net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/inet.h>
  50 #include <linux/in6.h>
  51 #include <linux/netdevice.h>
  52 #include <linux/if_addr.h>
  53 #include <linux/if_arp.h>
  54 #include <linux/if_arcnet.h>
  55 #include <linux/if_infiniband.h>
  56 #include <linux/route.h>
  57 #include <linux/inetdevice.h>
  58 #include <linux/init.h>
  59 #include <linux/slab.h>
  60 #ifdef CONFIG_SYSCTL
  61 #include <linux/sysctl.h>
  62 #endif
  63 #include <linux/capability.h>
  64 #include <linux/delay.h>
  65 #include <linux/notifier.h>
  66 #include <linux/string.h>
  67 #include <linux/hash.h>
  68
  69 #include <net/net_namespace.h>
  70 #include <net/sock.h>
  71 #include <net/snmp.h>
  72
  73 #include <net/af_ieee802154.h>
  74 #include <net/firewire.h>
  75 #include <net/ipv6.h>
  76 #include <net/protocol.h>
  77 #include <net/ndisc.h>
  78 #include <net/ip6_route.h>
  79 #include <net/addrconf.h>
  80 #include <net/tcp.h>
  81 #include <net/ip.h>
  82 #include <net/netlink.h>
  83 #include <net/pkt_sched.h>
  84 #include <linux/if_tunnel.h>
  85 #include <linux/rtnetlink.h>
  86 #include <linux/netconf.h>
  87 #include <linux/random.h>
  88 #include <linux/uaccess.h>
  89 #include <asm/unaligned.h>
  90
  91 #include <linux/proc_fs.h>
  92 #include <linux/seq_file.h>
  93 #include <linux/export.h>
  94
  95 /* Set to 3 to get tracing... */
  96 #define ACONF_DEBUG 2
  97
  98 #if ACONF_DEBUG >= 3
  99 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 100 #else
 101 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 102 #endif
 103
 104 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 105
 106 #define IPV6_MAX_STRLEN \
 107         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 108
 109 static inline u32 cstamp_delta(unsigned long cstamp)
 110 {
 111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112 }
 113
 114 #ifdef CONFIG_SYSCTL
 115 static int addrconf_sysctl_register(struct inet6_dev *idev);
 116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117 #else
 118 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 119 {
 120         return 0;
 121 }
 122
 123 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 124 {
 125 }
 126 #endif
 127
 128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130 static void ipv6_regen_rndid(unsigned long data);
 131
 132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 133 static int ipv6_count_addresses(struct inet6_dev *idev);
 134 static int ipv6_generate_stable_address(struct in6_addr *addr,
 135                                         u8 dad_count,
 136                                         const struct inet6_dev *idev);
 137
 138 /*
 139  *      Configured unicast address hash table
 140  */
 141 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 142 static DEFINE_SPINLOCK(addrconf_hash_lock);
 143
 144 static void addrconf_verify(void);
 145 static void addrconf_verify_rtnl(void);
 146 static void addrconf_verify_work(struct work_struct *);
 147
 148 static struct workqueue_struct *addrconf_wq;
 149 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 150
 151 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 152 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 153
 154 static void addrconf_type_change(struct net_device *dev,
 155                                  unsigned long event);
 156 static int addrconf_ifdown(struct net_device *dev, int how);
 157
 158 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 159                                                   int plen,
 160                                                   const struct net_device *dev,
 161                                                   u32 flags, u32 noflags);
 162
 163 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 164 static void addrconf_dad_work(struct work_struct *w);
 165 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 166 static void addrconf_dad_run(struct inet6_dev *idev);
 167 static void addrconf_rs_timer(unsigned long data);
 168 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 169 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 170
 171 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 172                                 struct prefix_info *pinfo);
 173 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 174                                struct net_device *dev);
 175
 176 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 177         .forwarding             = 0,
 178         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 179         .mtu6                   = IPV6_MIN_MTU,
 180         .accept_ra              = 1,
 181         .accept_redirects       = 1,
 182         .autoconf               = 1,
 183         .force_mld_version      = 0,
 184         .mldv1_unsolicited_report_interval = 10 * HZ,
 185         .mldv2_unsolicited_report_interval = HZ,
 186         .dad_transmits          = 1,
 187         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 188         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 189         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 190         .use_tempaddr           = 0,
 191         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 192         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 193         .regen_max_retry        = REGEN_MAX_RETRY,
 194         .max_desync_factor      = MAX_DESYNC_FACTOR,
 195         .max_addresses          = IPV6_MAX_ADDRESSES,
 196         .accept_ra_defrtr       = 1,
 197         .accept_ra_from_local   = 0,
 198         .accept_ra_pinfo        = 1,
 199 #ifdef CONFIG_IPV6_ROUTER_PREF
 200         .accept_ra_rtr_pref     = 1,
 201         .rtr_probe_interval     = 60 * HZ,
 202 #ifdef CONFIG_IPV6_ROUTE_INFO
 203         .accept_ra_rt_info_max_plen = 0,
 204 #endif
 205 #endif
 206         .proxy_ndp              = 0,
 207         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 208         .disable_ipv6           = 0,
 209         .accept_dad             = 1,
 210         .suppress_frag_ndisc    = 1,
 211         .accept_ra_mtu          = 1,
 212         .stable_secret          = {
 213                 .initialized = false,
 214         }
 215 };
 216
 217 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 218         .forwarding             = 0,
 219         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 220         .mtu6                   = IPV6_MIN_MTU,
 221         .accept_ra              = 1,
 222         .accept_redirects       = 1,
 223         .autoconf               = 1,
 224         .force_mld_version      = 0,
 225         .mldv1_unsolicited_report_interval = 10 * HZ,
 226         .mldv2_unsolicited_report_interval = HZ,
 227         .dad_transmits          = 1,
 228         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 229         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 230         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 231         .use_tempaddr           = 0,
 232         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 233         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 234         .regen_max_retry        = REGEN_MAX_RETRY,
 235         .max_desync_factor      = MAX_DESYNC_FACTOR,
 236         .max_addresses          = IPV6_MAX_ADDRESSES,
 237         .accept_ra_defrtr       = 1,
 238         .accept_ra_from_local   = 0,
 239         .accept_ra_pinfo        = 1,
 240 #ifdef CONFIG_IPV6_ROUTER_PREF
 241         .accept_ra_rtr_pref     = 1,
 242         .rtr_probe_interval     = 60 * HZ,
 243 #ifdef CONFIG_IPV6_ROUTE_INFO
 244         .accept_ra_rt_info_max_plen = 0,
 245 #endif
 246 #endif
 247         .proxy_ndp              = 0,
 248         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 249         .disable_ipv6           = 0,
 250         .accept_dad             = 1,
 251         .suppress_frag_ndisc    = 1,
 252         .accept_ra_mtu          = 1,
 253         .stable_secret          = {
 254                 .initialized = false,
 255         },
 256 };
 257
 258 /* Check if a valid qdisc is available */
 259 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 260 {
 261         return !qdisc_tx_is_noop(dev);
 262 }
 263
 264 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 265 {
 266         if (del_timer(&idev->rs_timer))
 267                 __in6_dev_put(idev);
 268 }
 269
 270 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 271 {
 272         if (cancel_delayed_work(&ifp->dad_work))
 273                 __in6_ifa_put(ifp);
 274 }
 275
 276 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 277                                   unsigned long when)
 278 {
 279         if (!timer_pending(&idev->rs_timer))
 280                 in6_dev_hold(idev);
 281         mod_timer(&idev->rs_timer, jiffies + when);
 282 }
 283
 284 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 285                                    unsigned long delay)
 286 {
 287         if (!delayed_work_pending(&ifp->dad_work))
 288                 in6_ifa_hold(ifp);
 289         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 290 }
 291
 292 static int snmp6_alloc_dev(struct inet6_dev *idev)
 293 {
 294         int i;
 295
 296         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 297         if (!idev->stats.ipv6)
 298                 goto err_ip;
 299
 300         for_each_possible_cpu(i) {
 301                 struct ipstats_mib *addrconf_stats;
 302                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 303                 u64_stats_init(&addrconf_stats->syncp);
 304         }
 305
 306
 307         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 308                                         GFP_KERNEL);
 309         if (!idev->stats.icmpv6dev)
 310                 goto err_icmp;
 311         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 312                                            GFP_KERNEL);
 313         if (!idev->stats.icmpv6msgdev)
 314                 goto err_icmpmsg;
 315
 316         return 0;
 317
 318 err_icmpmsg:
 319         kfree(idev->stats.icmpv6dev);
 320 err_icmp:
 321         free_percpu(idev->stats.ipv6);
 322 err_ip:
 323         return -ENOMEM;
 324 }
 325
 326 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 327 {
 328         struct inet6_dev *ndev;
 329         int err = -ENOMEM;
 330
 331         ASSERT_RTNL();
 332
 333         if (dev->mtu < IPV6_MIN_MTU)
 334                 return ERR_PTR(-EINVAL);
 335
 336         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 337         if (!ndev)
 338                 return ERR_PTR(err);
 339
 340         rwlock_init(&ndev->lock);
 341         ndev->dev = dev;
 342         INIT_LIST_HEAD(&ndev->addr_list);
 343         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 344                     (unsigned long)ndev);
 345         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 346         ndev->cnf.mtu6 = dev->mtu;
 347         ndev->cnf.sysctl = NULL;
 348         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 349         if (!ndev->nd_parms) {
 350                 kfree(ndev);
 351                 return ERR_PTR(err);
 352         }
 353         if (ndev->cnf.forwarding)
 354                 dev_disable_lro(dev);
 355         /* We refer to the device */
 356         dev_hold(dev);
 357
 358         if (snmp6_alloc_dev(ndev) < 0) {
 359                 ADBG(KERN_WARNING
 360                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 361                         __func__, dev->name);
 362                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 363                 dev_put(dev);
 364                 kfree(ndev);
 365                 return ERR_PTR(err);
 366         }
 367
 368         if (snmp6_register_dev(ndev) < 0) {
 369                 ADBG(KERN_WARNING
 370                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 371                         __func__, dev->name);
 372                 goto err_release;
 373         }
 374
 375         /* One reference from device.  We must do this before
 376          * we invoke __ipv6_regen_rndid().
 377          */
 378         in6_dev_hold(ndev);
 379
 380         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 381                 ndev->cnf.accept_dad = -1;
 382
 383 #if IS_ENABLED(CONFIG_IPV6_SIT)
 384         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 385                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 386                 ndev->cnf.rtr_solicits = 0;
 387         }
 388 #endif
 389
 390         INIT_LIST_HEAD(&ndev->tempaddr_list);
 391         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 392         if ((dev->flags&IFF_LOOPBACK) ||
 393             dev->type == ARPHRD_TUNNEL ||
 394             dev->type == ARPHRD_TUNNEL6 ||
 395             dev->type == ARPHRD_SIT ||
 396             dev->type == ARPHRD_NONE) {
 397                 ndev->cnf.use_tempaddr = -1;
 398         } else {
 399                 in6_dev_hold(ndev);
 400                 ipv6_regen_rndid((unsigned long) ndev);
 401         }
 402
 403         ndev->token = in6addr_any;
 404
 405         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 406                 ndev->if_flags |= IF_READY;
 407
 408         ipv6_mc_init_dev(ndev);
 409         ndev->tstamp = jiffies;
 410         err = addrconf_sysctl_register(ndev);
 411         if (err) {
 412                 ipv6_mc_destroy_dev(ndev);
 413                 del_timer(&ndev->regen_timer);
 414                 goto err_release;
 415         }
 416         /* protected by rtnl_lock */
 417         rcu_assign_pointer(dev->ip6_ptr, ndev);
 418
 419         /* Join interface-local all-node multicast group */
 420         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 421
 422         /* Join all-node multicast group */
 423         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 424
 425         /* Join all-router multicast group if forwarding is set */
 426         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 427                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 428
 429         return ndev;
 430
 431 err_release:
 432         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 433         ndev->dead = 1;
 434         in6_dev_finish_destroy(ndev);
 435         return ERR_PTR(err);
 436 }
 437
 438 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 439 {
 440         struct inet6_dev *idev;
 441
 442         ASSERT_RTNL();
 443
 444         idev = __in6_dev_get(dev);
 445         if (!idev) {
 446                 idev = ipv6_add_dev(dev);
 447                 if (IS_ERR(idev))
 448                         return NULL;
 449         }
 450
 451         if (dev->flags&IFF_UP)
 452                 ipv6_mc_up(idev);
 453         return idev;
 454 }
 455
 456 static int inet6_netconf_msgsize_devconf(int type)
 457 {
 458         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 459                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 460
 461         /* type -1 is used for ALL */
 462         if (type == -1 || type == NETCONFA_FORWARDING)
 463                 size += nla_total_size(4);
 464 #ifdef CONFIG_IPV6_MROUTE
 465         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 466                 size += nla_total_size(4);
 467 #endif
 468         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 469                 size += nla_total_size(4);
 470
 471         return size;
 472 }
 473
 474 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 475                                       struct ipv6_devconf *devconf, u32 portid,
 476                                       u32 seq, int event, unsigned int flags,
 477                                       int type)
 478 {
 479         struct nlmsghdr  *nlh;
 480         struct netconfmsg *ncm;
 481
 482         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 483                         flags);
 484         if (!nlh)
 485                 return -EMSGSIZE;
 486
 487         ncm = nlmsg_data(nlh);
 488         ncm->ncm_family = AF_INET6;
 489
 490         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 491                 goto nla_put_failure;
 492
 493         /* type -1 is used for ALL */
 494         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 495             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 496                 goto nla_put_failure;
 497 #ifdef CONFIG_IPV6_MROUTE
 498         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 499             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 500                         devconf->mc_forwarding) < 0)
 501                 goto nla_put_failure;
 502 #endif
 503         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 504             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 505                 goto nla_put_failure;
 506
 507         nlmsg_end(skb, nlh);
 508         return 0;
 509
 510 nla_put_failure:
 511         nlmsg_cancel(skb, nlh);
 512         return -EMSGSIZE;
 513 }
 514
 515 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 516                                   struct ipv6_devconf *devconf)
 517 {
 518         struct sk_buff *skb;
 519         int err = -ENOBUFS;
 520
 521         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 522         if (!skb)
 523                 goto errout;
 524
 525         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 526                                          RTM_NEWNETCONF, 0, type);
 527         if (err < 0) {
 528                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 529                 WARN_ON(err == -EMSGSIZE);
 530                 kfree_skb(skb);
 531                 goto errout;
 532         }
 533         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 534         return;
 535 errout:
 536         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 537 }
 538
 539 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 540         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 541         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 542         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 543 };
 544
 545 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 546                                      struct nlmsghdr *nlh)
 547 {
 548         struct net *net = sock_net(in_skb->sk);
 549         struct nlattr *tb[NETCONFA_MAX+1];
 550         struct netconfmsg *ncm;
 551         struct sk_buff *skb;
 552         struct ipv6_devconf *devconf;
 553         struct inet6_dev *in6_dev;
 554         struct net_device *dev;
 555         int ifindex;
 556         int err;
 557
 558         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 559                           devconf_ipv6_policy);
 560         if (err < 0)
 561                 goto errout;
 562
 563         err = EINVAL;
 564         if (!tb[NETCONFA_IFINDEX])
 565                 goto errout;
 566
 567         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 568         switch (ifindex) {
 569         case NETCONFA_IFINDEX_ALL:
 570                 devconf = net->ipv6.devconf_all;
 571                 break;
 572         case NETCONFA_IFINDEX_DEFAULT:
 573                 devconf = net->ipv6.devconf_dflt;
 574                 break;
 575         default:
 576                 dev = __dev_get_by_index(net, ifindex);
 577                 if (!dev)
 578                         goto errout;
 579                 in6_dev = __in6_dev_get(dev);
 580                 if (!in6_dev)
 581                         goto errout;
 582                 devconf = &in6_dev->cnf;
 583                 break;
 584         }
 585
 586         err = -ENOBUFS;
 587         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 588         if (!skb)
 589                 goto errout;
 590
 591         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 592                                          NETLINK_CB(in_skb).portid,
 593                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 594                                          -1);
 595         if (err < 0) {
 596                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 597                 WARN_ON(err == -EMSGSIZE);
 598                 kfree_skb(skb);
 599                 goto errout;
 600         }
 601         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 602 errout:
 603         return err;
 604 }
 605
 606 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 607                                       struct netlink_callback *cb)
 608 {
 609         struct net *net = sock_net(skb->sk);
 610         int h, s_h;
 611         int idx, s_idx;
 612         struct net_device *dev;
 613         struct inet6_dev *idev;
 614         struct hlist_head *head;
 615
 616         s_h = cb->args[0];
 617         s_idx = idx = cb->args[1];
 618
 619         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 620                 idx = 0;
 621                 head = &net->dev_index_head[h];
 622                 rcu_read_lock();
 623                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 624                           net->dev_base_seq;
 625                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 626                         if (idx < s_idx)
 627                                 goto cont;
 628                         idev = __in6_dev_get(dev);
 629                         if (!idev)
 630                                 goto cont;
 631
 632                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 633                                                        &idev->cnf,
 634                                                        NETLINK_CB(cb->skb).portid,
 635                                                        cb->nlh->nlmsg_seq,
 636                                                        RTM_NEWNETCONF,
 637                                                        NLM_F_MULTI,
 638                                                        -1) < 0) {
 639                                 rcu_read_unlock();
 640                                 goto done;
 641                         }
 642                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 643 cont:
 644                         idx++;
 645                 }
 646                 rcu_read_unlock();
 647         }
 648         if (h == NETDEV_HASHENTRIES) {
 649                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 650                                                net->ipv6.devconf_all,
 651                                                NETLINK_CB(cb->skb).portid,
 652                                                cb->nlh->nlmsg_seq,
 653                                                RTM_NEWNETCONF, NLM_F_MULTI,
 654                                                -1) < 0)
 655                         goto done;
 656                 else
 657                         h++;
 658         }
 659         if (h == NETDEV_HASHENTRIES + 1) {
 660                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 661                                                net->ipv6.devconf_dflt,
 662                                                NETLINK_CB(cb->skb).portid,
 663                                                cb->nlh->nlmsg_seq,
 664                                                RTM_NEWNETCONF, NLM_F_MULTI,
 665                                                -1) < 0)
 666                         goto done;
 667                 else
 668                         h++;
 669         }
 670 done:
 671         cb->args[0] = h;
 672         cb->args[1] = idx;
 673
 674         return skb->len;
 675 }
 676
 677 #ifdef CONFIG_SYSCTL
 678 static void dev_forward_change(struct inet6_dev *idev)
 679 {
 680         struct net_device *dev;
 681         struct inet6_ifaddr *ifa;
 682
 683         if (!idev)
 684                 return;
 685         dev = idev->dev;
 686         if (idev->cnf.forwarding)
 687                 dev_disable_lro(dev);
 688         if (dev->flags & IFF_MULTICAST) {
 689                 if (idev->cnf.forwarding) {
 690                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 691                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 692                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 693                 } else {
 694                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 695                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 696                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 697                 }
 698         }
 699
 700         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 701                 if (ifa->flags&IFA_F_TENTATIVE)
 702                         continue;
 703                 if (idev->cnf.forwarding)
 704                         addrconf_join_anycast(ifa);
 705                 else
 706                         addrconf_leave_anycast(ifa);
 707         }
 708         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 709                                      dev->ifindex, &idev->cnf);
 710 }
 711
 712
 713 static void addrconf_forward_change(struct net *net, __s32 newf)
 714 {
 715         struct net_device *dev;
 716         struct inet6_dev *idev;
 717
 718         for_each_netdev(net, dev) {
 719                 idev = __in6_dev_get(dev);
 720                 if (idev) {
 721                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 722                         idev->cnf.forwarding = newf;
 723                         if (changed)
 724                                 dev_forward_change(idev);
 725                 }
 726         }
 727 }
 728
 729 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 730 {
 731         struct net *net;
 732         int old;
 733
 734         if (!rtnl_trylock())
 735                 return restart_syscall();
 736
 737         net = (struct net *)table->extra2;
 738         old = *p;
 739         *p = newf;
 740
 741         if (p == &net->ipv6.devconf_dflt->forwarding) {
 742                 if ((!newf) ^ (!old))
 743                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 744                                                      NETCONFA_IFINDEX_DEFAULT,
 745                                                      net->ipv6.devconf_dflt);
 746                 rtnl_unlock();
 747                 return 0;
 748         }
 749
 750         if (p == &net->ipv6.devconf_all->forwarding) {
 751                 net->ipv6.devconf_dflt->forwarding = newf;
 752                 addrconf_forward_change(net, newf);
 753                 if ((!newf) ^ (!old))
 754                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 755                                                      NETCONFA_IFINDEX_ALL,
 756                                                      net->ipv6.devconf_all);
 757         } else if ((!newf) ^ (!old))
 758                 dev_forward_change((struct inet6_dev *)table->extra1);
 759         rtnl_unlock();
 760
 761         if (newf)
 762                 rt6_purge_dflt_routers(net);
 763         return 1;
 764 }
 765 #endif
 766
 767 /* Nobody refers to this ifaddr, destroy it */
 768 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 769 {
 770         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 771
 772 #ifdef NET_REFCNT_DEBUG
 773         pr_debug("%s\n", __func__);
 774 #endif
 775
 776         in6_dev_put(ifp->idev);
 777
 778         if (cancel_delayed_work(&ifp->dad_work))
 779                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 780                           ifp);
 781
 782         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 783                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 784                 return;
 785         }
 786         ip6_rt_put(ifp->rt);
 787
 788         kfree_rcu(ifp, rcu);
 789 }
 790
 791 static void
 792 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 793 {
 794         struct list_head *p;
 795         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 796
 797         /*
 798          * Each device address list is sorted in order of scope -
 799          * global before linklocal.
