
3577                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3578                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3579                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3580                                         if (!ifp->regen_count && ifpub) {
3581                                                 ifp->regen_count++;
3582                                                 in6_ifa_hold(ifp);
3583                                                 in6_ifa_hold(ifpub);
3584                                                 spin_unlock(&ifp->lock);
3585
3586                                                 spin_lock(&ifpub->lock);
3587                                                 ifpub->regen_count = 0;
3588                                                 spin_unlock(&ifpub->lock);
3589                                                 ipv6_create_tempaddr(ifpub, ifp);
3590                                                 in6_ifa_put(ifpub);
3591                                                 in6_ifa_put(ifp);
3592                                                 goto restart;
3593                                         }
3594                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3595                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3596                                 spin_unlock(&ifp->lock);
3597 #endif
3598                         } else {
3599                                 /* ifp->prefered_lft <= ifp->valid_lft */
3600                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3601                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3602                                 spin_unlock(&ifp->lock);
3603                         }
3604                 }
3605         }
3606
3607         next_sec = round_jiffies_up(next);
3608         next_sched = next;
3609
3610         /* If rounded timeout is accurate enough, accept it. */
3611         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3612                 next_sched = next_sec;
3613
3614         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3615         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3616                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3617
3618         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3619               now, next, next_sec, next_sched));
3620
3621         addr_chk_timer.expires = next_sched;
3622         add_timer(&addr_chk_timer);
3623         spin_unlock(&addrconf_verify_lock);
3624         rcu_read_unlock_bh();
3625 }
3626
3627 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3628                                      struct in6_addr **peer_pfx)
3629 {
3630         struct in6_addr *pfx = NULL;
3631
3632         *peer_pfx = NULL;
3633
3634         if (addr)
3635                 pfx = nla_data(addr);
3636
3637         if (local) {
3638                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3639                         *peer_pfx = pfx;
3640                 pfx = nla_data(local);
3641         }
3642
3643         return pfx;
3644 }
3645
3646 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3647         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3648         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3649         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3650 };
3651
3652 static int
3653 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3654 {
3655         struct net *net = sock_net(skb->sk);
3656         struct ifaddrmsg *ifm;
3657         struct nlattr *tb[IFA_MAX+1];
3658         struct in6_addr *pfx, *peer_pfx;
3659         int err;
3660
3661         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3662         if (err < 0)
3663                 return err;
3664
3665         ifm = nlmsg_data(nlh);
3666         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3667         if (pfx == NULL)
3668                 return -EINVAL;
3669
3670         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3671 }
3672
3673 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3674                              u32 prefered_lft, u32 valid_lft)
3675 {
3676         u32 flags;
3677         clock_t expires;
3678         unsigned long timeout;
3679
3680         if (!valid_lft || (prefered_lft > valid_lft))
3681                 return -EINVAL;
3682
3683         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3684         if (addrconf_finite_timeout(timeout)) {
3685                 expires = jiffies_to_clock_t(timeout * HZ);
3686                 valid_lft = timeout;
3687                 flags = RTF_EXPIRES;
3688         } else {
3689                 expires = 0;
3690                 flags = 0;
3691                 ifa_flags |= IFA_F_PERMANENT;
3692         }
3693
3694         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3695         if (addrconf_finite_timeout(timeout)) {
3696                 if (timeout == 0)
3697                         ifa_flags |= IFA_F_DEPRECATED;
3698                 prefered_lft = timeout;
3699         }
3700
3701         spin_lock_bh(&ifp->lock);
3702         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3703         ifp->tstamp = jiffies;
3704         ifp->valid_lft = valid_lft;
3705         ifp->prefered_lft = prefered_lft;
3706
3707         spin_unlock_bh(&ifp->lock);
3708         if (!(ifp->flags&IFA_F_TENTATIVE))
3709                 ipv6_ifa_notify(0, ifp);
3710
3711         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3712                               expires, flags);
3713         addrconf_verify(0);
3714
3715         return 0;
3716 }
3717
3718 static int
3719 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3720 {
3721         struct net *net = sock_net(skb->sk);
3722         struct ifaddrmsg *ifm;
3723         struct nlattr *tb[IFA_MAX+1];
3724         struct in6_addr *pfx, *peer_pfx;
3725         struct inet6_ifaddr *ifa;
3726         struct net_device *dev;
3727         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3728         u8 ifa_flags;
3729         int err;
3730
3731         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3732         if (err < 0)
3733                 return err;
3734
3735         ifm = nlmsg_data(nlh);
3736         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3737         if (pfx == NULL)
3738                 return -EINVAL;
3739
3740         if (tb[IFA_CACHEINFO]) {
3741                 struct ifa_cacheinfo *ci;
3742
3743                 ci = nla_data(tb[IFA_CACHEINFO]);
3744                 valid_lft = ci->ifa_valid;
3745                 preferred_lft = ci->ifa_prefered;
3746         } else {
3747                 preferred_lft = INFINITY_LIFE_TIME;
3748                 valid_lft = INFINITY_LIFE_TIME;
3749         }
3750
3751         dev =  __dev_get_by_index(net, ifm->ifa_index);
3752         if (dev == NULL)
3753                 return -ENODEV;
3754
3755         /* We ignore other flags so far. */
3756         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3757
3758         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3759         if (ifa == NULL) {
3760                 /*
3761                  * It would be best to check for !NLM_F_CREATE here but
3762                  * userspace alreay relies on not having to provide this.
