kernel/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         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2125                 if (rt->dst.dev->ifindex != dev->ifindex)
2126                         continue;
2127                 if ((rt->rt6i_flags & flags) != flags)
2128                         continue;
2129                 if ((rt->rt6i_flags & noflags) != 0)
2130                         continue;
2131                 dst_hold(&rt->dst);
2132                 break;
2133         }
2134 out:
2135         read_unlock_bh(&table->tb6_lock);
2136         return rt;
2137 }
2138
2139
2140 /* Create "default" multicast route to the interface */
2141
2142 static void addrconf_add_mroute(struct net_device *dev)
2143 {
2144         struct fib6_config cfg = {
2145                 .fc_table = RT6_TABLE_LOCAL,
2146                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2147                 .fc_ifindex = dev->ifindex,
2148                 .fc_dst_len = 8,
2149                 .fc_flags = RTF_UP,
2150                 .fc_nlinfo.nl_net = dev_net(dev),
2151         };
2152
2153         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2154
2155         ip6_route_add(&cfg);
2156 }
2157
2158 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2159 {
2160         struct inet6_dev *idev;
2161
2162         ASSERT_RTNL();
2163
2164         idev = ipv6_find_idev(dev);
2165         if (!idev)
2166                 return ERR_PTR(-ENOBUFS);
2167
2168         if (idev->cnf.disable_ipv6)
2169                 return ERR_PTR(-EACCES);
2170
2171         /* Add default multicast route */
2172         if (!(dev->flags & IFF_LOOPBACK))
2173                 addrconf_add_mroute(dev);
2174
2175         return idev;
2176 }
2177
2178 static void manage_tempaddrs(struct inet6_dev *idev,
2179                              struct inet6_ifaddr *ifp,
2180                              __u32 valid_lft, __u32 prefered_lft,
2181                              bool create, unsigned long now)
2182 {
2183         u32 flags;
2184         struct inet6_ifaddr *ift;
2185
2186         read_lock_bh(&idev->lock);
2187         /* update all temporary addresses in the list */
2188         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2189                 int age, max_valid, max_prefered;
2190
2191                 if (ifp != ift->ifpub)
2192                         continue;
2193
2194                 /* RFC 4941 section 3.3:
2195                  * If a received option will extend the lifetime of a public
2196                  * address, the lifetimes of temporary addresses should
2197                  * be extended, subject to the overall constraint that no
2198                  * temporary addresses should ever remain "valid" or "preferred"
2199                  * for a time longer than (TEMP_VALID_LIFETIME) or
2200                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2201                  */
2202                 age = (now - ift->cstamp) / HZ;
2203                 max_valid = idev->cnf.temp_valid_lft - age;
2204                 if (max_valid < 0)
2205                         max_valid = 0;
2206
2207                 max_prefered = idev->cnf.temp_prefered_lft -
2208                                idev->cnf.max_desync_factor - age;
2209                 if (max_prefered < 0)
2210                         max_prefered = 0;
2211
2212                 if (valid_lft > max_valid)
2213                         valid_lft = max_valid;
2214
2215                 if (prefered_lft > max_prefered)
2216                         prefered_lft = max_prefered;
2217
2218                 spin_lock(&ift->lock);
2219                 flags = ift->flags;
2220                 ift->valid_lft = valid_lft;
2221                 ift->prefered_lft = prefered_lft;
2222                 ift->tstamp = now;
2223                 if (prefered_lft > 0)
2224                         ift->flags &= ~IFA_F_DEPRECATED;
2225
2226                 spin_unlock(&ift->lock);
2227                 if (!(flags&IFA_F_TENTATIVE))
2228                         ipv6_ifa_notify(0, ift);
2229         }
2230
2231         if ((create || list_empty(&idev->tempaddr_list)) &&
2232             idev->cnf.use_tempaddr > 0) {
2233                 /* When a new public address is created as described
2234                  * in [ADDRCONF], also create a new temporary address.
2235                  * Also create a temporary address if it's enabled but
2236                  * no temporary address currently exists.
2237                  */
2238                 read_unlock_bh(&idev->lock);
2239                 ipv6_create_tempaddr(ifp, NULL);
2240         } else {
2241                 read_unlock_bh(&idev->lock);
2242         }
2243 }
2244
2245 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2246 {
2247         struct prefix_info *pinfo;
2248         __u32 valid_lft;
2249         __u32 prefered_lft;
2250         int addr_type;
2251         u32 addr_flags = 0;
2252         struct inet6_dev *in6_dev;
2253         struct net *net = dev_net(dev);
2254
2255         pinfo = (struct prefix_info *) opt;
2256
2257         if (len < sizeof(struct prefix_info)) {
2258                 ADBG("addrconf: prefix option too short\n");
2259                 return;
2260         }
2261
2262         /*
2263          *      Validation checks ([ADDRCONF], page 19)
2264          */
2265
2266         addr_type = ipv6_addr_type(&pinfo->prefix);
2267
2268         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2269                 return;
2270
2271         valid_lft = ntohl(pinfo->valid);
2272         prefered_lft = ntohl(pinfo->prefered);
2273
2274         if (prefered_lft > valid_lft) {
2275                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2276                 return;
2277         }
2278
2279         in6_dev = in6_dev_get(dev);
2280
2281         if (!in6_dev) {
2282                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2283                                     dev->name);
2284                 return;
2285         }
2286
2287         /*
2288          *      Two things going on here:
2289          *      1) Add routes for on-link prefixes
2290          *      2) Configure prefixes with the auto flag set
2291          */
2292
2293         if (pinfo->onlink) {
2294                 struct rt6_info *rt;
2295                 unsigned long rt_expires;
2296
2297                 /* Avoid arithmetic overflow. Really, we could
2298                  * save rt_expires in seconds, likely valid_lft,
2299                  * but it would require division in fib gc, that it
2300                  * not good.
2301                  */
2302                 if (HZ > USER_HZ)
2303                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2304                 else
2305                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2306
2307                 if (addrconf_finite_timeout(rt_expires))
2308                         rt_expires *= HZ;
2309
2310                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2311                                                pinfo->prefix_len,
2312                                                dev,
2313                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2314                                                RTF_GATEWAY | RTF_DEFAULT);
2315
2316                 if (rt) {
2317                         /* Autoconf prefix route */
2318                         if (valid_lft == 0) {
2319                                 ip6_del_rt(rt);
2320                                 rt = NULL;
2321                         } else if (addrconf_finite_timeout(rt_expires)) {
2322                                 /* not infinity */
2323                                 rt6_set_expires(rt, jiffies + rt_expires);
2324                         } else {
2325                                 rt6_clean_expires(rt);
2326                         }
2327                 } else if (valid_lft) {
2328                         clock_t expires = 0;
2329                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2330                         if (addrconf_finite_timeout(rt_expires)) {
2331                                 /* not infinity */
2332                                 flags |= RTF_EXPIRES;
2333                                 expires = jiffies_to_clock_t(rt_expires);
2334                         }
2335                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2336                                               dev, expires, flags);
2337                 }
2338                 ip6_rt_put(rt);
2339         }
2340
2341         /* Try to figure out our local address for this prefix */
2342
2343         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2344                 struct inet6_ifaddr *ifp;
2345                 struct in6_addr addr;
2346                 int create = 0, update_lft = 0;
2347                 bool tokenized = false;
2348
2349                 if (pinfo->prefix_len == 64) {
2350                         memcpy(&addr, &pinfo->prefix, 8);
2351
2352                         if (!ipv6_addr_any(&in6_dev->token)) {
2353                                 read_lock_bh(&in6_dev->lock);
2354                                 memcpy(addr.s6_addr + 8,
2355                                        in6_dev->token.s6_addr + 8, 8);
2356                                 read_unlock_bh(&in6_dev->lock);
2357                                 tokenized = true;
2358                         } else if (in6_dev->addr_gen_mode ==
2359                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2360                                    !ipv6_generate_stable_address(&addr, 0,
2361                                                                  in6_dev)) {
2362                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2363                                 goto ok;
2364                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2365                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2366                                 in6_dev_put(in6_dev);
2367                                 return;
2368                         }
2369                         goto ok;
2370                 }
2371                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2372                                     pinfo->prefix_len);
2373                 in6_dev_put(in6_dev);
2374                 return;
2375
2376 ok:
2377
2378                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2379
2380                 if (!ifp && valid_lft) {
2381                         int max_addresses = in6_dev->cnf.max_addresses;
2382
2383 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2384                         if (in6_dev->cnf.optimistic_dad &&
2385                             !net->ipv6.devconf_all->forwarding && sllao)
2386                                 addr_flags = IFA_F_OPTIMISTIC;
2387 #endif
2388
2389                         /* Do not allow to create too much of autoconfigured
2390                          * addresses; this would be too easy way to crash kernel.