 800          */
 801         list_for_each(p, &idev->addr_list) {
 802                 struct inet6_ifaddr *ifa
 803                         = list_entry(p, struct inet6_ifaddr, if_list);
 804                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 805                         break;
 806         }
 807
 808         list_add_tail(&ifp->if_list, p);
 809 }
 810
 811 static u32 inet6_addr_hash(const struct in6_addr *addr)
 812 {
 813         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 814 }
 815
 816 /* On success it returns ifp with increased reference count */
 817
 818 static struct inet6_ifaddr *
 819 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 820               const struct in6_addr *peer_addr, int pfxlen,
 821               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 822 {
 823         struct inet6_ifaddr *ifa = NULL;
 824         struct rt6_info *rt;
 825         unsigned int hash;
 826         int err = 0;
 827         int addr_type = ipv6_addr_type(addr);
 828
 829         if (addr_type == IPV6_ADDR_ANY ||
 830             addr_type & IPV6_ADDR_MULTICAST ||
 831             (!(idev->dev->flags & IFF_LOOPBACK) &&
 832              addr_type & IPV6_ADDR_LOOPBACK))
 833                 return ERR_PTR(-EADDRNOTAVAIL);
 834
 835         rcu_read_lock_bh();
 836         if (idev->dead) {
 837                 err = -ENODEV;                  /*XXX*/
 838                 goto out2;
 839         }
 840
 841         if (idev->cnf.disable_ipv6) {
 842                 err = -EACCES;
 843                 goto out2;
 844         }
 845
 846         spin_lock(&addrconf_hash_lock);
 847
 848         /* Ignore adding duplicate addresses on an interface */
 849         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 850                 ADBG("ipv6_add_addr: already assigned\n");
 851                 err = -EEXIST;
 852                 goto out;
 853         }
 854
 855         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 856
 857         if (!ifa) {
 858                 ADBG("ipv6_add_addr: malloc failed\n");
 859                 err = -ENOBUFS;
 860                 goto out;
 861         }
 862
 863         rt = addrconf_dst_alloc(idev, addr, false);
 864         if (IS_ERR(rt)) {
 865                 err = PTR_ERR(rt);
 866                 goto out;
 867         }
 868
 869         neigh_parms_data_state_setall(idev->nd_parms);
 870
 871         ifa->addr = *addr;
 872         if (peer_addr)
 873                 ifa->peer_addr = *peer_addr;
 874
 875         spin_lock_init(&ifa->lock);
 876         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 877         INIT_HLIST_NODE(&ifa->addr_lst);
 878         ifa->scope = scope;
 879         ifa->prefix_len = pfxlen;
 880         ifa->flags = flags | IFA_F_TENTATIVE;
 881         ifa->valid_lft = valid_lft;
 882         ifa->prefered_lft = prefered_lft;
 883         ifa->cstamp = ifa->tstamp = jiffies;
 884         ifa->tokenized = false;
 885
 886         ifa->rt = rt;
 887
 888         ifa->idev = idev;
 889         in6_dev_hold(idev);
 890         /* For caller */
 891         in6_ifa_hold(ifa);
 892
 893         /* Add to big hash table */
 894         hash = inet6_addr_hash(addr);
 895
 896         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 897         spin_unlock(&addrconf_hash_lock);
 898
 899         write_lock(&idev->lock);
 900         /* Add to inet6_dev unicast addr list. */
 901         ipv6_link_dev_addr(idev, ifa);
 902
 903         if (ifa->flags&IFA_F_TEMPORARY) {
 904                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 905                 in6_ifa_hold(ifa);
 906         }
 907
 908         in6_ifa_hold(ifa);
 909         write_unlock(&idev->lock);
 910 out2:
 911         rcu_read_unlock_bh();
 912
 913         if (likely(err == 0))
 914                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 915         else {
 916                 kfree(ifa);
 917                 ifa = ERR_PTR(err);
 918         }
 919
 920         return ifa;
 921 out:
 922         spin_unlock(&addrconf_hash_lock);
 923         goto out2;
 924 }
 925
 926 enum cleanup_prefix_rt_t {
 927         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
 928         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
 929         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
 930 };
 931
 932 /*
 933  * Check, whether the prefix for ifp would still need a prefix route
 934  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
 935  * constants.
 936  *
 937  * 1) we don't purge prefix if address was not permanent.
 938  *    prefix is managed by its own lifetime.
 939  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
 940  * 3) if there are no addresses, delete prefix.
 941  * 4) if there are still other permanent address(es),
 942  *    corresponding prefix is still permanent.
 943  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
 944  *    don't purge the prefix, assume user space is managing it.
 945  * 6) otherwise, update prefix lifetime to the
 946  *    longest valid lifetime among the corresponding
 947  *    addresses on the device.
 948  *    Note: subsequent RA will update lifetime.
 949  **/
 950 static enum cleanup_prefix_rt_t
 951 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
 952 {
 953         struct inet6_ifaddr *ifa;
 954         struct inet6_dev *idev = ifp->idev;
 955         unsigned long lifetime;
 956         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
 957
 958         *expires = jiffies;
 959
 960         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 961                 if (ifa == ifp)
 962                         continue;
 963                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 964                                        ifp->prefix_len))
 965                         continue;
 966                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
 967                         return CLEANUP_PREFIX_RT_NOP;
 968
 969                 action = CLEANUP_PREFIX_RT_EXPIRE;
 970
 971                 spin_lock(&ifa->lock);
 972
 973                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 974                 /*
 975                  * Note: Because this address is
 976                  * not permanent, lifetime <
 977                  * LONG_MAX / HZ here.
 978                  */
 979                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
 980                         *expires = ifa->tstamp + lifetime * HZ;
 981                 spin_unlock(&ifa->lock);
 982         }
 983
 984         return action;
 985 }
 986
 987 static void
 988 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
 989 {
 990         struct rt6_info *rt;
 991
 992         rt = addrconf_get_prefix_route(&ifp->addr,
 993                                        ifp->prefix_len,
 994                                        ifp->idev->dev,
 995                                        0, RTF_GATEWAY | RTF_DEFAULT);
 996         if (rt) {
 997                 if (del_rt)
 998                         ip6_del_rt(rt);
 999                 else {
1000                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1001                                 rt6_set_expires(rt, expires);
1002                         ip6_rt_put(rt);
1003                 }
1004         }
1005 }
1006
1007
1008 /* This function wants to get referenced ifp and releases it before return */
1009
1010 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1011 {
1012         int state;
1013         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1014         unsigned long expires;
1015
1016         ASSERT_RTNL();
1017
1018         spin_lock_bh(&ifp->lock);
1019         state = ifp->state;
1020         ifp->state = INET6_IFADDR_STATE_DEAD;
1021         spin_unlock_bh(&ifp->lock);
1022
1023         if (state == INET6_IFADDR_STATE_DEAD)
1024                 goto out;
1025
1026         spin_lock_bh(&addrconf_hash_lock);
1027         hlist_del_init_rcu(&ifp->addr_lst);
1028         spin_unlock_bh(&addrconf_hash_lock);
1029
1030         write_lock_bh(&ifp->idev->lock);
1031
1032         if (ifp->flags&IFA_F_TEMPORARY) {
1033                 list_del(&ifp->tmp_list);
1034                 if (ifp->ifpub) {
1035                         in6_ifa_put(ifp->ifpub);
1036                         ifp->ifpub = NULL;
1037                 }
1038                 __in6_ifa_put(ifp);
1039         }
1040
1041         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1042                 action = check_cleanup_prefix_route(ifp, &expires);
1043
1044         list_del_init(&ifp->if_list);
1045         __in6_ifa_put(ifp);
1046
1047         write_unlock_bh(&ifp->idev->lock);
1048
1049         addrconf_del_dad_work(ifp);
1050
1051         ipv6_ifa_notify(RTM_DELADDR, ifp);
1052
1053         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1054
1055         if (action != CLEANUP_PREFIX_RT_NOP) {
1056                 cleanup_prefix_route(ifp, expires,
1057                         action == CLEANUP_PREFIX_RT_DEL);
1058         }
1059
1060         /* clean up prefsrc entries */
1061         rt6_remove_prefsrc(ifp);
1062 out:
1063         in6_ifa_put(ifp);
1064 }
1065
1066 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1067 {
1068         struct inet6_dev *idev = ifp->idev;
1069         struct in6_addr addr, *tmpaddr;
1070         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1071         unsigned long regen_advance;
1072         int tmp_plen;
1073         int ret = 0;
1074         u32 addr_flags;
1075         unsigned long now = jiffies;
1076
1077         write_lock_bh(&idev->lock);
1078         if (ift) {
1079                 spin_lock_bh(&ift->lock);
1080                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1081                 spin_unlock_bh(&ift->lock);
1082                 tmpaddr = &addr;
1083         } else {
1084                 tmpaddr = NULL;
1085         }
1086 retry:
1087         in6_dev_hold(idev);
1088         if (idev->cnf.use_tempaddr <= 0) {
1089                 write_unlock_bh(&idev->lock);
1090                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1091                 in6_dev_put(idev);
1092                 ret = -1;
1093                 goto out;
1094         }
1095         spin_lock_bh(&ifp->lock);
1096         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1097                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1098                 spin_unlock_bh(&ifp->lock);
1099                 write_unlock_bh(&idev->lock);
1100                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1101                         __func__);
1102                 in6_dev_put(idev);
1103                 ret = -1;
1104                 goto out;
1105         }
1106         in6_ifa_hold(ifp);
1107         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1108         __ipv6_try_regen_rndid(idev, tmpaddr);
1109         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1110         age = (now - ifp->tstamp) / HZ;
1111         tmp_valid_lft = min_t(__u32,
1112                               ifp->valid_lft,
1113                               idev->cnf.temp_valid_lft + age);
1114         tmp_prefered_lft = min_t(__u32,
1115                                  ifp->prefered_lft,
1116                                  idev->cnf.temp_prefered_lft + age -
1117                                  idev->cnf.max_desync_factor);
1118         tmp_plen = ifp->prefix_len;
1119         tmp_tstamp = ifp->tstamp;
1120         spin_unlock_bh(&ifp->lock);
1121
1122         regen_advance = idev->cnf.regen_max_retry *
1123                         idev->cnf.dad_transmits *
1124                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1125         write_unlock_bh(&idev->lock);
1126
1127         /* A temporary address is created only if this calculated Preferred
1128          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1129          * an implementation must not create a temporary address with a zero
1130          * Preferred Lifetime.
1131          * Use age calculation as in addrconf_verify to avoid unnecessary
1132          * temporary addresses being generated.
1133          */
1134         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1135         if (tmp_prefered_lft <= regen_advance + age) {
1136                 in6_ifa_put(ifp);
1137                 in6_dev_put(idev);
1138                 ret = -1;
1139                 goto out;
1140         }
1141
1142         addr_flags = IFA_F_TEMPORARY;
1143         /* set in addrconf_prefix_rcv() */
1144         if (ifp->flags & IFA_F_OPTIMISTIC)
1145                 addr_flags |= IFA_F_OPTIMISTIC;
1146
1147         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1148                             ipv6_addr_scope(&addr), addr_flags,
1149                             tmp_valid_lft, tmp_prefered_lft);
1150         if (IS_ERR(ift)) {
1151                 in6_ifa_put(ifp);
1152                 in6_dev_put(idev);
1153                 pr_info("%s: retry temporary address regeneration\n", __func__);
1154                 tmpaddr = &addr;
1155                 write_lock_bh(&idev->lock);
1156                 goto retry;
1157         }
1158
1159         spin_lock_bh(&ift->lock);
1160         ift->ifpub = ifp;
1161         ift->cstamp = now;
1162         ift->tstamp = tmp_tstamp;
1163         spin_unlock_bh(&ift->lock);
1164
1165         addrconf_dad_start(ift);
1166         in6_ifa_put(ift);
1167         in6_dev_put(idev);
1168 out:
1169         return ret;
1170 }
1171
1172 /*
1173  *      Choose an appropriate source address (RFC3484)
1174  */
1175 enum {
1176         IPV6_SADDR_RULE_INIT = 0,
1177         IPV6_SADDR_RULE_LOCAL,
1178         IPV6_SADDR_RULE_SCOPE,
1179         IPV6_SADDR_RULE_PREFERRED,
1180 #ifdef CONFIG_IPV6_MIP6
1181         IPV6_SADDR_RULE_HOA,
1182 #endif
1183         IPV6_SADDR_RULE_OIF,
1184         IPV6_SADDR_RULE_LABEL,
1185         IPV6_SADDR_RULE_PRIVACY,
1186         IPV6_SADDR_RULE_ORCHID,
1187         IPV6_SADDR_RULE_PREFIX,
1188 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1189         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1190 #endif
1191         IPV6_SADDR_RULE_MAX
1192 };
1193
1194 struct ipv6_saddr_score {
1195         int                     rule;
1196         int                     addr_type;
1197         struct inet6_ifaddr     *ifa;
1198         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1199         int                     scopedist;
1200         int                     matchlen;
1201 };
1202
1203 struct ipv6_saddr_dst {
1204         const struct in6_addr *addr;
1205         int ifindex;
1206         int scope;
1207         int label;
1208         unsigned int prefs;
1209 };
1210
1211 static inline int ipv6_saddr_preferred(int type)
1212 {
1213         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1214                 return 1;
1215         return 0;
1216 }
1217
1218 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1219 {
1220 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1221         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1222 #else
1223         return false;
1224 #endif
1225 }
1226
1227 static int ipv6_get_saddr_eval(struct net *net,
1228                                struct ipv6_saddr_score *score,
1229                                struct ipv6_saddr_dst *dst,
1230                                int i)
1231 {
1232         int ret;
1233
1234         if (i <= score->rule) {
1235                 switch (i) {
1236                 case IPV6_SADDR_RULE_SCOPE:
1237                         ret = score->scopedist;
1238                         break;
1239                 case IPV6_SADDR_RULE_PREFIX:
1240                         ret = score->matchlen;
1241                         break;
1242                 default:
1243                         ret = !!test_bit(i, score->scorebits);
1244                 }
1245                 goto out;
1246         }
1247
1248         switch (i) {
1249         case IPV6_SADDR_RULE_INIT:
1250                 /* Rule 0: remember if hiscore is not ready yet */
1251                 ret = !!score->ifa;
1252                 break;
1253         case IPV6_SADDR_RULE_LOCAL:
1254                 /* Rule 1: Prefer same address */
1255                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1256                 break;
1257         case IPV6_SADDR_RULE_SCOPE:
1258                 /* Rule 2: Prefer appropriate scope
1259                  *
1260                  *      ret
1261                  *       ^
1262                  *    -1 |  d 15
1263                  *    ---+--+-+---> scope
1264                  *       |
1265                  *       |             d is scope of the destination.
1266                  *  B-d  |  \
1267                  *       |   \      <- smaller scope is better if
1268                  *  B-15 |    \        if scope is enough for destination.
1269                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1270                  * d-C-1 | /
1271                  *       |/         <- greater is better
1272                  *   -C  /             if scope is not enough for destination.
1273                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1274                  *
1275                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1276                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1277                  * Assume B = 0 and we get C > 29.
1278                  */
1279                 ret = __ipv6_addr_src_scope(score->addr_type);
1280                 if (ret >= dst->scope)
1281                         ret = -ret;
1282                 else
1283                         ret -= 128;     /* 30 is enough */
1284                 score->scopedist = ret;
1285                 break;
1286         case IPV6_SADDR_RULE_PREFERRED:
1287             {
1288                 /* Rule 3: Avoid deprecated and optimistic addresses */
1289                 u8 avoid = IFA_F_DEPRECATED;
1290
1291                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1292                         avoid |= IFA_F_OPTIMISTIC;
1293                 ret = ipv6_saddr_preferred(score->addr_type) ||
1294                       !(score->ifa->flags & avoid);
1295                 break;
1296             }
1297 #ifdef CONFIG_IPV6_MIP6
1298         case IPV6_SADDR_RULE_HOA:
1299             {
1300                 /* Rule 4: Prefer home address */
1301                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1302                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1303                 break;
1304             }
1305 #endif
1306         case IPV6_SADDR_RULE_OIF:
1307                 /* Rule 5: Prefer outgoing interface */
1308                 ret = (!dst->ifindex ||
1309                        dst->ifindex == score->ifa->idev->dev->ifindex);
1310                 break;
1311         case IPV6_SADDR_RULE_LABEL:
1312                 /* Rule 6: Prefer matching label */
1313                 ret = ipv6_addr_label(net,
1314                                       &score->ifa->addr, score->addr_type,
1315                                       score->ifa->idev->dev->ifindex) == dst->label;
1316                 break;
1317         case IPV6_SADDR_RULE_PRIVACY:
1318             {
1319                 /* Rule 7: Prefer public address
1320                  * Note: prefer temporary address if use_tempaddr >= 2
1321                  */
1322                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1323                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1324                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1325                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1326                 break;
1327             }
1328         case IPV6_SADDR_RULE_ORCHID:
1329                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1330                  *          non-ORCHID vs non-ORCHID
1331                  */
1332                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1333                         ipv6_addr_orchid(dst->addr));
1334                 break;
1335         case IPV6_SADDR_RULE_PREFIX:
1336                 /* Rule 8: Use longest matching prefix */
1337                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1338                 if (ret > score->ifa->prefix_len)
1339                         ret = score->ifa->prefix_len;
1340                 score->matchlen = ret;
1341                 break;
1342 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1343         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1344                 /* Optimistic addresses still have lower precedence than other
1345                  * preferred addresses.
1346                  */
1347                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1348                 break;
1349 #endif
1350         default:
1351                 ret = 0;
1352         }
1353
1354         if (ret)
1355                 __set_bit(i, score->scorebits);
1356         score->rule = i;
1357 out:
1358         return ret;
1359 }
1360
1361 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1362                        const struct in6_addr *daddr, unsigned int prefs,
1363                        struct in6_addr *saddr)
1364 {
1365         struct ipv6_saddr_score scores[2],
1366                                 *score = &scores[0], *hiscore = &scores[1];
1367         struct ipv6_saddr_dst dst;
1368         struct net_device *dev;
1369         int dst_type;
1370
1371         dst_type = __ipv6_addr_type(daddr);
1372         dst.addr = daddr;
1373         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1374         dst.scope = __ipv6_addr_src_scope(dst_type);
1375         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1376         dst.prefs = prefs;
1377
1378         hiscore->rule = -1;
1379         hiscore->ifa = NULL;
1380
1381         rcu_read_lock();
1382
1383         for_each_netdev_rcu(net, dev) {
1384                 struct inet6_dev *idev;
1385
1386                 /* Candidate Source Address (section 4)
1387                  *  - multicast and link-local destination address,
1388                  *    the set of candidate source address MUST only
1389                  *    include addresses assigned to interfaces
1390                  *    belonging to the same link as the outgoing
1391                  *    interface.
1392                  * (- For site-local destination addresses, the
1393                  *    set of candidate source addresses MUST only
1394                  *    include addresses assigned to interfaces
1395                  *    belonging to the same site as the outgoing
1396                  *    interface.)
1397                  */
1398                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1399                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1400                     dst.ifindex && dev->ifindex != dst.ifindex)
1401                         continue;
1402
1403                 idev = __in6_dev_get(dev);
1404                 if (!idev)
1405                         continue;
1406
1407                 read_lock_bh(&idev->lock);
1408                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1409                         int i;
1410
1411                         /*
1412                          * - Tentative Address (RFC2462 section 5.4)
1413                          *  - A tentative address is not considered
1414                          *    "assigned to an interface" in the traditional
1415                          *    sense, unless it is also flagged as optimistic.
1416                          * - Candidate Source Address (section 4)
1417                          *  - In any case, anycast addresses, multicast
1418                          *    addresses, and the unspecified address MUST
1419                          *    NOT be included in a candidate set.
1420                          */
1421                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1422                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1423                                 continue;
1424
1425                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1426
1427                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1428                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1429                                 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1430                                                     dev->name);
1431                                 continue;
1432                         }
1433
1434                         score->rule = -1;
1435                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1436
1437                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1438                                 int minihiscore, miniscore;
1439
1440                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1441                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1442
1443                                 if (minihiscore > miniscore) {
1444                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1445                                             score->scopedist > 0) {
1446                                                 /*
1447                                                  * special case:
1448                                                  * each remaining entry
1449                                                  * has too small (not enough)
1450                                                  * scope, because ifa entries
1451                                                  * are sorted by their scope
1452                                                  * values.