3763                  */
3764                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3765                                       ifm->ifa_prefixlen, ifa_flags,
3766                                       preferred_lft, valid_lft);
3767         }
3768
3769         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3770             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3771                 err = -EEXIST;
3772         else
3773                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3774
3775         in6_ifa_put(ifa);
3776
3777         return err;
3778 }
3779
3780 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3781                           u8 scope, int ifindex)
3782 {
3783         struct ifaddrmsg *ifm;
3784
3785         ifm = nlmsg_data(nlh);
3786         ifm->ifa_family = AF_INET6;
3787         ifm->ifa_prefixlen = prefixlen;
3788         ifm->ifa_flags = flags;
3789         ifm->ifa_scope = scope;
3790         ifm->ifa_index = ifindex;
3791 }
3792
3793 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3794                          unsigned long tstamp, u32 preferred, u32 valid)
3795 {
3796         struct ifa_cacheinfo ci;
3797
3798         ci.cstamp = cstamp_delta(cstamp);
3799         ci.tstamp = cstamp_delta(tstamp);
3800         ci.ifa_prefered = preferred;
3801         ci.ifa_valid = valid;
3802
3803         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3804 }
3805
3806 static inline int rt_scope(int ifa_scope)
3807 {
3808         if (ifa_scope & IFA_HOST)
3809                 return RT_SCOPE_HOST;
3810         else if (ifa_scope & IFA_LINK)
3811                 return RT_SCOPE_LINK;
3812         else if (ifa_scope & IFA_SITE)
3813                 return RT_SCOPE_SITE;
3814         else
3815                 return RT_SCOPE_UNIVERSE;
3816 }
3817
3818 static inline int inet6_ifaddr_msgsize(void)
3819 {
3820         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3821                + nla_total_size(16) /* IFA_LOCAL */
3822                + nla_total_size(16) /* IFA_ADDRESS */
3823                + nla_total_size(sizeof(struct ifa_cacheinfo));
3824 }
3825
3826 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3827                              u32 portid, u32 seq, int event, unsigned int flags)
3828 {
3829         struct nlmsghdr  *nlh;
3830         u32 preferred, valid;
3831
3832         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3833         if (nlh == NULL)
3834                 return -EMSGSIZE;
3835
3836         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3837                       ifa->idev->dev->ifindex);
3838
3839         if (!(ifa->flags&IFA_F_PERMANENT)) {
3840                 preferred = ifa->prefered_lft;
3841                 valid = ifa->valid_lft;
3842                 if (preferred != INFINITY_LIFE_TIME) {
3843                         long tval = (jiffies - ifa->tstamp)/HZ;
3844                         if (preferred > tval)
3845                                 preferred -= tval;
3846                         else
3847                                 preferred = 0;
3848                         if (valid != INFINITY_LIFE_TIME) {
3849                                 if (valid > tval)
3850                                         valid -= tval;
3851                                 else
3852                                         valid = 0;
3853                         }
3854                 }
3855         } else {
3856                 preferred = INFINITY_LIFE_TIME;
3857                 valid = INFINITY_LIFE_TIME;
3858         }
3859
3860         if (!ipv6_addr_any(&ifa->peer_addr)) {
3861                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3862                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3863                         goto error;
3864         } else
3865                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3866                         goto error;
3867
3868         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3869                 goto error;
3870
3871         return nlmsg_end(skb, nlh);
3872
3873 error:
3874         nlmsg_cancel(skb, nlh);
3875         return -EMSGSIZE;
3876 }
3877
3878 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3879                                 u32 portid, u32 seq, int event, u16 flags)
3880 {
3881         struct nlmsghdr  *nlh;
3882         u8 scope = RT_SCOPE_UNIVERSE;
3883         int ifindex = ifmca->idev->dev->ifindex;
3884
3885         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3886                 scope = RT_SCOPE_SITE;
3887
3888         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3889         if (nlh == NULL)
3890                 return -EMSGSIZE;
3891
3892         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3893         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3894             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3895                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3896                 nlmsg_cancel(skb, nlh);
3897                 return -EMSGSIZE;
3898         }
3899
3900         return nlmsg_end(skb, nlh);
3901 }
3902
3903 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3904                                 u32 portid, u32 seq, int event, unsigned int flags)
3905 {
3906         struct nlmsghdr  *nlh;
3907         u8 scope = RT_SCOPE_UNIVERSE;
3908         int ifindex = ifaca->aca_idev->dev->ifindex;
3909
3910         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3911                 scope = RT_SCOPE_SITE;
3912
3913         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3914         if (nlh == NULL)
3915                 return -EMSGSIZE;
3916
3917         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3918         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3919             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3920                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3921                 nlmsg_cancel(skb, nlh);
3922                 return -EMSGSIZE;
3923         }
3924
3925         return nlmsg_end(skb, nlh);
3926 }
3927
3928 enum addr_type_t {
3929         UNICAST_ADDR,
3930         MULTICAST_ADDR,
3931         ANYCAST_ADDR,
3932 };
3933
3934 /* called with rcu_read_lock() */
3935 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3936                           struct netlink_callback *cb, enum addr_type_t type,
3937                           int s_ip_idx, int *p_ip_idx)
3938 {
3939         struct ifmcaddr6 *ifmca;
3940         struct ifacaddr6 *ifaca;
3941         int err = 1;
3942         int ip_idx = *p_ip_idx;
3943
3944         read_lock_bh(&idev->lock);
3945         switch (type) {
3946         case UNICAST_ADDR: {
3947                 struct inet6_ifaddr *ifa;
3948
3949                 /* unicast address incl. temp addr */
3950                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3951                         if (++ip_idx < s_ip_idx)
3952                                 continue;
3953                         err = inet6_fill_ifaddr(skb, ifa,
3954                                                 NETLINK_CB(cb->skb).portid,
3955                                                 cb->nlh->nlmsg_seq,
3956                                                 RTM_NEWADDR,
3957                                                 NLM_F_MULTI);
3958                         if (err <= 0)
3959                                 break;
3960                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3961                 }
3962                 break;
3963         }
3964         case MULTICAST_ADDR:
3965                 /* multicast address */
3966                 for (ifmca = idev->mc_list; ifmca;
3967                      ifmca = ifmca->next, ip_idx++) {
3968                         if (ip_idx < s_ip_idx)
3969                                 continue;
3970                         err = inet6_fill_ifmcaddr(skb, ifmca,
3971                                                   NETLINK_CB(cb->skb).portid,
3972                                                   cb->nlh->nlmsg_seq,
3973                                                   RTM_GETMULTICAST,
3974                                                   NLM_F_MULTI);
3975                         if (err <= 0)
3976                                 break;
3977                 }