2391                          */
2392                         if (!max_addresses ||
2393                             ipv6_count_addresses(in6_dev) < max_addresses)
2394                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2395                                                     pinfo->prefix_len,
2396                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2397                                                     addr_flags, valid_lft,
2398                                                     prefered_lft);
2399
2400                         if (IS_ERR_OR_NULL(ifp)) {
2401                                 in6_dev_put(in6_dev);
2402                                 return;
2403                         }
2404
2405                         update_lft = 0;
2406                         create = 1;
2407                         spin_lock_bh(&ifp->lock);
2408                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2409                         ifp->cstamp = jiffies;
2410                         ifp->tokenized = tokenized;
2411                         spin_unlock_bh(&ifp->lock);
2412                         addrconf_dad_start(ifp);
2413                 }
2414
2415                 if (ifp) {
2416                         u32 flags;
2417                         unsigned long now;
2418                         u32 stored_lft;
2419
2420                         /* update lifetime (RFC2462 5.5.3 e) */
2421                         spin_lock_bh(&ifp->lock);
2422                         now = jiffies;
2423                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2424                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2425                         else
2426                                 stored_lft = 0;
2427                         if (!update_lft && !create && stored_lft) {
2428                                 const u32 minimum_lft = min_t(u32,
2429                                         stored_lft, MIN_VALID_LIFETIME);
2430                                 valid_lft = max(valid_lft, minimum_lft);
2431
2432                                 /* RFC4862 Section 5.5.3e:
2433                                  * "Note that the preferred lifetime of the
2434                                  *  corresponding address is always reset to
2435                                  *  the Preferred Lifetime in the received
2436                                  *  Prefix Information option, regardless of
2437                                  *  whether the valid lifetime is also reset or
2438                                  *  ignored."
2439                                  *
2440                                  * So we should always update prefered_lft here.
2441                                  */
2442                                 update_lft = 1;
2443                         }
2444
2445                         if (update_lft) {
2446                                 ifp->valid_lft = valid_lft;
2447                                 ifp->prefered_lft = prefered_lft;
2448                                 ifp->tstamp = now;
2449                                 flags = ifp->flags;
2450                                 ifp->flags &= ~IFA_F_DEPRECATED;
2451                                 spin_unlock_bh(&ifp->lock);
2452
2453                                 if (!(flags&IFA_F_TENTATIVE))
2454                                         ipv6_ifa_notify(0, ifp);
2455                         } else
2456                                 spin_unlock_bh(&ifp->lock);
2457
2458                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2459                                          create, now);
2460
2461                         in6_ifa_put(ifp);
2462                         addrconf_verify();
2463                 }
2464         }
2465         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2466         in6_dev_put(in6_dev);
2467 }
2468
2469 /*
2470  *      Set destination address.
2471  *      Special case for SIT interfaces where we create a new "virtual"
2472  *      device.
2473  */
2474 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2475 {
2476         struct in6_ifreq ireq;
2477         struct net_device *dev;
2478         int err = -EINVAL;
2479
2480         rtnl_lock();
2481
2482         err = -EFAULT;
2483         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2484                 goto err_exit;
2485
2486         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2487
2488         err = -ENODEV;
2489         if (!dev)
2490                 goto err_exit;
2491
2492 #if IS_ENABLED(CONFIG_IPV6_SIT)
2493         if (dev->type == ARPHRD_SIT) {
2494                 const struct net_device_ops *ops = dev->netdev_ops;
2495                 struct ifreq ifr;
2496                 struct ip_tunnel_parm p;
2497
2498                 err = -EADDRNOTAVAIL;
2499                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2500                         goto err_exit;
2501
2502                 memset(&p, 0, sizeof(p));
2503                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2504                 p.iph.saddr = 0;
2505                 p.iph.version = 4;
2506                 p.iph.ihl = 5;
2507                 p.iph.protocol = IPPROTO_IPV6;
2508                 p.iph.ttl = 64;
2509                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2510
2511                 if (ops->ndo_do_ioctl) {
2512                         mm_segment_t oldfs = get_fs();
2513
2514                         set_fs(KERNEL_DS);
2515                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2516                         set_fs(oldfs);
2517                 } else
2518                         err = -EOPNOTSUPP;
2519
2520                 if (err == 0) {
2521                         err = -ENOBUFS;
2522                         dev = __dev_get_by_name(net, p.name);
2523                         if (!dev)
2524                                 goto err_exit;
2525                         err = dev_open(dev);
2526                 }
2527         }
2528 #endif
2529
2530 err_exit:
2531         rtnl_unlock();
2532         return err;
2533 }
2534
2535 static int ipv6_mc_config(struct sock *sk, bool join,
2536                           const struct in6_addr *addr, int ifindex)
2537 {
2538         int ret;
2539
2540         ASSERT_RTNL();
2541
2542         lock_sock(sk);
2543         if (join)
2544                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2545         else
2546                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2547         release_sock(sk);
2548
2549         return ret;
2550 }
2551
2552 /*
2553  *      Manual configuration of address on an interface
2554  */
2555 static int inet6_addr_add(struct net *net, int ifindex,
2556                           const struct in6_addr *pfx,
2557                           const struct in6_addr *peer_pfx,
2558                           unsigned int plen, __u32 ifa_flags,
2559                           __u32 prefered_lft, __u32 valid_lft)
2560 {
2561         struct inet6_ifaddr *ifp;
2562         struct inet6_dev *idev;
2563         struct net_device *dev;
2564         unsigned long timeout;
2565         clock_t expires;
2566         int scope;
2567         u32 flags;
2568
2569         ASSERT_RTNL();
2570
2571         if (plen > 128)
2572                 return -EINVAL;
2573
2574         /* check the lifetime */
2575         if (!valid_lft || prefered_lft > valid_lft)
2576                 return -EINVAL;
2577
2578         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2579                 return -EINVAL;
2580
2581         dev = __dev_get_by_index(net, ifindex);
2582         if (!dev)
2583                 return -ENODEV;
2584
2585         idev = addrconf_add_dev(dev);
2586         if (IS_ERR(idev))
2587                 return PTR_ERR(idev);
2588
2589         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2590                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2591                                          true, pfx, ifindex);
2592
2593                 if (ret < 0)
2594                         return ret;
2595         }
2596
2597         scope = ipv6_addr_scope(pfx);
2598
2599         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2600         if (addrconf_finite_timeout(timeout)) {
2601                 expires = jiffies_to_clock_t(timeout * HZ);
2602                 valid_lft = timeout;
2603                 flags = RTF_EXPIRES;
2604         } else {
2605                 expires = 0;
2606                 flags = 0;
2607                 ifa_flags |= IFA_F_PERMANENT;
2608         }
2609
2610         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2611         if (addrconf_finite_timeout(timeout)) {
2612                 if (timeout == 0)
2613                         ifa_flags |= IFA_F_DEPRECATED;
2614                 prefered_lft = timeout;
2615         }
2616
2617         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2618                             valid_lft, prefered_lft);
2619
2620         if (!IS_ERR(ifp)) {
2621                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2622                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2623                                               expires, flags);
2624                 }
2625
2626                 /*
2627                  * Note that section 3.1 of RFC 4429 indicates
2628                  * that the Optimistic flag should not be set for
2629                  * manually configured addresses
2630                  */
2631                 addrconf_dad_start(ifp);
2632                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2633                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2634                                          true, jiffies);
2635                 in6_ifa_put(ifp);
2636                 addrconf_verify_rtnl();
2637                 return 0;
2638         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2639                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2640                                false, pfx, ifindex);
2641         }
2642
2643         return PTR_ERR(ifp);
2644 }
2645
2646 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2647                           const struct in6_addr *pfx, unsigned int plen)
2648 {
2649         struct inet6_ifaddr *ifp;
2650         struct inet6_dev *idev;
2651         struct net_device *dev;
2652
2653         if (plen > 128)
2654                 return -EINVAL;
2655
2656         dev = __dev_get_by_index(net, ifindex);
2657         if (!dev)
2658                 return -ENODEV;
2659
2660         idev = __in6_dev_get(dev);
2661         if (!idev)
2662                 return -ENXIO;
2663
2664         read_lock_bh(&idev->lock);
2665         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2666                 if (ifp->prefix_len == plen &&
2667                     ipv6_addr_equal(pfx, &ifp->addr)) {
2668                         in6_ifa_hold(ifp);
2669                         read_unlock_bh(&idev->lock);
2670
2671                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2672                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2673                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2674                                                  jiffies);