1453                                                  */
1454                                                 goto try_nextdev;
1455                                         }
1456                                         break;
1457                                 } else if (minihiscore < miniscore) {
1458                                         if (hiscore->ifa)
1459                                                 in6_ifa_put(hiscore->ifa);
1460
1461                                         in6_ifa_hold(score->ifa);
1462
1463                                         swap(hiscore, score);
1464
1465                                         /* restore our iterator */
1466                                         score->ifa = hiscore->ifa;
1467
1468                                         break;
1469                                 }
1470                         }
1471                 }
1472 try_nextdev:
1473                 read_unlock_bh(&idev->lock);
1474         }
1475         rcu_read_unlock();
1476
1477         if (!hiscore->ifa)
1478                 return -EADDRNOTAVAIL;
1479
1480         *saddr = hiscore->ifa->addr;
1481         in6_ifa_put(hiscore->ifa);
1482         return 0;
1483 }
1484 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1485
1486 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1487                       u32 banned_flags)
1488 {
1489         struct inet6_ifaddr *ifp;
1490         int err = -EADDRNOTAVAIL;
1491
1492         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1493                 if (ifp->scope > IFA_LINK)
1494                         break;
1495                 if (ifp->scope == IFA_LINK &&
1496                     !(ifp->flags & banned_flags)) {
1497                         *addr = ifp->addr;
1498                         err = 0;
1499                         break;
1500                 }
1501         }
1502         return err;
1503 }
1504
1505 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1506                     u32 banned_flags)
1507 {
1508         struct inet6_dev *idev;
1509         int err = -EADDRNOTAVAIL;
1510
1511         rcu_read_lock();
1512         idev = __in6_dev_get(dev);
1513         if (idev) {
1514                 read_lock_bh(&idev->lock);
1515                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1516                 read_unlock_bh(&idev->lock);
1517         }
1518         rcu_read_unlock();
1519         return err;
1520 }
1521
1522 static int ipv6_count_addresses(struct inet6_dev *idev)
1523 {
1524         int cnt = 0;
1525         struct inet6_ifaddr *ifp;
1526
1527         read_lock_bh(&idev->lock);
1528         list_for_each_entry(ifp, &idev->addr_list, if_list)
1529                 cnt++;
1530         read_unlock_bh(&idev->lock);
1531         return cnt;
1532 }
1533
1534 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1535                   const struct net_device *dev, int strict)
1536 {
1537         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1538 }
1539 EXPORT_SYMBOL(ipv6_chk_addr);
1540
1541 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1542                             const struct net_device *dev, int strict,
1543                             u32 banned_flags)
1544 {
1545         struct inet6_ifaddr *ifp;
1546         unsigned int hash = inet6_addr_hash(addr);
1547         u32 ifp_flags;
1548
1549         rcu_read_lock_bh();
1550         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1551                 if (!net_eq(dev_net(ifp->idev->dev), net))
1552                         continue;
1553                 /* Decouple optimistic from tentative for evaluation here.
1554                  * Ban optimistic addresses explicitly, when required.
1555                  */
1556                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1557                             ? (ifp->flags&~IFA_F_TENTATIVE)
1558                             : ifp->flags;
1559                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1560                     !(ifp_flags&banned_flags) &&
1561                     (!dev || ifp->idev->dev == dev ||
1562                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1563                         rcu_read_unlock_bh();
1564                         return 1;
1565                 }
1566         }
1567
1568         rcu_read_unlock_bh();
1569         return 0;
1570 }
1571 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1572
1573 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1574                                struct net_device *dev)
1575 {
1576         unsigned int hash = inet6_addr_hash(addr);
1577         struct inet6_ifaddr *ifp;
1578
1579         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1580                 if (!net_eq(dev_net(ifp->idev->dev), net))
1581                         continue;
1582                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1583                         if (!dev || ifp->idev->dev == dev)
1584                                 return true;
1585                 }
1586         }
1587         return false;
1588 }
1589
1590 /* Compares an address/prefix_len with addresses on device @dev.
1591  * If one is found it returns true.
1592  */
1593 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1594         const unsigned int prefix_len, struct net_device *dev)
1595 {
1596         struct inet6_dev *idev;
1597         struct inet6_ifaddr *ifa;
1598         bool ret = false;
1599
1600         rcu_read_lock();
1601         idev = __in6_dev_get(dev);
1602         if (idev) {
1603                 read_lock_bh(&idev->lock);
1604                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1605                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1606                         if (ret)
1607                                 break;
1608                 }
1609                 read_unlock_bh(&idev->lock);
1610         }
1611         rcu_read_unlock();
1612
1613         return ret;
1614 }
1615 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1616
1617 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1618 {
1619         struct inet6_dev *idev;
1620         struct inet6_ifaddr *ifa;
1621         int     onlink;
1622
1623         onlink = 0;
1624         rcu_read_lock();
1625         idev = __in6_dev_get(dev);
1626         if (idev) {
1627                 read_lock_bh(&idev->lock);
1628                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1629                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1630                                                    ifa->prefix_len);
1631                         if (onlink)
1632                                 break;
1633                 }
1634                 read_unlock_bh(&idev->lock);
1635         }
1636         rcu_read_unlock();
1637         return onlink;
1638 }
1639 EXPORT_SYMBOL(ipv6_chk_prefix);
1640
1641 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1642                                      struct net_device *dev, int strict)
1643 {
1644         struct inet6_ifaddr *ifp, *result = NULL;
1645         unsigned int hash = inet6_addr_hash(addr);
1646
1647         rcu_read_lock_bh();
1648         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1649                 if (!net_eq(dev_net(ifp->idev->dev), net))
1650                         continue;
1651                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1652                         if (!dev || ifp->idev->dev == dev ||
1653                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1654                                 result = ifp;
1655                                 in6_ifa_hold(ifp);
1656                                 break;
1657                         }
1658                 }
1659         }
1660         rcu_read_unlock_bh();
1661
1662         return result;
1663 }
1664
1665 /* Gets referenced address, destroys ifaddr */
1666
1667 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1668 {
1669         if (ifp->flags&IFA_F_PERMANENT) {
1670                 spin_lock_bh(&ifp->lock);
1671                 addrconf_del_dad_work(ifp);
1672                 ifp->flags |= IFA_F_TENTATIVE;
1673                 if (dad_failed)
1674                         ifp->flags |= IFA_F_DADFAILED;
1675                 spin_unlock_bh(&ifp->lock);
1676                 if (dad_failed)
1677                         ipv6_ifa_notify(0, ifp);
1678                 in6_ifa_put(ifp);
1679         } else if (ifp->flags&IFA_F_TEMPORARY) {
1680                 struct inet6_ifaddr *ifpub;
1681                 spin_lock_bh(&ifp->lock);
1682                 ifpub = ifp->ifpub;
1683                 if (ifpub) {
1684                         in6_ifa_hold(ifpub);
1685                         spin_unlock_bh(&ifp->lock);
1686                         ipv6_create_tempaddr(ifpub, ifp);
1687                         in6_ifa_put(ifpub);
1688                 } else {
1689                         spin_unlock_bh(&ifp->lock);
1690                 }
1691                 ipv6_del_addr(ifp);
1692         } else {
1693                 ipv6_del_addr(ifp);
1694         }
1695 }
1696
1697 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1698 {
1699         int err = -ENOENT;
1700
1701         spin_lock_bh(&ifp->lock);
1702         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1703                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1704                 err = 0;
1705         }
1706         spin_unlock_bh(&ifp->lock);
1707
1708         return err;
1709 }
1710
1711 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1712 {
1713         struct in6_addr addr;
1714         struct inet6_dev *idev = ifp->idev;
1715         struct net *net = dev_net(ifp->idev->dev);
1716
1717         if (addrconf_dad_end(ifp)) {
1718                 in6_ifa_put(ifp);
1719                 return;
1720         }
1721
1722         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1723                              ifp->idev->dev->name, &ifp->addr);
1724
1725         spin_lock_bh(&ifp->lock);
1726
1727         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1728                 int scope = ifp->scope;
1729                 u32 flags = ifp->flags;
1730                 struct in6_addr new_addr;
1731                 struct inet6_ifaddr *ifp2;
1732                 u32 valid_lft, preferred_lft;
1733                 int pfxlen = ifp->prefix_len;
1734                 int retries = ifp->stable_privacy_retry + 1;
1735
1736                 if (retries > net->ipv6.sysctl.idgen_retries) {
1737                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1738                                              ifp->idev->dev->name);
1739                         goto errdad;
1740                 }
1741
1742                 new_addr = ifp->addr;
1743                 if (ipv6_generate_stable_address(&new_addr, retries,
1744                                                  idev))
1745                         goto errdad;
1746
1747                 valid_lft = ifp->valid_lft;
1748                 preferred_lft = ifp->prefered_lft;
1749
1750                 spin_unlock_bh(&ifp->lock);
1751
1752                 if (idev->cnf.max_addresses &&
1753                     ipv6_count_addresses(idev) >=
1754                     idev->cnf.max_addresses)
1755                         goto lock_errdad;
1756
1757                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1758                                      ifp->idev->dev->name);
1759
1760                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1761                                      scope, flags, valid_lft,
1762                                      preferred_lft);
1763                 if (IS_ERR(ifp2))
1764                         goto lock_errdad;
1765
1766                 spin_lock_bh(&ifp2->lock);
1767                 ifp2->stable_privacy_retry = retries;
1768                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1769                 spin_unlock_bh(&ifp2->lock);
1770
1771                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1772                 in6_ifa_put(ifp2);
1773 lock_errdad:
1774                 spin_lock_bh(&ifp->lock);
1775         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1776                 addr.s6_addr32[0] = htonl(0xfe800000);
1777                 addr.s6_addr32[1] = 0;
1778
1779                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1780                     ipv6_addr_equal(&ifp->addr, &addr)) {
1781                         /* DAD failed for link-local based on MAC address */
1782                         idev->cnf.disable_ipv6 = 1;
1783
1784                         pr_info("%s: IPv6 being disabled!\n",
1785                                 ifp->idev->dev->name);
1786                 }
1787         }
1788
1789 errdad:
1790         /* transition from _POSTDAD to _ERRDAD */
1791         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1792         spin_unlock_bh(&ifp->lock);
1793
1794         addrconf_mod_dad_work(ifp, 0);
1795 }
1796
1797 /* Join to solicited addr multicast group.
1798  * caller must hold RTNL */
1799 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1800 {
1801         struct in6_addr maddr;
1802
1803         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1804                 return;
1805
1806         addrconf_addr_solict_mult(addr, &maddr);
1807         ipv6_dev_mc_inc(dev, &maddr);
1808 }
1809
1810 /* caller must hold RTNL */
1811 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1812 {
1813         struct in6_addr maddr;
1814
1815         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1816                 return;
1817
1818         addrconf_addr_solict_mult(addr, &maddr);
1819         __ipv6_dev_mc_dec(idev, &maddr);
1820 }
1821
1822 /* caller must hold RTNL */
1823 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1824 {
1825         struct in6_addr addr;
1826
1827         if (ifp->prefix_len >= 127) /* RFC 6164 */
1828                 return;
1829         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1830         if (ipv6_addr_any(&addr))
1831                 return;
1832         __ipv6_dev_ac_inc(ifp->idev, &addr);
1833 }
1834
1835 /* caller must hold RTNL */
1836 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1837 {
1838         struct in6_addr addr;
1839
1840         if (ifp->prefix_len >= 127) /* RFC 6164 */
1841                 return;
1842         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1843         if (ipv6_addr_any(&addr))
1844                 return;
1845         __ipv6_dev_ac_dec(ifp->idev, &addr);
1846 }
1847
1848 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1849 {
1850         if (dev->addr_len != ETH_ALEN)
1851                 return -1;
1852         memcpy(eui, dev->dev_addr, 3);
1853         memcpy(eui + 5, dev->dev_addr + 3, 3);
1854
1855         /*
1856          * The zSeries OSA network cards can be shared among various
1857          * OS instances, but the OSA cards have only one MAC address.
1858          * This leads to duplicate address conflicts in conjunction
1859          * with IPv6 if more than one instance uses the same card.
1860          *
1861          * The driver for these cards can deliver a unique 16-bit
1862          * identifier for each instance sharing the same card.  It is
1863          * placed instead of 0xFFFE in the interface identifier.  The
1864          * "u" bit of the interface identifier is not inverted in this
1865          * case.  Hence the resulting interface identifier has local
1866          * scope according to RFC2373.
1867          */
1868         if (dev->dev_id) {
1869                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1870                 eui[4] = dev->dev_id & 0xFF;
1871         } else {
1872                 eui[3] = 0xFF;
1873                 eui[4] = 0xFE;
1874                 eui[0] ^= 2;
1875         }
1876         return 0;
1877 }
1878
1879 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1880 {
1881         if (dev->addr_len != IEEE802154_ADDR_LEN)
1882                 return -1;
1883         memcpy(eui, dev->dev_addr, 8);
1884         eui[0] ^= 2;
1885         return 0;
1886 }
1887
1888 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1889 {
1890         union fwnet_hwaddr *ha;
1891
1892         if (dev->addr_len != FWNET_ALEN)
1893                 return -1;
1894
1895         ha = (union fwnet_hwaddr *)dev->dev_addr;
1896
1897         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1898         eui[0] ^= 2;
1899         return 0;
1900 }
1901
1902 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1903 {
1904         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1905         if (dev->addr_len != ARCNET_ALEN)
1906                 return -1;
1907         memset(eui, 0, 7);
1908         eui[7] = *(u8 *)dev->dev_addr;
1909         return 0;
1910 }
1911
1912 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1913 {
1914         if (dev->addr_len != INFINIBAND_ALEN)
1915                 return -1;
1916         memcpy(eui, dev->dev_addr + 12, 8);
1917         eui[0] |= 2;
1918         return 0;
1919 }
1920
1921 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1922 {
1923         if (addr == 0)
1924                 return -1;
1925         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1926                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1927                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1928                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1929                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1930                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1931         eui[1] = 0;
1932         eui[2] = 0x5E;
1933         eui[3] = 0xFE;
1934         memcpy(eui + 4, &addr, 4);
1935         return 0;
1936 }
1937
1938 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1939 {
1940         if (dev->priv_flags & IFF_ISATAP)
1941                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1942         return -1;
1943 }
1944
1945 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1946 {
1947         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1948 }
1949
1950 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1951 {
1952         memcpy(eui, dev->perm_addr, 3);
1953         memcpy(eui + 5, dev->perm_addr + 3, 3);
1954         eui[3] = 0xFF;
1955         eui[4] = 0xFE;
1956         eui[0] ^= 2;
1957         return 0;
1958 }
1959
1960 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1961 {
1962         switch (dev->type) {
1963         case ARPHRD_ETHER:
1964         case ARPHRD_FDDI:
1965                 return addrconf_ifid_eui48(eui, dev);
1966         case ARPHRD_ARCNET:
1967                 return addrconf_ifid_arcnet(eui, dev);
1968         case ARPHRD_INFINIBAND:
1969                 return addrconf_ifid_infiniband(eui, dev);
1970         case ARPHRD_SIT:
1971                 return addrconf_ifid_sit(eui, dev);
1972         case ARPHRD_IPGRE:
1973                 return addrconf_ifid_gre(eui, dev);
1974         case ARPHRD_6LOWPAN:
1975         case ARPHRD_IEEE802154:
1976                 return addrconf_ifid_eui64(eui, dev);
1977         case ARPHRD_IEEE1394:
1978                 return addrconf_ifid_ieee1394(eui, dev);
1979         case ARPHRD_TUNNEL6:
1980                 return addrconf_ifid_ip6tnl(eui, dev);
1981         }
1982         return -1;
1983 }
1984
1985 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1986 {
1987         int err = -1;
1988         struct inet6_ifaddr *ifp;
1989
1990         read_lock_bh(&idev->lock);
1991         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1992                 if (ifp->scope > IFA_LINK)
1993                         break;
1994                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1995                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1996                         err = 0;
1997                         break;
1998                 }
1999         }
2000         read_unlock_bh(&idev->lock);
2001         return err;
2002 }
2003
2004 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2005 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2006 {
2007 regen:
2008         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2009         idev->rndid[0] &= ~0x02;
2010
2011         /*
2012          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2013          * check if generated address is not inappropriate
2014          *
2015          *  - Reserved subnet anycast (RFC 2526)
2016          *      11111101 11....11 1xxxxxxx
2017          *  - ISATAP (RFC4214) 6.1
2018          *      00-00-5E-FE-xx-xx-xx-xx
2019          *  - value 0
2020          *  - XXX: already assigned to an address on the device
2021          */
2022         if (idev->rndid[0] == 0xfd &&
2023             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2024             (idev->rndid[7]&0x80))
2025                 goto regen;
2026         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2027                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2028                         goto regen;
2029                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2030                         goto regen;
2031         }
2032 }
2033
2034 static void ipv6_regen_rndid(unsigned long data)
2035 {
2036         struct inet6_dev *idev = (struct inet6_dev *) data;
2037         unsigned long expires;
2038
2039         rcu_read_lock_bh();
2040         write_lock_bh(&idev->lock);
2041
2042         if (idev->dead)
2043                 goto out;
2044
2045         __ipv6_regen_rndid(idev);
2046
2047         expires = jiffies +
2048                 idev->cnf.temp_prefered_lft * HZ -
2049                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2050                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2051                 idev->cnf.max_desync_factor * HZ;
2052         if (time_before(expires, jiffies)) {
2053                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2054                         __func__, idev->dev->name);
2055                 goto out;
2056         }
2057
2058         if (!mod_timer(&idev->regen_timer, expires))
2059                 in6_dev_hold(idev);
2060
2061 out:
2062         write_unlock_bh(&idev->lock);
2063         rcu_read_unlock_bh();
2064         in6_dev_put(idev);
2065 }
2066
2067 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2068 {
2069         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2070                 __ipv6_regen_rndid(idev);
2071 }
2072
2073 /*
2074  *      Add prefix route.
2075  */
2076
2077 static void
2078 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2079                       unsigned long expires, u32 flags)
2080 {
2081         struct fib6_config cfg = {
2082                 .fc_table = RT6_TABLE_PREFIX,
2083                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2084                 .fc_ifindex = dev->ifindex,
2085                 .fc_expires = expires,
2086                 .fc_dst_len = plen,
2087                 .fc_flags = RTF_UP | flags,
2088                 .fc_nlinfo.nl_net = dev_net(dev),
2089                 .fc_protocol = RTPROT_KERNEL,
2090         };
2091
2092         cfg.fc_dst = *pfx;
2093
2094         /* Prevent useless cloning on PtP SIT.
2095            This thing is done here expecting that the whole
2096            class of non-broadcast devices need not cloning.
2097          */
2098 #if IS_ENABLED(CONFIG_IPV6_SIT)
2099         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2100                 cfg.fc_flags |= RTF_NONEXTHOP;
2101 #endif
2102
2103         ip6_route_add(&cfg);
2104 }
2105
2106
2107 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2108                                                   int plen,
2109                                                   const struct net_device *dev,
2110                                                   u32 flags, u32 noflags)
2111 {
2112         struct fib6_node *fn;
2113         struct rt6_info *rt = NULL;
2114         struct fib6_table *table;
2115
2116         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2117         if (!table)
2118                 return NULL;
2119
2120         read_lock_bh(&table->tb6_lock);
2121         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2122         if (!fn)
2123                 goto out;
2124
2125         noflags |= RTF_CACHE;
2126         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2127                 if (rt->dst.dev->ifindex != dev->ifindex)
2128                         continue;
2129                 if ((rt->rt6i_flags & flags) != flags)
2130                         continue;
2131                 if ((rt->rt6i_flags & noflags) != 0)
2132                         continue;
2133                 dst_hold(&rt->dst);
2134                 break;
2135         }
2136 out:
2137         read_unlock_bh(&table->tb6_lock);
2138         return rt;
2139 }
2140
2141
2142 /* Create "default" multicast route to the interface */
2143
2144 static void addrconf_add_mroute(struct net_device *dev)
2145 {
2146         struct fib6_config cfg = {
2147                 .fc_table = RT6_TABLE_LOCAL,
2148                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2149                 .fc_ifindex = dev->ifindex,
2150                 .fc_dst_len = 8,
2151                 .fc_flags = RTF_UP,
2152                 .fc_nlinfo.nl_net = dev_net(dev),
2153         };
2154
2155         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2156
2157         ip6_route_add(&cfg);
2158 }
2159
2160 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2161 {
2162         struct inet6_dev *idev;
2163
2164         ASSERT_RTNL();
2165
2166         idev = ipv6_find_idev(dev);
2167         if (!idev)
2168                 return ERR_PTR(-ENOBUFS);
2169
2170         if (idev->cnf.disable_ipv6)
2171                 return ERR_PTR(-EACCES);
2172
2173         /* Add default multicast route */
2174         if (!(dev->flags & IFF_LOOPBACK))
2175                 addrconf_add_mroute(dev);
2176
2177         return idev;
2178 }
2179
2180 static void manage_tempaddrs(struct inet6_dev *idev,
2181                              struct inet6_ifaddr *ifp,
2182                              __u32 valid_lft, __u32 prefered_lft,
2183                              bool create, unsigned long now)
2184 {
2185         u32 flags;
2186         struct inet6_ifaddr *ift;
2187
2188         read_lock_bh(&idev->lock);
2189         /* update all temporary addresses in the list */
2190         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2191                 int age, max_valid, max_prefered;
2192
2193                 if (ifp != ift->ifpub)
2194                         continue;
2195
2196                 /* RFC 4941 section 3.3:
2197                  * If a received option will extend the lifetime of a public
2198                  * address, the lifetimes of temporary addresses should
2199                  * be extended, subject to the overall constraint that no
2200                  * temporary addresses should ever remain "valid" or "preferred"
2201                  * for a time longer than (TEMP_VALID_LIFETIME) or
2202                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2203                  */
2204                 age = (now - ift->cstamp) / HZ;
2205                 max_valid = idev->cnf.temp_valid_lft - age;
2206                 if (max_valid < 0)
2207                         max_valid = 0;
2208
2209                 max_prefered = idev->cnf.temp_prefered_lft -
2210                                idev->cnf.max_desync_factor - age;
2211                 if (max_prefered < 0)
2212                         max_prefered = 0;
2213
2214                 if (valid_lft > max_valid)
2215                         valid_lft = max_valid;
2216
2217                 if (prefered_lft > max_prefered)
2218                         prefered_lft = max_prefered;
2219
2220                 spin_lock(&ift->lock);
2221                 flags = ift->flags;
2222                 ift->valid_lft = valid_lft;
2223                 ift->prefered_lft = prefered_lft;
2224                 ift->tstamp = now;
2225                 if (prefered_lft > 0)
2226                         ift->flags &= ~IFA_F_DEPRECATED;
2227
2228                 spin_unlock(&ift->lock);
2229                 if (!(flags&IFA_F_TENTATIVE))
2230                         ipv6_ifa_notify(0, ift);
2231         }
2232
2233         if ((create || list_empty(&idev->tempaddr_list)) &&
2234             idev->cnf.use_tempaddr > 0) {
2235                 /* When a new public address is created as described
2236                  * in [ADDRCONF], also create a new temporary address.