3978                 break;
3979         case ANYCAST_ADDR:
3980                 /* anycast address */
3981                 for (ifaca = idev->ac_list; ifaca;
3982                      ifaca = ifaca->aca_next, ip_idx++) {
3983                         if (ip_idx < s_ip_idx)
3984                                 continue;
3985                         err = inet6_fill_ifacaddr(skb, ifaca,
3986                                                   NETLINK_CB(cb->skb).portid,
3987                                                   cb->nlh->nlmsg_seq,
3988                                                   RTM_GETANYCAST,
3989                                                   NLM_F_MULTI);
3990                         if (err <= 0)
3991                                 break;
3992                 }
3993                 break;
3994         default:
3995                 break;
3996         }
3997         read_unlock_bh(&idev->lock);
3998         *p_ip_idx = ip_idx;
3999         return err;
4000 }
4001
4002 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4003                            enum addr_type_t type)
4004 {
4005         struct net *net = sock_net(skb->sk);
4006         int h, s_h;
4007         int idx, ip_idx;
4008         int s_idx, s_ip_idx;
4009         struct net_device *dev;
4010         struct inet6_dev *idev;
4011         struct hlist_head *head;
4012
4013         s_h = cb->args[0];
4014         s_idx = idx = cb->args[1];
4015         s_ip_idx = ip_idx = cb->args[2];
4016
4017         rcu_read_lock();
4018         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4019         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4020                 idx = 0;
4021                 head = &net->dev_index_head[h];
4022                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4023                         if (idx < s_idx)
4024                                 goto cont;
4025                         if (h > s_h || idx > s_idx)
4026                                 s_ip_idx = 0;
4027                         ip_idx = 0;
4028                         idev = __in6_dev_get(dev);
4029                         if (!idev)
4030                                 goto cont;
4031
4032                         if (in6_dump_addrs(idev, skb, cb, type,
4033                                            s_ip_idx, &ip_idx) <= 0)
4034                                 goto done;
4035 cont:
4036                         idx++;
4037                 }
4038         }
4039 done:
4040         rcu_read_unlock();
4041         cb->args[0] = h;
4042         cb->args[1] = idx;
4043         cb->args[2] = ip_idx;
4044
4045         return skb->len;
4046 }
4047
4048 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4049 {
4050         enum addr_type_t type = UNICAST_ADDR;
4051
4052         return inet6_dump_addr(skb, cb, type);
4053 }
4054
4055 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4056 {
4057         enum addr_type_t type = MULTICAST_ADDR;
4058
4059         return inet6_dump_addr(skb, cb, type);
4060 }
4061
4062
4063 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4064 {
4065         enum addr_type_t type = ANYCAST_ADDR;
4066
4067         return inet6_dump_addr(skb, cb, type);
4068 }
4069
4070 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4071 {
4072         struct net *net = sock_net(in_skb->sk);
4073         struct ifaddrmsg *ifm;
4074         struct nlattr *tb[IFA_MAX+1];
4075         struct in6_addr *addr = NULL, *peer;
4076         struct net_device *dev = NULL;
4077         struct inet6_ifaddr *ifa;
4078         struct sk_buff *skb;
4079         int err;
4080
4081         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4082         if (err < 0)
4083                 goto errout;
4084
4085         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4086         if (addr == NULL) {
4087                 err = -EINVAL;
4088                 goto errout;
4089         }
4090
4091         ifm = nlmsg_data(nlh);
4092         if (ifm->ifa_index)
4093                 dev = __dev_get_by_index(net, ifm->ifa_index);
4094
4095         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4096         if (!ifa) {
4097                 err = -EADDRNOTAVAIL;
4098                 goto errout;
4099         }
4100
4101         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4102         if (!skb) {
4103                 err = -ENOBUFS;
4104                 goto errout_ifa;
4105         }
4106
4107         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4108                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4109         if (err < 0) {
4110                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4111                 WARN_ON(err == -EMSGSIZE);
4112                 kfree_skb(skb);
4113                 goto errout_ifa;
4114         }
4115         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4116 errout_ifa:
4117         in6_ifa_put(ifa);
4118 errout:
4119         return err;
4120 }
4121
4122 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4123 {
4124         struct sk_buff *skb;
4125         struct net *net = dev_net(ifa->idev->dev);
4126         int err = -ENOBUFS;
4127
4128         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4129         if (skb == NULL)
4130                 goto errout;
4131
4132         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4133         if (err < 0) {
4134                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4135                 WARN_ON(err == -EMSGSIZE);
4136                 kfree_skb(skb);
4137                 goto errout;
4138         }
4139         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4140         return;
4141 errout:
4142         if (err < 0)
4143                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4144 }
4145
4146 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4147                                 __s32 *array, int bytes)
4148 {
4149         BUG_ON(bytes < (DEVCONF_MAX * 4));
4150
4151         memset(array, 0, bytes);
4152         array[DEVCONF_FORWARDING] = cnf->forwarding;
4153         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4154         array[DEVCONF_MTU6] = cnf->mtu6;
4155         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4156         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4157         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4158         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4159         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4160         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4161                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4162         array[DEVCONF_RTR_SOLICIT_DELAY] =
4163                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4164         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4165 #ifdef CONFIG_IPV6_PRIVACY
4166         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4167         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4168         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4169         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4170         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4171 #endif
4172         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4173         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4174         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4175 #ifdef CONFIG_IPV6_ROUTER_PREF
4176         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4177         array[DEVCONF_RTR_PROBE_INTERVAL] =
4178                 jiffies_to_msecs(cnf->rtr_probe_interval);
4179 #ifdef CONFIG_IPV6_ROUTE_INFO
4180         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4181 #endif
4182 #endif
4183         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4184         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4185 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4186         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4187 #endif
4188 #ifdef CONFIG_IPV6_MROUTE
4189         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4190 #endif
4191         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4192         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4193         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4194         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4195 }