2675                         ipv6_del_addr(ifp);
2676                         addrconf_verify_rtnl();
2677                         if (ipv6_addr_is_multicast(pfx)) {
2678                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2679                                                false, pfx, dev->ifindex);
2680                         }
2681                         return 0;
2682                 }
2683         }
2684         read_unlock_bh(&idev->lock);
2685         return -EADDRNOTAVAIL;
2686 }
2687
2688
2689 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2690 {
2691         struct in6_ifreq ireq;
2692         int err;
2693
2694         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2695                 return -EPERM;
2696
2697         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2698                 return -EFAULT;
2699
2700         rtnl_lock();
2701         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2702                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2703                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2704         rtnl_unlock();
2705         return err;
2706 }
2707
2708 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2709 {
2710         struct in6_ifreq ireq;
2711         int err;
2712
2713         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2714                 return -EPERM;
2715
2716         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2717                 return -EFAULT;
2718
2719         rtnl_lock();
2720         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2721                              ireq.ifr6_prefixlen);
2722         rtnl_unlock();
2723         return err;
2724 }
2725
2726 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2727                      int plen, int scope)
2728 {
2729         struct inet6_ifaddr *ifp;
2730
2731         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2732                             scope, IFA_F_PERMANENT,
2733                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2734         if (!IS_ERR(ifp)) {
2735                 spin_lock_bh(&ifp->lock);
2736                 ifp->flags &= ~IFA_F_TENTATIVE;
2737                 spin_unlock_bh(&ifp->lock);
2738                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2739                 in6_ifa_put(ifp);
2740         }
2741 }
2742
2743 #if IS_ENABLED(CONFIG_IPV6_SIT)
2744 static void sit_add_v4_addrs(struct inet6_dev *idev)
2745 {
2746         struct in6_addr addr;
2747         struct net_device *dev;
2748         struct net *net = dev_net(idev->dev);
2749         int scope, plen;
2750         u32 pflags = 0;
2751
2752         ASSERT_RTNL();
2753
2754         memset(&addr, 0, sizeof(struct in6_addr));
2755         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2756
2757         if (idev->dev->flags&IFF_POINTOPOINT) {
2758                 addr.s6_addr32[0] = htonl(0xfe800000);
2759                 scope = IFA_LINK;
2760                 plen = 64;
2761         } else {
2762                 scope = IPV6_ADDR_COMPATv4;
2763                 plen = 96;
2764                 pflags |= RTF_NONEXTHOP;
2765         }
2766
2767         if (addr.s6_addr32[3]) {
2768                 add_addr(idev, &addr, plen, scope);
2769                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2770                 return;
2771         }
2772
2773         for_each_netdev(net, dev) {
2774                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2775                 if (in_dev && (dev->flags & IFF_UP)) {
2776                         struct in_ifaddr *ifa;
2777
2778                         int flag = scope;
2779
2780                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2781
2782                                 addr.s6_addr32[3] = ifa->ifa_local;
2783
2784                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2785                                         continue;
2786                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2787                                         if (idev->dev->flags&IFF_POINTOPOINT)
2788                                                 continue;
2789                                         flag |= IFA_HOST;
2790                                 }
2791
2792                                 add_addr(idev, &addr, plen, flag);
2793                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2794                                                       pflags);
2795                         }
2796                 }
2797         }
2798 }
2799 #endif
2800
2801 static void init_loopback(struct net_device *dev)
2802 {
2803         struct inet6_dev  *idev;
2804         struct net_device *sp_dev;
2805         struct inet6_ifaddr *sp_ifa;
2806         struct rt6_info *sp_rt;
2807
2808         /* ::1 */
2809
2810         ASSERT_RTNL();
2811
2812         idev = ipv6_find_idev(dev);
2813         if (!idev) {
2814                 pr_debug("%s: add_dev failed\n", __func__);
2815                 return;
2816         }
2817
2818         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2819
2820         /* Add routes to other interface's IPv6 addresses */
2821         for_each_netdev(dev_net(dev), sp_dev) {
2822                 if (!strcmp(sp_dev->name, dev->name))
2823                         continue;
2824
2825                 idev = __in6_dev_get(sp_dev);
2826                 if (!idev)
2827                         continue;
2828
2829                 read_lock_bh(&idev->lock);
2830                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2831
2832                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2833                                 continue;
2834
2835                         if (sp_ifa->rt) {
2836                                 /* This dst has been added to garbage list when
2837                                  * lo device down, release this obsolete dst and
2838                                  * reallocate a new router for ifa.
2839                                  */
2840                                 if (sp_ifa->rt->dst.obsolete > 0) {
2841                                         ip6_rt_put(sp_ifa->rt);
2842                                         sp_ifa->rt = NULL;
2843                                 } else {
2844                                         continue;
2845                                 }
2846                         }
2847
2848                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2849
2850                         /* Failure cases are ignored */
2851                         if (!IS_ERR(sp_rt)) {
2852                                 sp_ifa->rt = sp_rt;
2853                                 ip6_ins_rt(sp_rt);
2854                         }
2855                 }
2856                 read_unlock_bh(&idev->lock);
2857         }
2858 }
2859
2860 static void addrconf_add_linklocal(struct inet6_dev *idev,
2861                                    const struct in6_addr *addr, u32 flags)
2862 {
2863         struct inet6_ifaddr *ifp;
2864         u32 addr_flags = flags | IFA_F_PERMANENT;
2865
2866 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2867         if (idev->cnf.optimistic_dad &&
2868             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2869                 addr_flags |= IFA_F_OPTIMISTIC;
2870 #endif
2871
2872         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2873                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2874         if (!IS_ERR(ifp)) {
2875                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2876                 addrconf_dad_start(ifp);
2877                 in6_ifa_put(ifp);
2878         }
2879 }
2880
2881 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2882 {
2883         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2884                 return true;
2885
2886         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2887             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2888                 return true;
2889
2890         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2891             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2892                 return true;
2893
2894         return false;
2895 }
2896
2897 static int ipv6_generate_stable_address(struct in6_addr *address,
2898                                         u8 dad_count,
2899                                         const struct inet6_dev *idev)
2900 {
2901         static DEFINE_SPINLOCK(lock);
2902         static __u32 digest[SHA_DIGEST_WORDS];
2903         static __u32 workspace[SHA_WORKSPACE_WORDS];
2904
2905         static union {
2906                 char __data[SHA_MESSAGE_BYTES];
2907                 struct {
2908                         struct in6_addr secret;
2909                         __be32 prefix[2];
2910                         unsigned char hwaddr[MAX_ADDR_LEN];
2911                         u8 dad_count;
2912                 } __packed;
2913         } data;
2914
2915         struct in6_addr secret;
2916         struct in6_addr temp;
2917         struct net *net = dev_net(idev->dev);
2918
2919         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2920
2921         if (idev->cnf.stable_secret.initialized)
2922                 secret = idev->cnf.stable_secret.secret;
2923         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
2924                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
2925         else
2926                 return -1;
2927
2928 retry:
2929         spin_lock_bh(&lock);
2930
2931         sha_init(digest);
2932         memset(&data, 0, sizeof(data));
2933         memset(workspace, 0, sizeof(workspace));
2934         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
2935         data.prefix[0] = address->s6_addr32[0];
2936         data.prefix[1] = address->s6_addr32[1];
2937         data.secret = secret;
2938         data.dad_count = dad_count;
2939
2940         sha_transform(digest, data.__data, workspace);
2941
2942         temp = *address;
2943         temp.s6_addr32[2] = (__force __be32)digest[0];
2944         temp.s6_addr32[3] = (__force __be32)digest[1];
2945
2946         spin_unlock_bh(&lock);
2947
2948         if (ipv6_reserved_interfaceid(temp)) {
2949                 dad_count++;
2950                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
2951                         return -1;
2952                 goto retry;
2953         }
2954
2955         *address = temp;
2956         return 0;
2957 }
2958
2959 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2960 {
2961         struct in6_addr addr;
2962
2963         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2964
2965         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
2966                 if (!ipv6_generate_stable_address(&addr, 0, idev))
2967                         addrconf_add_linklocal(idev, &addr,
2968                                                IFA_F_STABLE_PRIVACY);
2969                 else if (prefix_route)
2970                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2971         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2972                 /* addrconf_add_linklocal also adds a prefix_route and we
2973                  * only need to care about prefix routes if ipv6_generate_eui64
2974                  * couldn't generate one.