2237                  * Also create a temporary address if it's enabled but
2238                  * no temporary address currently exists.
2239                  */
2240                 read_unlock_bh(&idev->lock);
2241                 ipv6_create_tempaddr(ifp, NULL);
2242         } else {
2243                 read_unlock_bh(&idev->lock);
2244         }
2245 }
2246
2247 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2248 {
2249         struct prefix_info *pinfo;
2250         __u32 valid_lft;
2251         __u32 prefered_lft;
2252         int addr_type;
2253         u32 addr_flags = 0;
2254         struct inet6_dev *in6_dev;
2255         struct net *net = dev_net(dev);
2256
2257         pinfo = (struct prefix_info *) opt;
2258
2259         if (len < sizeof(struct prefix_info)) {
2260                 ADBG("addrconf: prefix option too short\n");
2261                 return;
2262         }
2263
2264         /*
2265          *      Validation checks ([ADDRCONF], page 19)
2266          */
2267
2268         addr_type = ipv6_addr_type(&pinfo->prefix);
2269
2270         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2271                 return;
2272
2273         valid_lft = ntohl(pinfo->valid);
2274         prefered_lft = ntohl(pinfo->prefered);
2275
2276         if (prefered_lft > valid_lft) {
2277                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2278                 return;
2279         }
2280
2281         in6_dev = in6_dev_get(dev);
2282
2283         if (!in6_dev) {
2284                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2285                                     dev->name);
2286                 return;
2287         }
2288
2289         /*
2290          *      Two things going on here:
2291          *      1) Add routes for on-link prefixes
2292          *      2) Configure prefixes with the auto flag set
2293          */
2294
2295         if (pinfo->onlink) {
2296                 struct rt6_info *rt;
2297                 unsigned long rt_expires;
2298
2299                 /* Avoid arithmetic overflow. Really, we could
2300                  * save rt_expires in seconds, likely valid_lft,
2301                  * but it would require division in fib gc, that it
2302                  * not good.
2303                  */
2304                 if (HZ > USER_HZ)
2305                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2306                 else
2307                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2308
2309                 if (addrconf_finite_timeout(rt_expires))
2310                         rt_expires *= HZ;
2311
2312                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2313                                                pinfo->prefix_len,
2314                                                dev,
2315                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2316                                                RTF_GATEWAY | RTF_DEFAULT);
2317
2318                 if (rt) {
2319                         /* Autoconf prefix route */
2320                         if (valid_lft == 0) {
2321                                 ip6_del_rt(rt);
2322                                 rt = NULL;
2323                         } else if (addrconf_finite_timeout(rt_expires)) {
2324                                 /* not infinity */
2325                                 rt6_set_expires(rt, jiffies + rt_expires);
2326                         } else {
2327                                 rt6_clean_expires(rt);
2328                         }
2329                 } else if (valid_lft) {
2330                         clock_t expires = 0;
2331                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2332                         if (addrconf_finite_timeout(rt_expires)) {
2333                                 /* not infinity */
2334                                 flags |= RTF_EXPIRES;
2335                                 expires = jiffies_to_clock_t(rt_expires);
2336                         }
2337                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2338                                               dev, expires, flags);
2339                 }
2340                 ip6_rt_put(rt);
2341         }
2342
2343         /* Try to figure out our local address for this prefix */
2344
2345         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2346                 struct inet6_ifaddr *ifp;
2347                 struct in6_addr addr;
2348                 int create = 0, update_lft = 0;
2349                 bool tokenized = false;
2350
2351                 if (pinfo->prefix_len == 64) {
2352                         memcpy(&addr, &pinfo->prefix, 8);
2353
2354                         if (!ipv6_addr_any(&in6_dev->token)) {
2355                                 read_lock_bh(&in6_dev->lock);
2356                                 memcpy(addr.s6_addr + 8,
2357                                        in6_dev->token.s6_addr + 8, 8);
2358                                 read_unlock_bh(&in6_dev->lock);
2359                                 tokenized = true;
2360                         } else if (in6_dev->addr_gen_mode ==
2361                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2362                                    !ipv6_generate_stable_address(&addr, 0,
2363                                                                  in6_dev)) {
2364                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2365                                 goto ok;
2366                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2367                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2368                                 in6_dev_put(in6_dev);
2369                                 return;
2370                         }
2371                         goto ok;
2372                 }
2373                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2374                                     pinfo->prefix_len);
2375                 in6_dev_put(in6_dev);
2376                 return;
2377
2378 ok:
2379
2380                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2381
2382                 if (!ifp && valid_lft) {
2383                         int max_addresses = in6_dev->cnf.max_addresses;
2384
2385 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2386                         if (in6_dev->cnf.optimistic_dad &&
2387                             !net->ipv6.devconf_all->forwarding && sllao)
2388                                 addr_flags = IFA_F_OPTIMISTIC;
2389 #endif
2390
2391                         /* Do not allow to create too much of autoconfigured
2392                          * addresses; this would be too easy way to crash kernel.
2393                          */
2394                         if (!max_addresses ||
2395                             ipv6_count_addresses(in6_dev) < max_addresses)
2396                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2397                                                     pinfo->prefix_len,
2398                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2399                                                     addr_flags, valid_lft,
2400                                                     prefered_lft);
2401
2402                         if (IS_ERR_OR_NULL(ifp)) {
2403                                 in6_dev_put(in6_dev);
2404                                 return;
2405                         }
2406
2407                         update_lft = 0;
2408                         create = 1;
2409                         spin_lock_bh(&ifp->lock);
2410                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2411                         ifp->cstamp = jiffies;
2412                         ifp->tokenized = tokenized;
2413                         spin_unlock_bh(&ifp->lock);
2414                         addrconf_dad_start(ifp);
2415                 }
2416
2417                 if (ifp) {
2418                         u32 flags;
2419                         unsigned long now;
2420                         u32 stored_lft;
2421
2422                         /* update lifetime (RFC2462 5.5.3 e) */
2423                         spin_lock_bh(&ifp->lock);
2424                         now = jiffies;
2425                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2426                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2427                         else
2428                                 stored_lft = 0;
2429                         if (!update_lft && !create && stored_lft) {
2430                                 const u32 minimum_lft = min_t(u32,
2431                                         stored_lft, MIN_VALID_LIFETIME);
2432                                 valid_lft = max(valid_lft, minimum_lft);
2433
2434                                 /* RFC4862 Section 5.5.3e:
2435                                  * "Note that the preferred lifetime of the
2436                                  *  corresponding address is always reset to
2437                                  *  the Preferred Lifetime in the received
2438                                  *  Prefix Information option, regardless of
2439                                  *  whether the valid lifetime is also reset or
2440                                  *  ignored."
2441                                  *
2442                                  * So we should always update prefered_lft here.
2443                                  */
2444                                 update_lft = 1;
2445                         }
2446
2447                         if (update_lft) {
2448                                 ifp->valid_lft = valid_lft;
2449                                 ifp->prefered_lft = prefered_lft;
2450                                 ifp->tstamp = now;
2451                                 flags = ifp->flags;
2452                                 ifp->flags &= ~IFA_F_DEPRECATED;
2453                                 spin_unlock_bh(&ifp->lock);
2454
2455                                 if (!(flags&IFA_F_TENTATIVE))
2456                                         ipv6_ifa_notify(0, ifp);
2457                         } else
2458                                 spin_unlock_bh(&ifp->lock);
2459
2460                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2461                                          create, now);
2462
2463                         in6_ifa_put(ifp);
2464                         addrconf_verify();
2465                 }
2466         }
2467         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2468         in6_dev_put(in6_dev);
2469 }
2470
2471 /*
2472  *      Set destination address.
2473  *      Special case for SIT interfaces where we create a new "virtual"
2474  *      device.
2475  */
2476 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2477 {
2478         struct in6_ifreq ireq;
2479         struct net_device *dev;
2480         int err = -EINVAL;
2481
2482         rtnl_lock();
2483
2484         err = -EFAULT;
2485         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2486                 goto err_exit;
2487
2488         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2489
2490         err = -ENODEV;
2491         if (!dev)
2492                 goto err_exit;
2493
2494 #if IS_ENABLED(CONFIG_IPV6_SIT)
2495         if (dev->type == ARPHRD_SIT) {
2496                 const struct net_device_ops *ops = dev->netdev_ops;
2497                 struct ifreq ifr;
2498                 struct ip_tunnel_parm p;
2499
2500                 err = -EADDRNOTAVAIL;
2501                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2502                         goto err_exit;
2503
2504                 memset(&p, 0, sizeof(p));
2505                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2506                 p.iph.saddr = 0;
2507                 p.iph.version = 4;
2508                 p.iph.ihl = 5;
2509                 p.iph.protocol = IPPROTO_IPV6;
2510                 p.iph.ttl = 64;
2511                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2512
2513                 if (ops->ndo_do_ioctl) {
2514                         mm_segment_t oldfs = get_fs();
2515
2516                         set_fs(KERNEL_DS);
2517                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2518                         set_fs(oldfs);
2519                 } else
2520                         err = -EOPNOTSUPP;
2521
2522                 if (err == 0) {
2523                         err = -ENOBUFS;
2524                         dev = __dev_get_by_name(net, p.name);
2525                         if (!dev)
2526                                 goto err_exit;
2527                         err = dev_open(dev);
2528                 }
2529         }
2530 #endif
2531
2532 err_exit:
2533         rtnl_unlock();
2534         return err;
2535 }
2536
2537 static int ipv6_mc_config(struct sock *sk, bool join,
2538                           const struct in6_addr *addr, int ifindex)
2539 {
2540         int ret;
2541
2542         ASSERT_RTNL();
2543
2544         lock_sock(sk);
2545         if (join)
2546                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2547         else
2548                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2549         release_sock(sk);
2550
2551         return ret;
2552 }
2553
2554 /*
2555  *      Manual configuration of address on an interface
2556  */
2557 static int inet6_addr_add(struct net *net, int ifindex,
2558                           const struct in6_addr *pfx,
2559                           const struct in6_addr *peer_pfx,
2560                           unsigned int plen, __u32 ifa_flags,
2561                           __u32 prefered_lft, __u32 valid_lft)
2562 {
2563         struct inet6_ifaddr *ifp;
2564         struct inet6_dev *idev;
2565         struct net_device *dev;
2566         unsigned long timeout;
2567         clock_t expires;
2568         int scope;
2569         u32 flags;
2570
2571         ASSERT_RTNL();
2572
2573         if (plen > 128)
2574                 return -EINVAL;
2575
2576         /* check the lifetime */
2577         if (!valid_lft || prefered_lft > valid_lft)
2578                 return -EINVAL;
2579
2580         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2581                 return -EINVAL;
2582
2583         dev = __dev_get_by_index(net, ifindex);
2584         if (!dev)
2585                 return -ENODEV;
2586
2587         idev = addrconf_add_dev(dev);
2588         if (IS_ERR(idev))
2589                 return PTR_ERR(idev);
2590
2591         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2592                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2593                                          true, pfx, ifindex);
2594
2595                 if (ret < 0)
2596                         return ret;
2597         }
2598
2599         scope = ipv6_addr_scope(pfx);
2600
2601         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2602         if (addrconf_finite_timeout(timeout)) {
2603                 expires = jiffies_to_clock_t(timeout * HZ);
2604                 valid_lft = timeout;
2605                 flags = RTF_EXPIRES;
2606         } else {
2607                 expires = 0;
2608                 flags = 0;
2609                 ifa_flags |= IFA_F_PERMANENT;
2610         }
2611
2612         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2613         if (addrconf_finite_timeout(timeout)) {
2614                 if (timeout == 0)
2615                         ifa_flags |= IFA_F_DEPRECATED;
2616                 prefered_lft = timeout;
2617         }
2618
2619         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2620                             valid_lft, prefered_lft);
2621
2622         if (!IS_ERR(ifp)) {
2623                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2624                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2625                                               expires, flags);
2626                 }
2627
2628                 /*
2629                  * Note that section 3.1 of RFC 4429 indicates
2630                  * that the Optimistic flag should not be set for
2631                  * manually configured addresses
2632                  */
2633                 addrconf_dad_start(ifp);
2634                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2635                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2636                                          true, jiffies);
2637                 in6_ifa_put(ifp);
2638                 addrconf_verify_rtnl();
2639                 return 0;
2640         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2641                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2642                                false, pfx, ifindex);
2643         }
2644
2645         return PTR_ERR(ifp);
2646 }
2647
2648 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2649                           const struct in6_addr *pfx, unsigned int plen)
2650 {
2651         struct inet6_ifaddr *ifp;
2652         struct inet6_dev *idev;
2653         struct net_device *dev;
2654
2655         if (plen > 128)
2656                 return -EINVAL;
2657
2658         dev = __dev_get_by_index(net, ifindex);
2659         if (!dev)
2660                 return -ENODEV;
2661
2662         idev = __in6_dev_get(dev);
2663         if (!idev)
2664                 return -ENXIO;
2665
2666         read_lock_bh(&idev->lock);
2667         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2668                 if (ifp->prefix_len == plen &&
2669                     ipv6_addr_equal(pfx, &ifp->addr)) {
2670                         in6_ifa_hold(ifp);
2671                         read_unlock_bh(&idev->lock);
2672
2673                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2674                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2675                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2676                                                  jiffies);
2677                         ipv6_del_addr(ifp);
2678                         addrconf_verify_rtnl();
2679                         if (ipv6_addr_is_multicast(pfx)) {
2680                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2681                                                false, pfx, dev->ifindex);
2682                         }
2683                         return 0;
2684                 }
2685         }
2686         read_unlock_bh(&idev->lock);
2687         return -EADDRNOTAVAIL;
2688 }
2689
2690
2691 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2692 {
2693         struct in6_ifreq ireq;
2694         int err;
2695
2696         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2697                 return -EPERM;
2698
2699         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2700                 return -EFAULT;
2701
2702         rtnl_lock();
2703         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2704                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2705                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2706         rtnl_unlock();
2707         return err;
2708 }
2709
2710 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2711 {
2712         struct in6_ifreq ireq;
2713         int err;
2714
2715         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2716                 return -EPERM;
2717
2718         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2719                 return -EFAULT;
2720
2721         rtnl_lock();
2722         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2723                              ireq.ifr6_prefixlen);
2724         rtnl_unlock();
2725         return err;
2726 }
2727
2728 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2729                      int plen, int scope)
2730 {
2731         struct inet6_ifaddr *ifp;
2732
2733         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2734                             scope, IFA_F_PERMANENT,
2735                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2736         if (!IS_ERR(ifp)) {
2737                 spin_lock_bh(&ifp->lock);
2738                 ifp->flags &= ~IFA_F_TENTATIVE;
2739                 spin_unlock_bh(&ifp->lock);
2740                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2741                 in6_ifa_put(ifp);
2742         }
2743 }
2744
2745 #if IS_ENABLED(CONFIG_IPV6_SIT)
2746 static void sit_add_v4_addrs(struct inet6_dev *idev)
2747 {
2748         struct in6_addr addr;
2749         struct net_device *dev;
2750         struct net *net = dev_net(idev->dev);
2751         int scope, plen;
2752         u32 pflags = 0;
2753
2754         ASSERT_RTNL();
2755
2756         memset(&addr, 0, sizeof(struct in6_addr));
2757         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2758
2759         if (idev->dev->flags&IFF_POINTOPOINT) {
2760                 addr.s6_addr32[0] = htonl(0xfe800000);
2761                 scope = IFA_LINK;
2762                 plen = 64;
2763         } else {
2764                 scope = IPV6_ADDR_COMPATv4;
2765                 plen = 96;
2766                 pflags |= RTF_NONEXTHOP;
2767         }
2768
2769         if (addr.s6_addr32[3]) {
2770                 add_addr(idev, &addr, plen, scope);
2771                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2772                 return;
2773         }
2774
2775         for_each_netdev(net, dev) {
2776                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2777                 if (in_dev && (dev->flags & IFF_UP)) {
2778                         struct in_ifaddr *ifa;
2779
2780                         int flag = scope;
2781
2782                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2783
2784                                 addr.s6_addr32[3] = ifa->ifa_local;
2785
2786                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2787                                         continue;
2788                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2789                                         if (idev->dev->flags&IFF_POINTOPOINT)
2790                                                 continue;
2791                                         flag |= IFA_HOST;
2792                                 }
2793
2794                                 add_addr(idev, &addr, plen, flag);
2795                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2796                                                       pflags);
2797                         }
2798                 }
2799         }
2800 }
2801 #endif
2802
2803 static void init_loopback(struct net_device *dev)
2804 {
2805         struct inet6_dev  *idev;
2806         struct net_device *sp_dev;
2807         struct inet6_ifaddr *sp_ifa;
2808         struct rt6_info *sp_rt;
2809
2810         /* ::1 */
2811
2812         ASSERT_RTNL();
2813
2814         idev = ipv6_find_idev(dev);
2815         if (!idev) {
2816                 pr_debug("%s: add_dev failed\n", __func__);
2817                 return;
2818         }
2819
2820         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2821
2822         /* Add routes to other interface's IPv6 addresses */
2823         for_each_netdev(dev_net(dev), sp_dev) {
2824                 if (!strcmp(sp_dev->name, dev->name))
2825                         continue;
2826
2827                 idev = __in6_dev_get(sp_dev);
2828                 if (!idev)
2829                         continue;
2830
2831                 read_lock_bh(&idev->lock);
2832                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2833
2834                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2835                                 continue;
2836
2837                         if (sp_ifa->rt) {
2838                                 /* This dst has been added to garbage list when
2839                                  * lo device down, release this obsolete dst and
2840                                  * reallocate a new router for ifa.
2841                                  */
2842                                 if (sp_ifa->rt->dst.obsolete > 0) {
2843                                         ip6_rt_put(sp_ifa->rt);
2844                                         sp_ifa->rt = NULL;
2845                                 } else {
2846                                         continue;
2847                                 }
2848                         }
2849
2850                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2851
2852                         /* Failure cases are ignored */
2853                         if (!IS_ERR(sp_rt)) {
2854                                 sp_ifa->rt = sp_rt;
2855                                 ip6_ins_rt(sp_rt);
2856                         }
2857                 }
2858                 read_unlock_bh(&idev->lock);
2859         }
2860 }
2861
2862 static void addrconf_add_linklocal(struct inet6_dev *idev,
2863                                    const struct in6_addr *addr, u32 flags)
2864 {
2865         struct inet6_ifaddr *ifp;
2866         u32 addr_flags = flags | IFA_F_PERMANENT;
2867
2868 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2869         if (idev->cnf.optimistic_dad &&
2870             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2871                 addr_flags |= IFA_F_OPTIMISTIC;
2872 #endif
2873
2874         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2875                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2876         if (!IS_ERR(ifp)) {
2877                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2878                 addrconf_dad_start(ifp);
2879                 in6_ifa_put(ifp);
2880         }
2881 }
2882
2883 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2884 {
2885         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2886                 return true;
2887
2888         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2889             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2890                 return true;
2891
2892         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2893             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2894                 return true;
2895
2896         return false;
2897 }
2898
2899 static int ipv6_generate_stable_address(struct in6_addr *address,
2900                                         u8 dad_count,
2901                                         const struct inet6_dev *idev)
2902 {
2903         static DEFINE_SPINLOCK(lock);
2904         static __u32 digest[SHA_DIGEST_WORDS];
2905         static __u32 workspace[SHA_WORKSPACE_WORDS];
2906
2907         static union {
2908                 char __data[SHA_MESSAGE_BYTES];
2909                 struct {
2910                         struct in6_addr secret;
2911                         __be32 prefix[2];
2912                         unsigned char hwaddr[MAX_ADDR_LEN];
2913                         u8 dad_count;
2914                 } __packed;
2915         } data;
2916
2917         struct in6_addr secret;
2918         struct in6_addr temp;
2919         struct net *net = dev_net(idev->dev);
2920
2921         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2922
2923         if (idev->cnf.stable_secret.initialized)
2924                 secret = idev->cnf.stable_secret.secret;
2925         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
2926                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
2927         else
2928                 return -1;
2929
2930 retry:
2931         spin_lock_bh(&lock);
2932
2933         sha_init(digest);
2934         memset(&data, 0, sizeof(data));
2935         memset(workspace, 0, sizeof(workspace));
2936         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
2937         data.prefix[0] = address->s6_addr32[0];
2938         data.prefix[1] = address->s6_addr32[1];
2939         data.secret = secret;
2940         data.dad_count = dad_count;
2941
2942         sha_transform(digest, data.__data, workspace);
2943
2944         temp = *address;
2945         temp.s6_addr32[2] = (__force __be32)digest[0];
2946         temp.s6_addr32[3] = (__force __be32)digest[1];
2947
2948         spin_unlock_bh(&lock);
2949
2950         if (ipv6_reserved_interfaceid(temp)) {
2951                 dad_count++;
2952                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
2953                         return -1;
2954                 goto retry;
2955         }
2956
2957         *address = temp;
2958         return 0;
2959 }
2960
2961 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2962 {
2963         struct in6_addr addr;
2964
2965         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2966
2967         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
2968                 if (!ipv6_generate_stable_address(&addr, 0, idev))
2969                         addrconf_add_linklocal(idev, &addr,
2970                                                IFA_F_STABLE_PRIVACY);
2971                 else if (prefix_route)
2972                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2973         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2974                 /* addrconf_add_linklocal also adds a prefix_route and we
2975                  * only need to care about prefix routes if ipv6_generate_eui64
2976                  * couldn't generate one.