4196
4197 static inline size_t inet6_ifla6_size(void)
4198 {
4199         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4200              + nla_total_size(sizeof(struct ifla_cacheinfo))
4201              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4202              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4203              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4204              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4205 }
4206
4207 static inline size_t inet6_if_nlmsg_size(void)
4208 {
4209         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4210                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4211                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4212                + nla_total_size(4) /* IFLA_MTU */
4213                + nla_total_size(4) /* IFLA_LINK */
4214                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4215 }
4216
4217 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4218                                       int items, int bytes)
4219 {
4220         int i;
4221         int pad = bytes - sizeof(u64) * items;
4222         BUG_ON(pad < 0);
4223
4224         /* Use put_unaligned() because stats may not be aligned for u64. */
4225         put_unaligned(items, &stats[0]);
4226         for (i = 1; i < items; i++)
4227                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4228
4229         memset(&stats[items], 0, pad);
4230 }
4231
4232 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4233                                       int items, int bytes, size_t syncpoff)
4234 {
4235         int i;
4236         int pad = bytes - sizeof(u64) * items;
4237         BUG_ON(pad < 0);
4238
4239         /* Use put_unaligned() because stats may not be aligned for u64. */
4240         put_unaligned(items, &stats[0]);
4241         for (i = 1; i < items; i++)
4242                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4243
4244         memset(&stats[items], 0, pad);
4245 }
4246
4247 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4248                              int bytes)
4249 {
4250         switch (attrtype) {
4251         case IFLA_INET6_STATS:
4252                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4253                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4254                 break;
4255         case IFLA_INET6_ICMP6STATS:
4256                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4257                 break;
4258         }
4259 }
4260
4261 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4262 {
4263         struct nlattr *nla;
4264         struct ifla_cacheinfo ci;
4265
4266         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4267                 goto nla_put_failure;
4268         ci.max_reasm_len = IPV6_MAXPLEN;
4269         ci.tstamp = cstamp_delta(idev->tstamp);
4270         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4271         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4272         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4273                 goto nla_put_failure;
4274         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4275         if (nla == NULL)
4276                 goto nla_put_failure;
4277         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4278
4279         /* XXX - MC not implemented */
4280
4281         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4282         if (nla == NULL)
4283                 goto nla_put_failure;
4284         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4285
4286         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4287         if (nla == NULL)
4288                 goto nla_put_failure;
4289         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4290
4291         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4292         if (nla == NULL)
4293                 goto nla_put_failure;
4294         read_lock_bh(&idev->lock);
4295         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4296         read_unlock_bh(&idev->lock);
4297
4298         return 0;
4299
4300 nla_put_failure:
4301         return -EMSGSIZE;
4302 }
4303
4304 static size_t inet6_get_link_af_size(const struct net_device *dev)
4305 {
4306         if (!__in6_dev_get(dev))
4307                 return 0;
4308
4309         return inet6_ifla6_size();
4310 }
4311
4312 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4313 {
4314         struct inet6_dev *idev = __in6_dev_get(dev);
4315
4316         if (!idev)
4317                 return -ENODATA;
4318
4319         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4320                 return -EMSGSIZE;
4321
4322         return 0;
4323 }
4324
4325 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4326 {
4327         struct inet6_ifaddr *ifp;
4328         struct net_device *dev = idev->dev;
4329         bool update_rs = false;
4330
4331         if (token == NULL)
4332                 return -EINVAL;
4333         if (ipv6_addr_any(token))
4334                 return -EINVAL;
4335         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4336                 return -EINVAL;
4337         if (!ipv6_accept_ra(idev))
4338                 return -EINVAL;
4339         if (idev->cnf.rtr_solicits <= 0)
4340                 return -EINVAL;
4341
4342         write_lock_bh(&idev->lock);
4343
4344         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4345         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4346
4347         write_unlock_bh(&idev->lock);
4348
4349         if (!idev->dead && (idev->if_flags & IF_READY)) {
4350                 struct in6_addr ll_addr;
4351
4352                 ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4353                                 IFA_F_OPTIMISTIC);
4354
4355                 /* If we're not ready, then normal ifup will take care
4356                  * of this. Otherwise, we need to request our rs here.
4357                  */
4358                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4359                 update_rs = true;
4360         }
4361
4362         write_lock_bh(&idev->lock);
4363
4364         if (update_rs) {
4365                 idev->if_flags |= IF_RS_SENT;
4366                 idev->rs_probes = 1;
4367                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4368         }
4369
4370         /* Well, that's kinda nasty ... */
4371         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4372                 spin_lock(&ifp->lock);
4373                 if (ifp->tokenized) {
4374                         ifp->valid_lft = 0;
4375                         ifp->prefered_lft = 0;
4376                 }
4377                 spin_unlock(&ifp->lock);
4378         }
4379
4380         write_unlock_bh(&idev->lock);
4381         addrconf_verify(0);
4382         return 0;
4383 }
4384
4385 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4386 {
4387         int err = -EINVAL;
4388         struct inet6_dev *idev = __in6_dev_get(dev);
4389         struct nlattr *tb[IFLA_INET6_MAX + 1];
4390
4391         if (!idev)
4392                 return -EAFNOSUPPORT;
4393
4394         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4395                 BUG();
4396
4397         if (tb[IFLA_INET6_TOKEN])
4398                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4399
4400         return err;
4401 }
4402
4403 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4404                              u32 portid, u32 seq, int event, unsigned int flags)
4405 {
4406         struct net_device *dev = idev->dev;
4407         struct ifinfomsg *hdr;
4408         struct nlmsghdr *nlh;
4409         void *protoinfo;
4410
4411         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4412         if (nlh == NULL)
4413                 return -EMSGSIZE;
4414
4415         hdr = nlmsg_data(nlh);
4416         hdr->ifi_family = AF_INET6;
4417         hdr->__ifi_pad = 0;
4418         hdr->ifi_type = dev->type;
4419         hdr->ifi_index = dev->ifindex;
4420         hdr->ifi_flags = dev_get_flags(dev);
4421         hdr->ifi_change = 0;
4422
4423         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4424             (dev->addr_len &&