2975                  */
2976                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2977                         addrconf_add_linklocal(idev, &addr, 0);
2978                 else if (prefix_route)
2979                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2980         }
2981 }
2982
2983 static void addrconf_dev_config(struct net_device *dev)
2984 {
2985         struct inet6_dev *idev;
2986
2987         ASSERT_RTNL();
2988
2989         if ((dev->type != ARPHRD_ETHER) &&
2990             (dev->type != ARPHRD_FDDI) &&
2991             (dev->type != ARPHRD_ARCNET) &&
2992             (dev->type != ARPHRD_INFINIBAND) &&
2993             (dev->type != ARPHRD_IEEE802154) &&
2994             (dev->type != ARPHRD_IEEE1394) &&
2995             (dev->type != ARPHRD_TUNNEL6) &&
2996             (dev->type != ARPHRD_6LOWPAN)) {
2997                 /* Alas, we support only Ethernet autoconfiguration. */
2998                 return;
2999         }
3000
3001         idev = addrconf_add_dev(dev);
3002         if (IS_ERR(idev))
3003                 return;
3004
3005         addrconf_addr_gen(idev, false);
3006 }
3007
3008 #if IS_ENABLED(CONFIG_IPV6_SIT)
3009 static void addrconf_sit_config(struct net_device *dev)
3010 {
3011         struct inet6_dev *idev;
3012
3013         ASSERT_RTNL();
3014
3015         /*
3016          * Configure the tunnel with one of our IPv4
3017          * addresses... we should configure all of
3018          * our v4 addrs in the tunnel
3019          */
3020
3021         idev = ipv6_find_idev(dev);
3022         if (!idev) {
3023                 pr_debug("%s: add_dev failed\n", __func__);
3024                 return;
3025         }
3026
3027         if (dev->priv_flags & IFF_ISATAP) {
3028                 addrconf_addr_gen(idev, false);
3029                 return;
3030         }
3031
3032         sit_add_v4_addrs(idev);
3033
3034         if (dev->flags&IFF_POINTOPOINT)
3035                 addrconf_add_mroute(dev);
3036 }
3037 #endif
3038
3039 #if IS_ENABLED(CONFIG_NET_IPGRE)
3040 static void addrconf_gre_config(struct net_device *dev)
3041 {
3042         struct inet6_dev *idev;
3043
3044         ASSERT_RTNL();
3045
3046         idev = ipv6_find_idev(dev);
3047         if (!idev) {
3048                 pr_debug("%s: add_dev failed\n", __func__);
3049                 return;
3050         }
3051
3052         addrconf_addr_gen(idev, true);
3053 }
3054 #endif
3055
3056 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3057                            void *ptr)
3058 {
3059         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3060         struct inet6_dev *idev = __in6_dev_get(dev);
3061         int run_pending = 0;
3062         int err;
3063
3064         switch (event) {
3065         case NETDEV_REGISTER:
3066                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3067                         idev = ipv6_add_dev(dev);
3068                         if (IS_ERR(idev))
3069                                 return notifier_from_errno(PTR_ERR(idev));
3070                 }
3071                 break;
3072
3073         case NETDEV_UP:
3074         case NETDEV_CHANGE:
3075                 if (dev->flags & IFF_SLAVE)
3076                         break;
3077
3078                 if (idev && idev->cnf.disable_ipv6)
3079                         break;
3080
3081                 if (event == NETDEV_UP) {
3082                         if (!addrconf_qdisc_ok(dev)) {
3083                                 /* device is not ready yet. */
3084                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3085                                         dev->name);
3086                                 break;
3087                         }
3088
3089                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3090                                 idev = ipv6_add_dev(dev);
3091
3092                         if (!IS_ERR_OR_NULL(idev)) {
3093                                 idev->if_flags |= IF_READY;
3094                                 run_pending = 1;
3095                         }
3096                 } else {
3097                         if (!addrconf_qdisc_ok(dev)) {
3098                                 /* device is still not ready. */
3099                                 break;
3100                         }
3101
3102                         if (idev) {
3103                                 if (idev->if_flags & IF_READY)
3104                                         /* device is already configured. */
3105                                         break;
3106                                 idev->if_flags |= IF_READY;
3107                         }
3108
3109                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3110                                 dev->name);
3111
3112                         run_pending = 1;
3113                 }
3114
3115                 switch (dev->type) {
3116 #if IS_ENABLED(CONFIG_IPV6_SIT)
3117                 case ARPHRD_SIT:
3118                         addrconf_sit_config(dev);
3119                         break;
3120 #endif
3121 #if IS_ENABLED(CONFIG_NET_IPGRE)
3122                 case ARPHRD_IPGRE:
3123                         addrconf_gre_config(dev);
3124                         break;
3125 #endif
3126                 case ARPHRD_LOOPBACK:
3127                         init_loopback(dev);
3128                         break;
3129
3130                 default:
3131                         addrconf_dev_config(dev);
3132                         break;
3133                 }
3134
3135                 if (!IS_ERR_OR_NULL(idev)) {
3136                         if (run_pending)
3137                                 addrconf_dad_run(idev);
3138
3139                         /*
3140                          * If the MTU changed during the interface down,
3141                          * when the interface up, the changed MTU must be
3142                          * reflected in the idev as well as routers.
3143                          */
3144                         if (idev->cnf.mtu6 != dev->mtu &&
3145                             dev->mtu >= IPV6_MIN_MTU) {
3146                                 rt6_mtu_change(dev, dev->mtu);
3147                                 idev->cnf.mtu6 = dev->mtu;
3148                         }
3149                         idev->tstamp = jiffies;
3150                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3151
3152                         /*
3153                          * If the changed mtu during down is lower than
3154                          * IPV6_MIN_MTU stop IPv6 on this interface.
3155                          */
3156                         if (dev->mtu < IPV6_MIN_MTU)
3157                                 addrconf_ifdown(dev, 1);
3158                 }
3159                 break;
3160
3161         case NETDEV_CHANGEMTU:
3162                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3163                         rt6_mtu_change(dev, dev->mtu);
3164                         idev->cnf.mtu6 = dev->mtu;
3165                         break;
3166                 }
3167
3168                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3169                         idev = ipv6_add_dev(dev);
3170                         if (!IS_ERR(idev))
3171                                 break;
3172                 }
3173
3174                 /*
3175                  * if MTU under IPV6_MIN_MTU.
3176                  * Stop IPv6 on this interface.
3177                  */
3178
3179         case NETDEV_DOWN:
3180         case NETDEV_UNREGISTER:
3181                 /*
3182                  *      Remove all addresses from this interface.
3183                  */
3184                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3185                 break;
3186
3187         case NETDEV_CHANGENAME:
3188                 if (idev) {
3189                         snmp6_unregister_dev(idev);
3190                         addrconf_sysctl_unregister(idev);
3191                         err = addrconf_sysctl_register(idev);
3192                         if (err)
3193                                 return notifier_from_errno(err);
3194                         err = snmp6_register_dev(idev);
3195                         if (err) {
3196                                 addrconf_sysctl_unregister(idev);
3197                                 return notifier_from_errno(err);
3198                         }
3199                 }
3200                 break;
3201
3202         case NETDEV_PRE_TYPE_CHANGE:
3203         case NETDEV_POST_TYPE_CHANGE:
3204                 addrconf_type_change(dev, event);
3205                 break;
3206         }
3207
3208         return NOTIFY_OK;
3209 }
3210
3211 /*
3212  *      addrconf module should be notified of a device going up
3213  */
3214 static struct notifier_block ipv6_dev_notf = {
3215         .notifier_call = addrconf_notify,
3216 };
3217
3218 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3219 {
3220         struct inet6_dev *idev;
3221         ASSERT_RTNL();
3222
3223         idev = __in6_dev_get(dev);
3224
3225         if (event == NETDEV_POST_TYPE_CHANGE)
3226                 ipv6_mc_remap(idev);
3227         else if (event == NETDEV_PRE_TYPE_CHANGE)
3228                 ipv6_mc_unmap(idev);
3229 }
3230
3231 static int addrconf_ifdown(struct net_device *dev, int how)
3232 {
3233         struct net *net = dev_net(dev);
3234         struct inet6_dev *idev;
3235         struct inet6_ifaddr *ifa;
3236         int state, i;
3237
3238         ASSERT_RTNL();
3239
3240         rt6_ifdown(net, dev);
3241         neigh_ifdown(&nd_tbl, dev);
3242
3243         idev = __in6_dev_get(dev);
3244         if (!idev)
3245                 return -ENODEV;
3246
3247         /*
3248          * Step 1: remove reference to ipv6 device from parent device.
3249          *         Do not dev_put!
3250          */
3251         if (how) {
3252                 idev->dead = 1;
3253
3254                 /* protected by rtnl_lock */
3255                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3256
3257                 /* Step 1.5: remove snmp6 entry */
3258                 snmp6_unregister_dev(idev);
3259
3260         }
3261
3262         /* Step 2: clear hash table */
3263         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3264                 struct hlist_head *h = &inet6_addr_lst[i];
3265
3266                 spin_lock_bh(&addrconf_hash_lock);
3267 restart:
3268                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3269                         if (ifa->idev == idev) {
3270                                 hlist_del_init_rcu(&ifa->addr_lst);
3271                                 addrconf_del_dad_work(ifa);
3272                                 goto restart;
3273                         }
3274                 }
3275                 spin_unlock_bh(&addrconf_hash_lock);
3276         }
3277
3278         write_lock_bh(&idev->lock);
3279
3280         addrconf_del_rs_timer(idev);
3281
3282         /* Step 2: clear flags for stateless addrconf */
3283         if (!how)
3284                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3285
3286         if (how && del_timer(&idev->regen_timer))
3287                 in6_dev_put(idev);
3288
3289         /* Step 3: clear tempaddr list */
3290         while (!list_empty(&idev->tempaddr_list)) {
3291                 ifa = list_first_entry(&idev->tempaddr_list,
3292                                        struct inet6_ifaddr, tmp_list);
3293                 list_del(&ifa->tmp_list);
3294                 write_unlock_bh(&idev->lock);
3295                 spin_lock_bh(&ifa->lock);
3296
3297                 if (ifa->ifpub) {
3298                         in6_ifa_put(ifa->ifpub);
3299                         ifa->ifpub = NULL;
3300                 }
3301                 spin_unlock_bh(&ifa->lock);
3302                 in6_ifa_put(ifa);
3303                 write_lock_bh(&idev->lock);
3304         }
3305
3306         while (!list_empty(&idev->addr_list)) {
3307                 ifa = list_first_entry(&idev->addr_list,
3308                                        struct inet6_ifaddr, if_list);
3309                 addrconf_del_dad_work(ifa);
3310
3311                 list_del(&ifa->if_list);
3312
3313                 write_unlock_bh(&idev->lock);
3314
3315                 spin_lock_bh(&ifa->lock);
3316                 state = ifa->state;
3317                 ifa->state = INET6_IFADDR_STATE_DEAD;
3318                 spin_unlock_bh(&ifa->lock);
3319
3320                 if (state != INET6_IFADDR_STATE_DEAD) {
3321                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3322                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3323                 }
3324                 in6_ifa_put(ifa);
3325
3326                 write_lock_bh(&idev->lock);
3327         }
3328
3329         write_unlock_bh(&idev->lock);
3330
3331         /* Step 5: Discard anycast and multicast list */
3332         if (how) {
3333                 ipv6_ac_destroy_dev(idev);
3334                 ipv6_mc_destroy_dev(idev);
3335         } else {
3336                 ipv6_mc_down(idev);
3337         }
3338
3339         idev->tstamp = jiffies;
3340
3341         /* Last: Shot the device (if unregistered) */
3342         if (how) {
3343                 addrconf_sysctl_unregister(idev);
3344                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3345                 neigh_ifdown(&nd_tbl, dev);
3346                 in6_dev_put(idev);
3347         }
3348         return 0;
3349 }
3350
3351 static void addrconf_rs_timer(unsigned long data)
3352 {
3353         struct inet6_dev *idev = (struct inet6_dev *)data;
3354         struct net_device *dev = idev->dev;
3355         struct in6_addr lladdr;
3356
3357         write_lock(&idev->lock);
3358         if (idev->dead || !(idev->if_flags & IF_READY))
3359                 goto out;
3360
3361         if (!ipv6_accept_ra(idev))
3362                 goto out;
3363
3364         /* Announcement received after solicitation was sent */
3365         if (idev->if_flags & IF_RA_RCVD)
3366                 goto out;
3367
3368         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3369                 write_unlock(&idev->lock);
3370                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3371                         ndisc_send_rs(dev, &lladdr,
3372                                       &in6addr_linklocal_allrouters);
3373                 else
3374                         goto put;
3375
3376                 write_lock(&idev->lock);
3377                 /* The wait after the last probe can be shorter */
3378                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3379                                              idev->cnf.rtr_solicits) ?