2977                  */
2978                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2979                         addrconf_add_linklocal(idev, &addr, 0);
2980                 else if (prefix_route)
2981                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2982         }
2983 }
2984
2985 static void addrconf_dev_config(struct net_device *dev)
2986 {
2987         struct inet6_dev *idev;
2988
2989         ASSERT_RTNL();
2990
2991         if ((dev->type != ARPHRD_ETHER) &&
2992             (dev->type != ARPHRD_FDDI) &&
2993             (dev->type != ARPHRD_ARCNET) &&
2994             (dev->type != ARPHRD_INFINIBAND) &&
2995             (dev->type != ARPHRD_IEEE802154) &&
2996             (dev->type != ARPHRD_IEEE1394) &&
2997             (dev->type != ARPHRD_TUNNEL6) &&
2998             (dev->type != ARPHRD_6LOWPAN)) {
2999                 /* Alas, we support only Ethernet autoconfiguration. */
3000                 return;
3001         }
3002
3003         idev = addrconf_add_dev(dev);
3004         if (IS_ERR(idev))
3005                 return;
3006
3007         addrconf_addr_gen(idev, false);
3008 }
3009
3010 #if IS_ENABLED(CONFIG_IPV6_SIT)
3011 static void addrconf_sit_config(struct net_device *dev)
3012 {
3013         struct inet6_dev *idev;
3014
3015         ASSERT_RTNL();
3016
3017         /*
3018          * Configure the tunnel with one of our IPv4
3019          * addresses... we should configure all of
3020          * our v4 addrs in the tunnel
3021          */
3022
3023         idev = ipv6_find_idev(dev);
3024         if (!idev) {
3025                 pr_debug("%s: add_dev failed\n", __func__);
3026                 return;
3027         }
3028
3029         if (dev->priv_flags & IFF_ISATAP) {
3030                 addrconf_addr_gen(idev, false);
3031                 return;
3032         }
3033
3034         sit_add_v4_addrs(idev);
3035
3036         if (dev->flags&IFF_POINTOPOINT)
3037                 addrconf_add_mroute(dev);
3038 }
3039 #endif
3040
3041 #if IS_ENABLED(CONFIG_NET_IPGRE)
3042 static void addrconf_gre_config(struct net_device *dev)
3043 {
3044         struct inet6_dev *idev;
3045
3046         ASSERT_RTNL();
3047
3048         idev = ipv6_find_idev(dev);
3049         if (!idev) {
3050                 pr_debug("%s: add_dev failed\n", __func__);
3051                 return;
3052         }
3053
3054         addrconf_addr_gen(idev, true);
3055 }
3056 #endif
3057
3058 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3059                            void *ptr)
3060 {
3061         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3062         struct inet6_dev *idev = __in6_dev_get(dev);
3063         int run_pending = 0;
3064         int err;
3065
3066         switch (event) {
3067         case NETDEV_REGISTER:
3068                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3069                         idev = ipv6_add_dev(dev);
3070                         if (IS_ERR(idev))
3071                                 return notifier_from_errno(PTR_ERR(idev));
3072                 }
3073                 break;
3074
3075         case NETDEV_UP:
3076         case NETDEV_CHANGE:
3077                 if (dev->flags & IFF_SLAVE)
3078                         break;
3079
3080                 if (idev && idev->cnf.disable_ipv6)
3081                         break;
3082
3083                 if (event == NETDEV_UP) {
3084                         if (!addrconf_qdisc_ok(dev)) {
3085                                 /* device is not ready yet. */
3086                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3087                                         dev->name);
3088                                 break;
3089                         }
3090
3091                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3092                                 idev = ipv6_add_dev(dev);
3093
3094                         if (!IS_ERR_OR_NULL(idev)) {
3095                                 idev->if_flags |= IF_READY;
3096                                 run_pending = 1;
3097                         }
3098                 } else {
3099                         if (!addrconf_qdisc_ok(dev)) {
3100                                 /* device is still not ready. */
3101                                 break;
3102                         }
3103
3104                         if (idev) {
3105                                 if (idev->if_flags & IF_READY)
3106                                         /* device is already configured. */
3107                                         break;
3108                                 idev->if_flags |= IF_READY;
3109                         }
3110
3111                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3112                                 dev->name);
3113
3114                         run_pending = 1;
3115                 }
3116
3117                 switch (dev->type) {
3118 #if IS_ENABLED(CONFIG_IPV6_SIT)
3119                 case ARPHRD_SIT:
3120                         addrconf_sit_config(dev);
3121                         break;
3122 #endif
3123 #if IS_ENABLED(CONFIG_NET_IPGRE)
3124                 case ARPHRD_IPGRE:
3125                         addrconf_gre_config(dev);
3126                         break;
3127 #endif
3128                 case ARPHRD_LOOPBACK:
3129                         init_loopback(dev);
3130                         break;
3131
3132                 default:
3133                         addrconf_dev_config(dev);
3134                         break;
3135                 }
3136
3137                 if (!IS_ERR_OR_NULL(idev)) {
3138                         if (run_pending)
3139                                 addrconf_dad_run(idev);
3140
3141                         /*
3142                          * If the MTU changed during the interface down,
3143                          * when the interface up, the changed MTU must be
3144                          * reflected in the idev as well as routers.
3145                          */
3146                         if (idev->cnf.mtu6 != dev->mtu &&
3147                             dev->mtu >= IPV6_MIN_MTU) {
3148                                 rt6_mtu_change(dev, dev->mtu);
3149                                 idev->cnf.mtu6 = dev->mtu;
3150                         }
3151                         idev->tstamp = jiffies;
3152                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3153
3154                         /*
3155                          * If the changed mtu during down is lower than
3156                          * IPV6_MIN_MTU stop IPv6 on this interface.
3157                          */
3158                         if (dev->mtu < IPV6_MIN_MTU)
3159                                 addrconf_ifdown(dev, 1);
3160                 }
3161                 break;
3162
3163         case NETDEV_CHANGEMTU:
3164                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3165                         rt6_mtu_change(dev, dev->mtu);
3166                         idev->cnf.mtu6 = dev->mtu;
3167                         break;
3168                 }
3169
3170                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3171                         idev = ipv6_add_dev(dev);
3172                         if (!IS_ERR(idev))
3173                                 break;
3174                 }
3175
3176                 /*
3177                  * if MTU under IPV6_MIN_MTU.
3178                  * Stop IPv6 on this interface.
3179                  */
3180
3181         case NETDEV_DOWN:
3182         case NETDEV_UNREGISTER:
3183                 /*
3184                  *      Remove all addresses from this interface.
3185                  */
3186                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3187                 break;
3188
3189         case NETDEV_CHANGENAME:
3190                 if (idev) {
3191                         snmp6_unregister_dev(idev);
3192                         addrconf_sysctl_unregister(idev);
3193                         err = addrconf_sysctl_register(idev);
3194                         if (err)
3195                                 return notifier_from_errno(err);
3196                         err = snmp6_register_dev(idev);
3197                         if (err) {
3198                                 addrconf_sysctl_unregister(idev);
3199                                 return notifier_from_errno(err);
3200                         }
3201                 }
3202                 break;
3203
3204         case NETDEV_PRE_TYPE_CHANGE:
3205         case NETDEV_POST_TYPE_CHANGE:
3206                 addrconf_type_change(dev, event);
3207                 break;
3208         }
3209
3210         return NOTIFY_OK;
3211 }
3212
3213 /*
3214  *      addrconf module should be notified of a device going up
3215  */
3216 static struct notifier_block ipv6_dev_notf = {
3217         .notifier_call = addrconf_notify,
3218 };
3219
3220 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3221 {
3222         struct inet6_dev *idev;
3223         ASSERT_RTNL();
3224
3225         idev = __in6_dev_get(dev);
3226
3227         if (event == NETDEV_POST_TYPE_CHANGE)
3228                 ipv6_mc_remap(idev);
3229         else if (event == NETDEV_PRE_TYPE_CHANGE)
3230                 ipv6_mc_unmap(idev);
3231 }
3232
3233 static int addrconf_ifdown(struct net_device *dev, int how)
3234 {
3235         struct net *net = dev_net(dev);
3236         struct inet6_dev *idev;
3237         struct inet6_ifaddr *ifa;
3238         int state, i;
3239
3240         ASSERT_RTNL();
3241
3242         rt6_ifdown(net, dev);
3243         neigh_ifdown(&nd_tbl, dev);
3244
3245         idev = __in6_dev_get(dev);
3246         if (!idev)
3247                 return -ENODEV;
3248
3249         /*
3250          * Step 1: remove reference to ipv6 device from parent device.
3251          *         Do not dev_put!
3252          */
3253         if (how) {
3254                 idev->dead = 1;
3255
3256                 /* protected by rtnl_lock */
3257                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3258
3259                 /* Step 1.5: remove snmp6 entry */
3260                 snmp6_unregister_dev(idev);
3261
3262         }
3263
3264         /* Step 2: clear hash table */
3265         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3266                 struct hlist_head *h = &inet6_addr_lst[i];
3267
3268                 spin_lock_bh(&addrconf_hash_lock);
3269 restart:
3270                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3271                         if (ifa->idev == idev) {
3272                                 hlist_del_init_rcu(&ifa->addr_lst);
3273                                 addrconf_del_dad_work(ifa);
3274                                 goto restart;
3275                         }
3276                 }
3277                 spin_unlock_bh(&addrconf_hash_lock);
3278         }
3279
3280         write_lock_bh(&idev->lock);
3281
3282         addrconf_del_rs_timer(idev);
3283
3284         /* Step 2: clear flags for stateless addrconf */
3285         if (!how)
3286                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3287
3288         if (how && del_timer(&idev->regen_timer))
3289                 in6_dev_put(idev);
3290
3291         /* Step 3: clear tempaddr list */
3292         while (!list_empty(&idev->tempaddr_list)) {
3293                 ifa = list_first_entry(&idev->tempaddr_list,
3294                                        struct inet6_ifaddr, tmp_list);
3295                 list_del(&ifa->tmp_list);
3296                 write_unlock_bh(&idev->lock);
3297                 spin_lock_bh(&ifa->lock);
3298
3299                 if (ifa->ifpub) {
3300                         in6_ifa_put(ifa->ifpub);
3301                         ifa->ifpub = NULL;
3302                 }
3303                 spin_unlock_bh(&ifa->lock);
3304                 in6_ifa_put(ifa);
3305                 write_lock_bh(&idev->lock);
3306         }
3307
3308         while (!list_empty(&idev->addr_list)) {
3309                 ifa = list_first_entry(&idev->addr_list,
3310                                        struct inet6_ifaddr, if_list);
3311                 addrconf_del_dad_work(ifa);
3312
3313                 list_del(&ifa->if_list);
3314
3315                 write_unlock_bh(&idev->lock);
3316
3317                 spin_lock_bh(&ifa->lock);
3318                 state = ifa->state;
3319                 ifa->state = INET6_IFADDR_STATE_DEAD;
3320                 spin_unlock_bh(&ifa->lock);
3321
3322                 if (state != INET6_IFADDR_STATE_DEAD) {
3323                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3324                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3325                 }
3326                 in6_ifa_put(ifa);
3327
3328                 write_lock_bh(&idev->lock);
3329         }
3330
3331         write_unlock_bh(&idev->lock);
3332
3333         /* Step 5: Discard anycast and multicast list */
3334         if (how) {
3335                 ipv6_ac_destroy_dev(idev);
3336                 ipv6_mc_destroy_dev(idev);
3337         } else {
3338                 ipv6_mc_down(idev);
3339         }
3340
3341         idev->tstamp = jiffies;
3342
3343         /* Last: Shot the device (if unregistered) */
3344         if (how) {
3345                 addrconf_sysctl_unregister(idev);
3346                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3347                 neigh_ifdown(&nd_tbl, dev);
3348                 in6_dev_put(idev);
3349         }
3350         return 0;
3351 }
3352
3353 static void addrconf_rs_timer(unsigned long data)
3354 {
3355         struct inet6_dev *idev = (struct inet6_dev *)data;
3356         struct net_device *dev = idev->dev;
3357         struct in6_addr lladdr;
3358
3359         write_lock(&idev->lock);
3360         if (idev->dead || !(idev->if_flags & IF_READY))
3361                 goto out;
3362
3363         if (!ipv6_accept_ra(idev))
3364                 goto out;
3365
3366         /* Announcement received after solicitation was sent */
3367         if (idev->if_flags & IF_RA_RCVD)
3368                 goto out;
3369
3370         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3371                 write_unlock(&idev->lock);
3372                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3373                         ndisc_send_rs(dev, &lladdr,
3374                                       &in6addr_linklocal_allrouters);
3375                 else
3376                         goto put;
3377
3378                 write_lock(&idev->lock);
3379                 /* The wait after the last probe can be shorter */
3380                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3381                                              idev->cnf.rtr_solicits) ?
3382                                       idev->cnf.rtr_solicit_delay :
3383                                       idev->cnf.rtr_solicit_interval);
3384         } else {
3385                 /*
3386                  * Note: we do not support deprecated "all on-link"
3387                  * assumption any longer.
3388                  */
3389                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3390         }
3391
3392 out:
3393         write_unlock(&idev->lock);
3394 put:
3395         in6_dev_put(idev);
3396 }
3397
3398 /*
3399  *      Duplicate Address Detection
3400  */
3401 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3402 {
3403         unsigned long rand_num;
3404         struct inet6_dev *idev = ifp->idev;
3405
3406         if (ifp->flags & IFA_F_OPTIMISTIC)
3407                 rand_num = 0;
3408         else
3409                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3410
3411         ifp->dad_probes = idev->cnf.dad_transmits;
3412         addrconf_mod_dad_work(ifp, rand_num);
3413 }
3414
3415 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3416 {
3417         struct inet6_dev *idev = ifp->idev;
3418         struct net_device *dev = idev->dev;
3419
3420         addrconf_join_solict(dev, &ifp->addr);
3421
3422         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3423
3424         read_lock_bh(&idev->lock);
3425         spin_lock(&ifp->lock);
3426         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3427                 goto out;
3428
3429         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3430             idev->cnf.accept_dad < 1 ||
3431             !(ifp->flags&IFA_F_TENTATIVE) ||
3432             ifp->flags & IFA_F_NODAD) {
3433                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3434                 spin_unlock(&ifp->lock);
3435                 read_unlock_bh(&idev->lock);
3436
3437                 addrconf_dad_completed(ifp);
3438                 return;
3439         }
3440
3441         if (!(idev->if_flags & IF_READY)) {
3442                 spin_unlock(&ifp->lock);
3443                 read_unlock_bh(&idev->lock);
3444                 /*
3445                  * If the device is not ready:
3446                  * - keep it tentative if it is a permanent address.
3447                  * - otherwise, kill it.
3448                  */
3449                 in6_ifa_hold(ifp);
3450                 addrconf_dad_stop(ifp, 0);
3451                 return;
3452         }
3453
3454         /*
3455          * Optimistic nodes can start receiving
3456          * Frames right away
3457          */
3458         if (ifp->flags & IFA_F_OPTIMISTIC) {
3459                 ip6_ins_rt(ifp->rt);
3460                 if (ipv6_use_optimistic_addr(idev)) {
3461                         /* Because optimistic nodes can use this address,
3462                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3463                          */
3464                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3465                 }
3466         }
3467
3468         addrconf_dad_kick(ifp);
3469 out:
3470         spin_unlock(&ifp->lock);
3471         read_unlock_bh(&idev->lock);
3472 }
3473
3474 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3475 {
3476         bool begin_dad = false;
3477
3478         spin_lock_bh(&ifp->lock);
3479         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3480                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3481                 begin_dad = true;
3482         }
3483         spin_unlock_bh(&ifp->lock);
3484
3485         if (begin_dad)
3486                 addrconf_mod_dad_work(ifp, 0);
3487 }
3488
3489 static void addrconf_dad_work(struct work_struct *w)
3490 {
3491         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3492                                                 struct inet6_ifaddr,
3493                                                 dad_work);
3494         struct inet6_dev *idev = ifp->idev;
3495         struct in6_addr mcaddr;
3496
3497         enum {
3498                 DAD_PROCESS,
3499                 DAD_BEGIN,
3500                 DAD_ABORT,
3501         } action = DAD_PROCESS;
3502
3503         rtnl_lock();
3504
3505         spin_lock_bh(&ifp->lock);
3506         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3507                 action = DAD_BEGIN;
3508                 ifp->state = INET6_IFADDR_STATE_DAD;
3509         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3510                 action = DAD_ABORT;
3511                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3512         }
3513         spin_unlock_bh(&ifp->lock);
3514
3515         if (action == DAD_BEGIN) {
3516                 addrconf_dad_begin(ifp);
3517                 goto out;
3518         } else if (action == DAD_ABORT) {
3519                 addrconf_dad_stop(ifp, 1);
3520                 goto out;
3521         }
3522
3523         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3524                 goto out;
3525
3526         write_lock_bh(&idev->lock);
3527         if (idev->dead || !(idev->if_flags & IF_READY)) {
3528                 write_unlock_bh(&idev->lock);
3529                 goto out;
3530         }
3531
3532         spin_lock(&ifp->lock);
3533         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3534                 spin_unlock(&ifp->lock);
3535                 write_unlock_bh(&idev->lock);
3536                 goto out;
3537         }
3538
3539         if (ifp->dad_probes == 0) {
3540                 /*
3541                  * DAD was successful
3542                  */
3543
3544                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3545                 spin_unlock(&ifp->lock);
3546                 write_unlock_bh(&idev->lock);
3547
3548                 addrconf_dad_completed(ifp);
3549
3550                 goto out;
3551         }
3552
3553         ifp->dad_probes--;
3554         addrconf_mod_dad_work(ifp,
3555                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3556         spin_unlock(&ifp->lock);
3557         write_unlock_bh(&idev->lock);
3558
3559         /* send a neighbour solicitation for our addr */
3560         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3561         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3562 out:
3563         in6_ifa_put(ifp);
3564         rtnl_unlock();
3565 }
3566
3567 /* ifp->idev must be at least read locked */
3568 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3569 {
3570         struct inet6_ifaddr *ifpiter;
3571         struct inet6_dev *idev = ifp->idev;
3572
3573         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3574                 if (ifpiter->scope > IFA_LINK)
3575                         break;
3576                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3577                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3578                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3579                     IFA_F_PERMANENT)
3580                         return false;
3581         }
3582         return true;
3583 }
3584
3585 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3586 {
3587         struct net_device *dev = ifp->idev->dev;
3588         struct in6_addr lladdr;
3589         bool send_rs, send_mld;
3590
3591         addrconf_del_dad_work(ifp);
3592
3593         /*
3594          *      Configure the address for reception. Now it is valid.
3595          */
3596
3597         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3598
3599         /* If added prefix is link local and we are prepared to process
3600            router advertisements, start sending router solicitations.
3601          */
3602
3603         read_lock_bh(&ifp->idev->lock);
3604         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3605         send_rs = send_mld &&
3606                   ipv6_accept_ra(ifp->idev) &&
3607                   ifp->idev->cnf.rtr_solicits > 0 &&
3608                   (dev->flags&IFF_LOOPBACK) == 0;
3609         read_unlock_bh(&ifp->idev->lock);
3610
3611         /* While dad is in progress mld report's source address is in6_addrany.
3612          * Resend with proper ll now.