4425              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4426             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4427             (dev->ifindex != dev->iflink &&
4428              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4429                 goto nla_put_failure;
4430         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4431         if (protoinfo == NULL)
4432                 goto nla_put_failure;
4433
4434         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4435                 goto nla_put_failure;
4436
4437         nla_nest_end(skb, protoinfo);
4438         return nlmsg_end(skb, nlh);
4439
4440 nla_put_failure:
4441         nlmsg_cancel(skb, nlh);
4442         return -EMSGSIZE;
4443 }
4444
4445 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4446 {
4447         struct net *net = sock_net(skb->sk);
4448         int h, s_h;
4449         int idx = 0, s_idx;
4450         struct net_device *dev;
4451         struct inet6_dev *idev;
4452         struct hlist_head *head;
4453
4454         s_h = cb->args[0];
4455         s_idx = cb->args[1];
4456
4457         rcu_read_lock();
4458         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4459                 idx = 0;
4460                 head = &net->dev_index_head[h];
4461                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4462                         if (idx < s_idx)
4463                                 goto cont;
4464                         idev = __in6_dev_get(dev);
4465                         if (!idev)
4466                                 goto cont;
4467                         if (inet6_fill_ifinfo(skb, idev,
4468                                               NETLINK_CB(cb->skb).portid,
4469                                               cb->nlh->nlmsg_seq,
4470                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4471                                 goto out;
4472 cont:
4473                         idx++;
4474                 }
4475         }
4476 out:
4477         rcu_read_unlock();
4478         cb->args[1] = idx;
4479         cb->args[0] = h;
4480
4481         return skb->len;
4482 }
4483
4484 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4485 {
4486         struct sk_buff *skb;
4487         struct net *net = dev_net(idev->dev);
4488         int err = -ENOBUFS;
4489
4490         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4491         if (skb == NULL)
4492                 goto errout;
4493
4494         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4495         if (err < 0) {
4496                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4497                 WARN_ON(err == -EMSGSIZE);
4498                 kfree_skb(skb);
4499                 goto errout;
4500         }
4501         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4502         return;
4503 errout:
4504         if (err < 0)
4505                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4506 }
4507
4508 static inline size_t inet6_prefix_nlmsg_size(void)
4509 {
4510         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4511                + nla_total_size(sizeof(struct in6_addr))
4512                + nla_total_size(sizeof(struct prefix_cacheinfo));
4513 }
4514
4515 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4516                              struct prefix_info *pinfo, u32 portid, u32 seq,
4517                              int event, unsigned int flags)
4518 {
4519         struct prefixmsg *pmsg;
4520         struct nlmsghdr *nlh;
4521         struct prefix_cacheinfo ci;
4522
4523         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4524         if (nlh == NULL)
4525                 return -EMSGSIZE;
4526
4527         pmsg = nlmsg_data(nlh);
4528         pmsg->prefix_family = AF_INET6;
4529         pmsg->prefix_pad1 = 0;
4530         pmsg->prefix_pad2 = 0;
4531         pmsg->prefix_ifindex = idev->dev->ifindex;
4532         pmsg->prefix_len = pinfo->prefix_len;
4533         pmsg->prefix_type = pinfo->type;
4534         pmsg->prefix_pad3 = 0;
4535         pmsg->prefix_flags = 0;
4536         if (pinfo->onlink)
4537                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4538         if (pinfo->autoconf)
4539                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4540
4541         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4542                 goto nla_put_failure;
4543         ci.preferred_time = ntohl(pinfo->prefered);
4544         ci.valid_time = ntohl(pinfo->valid);
4545         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4546                 goto nla_put_failure;
4547         return nlmsg_end(skb, nlh);
4548
4549 nla_put_failure:
4550         nlmsg_cancel(skb, nlh);
4551         return -EMSGSIZE;
4552 }
4553
4554 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4555                          struct prefix_info *pinfo)
4556 {
4557         struct sk_buff *skb;
4558         struct net *net = dev_net(idev->dev);
4559         int err = -ENOBUFS;
4560
4561         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4562         if (skb == NULL)
4563                 goto errout;
4564
4565         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4566         if (err < 0) {
4567                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4568                 WARN_ON(err == -EMSGSIZE);
4569                 kfree_skb(skb);
4570                 goto errout;
4571         }
4572         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4573         return;
4574 errout:
4575         if (err < 0)
4576                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4577 }
4578
4579 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4580 {
4581         struct net *net = dev_net(ifp->idev->dev);
4582
4583         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4584
4585         switch (event) {
4586         case RTM_NEWADDR:
4587                 /*
4588                  * If the address was optimistic
4589                  * we inserted the route at the start of
4590                  * our DAD process, so we don't need
4591                  * to do it again
4592                  */
4593                 if (!(ifp->rt->rt6i_node))
4594                         ip6_ins_rt(ifp->rt);
4595                 if (ifp->idev->cnf.forwarding)
4596                         addrconf_join_anycast(ifp);
4597                 if (!ipv6_addr_any(&ifp->peer_addr))
4598                         addrconf_prefix_route(&ifp->peer_addr, 128,
4599                                               ifp->idev->dev, 0, 0);
4600                 break;
4601         case RTM_DELADDR:
4602                 if (ifp->idev->cnf.forwarding)
4603                         addrconf_leave_anycast(ifp);
4604                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4605                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4606                         struct rt6_info *rt;
4607                         struct net_device *dev = ifp->idev->dev;
4608
4609                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4610                                         dev->ifindex, 1);
4611                         if (rt) {
4612                                 dst_hold(&rt->dst);
4613                                 if (ip6_del_rt(rt))
4614                                         dst_free(&rt->dst);
4615                         }
4616                 }
4617                 dst_hold(&ifp->rt->dst);
4618
4619                 if (ip6_del_rt(ifp->rt))
4620                         dst_free(&ifp->rt->dst);
4621                 break;
4622         }
4623         atomic_inc(&net->ipv6.dev_addr_genid);
4624 }
4625
4626 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4627 {
4628         rcu_read_lock_bh();
4629         if (likely(ifp->idev->dead == 0))
4630                 __ipv6_ifa_notify(event, ifp);
4631         rcu_read_unlock_bh();
4632 }
4633
4634 #ifdef CONFIG_SYSCTL
4635
4636 static
4637 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4638                            void __user *buffer, size_t *lenp, loff_t *ppos)
4639 {
4640         int *valp = ctl->data;
4641         int val = *valp;
4642         loff_t pos = *ppos;
4643         struct ctl_table lctl;
4644         int ret;
4645
4646         /*
4647          * ctl->data points to idev->cnf.forwarding, we should
4648          * not modify it until we get the rtnl lock.