3380                                       idev->cnf.rtr_solicit_delay :
3381                                       idev->cnf.rtr_solicit_interval);
3382         } else {
3383                 /*
3384                  * Note: we do not support deprecated "all on-link"
3385                  * assumption any longer.
3386                  */
3387                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3388         }
3389
3390 out:
3391         write_unlock(&idev->lock);
3392 put:
3393         in6_dev_put(idev);
3394 }
3395
3396 /*
3397  *      Duplicate Address Detection
3398  */
3399 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3400 {
3401         unsigned long rand_num;
3402         struct inet6_dev *idev = ifp->idev;
3403
3404         if (ifp->flags & IFA_F_OPTIMISTIC)
3405                 rand_num = 0;
3406         else
3407                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3408
3409         ifp->dad_probes = idev->cnf.dad_transmits;
3410         addrconf_mod_dad_work(ifp, rand_num);
3411 }
3412
3413 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3414 {
3415         struct inet6_dev *idev = ifp->idev;
3416         struct net_device *dev = idev->dev;
3417
3418         addrconf_join_solict(dev, &ifp->addr);
3419
3420         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3421
3422         read_lock_bh(&idev->lock);
3423         spin_lock(&ifp->lock);
3424         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3425                 goto out;
3426
3427         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3428             idev->cnf.accept_dad < 1 ||
3429             !(ifp->flags&IFA_F_TENTATIVE) ||
3430             ifp->flags & IFA_F_NODAD) {
3431                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3432                 spin_unlock(&ifp->lock);
3433                 read_unlock_bh(&idev->lock);
3434
3435                 addrconf_dad_completed(ifp);
3436                 return;
3437         }
3438
3439         if (!(idev->if_flags & IF_READY)) {
3440                 spin_unlock(&ifp->lock);
3441                 read_unlock_bh(&idev->lock);
3442                 /*
3443                  * If the device is not ready:
3444                  * - keep it tentative if it is a permanent address.
3445                  * - otherwise, kill it.
3446                  */
3447                 in6_ifa_hold(ifp);
3448                 addrconf_dad_stop(ifp, 0);
3449                 return;
3450         }
3451
3452         /*
3453          * Optimistic nodes can start receiving
3454          * Frames right away
3455          */
3456         if (ifp->flags & IFA_F_OPTIMISTIC) {
3457                 ip6_ins_rt(ifp->rt);
3458                 if (ipv6_use_optimistic_addr(idev)) {
3459                         /* Because optimistic nodes can use this address,
3460                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3461                          */
3462                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3463                 }
3464         }
3465
3466         addrconf_dad_kick(ifp);
3467 out:
3468         spin_unlock(&ifp->lock);
3469         read_unlock_bh(&idev->lock);
3470 }
3471
3472 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3473 {
3474         bool begin_dad = false;
3475
3476         spin_lock_bh(&ifp->lock);
3477         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3478                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3479                 begin_dad = true;
3480         }
3481         spin_unlock_bh(&ifp->lock);
3482
3483         if (begin_dad)
3484                 addrconf_mod_dad_work(ifp, 0);
3485 }
3486
3487 static void addrconf_dad_work(struct work_struct *w)
3488 {
3489         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3490                                                 struct inet6_ifaddr,
3491                                                 dad_work);
3492         struct inet6_dev *idev = ifp->idev;
3493         struct in6_addr mcaddr;
3494
3495         enum {
3496                 DAD_PROCESS,
3497                 DAD_BEGIN,
3498                 DAD_ABORT,
3499         } action = DAD_PROCESS;
3500
3501         rtnl_lock();
3502
3503         spin_lock_bh(&ifp->lock);
3504         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3505                 action = DAD_BEGIN;
3506                 ifp->state = INET6_IFADDR_STATE_DAD;
3507         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3508                 action = DAD_ABORT;
3509                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3510         }
3511         spin_unlock_bh(&ifp->lock);
3512
3513         if (action == DAD_BEGIN) {
3514                 addrconf_dad_begin(ifp);
3515                 goto out;
3516         } else if (action == DAD_ABORT) {
3517                 addrconf_dad_stop(ifp, 1);
3518                 goto out;
3519         }
3520
3521         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3522                 goto out;
3523
3524         write_lock_bh(&idev->lock);
3525         if (idev->dead || !(idev->if_flags & IF_READY)) {
3526                 write_unlock_bh(&idev->lock);
3527                 goto out;
3528         }
3529
3530         spin_lock(&ifp->lock);
3531         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3532                 spin_unlock(&ifp->lock);
3533                 write_unlock_bh(&idev->lock);
3534                 goto out;
3535         }
3536
3537         if (ifp->dad_probes == 0) {
3538                 /*
3539                  * DAD was successful
3540                  */
3541
3542                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3543                 spin_unlock(&ifp->lock);
3544                 write_unlock_bh(&idev->lock);
3545
3546                 addrconf_dad_completed(ifp);
3547
3548                 goto out;
3549         }
3550
3551         ifp->dad_probes--;
3552         addrconf_mod_dad_work(ifp,
3553                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3554         spin_unlock(&ifp->lock);
3555         write_unlock_bh(&idev->lock);
3556
3557         /* send a neighbour solicitation for our addr */
3558         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3559         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3560 out:
3561         in6_ifa_put(ifp);
3562         rtnl_unlock();
3563 }
3564
3565 /* ifp->idev must be at least read locked */
3566 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3567 {
3568         struct inet6_ifaddr *ifpiter;
3569         struct inet6_dev *idev = ifp->idev;
3570
3571         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3572                 if (ifpiter->scope > IFA_LINK)
3573                         break;
3574                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3575                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3576                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3577                     IFA_F_PERMANENT)
3578                         return false;
3579         }
3580         return true;
3581 }
3582
3583 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3584 {
3585         struct net_device *dev = ifp->idev->dev;
3586         struct in6_addr lladdr;
3587         bool send_rs, send_mld;
3588
3589         addrconf_del_dad_work(ifp);
3590
3591         /*
3592          *      Configure the address for reception. Now it is valid.
3593          */
3594
3595         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3596
3597         /* If added prefix is link local and we are prepared to process
3598            router advertisements, start sending router solicitations.
3599          */
3600
3601         read_lock_bh(&ifp->idev->lock);
3602         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3603         send_rs = send_mld &&
3604                   ipv6_accept_ra(ifp->idev) &&
3605                   ifp->idev->cnf.rtr_solicits > 0 &&
3606                   (dev->flags&IFF_LOOPBACK) == 0;
3607         read_unlock_bh(&ifp->idev->lock);
3608
3609         /* While dad is in progress mld report's source address is in6_addrany.
3610          * Resend with proper ll now.