3613          */
3614         if (send_mld)
3615                 ipv6_mc_dad_complete(ifp->idev);
3616
3617         if (send_rs) {
3618                 /*
3619                  *      If a host as already performed a random delay
3620                  *      [...] as part of DAD [...] there is no need
3621                  *      to delay again before sending the first RS
3622                  */
3623                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3624                         return;
3625                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3626
3627                 write_lock_bh(&ifp->idev->lock);
3628                 spin_lock(&ifp->lock);
3629                 ifp->idev->rs_probes = 1;
3630                 ifp->idev->if_flags |= IF_RS_SENT;
3631                 addrconf_mod_rs_timer(ifp->idev,
3632                                       ifp->idev->cnf.rtr_solicit_interval);
3633                 spin_unlock(&ifp->lock);
3634                 write_unlock_bh(&ifp->idev->lock);
3635         }
3636 }
3637
3638 static void addrconf_dad_run(struct inet6_dev *idev)
3639 {
3640         struct inet6_ifaddr *ifp;
3641
3642         read_lock_bh(&idev->lock);
3643         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3644                 spin_lock(&ifp->lock);
3645                 if (ifp->flags & IFA_F_TENTATIVE &&
3646                     ifp->state == INET6_IFADDR_STATE_DAD)
3647                         addrconf_dad_kick(ifp);
3648                 spin_unlock(&ifp->lock);
3649         }
3650         read_unlock_bh(&idev->lock);
3651 }
3652
3653 #ifdef CONFIG_PROC_FS
3654 struct if6_iter_state {
3655         struct seq_net_private p;
3656         int bucket;
3657         int offset;
3658 };
3659
3660 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3661 {
3662         struct inet6_ifaddr *ifa = NULL;
3663         struct if6_iter_state *state = seq->private;
3664         struct net *net = seq_file_net(seq);
3665         int p = 0;
3666
3667         /* initial bucket if pos is 0 */
3668         if (pos == 0) {
3669                 state->bucket = 0;
3670                 state->offset = 0;
3671         }
3672
3673         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3674                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3675                                          addr_lst) {
3676                         if (!net_eq(dev_net(ifa->idev->dev), net))
3677                                 continue;
3678                         /* sync with offset */
3679                         if (p < state->offset) {
3680                                 p++;
3681                                 continue;
3682                         }
3683                         state->offset++;
3684                         return ifa;
3685                 }
3686
3687                 /* prepare for next bucket */
3688                 state->offset = 0;
3689                 p = 0;
3690         }
3691         return NULL;
3692 }
3693
3694 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3695                                          struct inet6_ifaddr *ifa)
3696 {
3697         struct if6_iter_state *state = seq->private;
3698         struct net *net = seq_file_net(seq);
3699
3700         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3701                 if (!net_eq(dev_net(ifa->idev->dev), net))
3702                         continue;
3703                 state->offset++;
3704                 return ifa;
3705         }
3706
3707         while (++state->bucket < IN6_ADDR_HSIZE) {
3708                 state->offset = 0;
3709                 hlist_for_each_entry_rcu_bh(ifa,
3710                                      &inet6_addr_lst[state->bucket], addr_lst) {
3711                         if (!net_eq(dev_net(ifa->idev->dev), net))
3712                                 continue;
3713                         state->offset++;
3714                         return ifa;
3715                 }
3716         }
3717
3718         return NULL;
3719 }
3720
3721 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3722         __acquires(rcu_bh)
3723 {
3724         rcu_read_lock_bh();
3725         return if6_get_first(seq, *pos);
3726 }
3727
3728 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3729 {
3730         struct inet6_ifaddr *ifa;
3731
3732         ifa = if6_get_next(seq, v);
3733         ++*pos;
3734         return ifa;
3735 }
3736
3737 static void if6_seq_stop(struct seq_file *seq, void *v)
3738         __releases(rcu_bh)
3739 {
3740         rcu_read_unlock_bh();
3741 }
3742
3743 static int if6_seq_show(struct seq_file *seq, void *v)
3744 {
3745         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3746         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3747                    &ifp->addr,
3748                    ifp->idev->dev->ifindex,
3749                    ifp->prefix_len,
3750                    ifp->scope,
3751                    (u8) ifp->flags,
3752                    ifp->idev->dev->name);
3753         return 0;
3754 }
3755
3756 static const struct seq_operations if6_seq_ops = {
3757         .start  = if6_seq_start,
3758         .next   = if6_seq_next,
3759         .show   = if6_seq_show,
3760         .stop   = if6_seq_stop,
3761 };
3762
3763 static int if6_seq_open(struct inode *inode, struct file *file)
3764 {
3765         return seq_open_net(inode, file, &if6_seq_ops,
3766                             sizeof(struct if6_iter_state));
3767 }
3768
3769 static const struct file_operations if6_fops = {
3770         .owner          = THIS_MODULE,
3771         .open           = if6_seq_open,
3772         .read           = seq_read,
3773         .llseek         = seq_lseek,
3774         .release        = seq_release_net,
3775 };
3776
3777 static int __net_init if6_proc_net_init(struct net *net)
3778 {
3779         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3780                 return -ENOMEM;
3781         return 0;
3782 }
3783
3784 static void __net_exit if6_proc_net_exit(struct net *net)
3785 {
3786         remove_proc_entry("if_inet6", net->proc_net);
3787 }
3788
3789 static struct pernet_operations if6_proc_net_ops = {
3790         .init = if6_proc_net_init,
3791         .exit = if6_proc_net_exit,
3792 };
3793
3794 int __init if6_proc_init(void)
3795 {
3796         return register_pernet_subsys(&if6_proc_net_ops);
3797 }
3798
3799 void if6_proc_exit(void)
3800 {
3801         unregister_pernet_subsys(&if6_proc_net_ops);
3802 }
3803 #endif  /* CONFIG_PROC_FS */
3804
3805 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3806 /* Check if address is a home address configured on any interface. */
3807 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3808 {
3809         int ret = 0;
3810         struct inet6_ifaddr *ifp = NULL;
3811         unsigned int hash = inet6_addr_hash(addr);
3812
3813         rcu_read_lock_bh();
3814         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3815                 if (!net_eq(dev_net(ifp->idev->dev), net))
3816                         continue;
3817                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3818                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3819                         ret = 1;
3820                         break;
3821                 }
3822         }
3823         rcu_read_unlock_bh();
3824         return ret;
3825 }
3826 #endif
3827
3828 /*
3829  *      Periodic address status verification
3830  */
3831
3832 static void addrconf_verify_rtnl(void)
3833 {
3834         unsigned long now, next, next_sec, next_sched;
3835         struct inet6_ifaddr *ifp;
3836         int i;
3837
3838         ASSERT_RTNL();
3839
3840         rcu_read_lock_bh();
3841         now = jiffies;
3842         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3843
3844         cancel_delayed_work(&addr_chk_work);
3845
3846         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3847 restart:
3848                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3849                         unsigned long age;
3850
3851                         /* When setting preferred_lft to a value not zero or
3852                          * infinity, while valid_lft is infinity
3853                          * IFA_F_PERMANENT has a non-infinity life time.
3854                          */
3855                         if ((ifp->flags & IFA_F_PERMANENT) &&
3856                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3857                                 continue;
3858
3859                         spin_lock(&ifp->lock);
3860                         /* We try to batch several events at once. */
3861                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3862
3863                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3864                             age >= ifp->valid_lft) {
3865                                 spin_unlock(&ifp->lock);
3866                                 in6_ifa_hold(ifp);
3867                                 ipv6_del_addr(ifp);
3868                                 goto restart;
3869                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3870                                 spin_unlock(&ifp->lock);
3871                                 continue;
3872                         } else if (age >= ifp->prefered_lft) {
3873                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3874                                 int deprecate = 0;
3875
3876                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3877                                         deprecate = 1;
3878                                         ifp->flags |= IFA_F_DEPRECATED;
3879                                 }
3880
3881                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3882                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3883                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3884
3885                                 spin_unlock(&ifp->lock);
3886
3887                                 if (deprecate) {
3888                                         in6_ifa_hold(ifp);
3889
3890                                         ipv6_ifa_notify(0, ifp);
3891                                         in6_ifa_put(ifp);
3892                                         goto restart;
3893                                 }
3894                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3895                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3896                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3897                                         ifp->idev->cnf.dad_transmits *
3898                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3899
3900                                 if (age >= ifp->prefered_lft - regen_advance) {
3901                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3902                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3903                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3904                                         if (!ifp->regen_count && ifpub) {
3905                                                 ifp->regen_count++;
3906                                                 in6_ifa_hold(ifp);
3907                                                 in6_ifa_hold(ifpub);
3908                                                 spin_unlock(&ifp->lock);
3909
3910                                                 spin_lock(&ifpub->lock);
3911                                                 ifpub->regen_count = 0;
3912                                                 spin_unlock(&ifpub->lock);
3913                                                 ipv6_create_tempaddr(ifpub, ifp);
3914                                                 in6_ifa_put(ifpub);
3915                                                 in6_ifa_put(ifp);
3916                                                 goto restart;
3917                                         }
3918                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3919                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3920                                 spin_unlock(&ifp->lock);
3921                         } else {
3922                                 /* ifp->prefered_lft <= ifp->valid_lft */
3923                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3924                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3925                                 spin_unlock(&ifp->lock);
3926                         }
3927                 }
3928         }
3929
3930         next_sec = round_jiffies_up(next);
3931         next_sched = next;
3932
3933         /* If rounded timeout is accurate enough, accept it. */
3934         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3935                 next_sched = next_sec;
3936
3937         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3938         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3939                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3940
3941         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3942               now, next, next_sec, next_sched);
3943         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3944         rcu_read_unlock_bh();
3945 }
3946
3947 static void addrconf_verify_work(struct work_struct *w)
3948 {
3949         rtnl_lock();
3950         addrconf_verify_rtnl();
3951         rtnl_unlock();
3952 }
3953
3954 static void addrconf_verify(void)
3955 {
3956         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3957 }
3958
3959 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3960                                      struct in6_addr **peer_pfx)
3961 {
3962         struct in6_addr *pfx = NULL;
3963
3964         *peer_pfx = NULL;
3965
3966         if (addr)
3967                 pfx = nla_data(addr);
3968
3969         if (local) {
3970                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3971                         *peer_pfx = pfx;
3972                 pfx = nla_data(local);
3973         }
3974
3975         return pfx;
3976 }
3977
3978 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3979         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3980         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3981         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3982         [IFA_FLAGS]             = { .len = sizeof(u32) },
3983 };
3984
3985 static int
3986 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3987 {
3988         struct net *net = sock_net(skb->sk);
3989         struct ifaddrmsg *ifm;
3990         struct nlattr *tb[IFA_MAX+1];
3991         struct in6_addr *pfx, *peer_pfx;
3992         u32 ifa_flags;
3993         int err;
3994
3995         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3996         if (err < 0)
3997                 return err;
3998
3999         ifm = nlmsg_data(nlh);
4000         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4001         if (!pfx)
4002                 return -EINVAL;
4003
4004         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4005
4006         /* We ignore other flags so far. */
4007         ifa_flags &= IFA_F_MANAGETEMPADDR;
4008
4009         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4010                               ifm->ifa_prefixlen);
4011 }
4012
4013 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4014                              u32 prefered_lft, u32 valid_lft)
4015 {
4016         u32 flags;
4017         clock_t expires;
4018         unsigned long timeout;
4019         bool was_managetempaddr;
4020         bool had_prefixroute;
4021
4022         ASSERT_RTNL();
4023
4024         if (!valid_lft || (prefered_lft > valid_lft))
4025                 return -EINVAL;
4026
4027         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4028             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4029                 return -EINVAL;
4030
4031         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4032         if (addrconf_finite_timeout(timeout)) {
4033                 expires = jiffies_to_clock_t(timeout * HZ);
4034                 valid_lft = timeout;
4035                 flags = RTF_EXPIRES;
4036         } else {
4037                 expires = 0;
4038                 flags = 0;
4039                 ifa_flags |= IFA_F_PERMANENT;
4040         }
4041
4042         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4043         if (addrconf_finite_timeout(timeout)) {
4044                 if (timeout == 0)
4045                         ifa_flags |= IFA_F_DEPRECATED;
4046                 prefered_lft = timeout;
4047         }
4048
4049         spin_lock_bh(&ifp->lock);
4050         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4051         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4052                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4053         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4054                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4055                         IFA_F_NOPREFIXROUTE);
4056         ifp->flags |= ifa_flags;
4057         ifp->tstamp = jiffies;
4058         ifp->valid_lft = valid_lft;
4059         ifp->prefered_lft = prefered_lft;
4060
4061         spin_unlock_bh(&ifp->lock);
4062         if (!(ifp->flags&IFA_F_TENTATIVE))
4063                 ipv6_ifa_notify(0, ifp);
4064
4065         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4066                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4067                                       expires, flags);
4068         } else if (had_prefixroute) {
4069                 enum cleanup_prefix_rt_t action;
4070                 unsigned long rt_expires;
4071
4072                 write_lock_bh(&ifp->idev->lock);
4073                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4074                 write_unlock_bh(&ifp->idev->lock);
4075
4076                 if (action != CLEANUP_PREFIX_RT_NOP) {
4077                         cleanup_prefix_route(ifp, rt_expires,
4078                                 action == CLEANUP_PREFIX_RT_DEL);
4079                 }
4080         }
4081
4082         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4083                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4084                         valid_lft = prefered_lft = 0;
4085                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4086                                  !was_managetempaddr, jiffies);
4087         }
4088
4089         addrconf_verify_rtnl();
4090
4091         return 0;
4092 }
4093
4094 static int
4095 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4096 {
4097         struct net *net = sock_net(skb->sk);
4098         struct ifaddrmsg *ifm;
4099         struct nlattr *tb[IFA_MAX+1];
4100         struct in6_addr *pfx, *peer_pfx;
4101         struct inet6_ifaddr *ifa;
4102         struct net_device *dev;
4103         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4104         u32 ifa_flags;
4105         int err;
4106
4107         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4108         if (err < 0)
4109                 return err;
4110
4111         ifm = nlmsg_data(nlh);
4112         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4113         if (!pfx)
4114                 return -EINVAL;
4115
4116         if (tb[IFA_CACHEINFO]) {
4117                 struct ifa_cacheinfo *ci;
4118
4119                 ci = nla_data(tb[IFA_CACHEINFO]);
4120                 valid_lft = ci->ifa_valid;
4121                 preferred_lft = ci->ifa_prefered;
4122         } else {
4123                 preferred_lft = INFINITY_LIFE_TIME;
4124                 valid_lft = INFINITY_LIFE_TIME;
4125         }
4126
4127         dev =  __dev_get_by_index(net, ifm->ifa_index);
4128         if (!dev)
4129                 return -ENODEV;
4130
4131         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4132
4133         /* We ignore other flags so far. */
4134         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4135                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4136
4137         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4138         if (!ifa) {
4139                 /*
4140                  * It would be best to check for !NLM_F_CREATE here but
4141                  * userspace already relies on not having to provide this.
4142                  */
4143                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4144                                       ifm->ifa_prefixlen, ifa_flags,
4145                                       preferred_lft, valid_lft);
4146         }
4147
4148         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4149             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4150                 err = -EEXIST;
4151         else
4152                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4153
4154         in6_ifa_put(ifa);
4155
4156         return err;
4157 }
4158
4159 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4160                           u8 scope, int ifindex)
4161 {
4162         struct ifaddrmsg *ifm;
4163
4164         ifm = nlmsg_data(nlh);
4165         ifm->ifa_family = AF_INET6;
4166         ifm->ifa_prefixlen = prefixlen;
4167         ifm->ifa_flags = flags;
4168         ifm->ifa_scope = scope;
4169         ifm->ifa_index = ifindex;
4170 }
4171
4172 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4173                          unsigned long tstamp, u32 preferred, u32 valid)
4174 {
4175         struct ifa_cacheinfo ci;
4176
4177         ci.cstamp = cstamp_delta(cstamp);
4178         ci.tstamp = cstamp_delta(tstamp);
4179         ci.ifa_prefered = preferred;
4180         ci.ifa_valid = valid;
4181
4182         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4183 }
4184
4185 static inline int rt_scope(int ifa_scope)
4186 {
4187         if (ifa_scope & IFA_HOST)
4188                 return RT_SCOPE_HOST;
4189         else if (ifa_scope & IFA_LINK)
4190                 return RT_SCOPE_LINK;
4191         else if (ifa_scope & IFA_SITE)
4192                 return RT_SCOPE_SITE;
4193         else
4194                 return RT_SCOPE_UNIVERSE;
4195 }
4196
4197 static inline int inet6_ifaddr_msgsize(void)
4198 {
4199         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4200                + nla_total_size(16) /* IFA_LOCAL */
4201                + nla_total_size(16) /* IFA_ADDRESS */
4202                + nla_total_size(sizeof(struct ifa_cacheinfo))
4203                + nla_total_size(4)  /* IFA_FLAGS */;
4204 }
4205
4206 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4207                              u32 portid, u32 seq, int event, unsigned int flags)
4208 {
4209         struct nlmsghdr  *nlh;
4210         u32 preferred, valid;
4211
4212         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4213         if (!nlh)
4214                 return -EMSGSIZE;
4215
4216         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4217                       ifa->idev->dev->ifindex);
4218
4219         if (!((ifa->flags&IFA_F_PERMANENT) &&
4220               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4221                 preferred = ifa->prefered_lft;
4222                 valid = ifa->valid_lft;
4223                 if (preferred != INFINITY_LIFE_TIME) {
4224                         long tval = (jiffies - ifa->tstamp)/HZ;
4225                         if (preferred > tval)
4226                                 preferred -= tval;
4227                         else
4228                                 preferred = 0;
4229                         if (valid != INFINITY_LIFE_TIME) {
4230                                 if (valid > tval)
4231                                         valid -= tval;
4232                                 else
4233                                         valid = 0;
4234                         }
4235                 }
4236         } else {
4237                 preferred = INFINITY_LIFE_TIME;
4238                 valid = INFINITY_LIFE_TIME;
4239         }
4240
4241         if (!ipv6_addr_any(&ifa->peer_addr)) {
4242                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4243                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4244                         goto error;
4245         } else
4246                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4247                         goto error;
4248
4249         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4250                 goto error;
4251
4252         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4253                 goto error;
4254
4255         nlmsg_end(skb, nlh);
4256         return 0;
4257
4258 error:
4259         nlmsg_cancel(skb, nlh);
4260         return -EMSGSIZE;
4261 }
4262
4263 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4264                                 u32 portid, u32 seq, int event, u16 flags)
4265 {
4266         struct nlmsghdr  *nlh;
4267         u8 scope = RT_SCOPE_UNIVERSE;
4268         int ifindex = ifmca->idev->dev->ifindex;
4269
4270         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4271                 scope = RT_SCOPE_SITE;
4272
4273         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4274         if (!nlh)
4275                 return -EMSGSIZE;
4276
4277         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4278         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4279             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4280                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4281                 nlmsg_cancel(skb, nlh);
4282                 return -EMSGSIZE;
4283         }
4284
4285         nlmsg_end(skb, nlh);
4286         return 0;
4287 }
4288
4289 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4290                                 u32 portid, u32 seq, int event, unsigned int flags)
4291 {
4292         struct nlmsghdr  *nlh;
4293         u8 scope = RT_SCOPE_UNIVERSE;
4294         int ifindex = ifaca->aca_idev->dev->ifindex;
4295
4296         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4297                 scope = RT_SCOPE_SITE;
4298
4299         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4300         if (!nlh)
4301                 return -EMSGSIZE;
4302
4303         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4304         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4305             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4306                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4307                 nlmsg_cancel(skb, nlh);
4308                 return -EMSGSIZE;
4309         }
4310
4311         nlmsg_end(skb, nlh);
4312         return 0;
4313 }
4314
4315 enum addr_type_t {
4316         UNICAST_ADDR,
4317         MULTICAST_ADDR,
4318         ANYCAST_ADDR,
4319 };
4320
4321 /* called with rcu_read_lock() */
4322 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4323                           struct netlink_callback *cb, enum addr_type_t type,
4324                           int s_ip_idx, int *p_ip_idx)
4325 {
4326         struct ifmcaddr6 *ifmca;
4327         struct ifacaddr6 *ifaca;
4328         int err = 1;
4329         int ip_idx = *p_ip_idx;
4330
4331         read_lock_bh(&idev->lock);
4332         switch (type) {
4333         case UNICAST_ADDR: {
4334                 struct inet6_ifaddr *ifa;
4335
4336                 /* unicast address incl. temp addr */
4337                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4338                         if (++ip_idx < s_ip_idx)
4339                                 continue;
4340                         err = inet6_fill_ifaddr(skb, ifa,
4341                                                 NETLINK_CB(cb->skb).portid,
4342                                                 cb->nlh->nlmsg_seq,
4343                                                 RTM_NEWADDR,
4344                                                 NLM_F_MULTI);
4345                         if (err < 0)
4346                                 break;
4347                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4348                 }
4349                 break;
4350         }
4351         case MULTICAST_ADDR:
4352                 /* multicast address */
4353                 for (ifmca = idev->mc_list; ifmca;
4354                      ifmca = ifmca->next, ip_idx++) {
4355                         if (ip_idx < s_ip_idx)
4356                                 continue;
4357                         err = inet6_fill_ifmcaddr(skb, ifmca,
4358                                                   NETLINK_CB(cb->skb).portid,
4359                                                   cb->nlh->nlmsg_seq,
4360                                                   RTM_GETMULTICAST,
4361                                                   NLM_F_MULTI);
4362                         if (err < 0)
4363                                 break;
4364                 }
4365                 break;
4366         case ANYCAST_ADDR:
4367                 /* anycast address */
4368                 for (ifaca = idev->ac_list; ifaca;
4369                      ifaca = ifaca->aca_next, ip_idx++) {
4370                         if (ip_idx < s_ip_idx)
4371                                 continue;
4372                         err = inet6_fill_ifacaddr(skb, ifaca,
4373                                                   NETLINK_CB(cb->skb).portid,
4374                                                   cb->nlh->nlmsg_seq,
4375                                                   RTM_GETANYCAST,
4376                                                   NLM_F_MULTI);