4649          */
4650         lctl = *ctl;
4651         lctl.data = &val;
4652
4653         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4654
4655         if (write)
4656                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4657         if (ret)
4658                 *ppos = pos;
4659         return ret;
4660 }
4661
4662 static void dev_disable_change(struct inet6_dev *idev)
4663 {
4664         struct netdev_notifier_info info;
4665
4666         if (!idev || !idev->dev)
4667                 return;
4668
4669         netdev_notifier_info_init(&info, idev->dev);
4670         if (idev->cnf.disable_ipv6)
4671                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4672         else
4673                 addrconf_notify(NULL, NETDEV_UP, &info);
4674 }
4675
4676 static void addrconf_disable_change(struct net *net, __s32 newf)
4677 {
4678         struct net_device *dev;
4679         struct inet6_dev *idev;
4680
4681         rcu_read_lock();
4682         for_each_netdev_rcu(net, dev) {
4683                 idev = __in6_dev_get(dev);
4684                 if (idev) {
4685                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4686                         idev->cnf.disable_ipv6 = newf;
4687                         if (changed)
4688                                 dev_disable_change(idev);
4689                 }
4690         }
4691         rcu_read_unlock();
4692 }
4693
4694 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4695 {
4696         struct net *net;
4697         int old;
4698
4699         if (!rtnl_trylock())
4700                 return restart_syscall();
4701
4702         net = (struct net *)table->extra2;
4703         old = *p;
4704         *p = newf;
4705
4706         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4707                 rtnl_unlock();
4708                 return 0;
4709         }
4710
4711         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4712                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4713                 addrconf_disable_change(net, newf);
4714         } else if ((!newf) ^ (!old))
4715                 dev_disable_change((struct inet6_dev *)table->extra1);
4716
4717         rtnl_unlock();
4718         return 0;
4719 }
4720
4721 static
4722 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4723                             void __user *buffer, size_t *lenp, loff_t *ppos)
4724 {
4725         int *valp = ctl->data;
4726         int val = *valp;
4727         loff_t pos = *ppos;
4728         struct ctl_table lctl;
4729         int ret;
4730
4731         /*
4732          * ctl->data points to idev->cnf.disable_ipv6, we should
4733          * not modify it until we get the rtnl lock.
4734          */
4735         lctl = *ctl;
4736         lctl.data = &val;
4737
4738         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4739
4740         if (write)
4741                 ret = addrconf_disable_ipv6(ctl, valp, val);
4742         if (ret)
4743                 *ppos = pos;
4744         return ret;
4745 }
4746
4747 static struct addrconf_sysctl_table
4748 {
4749         struct ctl_table_header *sysctl_header;
4750         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4751 } addrconf_sysctl __read_mostly = {
4752         .sysctl_header = NULL,
4753         .addrconf_vars = {
4754                 {
4755                         .procname       = "forwarding",
4756                         .data           = &ipv6_devconf.forwarding,
4757                         .maxlen         = sizeof(int),
4758                         .mode           = 0644,
4759                         .proc_handler   = addrconf_sysctl_forward,
4760                 },
4761                 {
4762                         .procname       = "hop_limit",
4763                         .data           = &ipv6_devconf.hop_limit,
4764                         .maxlen         = sizeof(int),
4765                         .mode           = 0644,
4766                         .proc_handler   = proc_dointvec,
4767                 },
4768                 {
4769                         .procname       = "mtu",
4770                         .data           = &ipv6_devconf.mtu6,
4771                         .maxlen         = sizeof(int),
4772                         .mode           = 0644,
4773                         .proc_handler   = proc_dointvec,
4774                 },
4775                 {
4776                         .procname       = "accept_ra",
4777                         .data           = &ipv6_devconf.accept_ra,
4778                         .maxlen         = sizeof(int),
4779                         .mode           = 0644,
4780                         .proc_handler   = proc_dointvec,
4781                 },
4782                 {
4783                         .procname       = "accept_redirects",
4784                         .data           = &ipv6_devconf.accept_redirects,
4785                         .maxlen         = sizeof(int),
4786                         .mode           = 0644,
4787                         .proc_handler   = proc_dointvec,
4788                 },
4789                 {
4790                         .procname       = "autoconf",
4791                         .data           = &ipv6_devconf.autoconf,
4792                         .maxlen         = sizeof(int),
4793                         .mode           = 0644,
4794                         .proc_handler   = proc_dointvec,
4795                 },
4796                 {
4797                         .procname       = "dad_transmits",
4798                         .data           = &ipv6_devconf.dad_transmits,
4799                         .maxlen         = sizeof(int),
4800                         .mode           = 0644,
4801                         .proc_handler   = proc_dointvec,
4802                 },
4803                 {
4804                         .procname       = "router_solicitations",
4805                         .data           = &ipv6_devconf.rtr_solicits,
4806                         .maxlen         = sizeof(int),
4807                         .mode           = 0644,
4808                         .proc_handler   = proc_dointvec,
4809                 },
4810                 {
4811                         .procname       = "router_solicitation_interval",
4812                         .data           = &ipv6_devconf.rtr_solicit_interval,
4813                         .maxlen         = sizeof(int),
4814                         .mode           = 0644,
4815                         .proc_handler   = proc_dointvec_jiffies,
4816                 },
4817                 {
4818                         .procname       = "router_solicitation_delay",
4819                         .data           = &ipv6_devconf.rtr_solicit_delay,
4820                         .maxlen         = sizeof(int),
4821                         .mode           = 0644,
4822                         .proc_handler   = proc_dointvec_jiffies,
4823                 },
4824                 {
4825                         .procname       = "force_mld_version",
4826                         .data           = &ipv6_devconf.force_mld_version,
4827                         .maxlen         = sizeof(int),
4828                         .mode           = 0644,
4829                         .proc_handler   = proc_dointvec,
4830                 },
4831 #ifdef CONFIG_IPV6_PRIVACY
4832                 {
4833                         .procname       = "use_tempaddr",
4834                         .data           = &ipv6_devconf.use_tempaddr,
4835                         .maxlen         = sizeof(int),
4836                         .mode           = 0644,
4837                         .proc_handler   = proc_dointvec,