3611          */
3612         if (send_mld)
3613                 ipv6_mc_dad_complete(ifp->idev);
3614
3615         if (send_rs) {
3616                 /*
3617                  *      If a host as already performed a random delay
3618                  *      [...] as part of DAD [...] there is no need
3619                  *      to delay again before sending the first RS
3620                  */
3621                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3622                         return;
3623                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3624
3625                 write_lock_bh(&ifp->idev->lock);
3626                 spin_lock(&ifp->lock);
3627                 ifp->idev->rs_probes = 1;
3628                 ifp->idev->if_flags |= IF_RS_SENT;
3629                 addrconf_mod_rs_timer(ifp->idev,
3630                                       ifp->idev->cnf.rtr_solicit_interval);
3631                 spin_unlock(&ifp->lock);
3632                 write_unlock_bh(&ifp->idev->lock);
3633         }
3634 }
3635
3636 static void addrconf_dad_run(struct inet6_dev *idev)
3637 {
3638         struct inet6_ifaddr *ifp;
3639
3640         read_lock_bh(&idev->lock);
3641         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3642                 spin_lock(&ifp->lock);
3643                 if (ifp->flags & IFA_F_TENTATIVE &&
3644                     ifp->state == INET6_IFADDR_STATE_DAD)
3645                         addrconf_dad_kick(ifp);
3646                 spin_unlock(&ifp->lock);
3647         }
3648         read_unlock_bh(&idev->lock);
3649 }
3650
3651 #ifdef CONFIG_PROC_FS
3652 struct if6_iter_state {
3653         struct seq_net_private p;
3654         int bucket;
3655         int offset;
3656 };
3657
3658 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3659 {
3660         struct inet6_ifaddr *ifa = NULL;
3661         struct if6_iter_state *state = seq->private;
3662         struct net *net = seq_file_net(seq);
3663         int p = 0;
3664
3665         /* initial bucket if pos is 0 */
3666         if (pos == 0) {
3667                 state->bucket = 0;
3668                 state->offset = 0;
3669         }
3670
3671         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3672                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3673                                          addr_lst) {
3674                         if (!net_eq(dev_net(ifa->idev->dev), net))
3675                                 continue;
3676                         /* sync with offset */
3677                         if (p < state->offset) {
3678                                 p++;
3679                                 continue;
3680                         }
3681                         state->offset++;
3682                         return ifa;
3683                 }
3684
3685                 /* prepare for next bucket */
3686                 state->offset = 0;
3687                 p = 0;
3688         }
3689         return NULL;
3690 }
3691
3692 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3693                                          struct inet6_ifaddr *ifa)
3694 {
3695         struct if6_iter_state *state = seq->private;
3696         struct net *net = seq_file_net(seq);
3697
3698         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3699                 if (!net_eq(dev_net(ifa->idev->dev), net))
3700                         continue;
3701                 state->offset++;
3702                 return ifa;
3703         }
3704
3705         while (++state->bucket < IN6_ADDR_HSIZE) {
3706                 state->offset = 0;
3707                 hlist_for_each_entry_rcu_bh(ifa,
3708                                      &inet6_addr_lst[state->bucket], addr_lst) {
3709                         if (!net_eq(dev_net(ifa->idev->dev), net))
3710                                 continue;
3711                         state->offset++;
3712                         return ifa;
3713                 }
3714         }
3715
3716         return NULL;
3717 }
3718
3719 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3720         __acquires(rcu_bh)
3721 {
3722         rcu_read_lock_bh();
3723         return if6_get_first(seq, *pos);
3724 }
3725
3726 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3727 {
3728         struct inet6_ifaddr *ifa;
3729
3730         ifa = if6_get_next(seq, v);
3731         ++*pos;
3732         return ifa;
3733 }
3734
3735 static void if6_seq_stop(struct seq_file *seq, void *v)
3736         __releases(rcu_bh)
3737 {
3738         rcu_read_unlock_bh();
3739 }
3740
3741 static int if6_seq_show(struct seq_file *seq, void *v)
3742 {
3743         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3744         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3745                    &ifp->addr,
3746                    ifp->idev->dev->ifindex,
3747                    ifp->prefix_len,
3748                    ifp->scope,
3749                    (u8) ifp->flags,
3750                    ifp->idev->dev->name);
3751         return 0;
3752 }
3753
3754 static const struct seq_operations if6_seq_ops = {
3755         .start  = if6_seq_start,
3756         .next   = if6_seq_next,
3757         .show   = if6_seq_show,
3758         .stop   = if6_seq_stop,
3759 };
3760
3761 static int if6_seq_open(struct inode *inode, struct file *file)
3762 {
3763         return seq_open_net(inode, file, &if6_seq_ops,
3764                             sizeof(struct if6_iter_state));
3765 }
3766
3767 static const struct file_operations if6_fops = {
3768         .owner          = THIS_MODULE,
3769         .open           = if6_seq_open,
3770         .read           = seq_read,
3771         .llseek         = seq_lseek,
3772         .release        = seq_release_net,
3773 };
3774
3775 static int __net_init if6_proc_net_init(struct net *net)
3776 {
3777         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3778                 return -ENOMEM;
3779         return 0;
3780 }
3781
3782 static void __net_exit if6_proc_net_exit(struct net *net)
3783 {
3784         remove_proc_entry("if_inet6", net->proc_net);
3785 }
3786
3787 static struct pernet_operations if6_proc_net_ops = {
3788         .init = if6_proc_net_init,
3789         .exit = if6_proc_net_exit,
3790 };
3791
3792 int __init if6_proc_init(void)
3793 {
3794         return register_pernet_subsys(&if6_proc_net_ops);
3795 }
3796
3797 void if6_proc_exit(void)
3798 {
3799         unregister_pernet_subsys(&if6_proc_net_ops);
3800 }
3801 #endif  /* CONFIG_PROC_FS */
3802
3803 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3804 /* Check if address is a home address configured on any interface. */
3805 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3806 {
3807         int ret = 0;
3808         struct inet6_ifaddr *ifp = NULL;
3809         unsigned int hash = inet6_addr_hash(addr);
3810
3811         rcu_read_lock_bh();
3812         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3813                 if (!net_eq(dev_net(ifp->idev->dev), net))
3814                         continue;
3815                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3816                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3817                         ret = 1;
3818                         break;
3819                 }
3820         }
3821         rcu_read_unlock_bh();
3822         return ret;
3823 }
3824 #endif
3825
3826 /*
3827  *      Periodic address status verification
3828  */
3829
3830 static void addrconf_verify_rtnl(void)
3831 {
3832         unsigned long now, next, next_sec, next_sched;
3833         struct inet6_ifaddr *ifp;
3834         int i;
3835
3836         ASSERT_RTNL();
3837
3838         rcu_read_lock_bh();
3839         now = jiffies;
3840         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3841
3842         cancel_delayed_work(&addr_chk_work);
3843
3844         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3845 restart:
3846                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3847                         unsigned long age;
3848
3849                         /* When setting preferred_lft to a value not zero or
3850                          * infinity, while valid_lft is infinity
3851                          * IFA_F_PERMANENT has a non-infinity life time.