4377                         if (err < 0)
4378                                 break;
4379                 }
4380                 break;
4381         default:
4382                 break;
4383         }
4384         read_unlock_bh(&idev->lock);
4385         *p_ip_idx = ip_idx;
4386         return err;
4387 }
4388
4389 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4390                            enum addr_type_t type)
4391 {
4392         struct net *net = sock_net(skb->sk);
4393         int h, s_h;
4394         int idx, ip_idx;
4395         int s_idx, s_ip_idx;
4396         struct net_device *dev;
4397         struct inet6_dev *idev;
4398         struct hlist_head *head;
4399
4400         s_h = cb->args[0];
4401         s_idx = idx = cb->args[1];
4402         s_ip_idx = ip_idx = cb->args[2];
4403
4404         rcu_read_lock();
4405         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4406         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4407                 idx = 0;
4408                 head = &net->dev_index_head[h];
4409                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4410                         if (idx < s_idx)
4411                                 goto cont;
4412                         if (h > s_h || idx > s_idx)
4413                                 s_ip_idx = 0;
4414                         ip_idx = 0;
4415                         idev = __in6_dev_get(dev);
4416                         if (!idev)
4417                                 goto cont;
4418
4419                         if (in6_dump_addrs(idev, skb, cb, type,
4420                                            s_ip_idx, &ip_idx) < 0)
4421                                 goto done;
4422 cont:
4423                         idx++;
4424                 }
4425         }
4426 done:
4427         rcu_read_unlock();
4428         cb->args[0] = h;
4429         cb->args[1] = idx;
4430         cb->args[2] = ip_idx;
4431
4432         return skb->len;
4433 }
4434
4435 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4436 {
4437         enum addr_type_t type = UNICAST_ADDR;
4438
4439         return inet6_dump_addr(skb, cb, type);
4440 }
4441
4442 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4443 {
4444         enum addr_type_t type = MULTICAST_ADDR;
4445
4446         return inet6_dump_addr(skb, cb, type);
4447 }
4448
4449
4450 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4451 {
4452         enum addr_type_t type = ANYCAST_ADDR;
4453
4454         return inet6_dump_addr(skb, cb, type);
4455 }
4456
4457 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4458 {
4459         struct net *net = sock_net(in_skb->sk);
4460         struct ifaddrmsg *ifm;
4461         struct nlattr *tb[IFA_MAX+1];
4462         struct in6_addr *addr = NULL, *peer;
4463         struct net_device *dev = NULL;
4464         struct inet6_ifaddr *ifa;
4465         struct sk_buff *skb;
4466         int err;
4467
4468         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4469         if (err < 0)
4470                 goto errout;
4471
4472         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4473         if (!addr) {
4474                 err = -EINVAL;
4475                 goto errout;
4476         }
4477
4478         ifm = nlmsg_data(nlh);
4479         if (ifm->ifa_index)
4480                 dev = __dev_get_by_index(net, ifm->ifa_index);
4481
4482         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4483         if (!ifa) {
4484                 err = -EADDRNOTAVAIL;
4485                 goto errout;
4486         }
4487
4488         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4489         if (!skb) {
4490                 err = -ENOBUFS;
4491                 goto errout_ifa;
4492         }
4493
4494         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4495                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4496         if (err < 0) {
4497                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4498                 WARN_ON(err == -EMSGSIZE);
4499                 kfree_skb(skb);
4500                 goto errout_ifa;
4501         }
4502         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4503 errout_ifa:
4504         in6_ifa_put(ifa);
4505 errout:
4506         return err;
4507 }
4508
4509 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4510 {
4511         struct sk_buff *skb;
4512         struct net *net = dev_net(ifa->idev->dev);
4513         int err = -ENOBUFS;
4514
4515         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4516         if (!skb)
4517                 goto errout;
4518
4519         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4520         if (err < 0) {
4521                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4522                 WARN_ON(err == -EMSGSIZE);
4523                 kfree_skb(skb);
4524                 goto errout;
4525         }
4526         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4527         return;
4528 errout:
4529         if (err < 0)
4530                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4531 }
4532
4533 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4534                                 __s32 *array, int bytes)
4535 {
4536         BUG_ON(bytes < (DEVCONF_MAX * 4));
4537
4538         memset(array, 0, bytes);
4539         array[DEVCONF_FORWARDING] = cnf->forwarding;
4540         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4541         array[DEVCONF_MTU6] = cnf->mtu6;
4542         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4543         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4544         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4545         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4546         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4547         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4548                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4549         array[DEVCONF_RTR_SOLICIT_DELAY] =
4550                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4551         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4552         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4553                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4554         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4555                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4556         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4557         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4558         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4559         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4560         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4561         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4562         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4563         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4564 #ifdef CONFIG_IPV6_ROUTER_PREF
4565         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4566         array[DEVCONF_RTR_PROBE_INTERVAL] =
4567                 jiffies_to_msecs(cnf->rtr_probe_interval);
4568 #ifdef CONFIG_IPV6_ROUTE_INFO
4569         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4570 #endif
4571 #endif
4572         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4573         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4574 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4575         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4576         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4577 #endif
4578 #ifdef CONFIG_IPV6_MROUTE
4579         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4580 #endif
4581         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4582         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4583         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4584         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4585         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4586         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4587         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4588         /* we omit DEVCONF_STABLE_SECRET for now */
4589 }
4590
4591 static inline size_t inet6_ifla6_size(void)
4592 {
4593         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4594              + nla_total_size(sizeof(struct ifla_cacheinfo))
4595              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4596              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4597              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4598              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4599 }
4600
4601 static inline size_t inet6_if_nlmsg_size(void)
4602 {
4603         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4604                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4605                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4606                + nla_total_size(4) /* IFLA_MTU */
4607                + nla_total_size(4) /* IFLA_LINK */
4608                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4609 }
4610
4611 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4612                                       int items, int bytes)
4613 {
4614         int i;
4615         int pad = bytes - sizeof(u64) * items;
4616         BUG_ON(pad < 0);
4617
4618         /* Use put_unaligned() because stats may not be aligned for u64. */
4619         put_unaligned(items, &stats[0]);
4620         for (i = 1; i < items; i++)
4621                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4622
4623         memset(&stats[items], 0, pad);
4624 }
4625
4626 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4627                                       int items, int bytes, size_t syncpoff)
4628 {
4629         int i;
4630         int pad = bytes - sizeof(u64) * items;
4631         BUG_ON(pad < 0);
4632
4633         /* Use put_unaligned() because stats may not be aligned for u64. */
4634         put_unaligned(items, &stats[0]);
4635         for (i = 1; i < items; i++)
4636                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4637
4638         memset(&stats[items], 0, pad);
4639 }
4640
4641 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4642                              int bytes)
4643 {
4644         switch (attrtype) {
4645         case IFLA_INET6_STATS:
4646                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4647                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4648                 break;
4649         case IFLA_INET6_ICMP6STATS:
4650                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4651                 break;
4652         }
4653 }
4654
4655 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4656 {
4657         struct nlattr *nla;
4658         struct ifla_cacheinfo ci;
4659
4660         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4661                 goto nla_put_failure;
4662         ci.max_reasm_len = IPV6_MAXPLEN;
4663         ci.tstamp = cstamp_delta(idev->tstamp);
4664         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4665         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4666         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4667                 goto nla_put_failure;
4668         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4669         if (!nla)
4670                 goto nla_put_failure;
4671         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4672
4673         /* XXX - MC not implemented */
4674
4675         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4676         if (!nla)
4677                 goto nla_put_failure;
4678         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4679
4680         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4681         if (!nla)
4682                 goto nla_put_failure;
4683         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4684
4685         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4686         if (!nla)
4687                 goto nla_put_failure;
4688
4689         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4690                 goto nla_put_failure;
4691
4692         read_lock_bh(&idev->lock);
4693         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4694         read_unlock_bh(&idev->lock);
4695
4696         return 0;
4697
4698 nla_put_failure:
4699         return -EMSGSIZE;
4700 }
4701
4702 static size_t inet6_get_link_af_size(const struct net_device *dev)
4703 {
4704         if (!__in6_dev_get(dev))
4705                 return 0;
4706
4707         return inet6_ifla6_size();
4708 }
4709
4710 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4711 {
4712         struct inet6_dev *idev = __in6_dev_get(dev);
4713
4714         if (!idev)
4715                 return -ENODATA;
4716
4717         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4718                 return -EMSGSIZE;
4719
4720         return 0;
4721 }
4722
4723 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4724 {
4725         struct inet6_ifaddr *ifp;
4726         struct net_device *dev = idev->dev;
4727         bool update_rs = false;
4728         struct in6_addr ll_addr;
4729
4730         ASSERT_RTNL();
4731
4732         if (!token)
4733                 return -EINVAL;
4734         if (ipv6_addr_any(token))
4735                 return -EINVAL;
4736         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4737                 return -EINVAL;
4738         if (!ipv6_accept_ra(idev))
4739                 return -EINVAL;
4740         if (idev->cnf.rtr_solicits <= 0)
4741                 return -EINVAL;
4742
4743         write_lock_bh(&idev->lock);
4744
4745         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4746         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4747
4748         write_unlock_bh(&idev->lock);
4749
4750         if (!idev->dead && (idev->if_flags & IF_READY) &&
4751             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4752                              IFA_F_OPTIMISTIC)) {
4753
4754                 /* If we're not ready, then normal ifup will take care
4755                  * of this. Otherwise, we need to request our rs here.
4756                  */
4757                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4758                 update_rs = true;
4759         }
4760
4761         write_lock_bh(&idev->lock);
4762
4763         if (update_rs) {
4764                 idev->if_flags |= IF_RS_SENT;
4765                 idev->rs_probes = 1;
4766                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4767         }
4768
4769         /* Well, that's kinda nasty ... */
4770         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4771                 spin_lock(&ifp->lock);
4772                 if (ifp->tokenized) {
4773                         ifp->valid_lft = 0;
4774                         ifp->prefered_lft = 0;
4775                 }
4776                 spin_unlock(&ifp->lock);
4777         }
4778
4779         write_unlock_bh(&idev->lock);
4780         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4781         addrconf_verify_rtnl();
4782         return 0;
4783 }
4784
4785 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4786         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4787         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4788 };
4789
4790 static int inet6_validate_link_af(const struct net_device *dev,
4791                                   const struct nlattr *nla)
4792 {
4793         struct nlattr *tb[IFLA_INET6_MAX + 1];
4794
4795         if (dev && !__in6_dev_get(dev))
4796                 return -EAFNOSUPPORT;
4797
4798         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4799 }
4800
4801 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4802 {
4803         int err = -EINVAL;
4804         struct inet6_dev *idev = __in6_dev_get(dev);
4805         struct nlattr *tb[IFLA_INET6_MAX + 1];
4806
4807         if (!idev)
4808                 return -EAFNOSUPPORT;
4809
4810         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4811                 BUG();
4812
4813         if (tb[IFLA_INET6_TOKEN]) {
4814                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4815                 if (err)
4816                         return err;
4817         }
4818
4819         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4820                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4821
4822                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4823                     mode != IN6_ADDR_GEN_MODE_NONE &&
4824                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4825                         return -EINVAL;
4826
4827                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4828                     !idev->cnf.stable_secret.initialized &&
4829                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4830                         return -EINVAL;
4831
4832                 idev->addr_gen_mode = mode;
4833                 err = 0;
4834         }
4835
4836         return err;
4837 }
4838
4839 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4840                              u32 portid, u32 seq, int event, unsigned int flags)
4841 {
4842         struct net_device *dev = idev->dev;
4843         struct ifinfomsg *hdr;
4844         struct nlmsghdr *nlh;
4845         void *protoinfo;
4846
4847         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4848         if (!nlh)
4849                 return -EMSGSIZE;
4850
4851         hdr = nlmsg_data(nlh);
4852         hdr->ifi_family = AF_INET6;
4853         hdr->__ifi_pad = 0;
4854         hdr->ifi_type = dev->type;
4855         hdr->ifi_index = dev->ifindex;
4856         hdr->ifi_flags = dev_get_flags(dev);
4857         hdr->ifi_change = 0;
4858
4859         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4860             (dev->addr_len &&
4861              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4862             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4863             (dev->ifindex != dev_get_iflink(dev) &&
4864              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4865                 goto nla_put_failure;
4866         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4867         if (!protoinfo)
4868                 goto nla_put_failure;
4869
4870         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4871                 goto nla_put_failure;
4872
4873         nla_nest_end(skb, protoinfo);
4874         nlmsg_end(skb, nlh);
4875         return 0;
4876
4877 nla_put_failure:
4878         nlmsg_cancel(skb, nlh);
4879         return -EMSGSIZE;
4880 }
4881
4882 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4883 {
4884         struct net *net = sock_net(skb->sk);
4885         int h, s_h;
4886         int idx = 0, s_idx;
4887         struct net_device *dev;
4888         struct inet6_dev *idev;
4889         struct hlist_head *head;
4890
4891         s_h = cb->args[0];
4892         s_idx = cb->args[1];
4893
4894         rcu_read_lock();
4895         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4896                 idx = 0;
4897                 head = &net->dev_index_head[h];
4898                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4899                         if (idx < s_idx)
4900                                 goto cont;
4901                         idev = __in6_dev_get(dev);
4902                         if (!idev)
4903                                 goto cont;
4904                         if (inet6_fill_ifinfo(skb, idev,
4905                                               NETLINK_CB(cb->skb).portid,
4906                                               cb->nlh->nlmsg_seq,
4907                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
4908                                 goto out;
4909 cont:
4910                         idx++;
4911                 }
4912         }
4913 out:
4914         rcu_read_unlock();
4915         cb->args[1] = idx;
4916         cb->args[0] = h;
4917
4918         return skb->len;
4919 }
4920
4921 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4922 {
4923         struct sk_buff *skb;
4924         struct net *net = dev_net(idev->dev);
4925         int err = -ENOBUFS;
4926
4927         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4928         if (!skb)
4929                 goto errout;
4930
4931         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4932         if (err < 0) {
4933                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4934                 WARN_ON(err == -EMSGSIZE);
4935                 kfree_skb(skb);
4936                 goto errout;
4937         }
4938         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4939         return;
4940 errout:
4941         if (err < 0)
4942                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4943 }
4944
4945 static inline size_t inet6_prefix_nlmsg_size(void)
4946 {
4947         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4948                + nla_total_size(sizeof(struct in6_addr))
4949                + nla_total_size(sizeof(struct prefix_cacheinfo));
4950 }
4951
4952 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4953                              struct prefix_info *pinfo, u32 portid, u32 seq,
4954                              int event, unsigned int flags)
4955 {
4956         struct prefixmsg *pmsg;
4957         struct nlmsghdr *nlh;
4958         struct prefix_cacheinfo ci;
4959
4960         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4961         if (!nlh)
4962                 return -EMSGSIZE;
4963
4964         pmsg = nlmsg_data(nlh);
4965         pmsg->prefix_family = AF_INET6;
4966         pmsg->prefix_pad1 = 0;
4967         pmsg->prefix_pad2 = 0;
4968         pmsg->prefix_ifindex = idev->dev->ifindex;
4969         pmsg->prefix_len = pinfo->prefix_len;
4970         pmsg->prefix_type = pinfo->type;
4971         pmsg->prefix_pad3 = 0;
4972         pmsg->prefix_flags = 0;
4973         if (pinfo->onlink)
4974                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4975         if (pinfo->autoconf)
4976                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4977
4978         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4979                 goto nla_put_failure;
4980         ci.preferred_time = ntohl(pinfo->prefered);
4981         ci.valid_time = ntohl(pinfo->valid);
4982         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4983                 goto nla_put_failure;
4984         nlmsg_end(skb, nlh);
4985         return 0;
4986
4987 nla_put_failure:
4988         nlmsg_cancel(skb, nlh);
4989         return -EMSGSIZE;
4990 }
4991
4992 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4993                          struct prefix_info *pinfo)
4994 {
4995         struct sk_buff *skb;
4996         struct net *net = dev_net(idev->dev);
4997         int err = -ENOBUFS;
4998
4999         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5000         if (!skb)
5001                 goto errout;
5002
5003         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5004         if (err < 0) {
5005                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5006                 WARN_ON(err == -EMSGSIZE);
5007                 kfree_skb(skb);
5008                 goto errout;
5009         }
5010         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5011         return;
5012 errout:
5013         if (err < 0)
5014                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5015 }
5016
5017 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5018 {
5019         struct net *net = dev_net(ifp->idev->dev);
5020
5021         if (event)
5022                 ASSERT_RTNL();
5023
5024         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5025
5026         switch (event) {
5027         case RTM_NEWADDR:
5028                 /*
5029                  * If the address was optimistic
5030                  * we inserted the route at the start of
5031                  * our DAD process, so we don't need
5032                  * to do it again
5033                  */
5034                 if (!(ifp->rt->rt6i_node))
5035                         ip6_ins_rt(ifp->rt);
5036                 if (ifp->idev->cnf.forwarding)
5037                         addrconf_join_anycast(ifp);
5038                 if (!ipv6_addr_any(&ifp->peer_addr))
5039                         addrconf_prefix_route(&ifp->peer_addr, 128,
5040                                               ifp->idev->dev, 0, 0);
5041                 break;
5042         case RTM_DELADDR:
5043                 if (ifp->idev->cnf.forwarding)
5044                         addrconf_leave_anycast(ifp);
5045                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5046                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5047                         struct rt6_info *rt;
5048
5049                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5050                                                        ifp->idev->dev, 0, 0);
5051                         if (rt && ip6_del_rt(rt))
5052                                 dst_free(&rt->dst);
5053                 }
5054                 dst_hold(&ifp->rt->dst);
5055
5056                 if (ip6_del_rt(ifp->rt))
5057                         dst_free(&ifp->rt->dst);
5058
5059                 rt_genid_bump_ipv6(net);
5060                 break;
5061         }
5062         atomic_inc(&net->ipv6.dev_addr_genid);
5063 }
5064
5065 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5066 {
5067         rcu_read_lock_bh();
5068         if (likely(ifp->idev->dead == 0))
5069                 __ipv6_ifa_notify(event, ifp);
5070         rcu_read_unlock_bh();
5071 }
5072
5073 #ifdef CONFIG_SYSCTL
5074
5075 static
5076 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5077                            void __user *buffer, size_t *lenp, loff_t *ppos)
5078 {
5079         int *valp = ctl->data;
5080         int val = *valp;
5081         loff_t pos = *ppos;
5082         struct ctl_table lctl;
5083         int ret;
5084
5085         /*
5086          * ctl->data points to idev->cnf.forwarding, we should
5087          * not modify it until we get the rtnl lock.
5088          */
5089         lctl = *ctl;
5090         lctl.data = &val;
5091
5092         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5093
5094         if (write)
5095                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5096         if (ret)
5097                 *ppos = pos;
5098         return ret;
5099 }
5100
5101 static
5102 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5103                         void __user *buffer, size_t *lenp, loff_t *ppos)
5104 {
5105         struct inet6_dev *idev = ctl->extra1;
5106         int min_mtu = IPV6_MIN_MTU;
5107         struct ctl_table lctl;
5108
5109         lctl = *ctl;
5110         lctl.extra1 = &min_mtu;
5111         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5112
5113         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5114 }
5115
5116 static void dev_disable_change(struct inet6_dev *idev)
5117 {
5118         struct netdev_notifier_info info;
5119
5120         if (!idev || !idev->dev)
5121                 return;
5122
5123         netdev_notifier_info_init(&info, idev->dev);
5124         if (idev->cnf.disable_ipv6)
5125                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5126         else
5127                 addrconf_notify(NULL, NETDEV_UP, &info);
5128 }
5129
5130 static void addrconf_disable_change(struct net *net, __s32 newf)
5131 {
5132         struct net_device *dev;
5133         struct inet6_dev *idev;
5134
5135         rcu_read_lock();
5136         for_each_netdev_rcu(net, dev) {
5137                 idev = __in6_dev_get(dev);
5138                 if (idev) {
5139                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5140                         idev->cnf.disable_ipv6 = newf;
5141                         if (changed)
5142                                 dev_disable_change(idev);
5143                 }
5144         }
5145         rcu_read_unlock();
5146 }
5147
5148 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5149 {
5150         struct net *net;
5151         int old;
5152
5153         if (!rtnl_trylock())
5154                 return restart_syscall();
5155
5156         net = (struct net *)table->extra2;
5157         old = *p;
5158         *p = newf;
5159
5160         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5161                 rtnl_unlock();
5162                 return 0;
5163         }
5164
5165         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5166                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5167                 addrconf_disable_change(net, newf);
5168         } else if ((!newf) ^ (!old))
5169                 dev_disable_change((struct inet6_dev *)table->extra1);
5170
5171         rtnl_unlock();
5172         return 0;
5173 }
5174
5175 static
5176 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5177                             void __user *buffer, size_t *lenp, loff_t *ppos)
5178 {
5179         int *valp = ctl->data;
5180         int val = *valp;
5181         loff_t pos = *ppos;
5182         struct ctl_table lctl;
5183         int ret;
5184
5185         /*
5186          * ctl->data points to idev->cnf.disable_ipv6, we should
5187          * not modify it until we get the rtnl lock.