4838                 },
4839                 {
4840                         .procname       = "temp_valid_lft",
4841                         .data           = &ipv6_devconf.temp_valid_lft,
4842                         .maxlen         = sizeof(int),
4843                         .mode           = 0644,
4844                         .proc_handler   = proc_dointvec,
4845                 },
4846                 {
4847                         .procname       = "temp_prefered_lft",
4848                         .data           = &ipv6_devconf.temp_prefered_lft,
4849                         .maxlen         = sizeof(int),
4850                         .mode           = 0644,
4851                         .proc_handler   = proc_dointvec,
4852                 },
4853                 {
4854                         .procname       = "regen_max_retry",
4855                         .data           = &ipv6_devconf.regen_max_retry,
4856                         .maxlen         = sizeof(int),
4857                         .mode           = 0644,
4858                         .proc_handler   = proc_dointvec,
4859                 },
4860                 {
4861                         .procname       = "max_desync_factor",
4862                         .data           = &ipv6_devconf.max_desync_factor,
4863                         .maxlen         = sizeof(int),
4864                         .mode           = 0644,
4865                         .proc_handler   = proc_dointvec,
4866                 },
4867 #endif
4868                 {
4869                         .procname       = "max_addresses",
4870                         .data           = &ipv6_devconf.max_addresses,
4871                         .maxlen         = sizeof(int),
4872                         .mode           = 0644,
4873                         .proc_handler   = proc_dointvec,
4874                 },
4875                 {
4876                         .procname       = "accept_ra_defrtr",
4877                         .data           = &ipv6_devconf.accept_ra_defrtr,
4878                         .maxlen         = sizeof(int),
4879                         .mode           = 0644,
4880                         .proc_handler   = proc_dointvec,
4881                 },
4882                 {
4883                         .procname       = "accept_ra_pinfo",
4884                         .data           = &ipv6_devconf.accept_ra_pinfo,
4885                         .maxlen         = sizeof(int),
4886                         .mode           = 0644,
4887                         .proc_handler   = proc_dointvec,
4888                 },
4889 #ifdef CONFIG_IPV6_ROUTER_PREF
4890                 {
4891                         .procname       = "accept_ra_rtr_pref",
4892                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4893                         .maxlen         = sizeof(int),
4894                         .mode           = 0644,
4895                         .proc_handler   = proc_dointvec,
4896                 },
4897                 {
4898                         .procname       = "router_probe_interval",
4899                         .data           = &ipv6_devconf.rtr_probe_interval,
4900                         .maxlen         = sizeof(int),
4901                         .mode           = 0644,
4902                         .proc_handler   = proc_dointvec_jiffies,
4903                 },
4904 #ifdef CONFIG_IPV6_ROUTE_INFO
4905                 {
4906                         .procname       = "accept_ra_rt_info_max_plen",
4907                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4908                         .maxlen         = sizeof(int),
4909                         .mode           = 0644,
4910                         .proc_handler   = proc_dointvec,
4911                 },
4912 #endif
4913 #endif
4914                 {
4915                         .procname       = "proxy_ndp",
4916                         .data           = &ipv6_devconf.proxy_ndp,
4917                         .maxlen         = sizeof(int),
4918                         .mode           = 0644,
4919                         .proc_handler   = proc_dointvec,
4920                 },
4921                 {
4922                         .procname       = "accept_source_route",
4923                         .data           = &ipv6_devconf.accept_source_route,
4924                         .maxlen         = sizeof(int),
4925                         .mode           = 0644,
4926                         .proc_handler   = proc_dointvec,
4927                 },
4928 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4929                 {
4930                         .procname       = "optimistic_dad",
4931                         .data           = &ipv6_devconf.optimistic_dad,
4932                         .maxlen         = sizeof(int),
4933                         .mode           = 0644,
4934                         .proc_handler   = proc_dointvec,
4935
4936                 },
4937 #endif
4938 #ifdef CONFIG_IPV6_MROUTE
4939                 {
4940                         .procname       = "mc_forwarding",
4941                         .data           = &ipv6_devconf.mc_forwarding,
4942                         .maxlen         = sizeof(int),
4943                         .mode           = 0444,
4944                         .proc_handler   = proc_dointvec,
4945                 },
4946 #endif
4947                 {
4948                         .procname       = "disable_ipv6",
4949                         .data           = &ipv6_devconf.disable_ipv6,
4950                         .maxlen         = sizeof(int),
4951                         .mode           = 0644,
4952                         .proc_handler   = addrconf_sysctl_disable,
4953                 },
4954                 {
4955                         .procname       = "accept_dad",
4956                         .data           = &ipv6_devconf.accept_dad,
4957                         .maxlen         = sizeof(int),
4958                         .mode           = 0644,
4959                         .proc_handler   = proc_dointvec,
4960                 },
4961                 {
4962                         .procname       = "force_tllao",
4963                         .data           = &ipv6_devconf.force_tllao,
4964                         .maxlen         = sizeof(int),
4965                         .mode           = 0644,
4966                         .proc_handler   = proc_dointvec
4967                 },
4968                 {
4969                         .procname       = "ndisc_notify",
4970                         .data           = &ipv6_devconf.ndisc_notify,
4971                         .maxlen         = sizeof(int),
4972                         .mode           = 0644,
4973                         .proc_handler   = proc_dointvec
4974                 },
4975                 {
4976                         /* sentinel */
4977                 }
4978         },
4979 };
4980
4981 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4982                 struct inet6_dev *idev, struct ipv6_devconf *p)
4983 {
4984         int i;
4985         struct addrconf_sysctl_table *t;
4986         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4987
4988         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4989         if (t == NULL)
4990                 goto out;
4991
4992         for (i = 0; t->addrconf_vars[i].data; i++) {
4993                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4994                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4995                 t->addrconf_vars[i].extra2 = net;
4996         }
4997
4998         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4999