3852                          */
3853                         if ((ifp->flags & IFA_F_PERMANENT) &&
3854                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3855                                 continue;
3856
3857                         spin_lock(&ifp->lock);
3858                         /* We try to batch several events at once. */
3859                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3860
3861                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3862                             age >= ifp->valid_lft) {
3863                                 spin_unlock(&ifp->lock);
3864                                 in6_ifa_hold(ifp);
3865                                 ipv6_del_addr(ifp);
3866                                 goto restart;
3867                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3868                                 spin_unlock(&ifp->lock);
3869                                 continue;
3870                         } else if (age >= ifp->prefered_lft) {
3871                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3872                                 int deprecate = 0;
3873
3874                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3875                                         deprecate = 1;
3876                                         ifp->flags |= IFA_F_DEPRECATED;
3877                                 }
3878
3879                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3880                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3881                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3882
3883                                 spin_unlock(&ifp->lock);
3884
3885                                 if (deprecate) {
3886                                         in6_ifa_hold(ifp);
3887
3888                                         ipv6_ifa_notify(0, ifp);
3889                                         in6_ifa_put(ifp);
3890                                         goto restart;
3891                                 }
3892                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3893                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3894                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3895                                         ifp->idev->cnf.dad_transmits *
3896                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3897
3898                                 if (age >= ifp->prefered_lft - regen_advance) {
3899                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3900                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3901                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3902                                         if (!ifp->regen_count && ifpub) {
3903                                                 ifp->regen_count++;
3904                                                 in6_ifa_hold(ifp);
3905                                                 in6_ifa_hold(ifpub);
3906                                                 spin_unlock(&ifp->lock);
3907
3908                                                 spin_lock(&ifpub->lock);
3909                                                 ifpub->regen_count = 0;
3910                                                 spin_unlock(&ifpub->lock);
3911                                                 ipv6_create_tempaddr(ifpub, ifp);
3912                                                 in6_ifa_put(ifpub);
3913                                                 in6_ifa_put(ifp);
3914                                                 goto restart;
3915                                         }
3916                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3917                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3918                                 spin_unlock(&ifp->lock);
3919                         } else {
3920                                 /* ifp->prefered_lft <= ifp->valid_lft */
3921                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3922                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3923                                 spin_unlock(&ifp->lock);
3924                         }
3925                 }
3926         }
3927
3928         next_sec = round_jiffies_up(next);
3929         next_sched = next;
3930
3931         /* If rounded timeout is accurate enough, accept it. */
3932         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3933                 next_sched = next_sec;
3934
3935         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3936         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3937                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3938
3939         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3940               now, next, next_sec, next_sched);
3941         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3942         rcu_read_unlock_bh();
3943 }
3944
3945 static void addrconf_verify_work(struct work_struct *w)
3946 {
3947         rtnl_lock();
3948         addrconf_verify_rtnl();
3949         rtnl_unlock();
3950 }
3951
3952 static void addrconf_verify(void)
3953 {
3954         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3955 }
3956
3957 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3958                                      struct in6_addr **peer_pfx)
3959 {
3960         struct in6_addr *pfx = NULL;
3961
3962         *peer_pfx = NULL;
3963
3964         if (addr)
3965                 pfx = nla_data(addr);
3966
3967         if (local) {
3968                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3969                         *peer_pfx = pfx;
3970                 pfx = nla_data(local);
3971         }
3972
3973         return pfx;
3974 }
3975
3976 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3977         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3978         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3979         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3980         [IFA_FLAGS]             = { .len = sizeof(u32) },
3981 };
3982
3983 static int
3984 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3985 {
3986         struct net *net = sock_net(skb->sk);
3987         struct ifaddrmsg *ifm;
3988         struct nlattr *tb[IFA_MAX+1];
3989         struct in6_addr *pfx, *peer_pfx;
3990         u32 ifa_flags;
3991         int err;
3992
3993         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3994         if (err < 0)
3995                 return err;
3996
3997         ifm = nlmsg_data(nlh);
3998         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3999         if (!pfx)
4000                 return -EINVAL;
4001
4002         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4003
4004         /* We ignore other flags so far. */
4005         ifa_flags &= IFA_F_MANAGETEMPADDR;
4006
4007         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4008                               ifm->ifa_prefixlen);
4009 }
4010
4011 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4012                              u32 prefered_lft, u32 valid_lft)
4013 {
4014         u32 flags;
4015         clock_t expires;
4016         unsigned long timeout;
4017         bool was_managetempaddr;
4018         bool had_prefixroute;
4019
4020         ASSERT_RTNL();
4021
4022         if (!valid_lft || (prefered_lft > valid_lft))
4023                 return -EINVAL;
4024
4025         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4026             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4027                 return -EINVAL;
4028
4029         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4030         if (addrconf_finite_timeout(timeout)) {
4031                 expires = jiffies_to_clock_t(timeout * HZ);
4032                 valid_lft = timeout;
4033                 flags = RTF_EXPIRES;
4034         } else {
4035                 expires = 0;
4036                 flags = 0;
4037                 ifa_flags |= IFA_F_PERMANENT;
4038         }
4039
4040         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4041         if (addrconf_finite_timeout(timeout)) {
4042                 if (timeout == 0)
4043                         ifa_flags |= IFA_F_DEPRECATED;
4044                 prefered_lft = timeout;
4045         }
4046
4047         spin_lock_bh(&ifp->lock);
4048         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4049         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4050                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4051         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4052                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4053                         IFA_F_NOPREFIXROUTE);
4054         ifp->flags |= ifa_flags;
4055         ifp->tstamp = jiffies;
4056         ifp->valid_lft = valid_lft;
4057         ifp->prefered_lft = prefered_lft;
4058
4059         spin_unlock_bh(&ifp->lock);
4060         if (!(ifp->flags&IFA_F_TENTATIVE))
4061                 ipv6_ifa_notify(0, ifp);
4062
4063         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4064                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4065                                       expires, flags);
4066         } else if (had_prefixroute) {
4067                 enum cleanup_prefix_rt_t action;
4068                 unsigned long rt_expires;
4069
4070                 write_lock_bh(&ifp->idev->lock);
4071                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4072                 write_unlock_bh(&ifp->idev->lock);
4073
4074                 if (action != CLEANUP_PREFIX_RT_NOP) {
4075                         cleanup_prefix_route(ifp, rt_expires,
4076                                 action == CLEANUP_PREFIX_RT_DEL);
4077                 }
4078         }
4079
4080         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4081                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4082                         valid_lft = prefered_lft = 0;
4083                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4084                                  !was_managetempaddr, jiffies);
4085         }
4086
4087         addrconf_verify_rtnl();
4088
4089         return 0;
4090 }
4091
4092 static int
4093 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4094 {
4095         struct net *net = sock_net(skb->sk);
4096         struct ifaddrmsg *ifm;
4097         struct nlattr *tb[IFA_MAX+1];
4098         struct in6_addr *pfx, *peer_pfx;
4099         struct inet6_ifaddr *ifa;
4100         struct net_device *dev;
4101         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4102         u32 ifa_flags;
4103         int err;
4104
4105         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4106         if (err < 0)
4107                 return err;
4108
4109         ifm = nlmsg_data(nlh);
4110         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4111         if (!pfx)
4112                 return -EINVAL;
4113
4114         if (tb[IFA_CACHEINFO]) {
4115                 struct ifa_cacheinfo *ci;
4116
4117                 ci = nla_data(tb[IFA_CACHEINFO]);
4118                 valid_lft = ci->ifa_valid;
4119                 preferred_lft = ci->ifa_prefered;
4120         } else {
4121                 preferred_lft = INFINITY_LIFE_TIME;
4122                 valid_lft = INFINITY_LIFE_TIME;
4123         }
4124
4125         dev =  __dev_get_by_index(net, ifm->ifa_index);
4126         if (!dev)
4127                 return -ENODEV;
4128
4129         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4130
4131         /* We ignore other flags so far. */
4132         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4133                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4134
4135         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4136         if (!ifa) {
4137                 /*
4138                  * It would be best to check for !NLM_F_CREATE here but
4139                  * userspace already relies on not having to provide this.
4140                  */
4141                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4142                                       ifm->ifa_prefixlen, ifa_flags,
4143                                       preferred_lft, valid_lft);
4144         }
4145
4146         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4147             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4148                 err = -EEXIST;
4149         else
4150                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4151
4152         in6_ifa_put(ifa);
4153
4154         return err;
4155 }
4156
4157 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4158                           u8 scope, int ifindex)
4159 {
4160         struct ifaddrmsg *ifm;
4161
4162         ifm = nlmsg_data(nlh);
4163         ifm->ifa_family = AF_INET6;
4164         ifm->ifa_prefixlen = prefixlen;
4165         ifm->ifa_flags = flags;
4166         ifm->ifa_scope = scope;
4167         ifm->ifa_index = ifindex;
4168 }
4169
4170 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4171                          unsigned long tstamp, u32 preferred, u32 valid)
4172 {
4173         struct ifa_cacheinfo ci;
4174
4175         ci.cstamp = cstamp_delta(cstamp);
4176         ci.tstamp = cstamp_delta(tstamp);
4177         ci.ifa_prefered = preferred;
4178         ci.ifa_valid = valid;
4179
4180         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4181 }
4182
4183 static inline int rt_scope(int ifa_scope)
4184 {
4185         if (ifa_scope & IFA_HOST)
4186                 return RT_SCOPE_HOST;
4187         else if (ifa_scope & IFA_LINK)
4188                 return RT_SCOPE_LINK;
4189         else if (ifa_scope & IFA_SITE)
4190                 return RT_SCOPE_SITE;
4191         else
4192                 return RT_SCOPE_UNIVERSE;
4193 }
4194
4195 static inline int inet6_ifaddr_msgsize(void)
4196 {
4197         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4198                + nla_total_size(16) /* IFA_LOCAL */
4199                + nla_total_size(16) /* IFA_ADDRESS */
4200                + nla_total_size(sizeof(struct ifa_cacheinfo))
4201                + nla_total_size(4)  /* IFA_FLAGS */;
4202 }
4203
4204 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4205                              u32 portid, u32 seq, int event, unsigned int flags)
4206 {
4207         struct nlmsghdr  *nlh;
4208         u32 preferred, valid;
4209