5188          */
5189         lctl = *ctl;
5190         lctl.data = &val;
5191
5192         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5193
5194         if (write)
5195                 ret = addrconf_disable_ipv6(ctl, valp, val);
5196         if (ret)
5197                 *ppos = pos;
5198         return ret;
5199 }
5200
5201 static
5202 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5203                               void __user *buffer, size_t *lenp, loff_t *ppos)
5204 {
5205         int *valp = ctl->data;
5206         int ret;
5207         int old, new;
5208
5209         old = *valp;
5210         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5211         new = *valp;
5212
5213         if (write && old != new) {
5214                 struct net *net = ctl->extra2;
5215
5216                 if (!rtnl_trylock())
5217                         return restart_syscall();
5218
5219                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5220                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5221                                                      NETCONFA_IFINDEX_DEFAULT,
5222                                                      net->ipv6.devconf_dflt);
5223                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5224                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5225                                                      NETCONFA_IFINDEX_ALL,
5226                                                      net->ipv6.devconf_all);
5227                 else {
5228                         struct inet6_dev *idev = ctl->extra1;
5229
5230                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5231                                                      idev->dev->ifindex,
5232                                                      &idev->cnf);
5233                 }
5234                 rtnl_unlock();
5235         }
5236
5237         return ret;
5238 }
5239
5240 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5241                                          void __user *buffer, size_t *lenp,
5242                                          loff_t *ppos)
5243 {
5244         int err;
5245         struct in6_addr addr;
5246         char str[IPV6_MAX_STRLEN];
5247         struct ctl_table lctl = *ctl;
5248         struct net *net = ctl->extra2;
5249         struct ipv6_stable_secret *secret = ctl->data;
5250
5251         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5252                 return -EIO;
5253
5254         lctl.maxlen = IPV6_MAX_STRLEN;
5255         lctl.data = str;
5256
5257         if (!rtnl_trylock())
5258                 return restart_syscall();
5259
5260         if (!write && !secret->initialized) {
5261                 err = -EIO;
5262                 goto out;
5263         }
5264
5265         if (!write) {
5266                 err = snprintf(str, sizeof(str), "%pI6",
5267                                &secret->secret);
5268                 if (err >= sizeof(str)) {
5269                         err = -EIO;
5270                         goto out;
5271                 }
5272         }
5273
5274         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5275         if (err || !write)
5276                 goto out;
5277
5278         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5279                 err = -EIO;
5280                 goto out;
5281         }
5282
5283         secret->initialized = true;
5284         secret->secret = addr;
5285
5286         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5287                 struct net_device *dev;
5288
5289                 for_each_netdev(net, dev) {
5290                         struct inet6_dev *idev = __in6_dev_get(dev);
5291
5292                         if (idev) {
5293                                 idev->addr_gen_mode =
5294                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5295                         }
5296                 }
5297         } else {
5298                 struct inet6_dev *idev = ctl->extra1;
5299
5300                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5301         }
5302
5303 out:
5304         rtnl_unlock();
5305
5306         return err;
5307 }
5308
5309 static struct addrconf_sysctl_table
5310 {
5311         struct ctl_table_header *sysctl_header;
5312         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5313 } addrconf_sysctl __read_mostly = {
5314         .sysctl_header = NULL,
5315         .addrconf_vars = {
5316                 {
5317                         .procname       = "forwarding",
5318                         .data           = &ipv6_devconf.forwarding,
5319                         .maxlen         = sizeof(int),
5320                         .mode           = 0644,
5321                         .proc_handler   = addrconf_sysctl_forward,
5322                 },
5323                 {
5324                         .procname       = "hop_limit",
5325                         .data           = &ipv6_devconf.hop_limit,
5326                         .maxlen         = sizeof(int),
5327                         .mode           = 0644,
5328                         .proc_handler   = proc_dointvec,
5329                 },
5330                 {
5331                         .procname       = "mtu",
5332                         .data           = &ipv6_devconf.mtu6,
5333                         .maxlen         = sizeof(int),
5334                         .mode           = 0644,
5335                         .proc_handler   = addrconf_sysctl_mtu,
5336                 },
5337                 {
5338                         .procname       = "accept_ra",
5339                         .data           = &ipv6_devconf.accept_ra,
5340                         .maxlen         = sizeof(int),
5341                         .mode           = 0644,
5342                         .proc_handler   = proc_dointvec,
5343                 },
5344                 {
5345                         .procname       = "accept_redirects",
5346                         .data           = &ipv6_devconf.accept_redirects,
5347                         .maxlen         = sizeof(int),
5348                         .mode           = 0644,
5349                         .proc_handler   = proc_dointvec,
5350                 },
5351                 {
5352                         .procname       = "autoconf",
5353                         .data           = &ipv6_devconf.autoconf,
5354                         .maxlen         = sizeof(int),
5355                         .mode           = 0644,
5356                         .proc_handler   = proc_dointvec,
5357                 },
5358                 {
5359                         .procname       = "dad_transmits",
5360                         .data           = &ipv6_devconf.dad_transmits,
5361                         .maxlen         = sizeof(int),
5362                         .mode           = 0644,
5363                         .proc_handler   = proc_dointvec,
5364                 },
5365                 {
5366                         .procname       = "router_solicitations",
5367                         .data           = &ipv6_devconf.rtr_solicits,
5368                         .maxlen         = sizeof(int),
5369                         .mode           = 0644,
5370                         .proc_handler   = proc_dointvec,
5371                 },
5372                 {
5373                         .procname       = "router_solicitation_interval",
5374                         .data           = &ipv6_devconf.rtr_solicit_interval,
5375                         .maxlen         = sizeof(int),
5376                         .mode           = 0644,
5377                         .proc_handler   = proc_dointvec_jiffies,
5378                 },
5379                 {
5380                         .procname       = "router_solicitation_delay",
5381                         .data           = &ipv6_devconf.rtr_solicit_delay,
5382                         .maxlen         = sizeof(int),
5383                         .mode           = 0644,
5384                         .proc_handler   = proc_dointvec_jiffies,
5385                 },
5386                 {
5387                         .procname       = "force_mld_version",
5388                         .data           = &ipv6_devconf.force_mld_version,
5389                         .maxlen         = sizeof(int),
5390                         .mode           = 0644,
5391                         .proc_handler   = proc_dointvec,
5392                 },
5393                 {
5394                         .procname       = "mldv1_unsolicited_report_interval",
5395                         .data           =
5396                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5397                         .maxlen         = sizeof(int),
5398                         .mode           = 0644,
5399                         .proc_handler   = proc_dointvec_ms_jiffies,
5400                 },
5401                 {
5402                         .procname       = "mldv2_unsolicited_report_interval",
5403                         .data           =
5404                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5405                         .maxlen         = sizeof(int),
5406                         .mode           = 0644,
5407                         .proc_handler   = proc_dointvec_ms_jiffies,
5408                 },
5409                 {
5410                         .procname       = "use_tempaddr",
5411                         .data           = &ipv6_devconf.use_tempaddr,
5412                         .maxlen         = sizeof(int),
5413                         .mode           = 0644,
5414                         .proc_handler   = proc_dointvec,
5415                 },
5416                 {
5417                         .procname       = "temp_valid_lft",
5418                         .data           = &ipv6_devconf.temp_valid_lft,
5419                         .maxlen         = sizeof(int),
5420                         .mode           = 0644,
5421                         .proc_handler   = proc_dointvec,
5422                 },
5423                 {
5424                         .procname       = "temp_prefered_lft",
5425                         .data           = &ipv6_devconf.temp_prefered_lft,
5426                         .maxlen         = sizeof(int),
5427                         .mode           = 0644,
5428                         .proc_handler   = proc_dointvec,
5429                 },
5430                 {
5431                         .procname       = "regen_max_retry",
5432                         .data           = &ipv6_devconf.regen_max_retry,
5433                         .maxlen         = sizeof(int),
5434                         .mode           = 0644,
5435                         .proc_handler   = proc_dointvec,
5436                 },
5437                 {
5438                         .procname       = "max_desync_factor",
5439                         .data           = &ipv6_devconf.max_desync_factor,
5440                         .maxlen         = sizeof(int),
5441                         .mode           = 0644,
5442                         .proc_handler   = proc_dointvec,
5443                 },
5444                 {
5445                         .procname       = "max_addresses",
5446                         .data           = &ipv6_devconf.max_addresses,
5447                         .maxlen         = sizeof(int),
5448                         .mode           = 0644,
5449                         .proc_handler   = proc_dointvec,
5450                 },
5451                 {
5452                         .procname       = "accept_ra_defrtr",
5453                         .data           = &ipv6_devconf.accept_ra_defrtr,
5454                         .maxlen         = sizeof(int),
5455                         .mode           = 0644,
5456                         .proc_handler   = proc_dointvec,
5457                 },
5458                 {
5459                         .procname       = "accept_ra_pinfo",
5460                         .data           = &ipv6_devconf.accept_ra_pinfo,
5461                         .maxlen         = sizeof(int),
5462                         .mode           = 0644,
5463                         .proc_handler   = proc_dointvec,
5464                 },
5465 #ifdef CONFIG_IPV6_ROUTER_PREF
5466                 {
5467                         .procname       = "accept_ra_rtr_pref",
5468                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5469                         .maxlen         = sizeof(int),
5470                         .mode           = 0644,
5471                         .proc_handler   = proc_dointvec,
5472                 },
5473                 {
5474                         .procname       = "router_probe_interval",
5475                         .data           = &ipv6_devconf.rtr_probe_interval,
5476                         .maxlen         = sizeof(int),
5477                         .mode           = 0644,
5478                         .proc_handler   = proc_dointvec_jiffies,
5479                 },
5480 #ifdef CONFIG_IPV6_ROUTE_INFO
5481                 {
5482                         .procname       = "accept_ra_rt_info_max_plen",
5483                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5484                         .maxlen         = sizeof(int),
5485                         .mode           = 0644,
5486                         .proc_handler   = proc_dointvec,
5487                 },
5488 #endif
5489 #endif
5490                 {
5491                         .procname       = "proxy_ndp",
5492                         .data           = &ipv6_devconf.proxy_ndp,
5493                         .maxlen         = sizeof(int),
5494                         .mode           = 0644,
5495                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5496                 },
5497                 {
5498                         .procname       = "accept_source_route",
5499                         .data           = &ipv6_devconf.accept_source_route,
5500                         .maxlen         = sizeof(int),
5501                         .mode           = 0644,
5502                         .proc_handler   = proc_dointvec,
5503                 },
5504 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5505                 {
5506                         .procname       = "optimistic_dad",
5507                         .data           = &ipv6_devconf.optimistic_dad,
5508                         .maxlen         = sizeof(int),
5509                         .mode           = 0644,
5510                         .proc_handler   = proc_dointvec,
5511
5512                 },
5513                 {
5514                         .procname       = "use_optimistic",
5515                         .data           = &ipv6_devconf.use_optimistic,
5516                         .maxlen         = sizeof(int),
5517                         .mode           = 0644,
5518                         .proc_handler   = proc_dointvec,
5519
5520                 },
5521 #endif
5522 #ifdef CONFIG_IPV6_MROUTE
5523                 {
5524                         .procname       = "mc_forwarding",
5525                         .data           = &ipv6_devconf.mc_forwarding,
5526                         .maxlen         = sizeof(int),
5527                         .mode           = 0444,
5528                         .proc_handler   = proc_dointvec,
5529                 },
5530 #endif
5531                 {
5532                         .procname       = "disable_ipv6",
5533                         .data           = &ipv6_devconf.disable_ipv6,
5534                         .maxlen         = sizeof(int),
5535                         .mode           = 0644,
5536                         .proc_handler   = addrconf_sysctl_disable,
5537                 },
5538                 {
5539                         .procname       = "accept_dad",
5540                         .data           = &ipv6_devconf.accept_dad,
5541                         .maxlen         = sizeof(int),
5542                         .mode           = 0644,
5543                         .proc_handler   = proc_dointvec,
5544                 },
5545                 {
5546                         .procname       = "force_tllao",
5547                         .data           = &ipv6_devconf.force_tllao,
5548                         .maxlen         = sizeof(int),
5549                         .mode           = 0644,
5550                         .proc_handler   = proc_dointvec
5551                 },
5552                 {
5553                         .procname       = "ndisc_notify",
5554                         .data           = &ipv6_devconf.ndisc_notify,
5555                         .maxlen         = sizeof(int),
5556                         .mode           = 0644,
5557                         .proc_handler   = proc_dointvec
5558                 },
5559                 {
5560                         .procname       = "suppress_frag_ndisc",
5561                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5562                         .maxlen         = sizeof(int),
5563                         .mode           = 0644,
5564                         .proc_handler   = proc_dointvec
5565                 },
5566                 {
5567                         .procname       = "accept_ra_from_local",
5568                         .data           = &ipv6_devconf.accept_ra_from_local,
5569                         .maxlen         = sizeof(int),
5570                         .mode           = 0644,
5571                         .proc_handler   = proc_dointvec,
5572                 },
5573                 {
5574                         .procname       = "accept_ra_mtu",
5575                         .data           = &ipv6_devconf.accept_ra_mtu,
5576                         .maxlen         = sizeof(int),
5577                         .mode           = 0644,
5578                         .proc_handler   = proc_dointvec,
5579                 },
5580                 {
5581                         .procname       = "stable_secret",
5582                         .data           = &ipv6_devconf.stable_secret,
5583                         .maxlen         = IPV6_MAX_STRLEN,
5584                         .mode           = 0600,
5585                         .proc_handler   = addrconf_sysctl_stable_secret,
5586                 },
5587                 {
5588                         /* sentinel */
5589                 }
5590         },
5591 };
5592
5593 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5594                 struct inet6_dev *idev, struct ipv6_devconf *p)
5595 {
5596         int i;
5597         struct addrconf_sysctl_table *t;
5598         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5599
5600         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5601         if (!t)
5602                 goto out;
5603
5604         for (i = 0; t->addrconf_vars[i].data; i++) {
5605                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5606                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5607                 t->addrconf_vars[i].extra2 = net;
5608         }
5609
5610         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5611
5612         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5613         if (!t->sysctl_header)
5614                 goto free;
5615
5616         p->sysctl = t;
5617         return 0;
5618
5619 free:
5620         kfree(t);
5621 out:
5622         return -ENOBUFS;
5623 }
5624
5625 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5626 {
5627         struct addrconf_sysctl_table *t;
5628
5629         if (!p->sysctl)
5630                 return;
5631
5632         t = p->sysctl;
5633         p->sysctl = NULL;
5634         unregister_net_sysctl_table(t->sysctl_header);
5635         kfree(t);
5636 }
5637
5638 static int addrconf_sysctl_register(struct inet6_dev *idev)
5639 {
5640         int err;
5641
5642         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5643                 return -EINVAL;
5644
5645         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5646                                     &ndisc_ifinfo_sysctl_change);
5647         if (err)
5648                 return err;
5649         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5650                                          idev, &idev->cnf);
5651         if (err)
5652                 neigh_sysctl_unregister(idev->nd_parms);
5653
5654         return err;
5655 }
5656
5657 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5658 {
5659         __addrconf_sysctl_unregister(&idev->cnf);
5660         neigh_sysctl_unregister(idev->nd_parms);
5661 }
5662
5663
5664 #endif
5665
5666 static int __net_init addrconf_init_net(struct net *net)
5667 {
5668         int err = -ENOMEM;
5669         struct ipv6_devconf *all, *dflt;
5670
5671         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5672         if (!all)
5673                 goto err_alloc_all;
5674
5675         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5676         if (!dflt)
5677                 goto err_alloc_dflt;
5678
5679         /* these will be inherited by all namespaces */
5680         dflt->autoconf = ipv6_defaults.autoconf;
5681         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5682
5683         dflt->stable_secret.initialized = false;
5684         all->stable_secret.initialized = false;
5685
5686         net->ipv6.devconf_all = all;
5687         net->ipv6.devconf_dflt = dflt;
5688
5689 #ifdef CONFIG_SYSCTL
5690         err = __addrconf_sysctl_register(net, "all", NULL, all);
5691         if (err < 0)
5692                 goto err_reg_all;
5693
5694         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5695         if (err < 0)
5696                 goto err_reg_dflt;
5697 #endif
5698         return 0;
5699
5700 #ifdef CONFIG_SYSCTL
5701 err_reg_dflt:
5702         __addrconf_sysctl_unregister(all);
5703 err_reg_all:
5704         kfree(dflt);
5705 #endif
5706 err_alloc_dflt:
5707         kfree(all);
5708 err_alloc_all:
5709         return err;
5710 }
5711
5712 static void __net_exit addrconf_exit_net(struct net *net)
5713 {
5714 #ifdef CONFIG_SYSCTL
5715         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5716         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5717 #endif
5718         kfree(net->ipv6.devconf_dflt);
5719         kfree(net->ipv6.devconf_all);
5720 }
5721
5722 static struct pernet_operations addrconf_ops = {
5723         .init = addrconf_init_net,
5724         .exit = addrconf_exit_net,
5725 };
5726
5727 static struct rtnl_af_ops inet6_ops __read_mostly = {
5728         .family           = AF_INET6,
5729         .fill_link_af     = inet6_fill_link_af,
5730         .get_link_af_size = inet6_get_link_af_size,
5731         .validate_link_af = inet6_validate_link_af,
5732         .set_link_af      = inet6_set_link_af,
5733 };
5734
5735 /*
5736  *      Init / cleanup code
5737  */
5738
5739 int __init addrconf_init(void)
5740 {
5741         struct inet6_dev *idev;
5742         int i, err;
5743
5744         err = ipv6_addr_label_init();
5745         if (err < 0) {
5746                 pr_crit("%s: cannot initialize default policy table: %d\n",
5747                         __func__, err);
5748                 goto out;
5749         }
5750
5751         err = register_pernet_subsys(&addrconf_ops);
5752         if (err < 0)
5753                 goto out_addrlabel;
5754
5755         addrconf_wq = create_workqueue("ipv6_addrconf");
5756         if (!addrconf_wq) {
5757                 err = -ENOMEM;
5758                 goto out_nowq;
5759         }
5760
5761         /* The addrconf netdev notifier requires that loopback_dev
5762          * has it's ipv6 private information allocated and setup
5763          * before it can bring up and give link-local addresses
5764          * to other devices which are up.
5765          *
5766          * Unfortunately, loopback_dev is not necessarily the first
5767          * entry in the global dev_base list of net devices.  In fact,
5768          * it is likely to be the very last entry on that list.
5769          * So this causes the notifier registry below to try and
5770          * give link-local addresses to all devices besides loopback_dev
5771          * first, then loopback_dev, which cases all the non-loopback_dev
5772          * devices to fail to get a link-local address.
5773          *
5774          * So, as a temporary fix, allocate the ipv6 structure for
5775          * loopback_dev first by hand.
5776          * Longer term, all of the dependencies ipv6 has upon the loopback
5777          * device and it being up should be removed.
5778          */
5779         rtnl_lock();
5780         idev = ipv6_add_dev(init_net.loopback_dev);
5781         rtnl_unlock();
5782         if (IS_ERR(idev)) {
5783                 err = PTR_ERR(idev);
5784                 goto errlo;
5785         }
5786
5787         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5788                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5789
5790         register_netdevice_notifier(&ipv6_dev_notf);
5791
5792         addrconf_verify();
5793
5794         rtnl_af_register(&inet6_ops);
5795
5796         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5797                               NULL);
5798         if (err < 0)
5799                 goto errout;
5800
5801         /* Only the first call to __rtnl_register can fail */
5802         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5803         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5804         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5805                         inet6_dump_ifaddr, NULL);
5806         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5807                         inet6_dump_ifmcaddr, NULL);
5808         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5809                         inet6_dump_ifacaddr, NULL);
5810         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5811                         inet6_netconf_dump_devconf, NULL);
5812
5813         ipv6_addr_label_rtnl_register();
5814
5815         return 0;
5816 errout:
5817         rtnl_af_unregister(&inet6_ops);
5818         unregister_netdevice_notifier(&ipv6_dev_notf);
5819 errlo:
5820         destroy_workqueue(addrconf_wq);
5821 out_nowq:
5822         unregister_pernet_subsys(&addrconf_ops);
5823 out_addrlabel:
5824         ipv6_addr_label_cleanup();
5825 out:
5826         return err;
5827 }
5828
5829 void addrconf_cleanup(void)
5830 {
5831         struct net_device *dev;
5832         int i;
5833
5834         unregister_netdevice_notifier(&ipv6_dev_notf);
5835         unregister_pernet_subsys(&addrconf_ops);
5836         ipv6_addr_label_cleanup();
5837
5838         rtnl_lock();
5839
5840         __rtnl_af_unregister(&inet6_ops);
5841
5842         /* clean dev list */
5843         for_each_netdev(&init_net, dev) {
5844                 if (__in6_dev_get(dev) == NULL)
5845                         continue;
5846                 addrconf_ifdown(dev, 1);
5847         }
5848         addrconf_ifdown(init_net.loopback_dev, 2);
5849
5850         /*
5851          *      Check hash table.
5852          */
5853         spin_lock_bh(&addrconf_hash_lock);
5854         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5855                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5856         spin_unlock_bh(&addrconf_hash_lock);
5857         cancel_delayed_work(&addr_chk_work);
5858         rtnl_unlock();
5859
5860         destroy_workqueue(addrconf_wq);
5861 }