5000         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5001         if (t->sysctl_header == NULL)
5002                 goto free;
5003
5004         p->sysctl = t;
5005         return 0;
5006
5007 free:
5008         kfree(t);
5009 out:
5010         return -ENOBUFS;
5011 }
5012
5013 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5014 {
5015         struct addrconf_sysctl_table *t;
5016
5017         if (p->sysctl == NULL)
5018                 return;
5019
5020         t = p->sysctl;
5021         p->sysctl = NULL;
5022         unregister_net_sysctl_table(t->sysctl_header);
5023         kfree(t);
5024 }
5025
5026 static void addrconf_sysctl_register(struct inet6_dev *idev)
5027 {
5028         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5029                               &ndisc_ifinfo_sysctl_change);
5030         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5031                                         idev, &idev->cnf);
5032 }
5033
5034 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5035 {
5036         __addrconf_sysctl_unregister(&idev->cnf);
5037         neigh_sysctl_unregister(idev->nd_parms);
5038 }
5039
5040
5041 #endif
5042
5043 static int __net_init addrconf_init_net(struct net *net)
5044 {
5045         int err = -ENOMEM;
5046         struct ipv6_devconf *all, *dflt;
5047
5048         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5049         if (all == NULL)
5050                 goto err_alloc_all;
5051
5052         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5053         if (dflt == NULL)
5054                 goto err_alloc_dflt;
5055
5056         /* these will be inherited by all namespaces */
5057         dflt->autoconf = ipv6_defaults.autoconf;
5058         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5059
5060         net->ipv6.devconf_all = all;
5061         net->ipv6.devconf_dflt = dflt;
5062
5063 #ifdef CONFIG_SYSCTL
5064         err = __addrconf_sysctl_register(net, "all", NULL, all);
5065         if (err < 0)
5066                 goto err_reg_all;
5067
5068         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5069         if (err < 0)
5070                 goto err_reg_dflt;
5071 #endif
5072         return 0;
5073
5074 #ifdef CONFIG_SYSCTL
5075 err_reg_dflt:
5076         __addrconf_sysctl_unregister(all);
5077 err_reg_all:
5078         kfree(dflt);
5079 #endif
5080 err_alloc_dflt:
5081         kfree(all);
5082 err_alloc_all:
5083         return err;
5084 }
5085
5086 static void __net_exit addrconf_exit_net(struct net *net)
5087 {
5088 #ifdef CONFIG_SYSCTL
5089         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5090         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5091 #endif
5092         if (!net_eq(net, &init_net)) {
5093                 kfree(net->ipv6.devconf_dflt);
5094                 kfree(net->ipv6.devconf_all);
5095         }
5096 }
5097
5098 static struct pernet_operations addrconf_ops = {
5099         .init = addrconf_init_net,
5100         .exit = addrconf_exit_net,
5101 };
5102
5103 static struct rtnl_af_ops inet6_ops = {
5104         .family           = AF_INET6,
5105         .fill_link_af     = inet6_fill_link_af,
5106         .get_link_af_size = inet6_get_link_af_size,
5107         .set_link_af      = inet6_set_link_af,
5108 };
5109
5110 /*
5111  *      Init / cleanup code
5112  */
5113
5114 int __init addrconf_init(void)
5115 {
5116         int i, err;
5117
5118         err = ipv6_addr_label_init();
5119         if (err < 0) {
5120                 pr_crit("%s: cannot initialize default policy table: %d\n",
5121                         __func__, err);
5122                 goto out;
5123         }
5124
5125         err = register_pernet_subsys(&addrconf_ops);
5126         if (err < 0)
5127                 goto out_addrlabel;
5128
5129         /* The addrconf netdev notifier requires that loopback_dev
5130          * has it's ipv6 private information allocated and setup
5131          * before it can bring up and give link-local addresses
5132          * to other devices which are up.
5133          *
5134          * Unfortunately, loopback_dev is not necessarily the first
5135          * entry in the global dev_base list of net devices.  In fact,
5136          * it is likely to be the very last entry on that list.
5137          * So this causes the notifier registry below to try and
5138          * give link-local addresses to all devices besides loopback_dev
5139          * first, then loopback_dev, which cases all the non-loopback_dev
5140          * devices to fail to get a link-local address.
5141          *
5142          * So, as a temporary fix, allocate the ipv6 structure for
5143          * loopback_dev first by hand.
5144          * Longer term, all of the dependencies ipv6 has upon the loopback
5145          * device and it being up should be removed.
5146          */
5147         rtnl_lock();
5148         if (!ipv6_add_dev(init_net.loopback_dev))
5149                 err = -ENOMEM;
5150         rtnl_unlock();
5151         if (err)
5152                 goto errlo;
5153
5154         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5155                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5156
5157         register_netdevice_notifier(&ipv6_dev_notf);
5158
5159         addrconf_verify(0);
5160
5161         err = rtnl_af_register(&inet6_ops);
5162         if (err < 0)
5163                 goto errout_af;
5164
5165         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5166                               NULL);
5167         if (err < 0)
5168                 goto errout;
5169
5170         /* Only the first call to __rtnl_register can fail */
5171         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5172         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5173         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5174                         inet6_dump_ifaddr, NULL);
5175         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5176                         inet6_dump_ifmcaddr, NULL);
5177         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5178                         inet6_dump_ifacaddr, NULL);
5179         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5180                         inet6_netconf_dump_devconf, NULL);
5181
5182         ipv6_addr_label_rtnl_register();
5183
5184         return 0;
5185 errout:
5186         rtnl_af_unregister(&inet6_ops);
5187 errout_af:
5188         unregister_netdevice_notifier(&ipv6_dev_notf);
5189 errlo:
5190         unregister_pernet_subsys(&addrconf_ops);
5191 out_addrlabel:
5192         ipv6_addr_label_cleanup();
5193 out:
5194         return err;
5195 }
5196
5197 void addrconf_cleanup(void)
5198 {
5199         struct net_device *dev;
5200         int i;
5201
5202         unregister_netdevice_notifier(&ipv6_dev_notf);
5203         unregister_pernet_subsys(&addrconf_ops);
5204         ipv6_addr_label_cleanup();
5205
5206         rtnl_lock();
5207
5208         __rtnl_af_unregister(&inet6_ops);
5209
5210         /* clean dev list */
5211         for_each_netdev(&init_net, dev) {
5212                 if (__in6_dev_get(dev) == NULL)
5213                         continue;
5214                 addrconf_ifdown(dev, 1);
5215         }
5216         addrconf_ifdown(init_net.loopback_dev, 2);
5217
5218         /*
5219          *      Check hash table.
5220          */
5221         spin_lock_bh(&addrconf_hash_lock);
5222         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5223                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5224         spin_unlock_bh(&addrconf_hash_lock);
5225
5226         del_timer(&addr_chk_timer);
5227         rtnl_unlock();
5228 }