4210         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4211         if (!nlh)
4212                 return -EMSGSIZE;
4213
4214         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4215                       ifa->idev->dev->ifindex);
4216
4217         if (!((ifa->flags&IFA_F_PERMANENT) &&
4218               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4219                 preferred = ifa->prefered_lft;
4220                 valid = ifa->valid_lft;
4221                 if (preferred != INFINITY_LIFE_TIME) {
4222                         long tval = (jiffies - ifa->tstamp)/HZ;
4223                         if (preferred > tval)
4224                                 preferred -= tval;
4225                         else
4226                                 preferred = 0;
4227                         if (valid != INFINITY_LIFE_TIME) {
4228                                 if (valid > tval)
4229                                         valid -= tval;
4230                                 else
4231                                         valid = 0;
4232                         }
4233                 }
4234         } else {
4235                 preferred = INFINITY_LIFE_TIME;
4236                 valid = INFINITY_LIFE_TIME;
4237         }
4238
4239         if (!ipv6_addr_any(&ifa->peer_addr)) {
4240                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4241                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4242                         goto error;
4243         } else
4244                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4245                         goto error;
4246
4247         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4248                 goto error;
4249
4250         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4251                 goto error;
4252
4253         nlmsg_end(skb, nlh);
4254         return 0;
4255
4256 error:
4257         nlmsg_cancel(skb, nlh);
4258         return -EMSGSIZE;
4259 }
4260
4261 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4262                                 u32 portid, u32 seq, int event, u16 flags)
4263 {
4264         struct nlmsghdr  *nlh;
4265         u8 scope = RT_SCOPE_UNIVERSE;
4266         int ifindex = ifmca->idev->dev->ifindex;
4267
4268         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4269                 scope = RT_SCOPE_SITE;
4270
4271         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4272         if (!nlh)
4273                 return -EMSGSIZE;
4274
4275         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4276         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4277             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4278                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4279                 nlmsg_cancel(skb, nlh);
4280                 return -EMSGSIZE;
4281         }
4282
4283         nlmsg_end(skb, nlh);
4284         return 0;
4285 }
4286
4287 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4288                                 u32 portid, u32 seq, int event, unsigned int flags)
4289 {
4290         struct nlmsghdr  *nlh;
4291         u8 scope = RT_SCOPE_UNIVERSE;
4292         int ifindex = ifaca->aca_idev->dev->ifindex;
4293
4294         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4295                 scope = RT_SCOPE_SITE;
4296
4297         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4298         if (!nlh)
4299                 return -EMSGSIZE;
4300
4301         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4302         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4303             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4304                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4305                 nlmsg_cancel(skb, nlh);
4306                 return -EMSGSIZE;
4307         }
4308
4309         nlmsg_end(skb, nlh);
4310         return 0;
4311 }
4312
4313 enum addr_type_t {
4314         UNICAST_ADDR,
4315         MULTICAST_ADDR,
4316         ANYCAST_ADDR,
4317 };
4318
4319 /* called with rcu_read_lock() */
4320 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4321                           struct netlink_callback *cb, enum addr_type_t type,
4322                           int s_ip_idx, int *p_ip_idx)
4323 {
4324         struct ifmcaddr6 *ifmca;
4325         struct ifacaddr6 *ifaca;
4326         int err = 1;
4327         int ip_idx = *p_ip_idx;
4328
4329         read_lock_bh(&idev->lock);
4330         switch (type) {
4331         case UNICAST_ADDR: {
4332                 struct inet6_ifaddr *ifa;
4333
4334                 /* unicast address incl. temp addr */
4335                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4336                         if (++ip_idx < s_ip_idx)
4337                                 continue;
4338                         err = inet6_fill_ifaddr(skb, ifa,
4339                                                 NETLINK_CB(cb->skb).portid,
4340                                                 cb->nlh->nlmsg_seq,
4341                                                 RTM_NEWADDR,
4342                                                 NLM_F_MULTI);
4343                         if (err < 0)
4344                                 break;
4345                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4346                 }
4347                 break;
4348         }
4349         case MULTICAST_ADDR:
4350                 /* multicast address */
4351                 for (ifmca = idev->mc_list; ifmca;
4352                      ifmca = ifmca->next, ip_idx++) {
4353                         if (ip_idx < s_ip_idx)
4354                                 continue;
4355                         err = inet6_fill_ifmcaddr(skb, ifmca,
4356                                                   NETLINK_CB(cb->skb).portid,
4357                                                   cb->nlh->nlmsg_seq,
4358                                                   RTM_GETMULTICAST,
4359                                                   NLM_F_MULTI);
4360                         if (err < 0)
4361                                 break;
4362                 }
4363                 break;
4364         case ANYCAST_ADDR:
4365                 /* anycast address */
4366                 for (ifaca = idev->ac_list; ifaca;
4367                      ifaca = ifaca->aca_next, ip_idx++) {
4368                         if (ip_idx < s_ip_idx)
4369                                 continue;
4370                         err = inet6_fill_ifacaddr(skb, ifaca,
4371                                                   NETLINK_CB(cb->skb).portid,
4372                                                   cb->nlh->nlmsg_seq,
4373                                                   RTM_GETANYCAST,
4374                                                   NLM_F_MULTI);
4375                         if (err < 0)
4376                                 break;
4377                 }
4378                 break;
4379         default:
4380                 break;
4381         }
4382         read_unlock_bh(&idev->lock);
4383         *p_ip_idx = ip_idx;
4384         return err;
4385 }
4386
4387 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4388                            enum addr_type_t type)
4389 {
4390         struct net *net = sock_net(skb->sk);
4391         int h, s_h;
4392         int idx, ip_idx;
4393         int s_idx, s_ip_idx;
4394         struct net_device *dev;
4395         struct inet6_dev *idev;
4396         struct hlist_head *head;
4397
4398         s_h = cb->args[0];
4399         s_idx = idx = cb->args[1];
4400         s_ip_idx = ip_idx = cb->args[2];
4401
4402         rcu_read_lock();
4403         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4404         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4405                 idx = 0;
4406                 head = &net->dev_index_head[h];
4407                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4408                         if (idx < s_idx)
4409                                 goto cont;
4410                         if (h > s_h || idx > s_idx)
4411                                 s_ip_idx = 0;
4412                         ip_idx = 0;
4413                         idev = __in6_dev_get(dev);
4414                         if (!idev)
4415                                 goto cont;
4416
4417                         if (in6_dump_addrs(idev, skb, cb, type,
4418                                            s_ip_idx, &ip_idx) < 0)
4419                                 goto done;
4420 cont:
4421                         idx++;
4422                 }
4423         }
4424 done:
4425         rcu_read_unlock();
4426         cb->args[0] = h;
4427         cb->args[1] = idx;
4428         cb->args[2] = ip_idx;
4429
4430         return skb->len;
4431 }
4432
4433 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4434 {
4435         enum addr_type_t type = UNICAST_ADDR;
4436
4437         return inet6_dump_addr(skb, cb, type);
4438 }
4439
4440 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4441 {
4442         enum addr_type_t type = MULTICAST_ADDR;
4443
4444         return inet6_dump_addr(skb, cb, type);
4445 }
4446
4447
4448 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4449 {
4450         enum addr_type_t type = ANYCAST_ADDR;
4451
4452         return inet6_dump_addr(skb, cb, type);
4453 }
4454
4455 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4456 {
4457         struct net *net = sock_net(in_skb->sk);
4458         struct ifaddrmsg *ifm;
4459         struct nlattr *tb[IFA_MAX+1];
4460         struct in6_addr *addr = NULL, *peer;
4461         struct net_device *dev = NULL;
4462         struct inet6_ifaddr *ifa;
4463         struct sk_buff *skb;
4464         int err;
4465
4466         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4467         if (err < 0)
4468                 goto errout;
4469
4470         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4471         if (!addr) {
4472                 err = -EINVAL;
4473                 goto errout;
4474         }
4475
4476         ifm = nlmsg_data(nlh);
4477         if (ifm->ifa_index)
4478                 dev = __dev_get_by_index(net, ifm->ifa_index);
4479
4480         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4481         if (!ifa) {
4482                 err = -EADDRNOTAVAIL;
4483                 goto errout;
4484         }
4485
4486         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4487         if (!skb) {
4488                 err = -ENOBUFS;
4489                 goto errout_ifa;
4490         }
4491
4492         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4493                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4494         if (err < 0) {
4495                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4496                 WARN_ON(err == -EMSGSIZE);
4497                 kfree_skb(skb);
4498                 goto errout_ifa;
4499         }
4500         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4501 errout_ifa:
4502         in6_ifa_put(ifa);
4503 errout:
4504         return err;
4505 }
4506
4507 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4508 {
4509         struct sk_buff *skb;
4510         struct net *net = dev_net(ifa->idev->dev);
4511         int err = -ENOBUFS;
4512
4513         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4514         if (!skb)
4515                 goto errout;
4516
4517         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4518         if (err < 0) {
4519                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4520                 WARN_ON(err == -EMSGSIZE);
4521                 kfree_skb(skb);
4522                 goto errout;
4523         }
4524         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4525         return;
4526 errout:
4527         if (err < 0)
4528                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4529 }
4530
4531 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4532                                 __s32 *array, int bytes)
4533 {
4534         BUG_ON(bytes < (DEVCONF_MAX * 4));
4535
4536         memset(array, 0, bytes);
4537         array[DEVCONF_FORWARDING] = cnf->forwarding;
4538         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4539         array[DEVCONF_MTU6] = cnf->mtu6;
4540         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4541         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4542         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4543         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4544         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4545         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4546                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4547         array[DEVCONF_RTR_SOLICIT_DELAY] =
4548                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4549         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4550         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4551                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4552         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4553                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4554         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4555         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4556         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4557         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4558         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4559         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4560         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4561         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4562 #ifdef CONFIG_IPV6_ROUTER_PREF
4563         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4564         array[DEVCONF_RTR_PROBE_INTERVAL] =
4565                 jiffies_to_msecs(cnf->rtr_probe_interval);
4566 #ifdef CONFIG_IPV6_ROUTE_INFO
4567         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4568 #endif
4569 #endif
4570         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4571         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4572 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4573         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4574         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4575 #endif
4576 #ifdef CONFIG_IPV6_MROUTE
4577         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4578 #endif
4579         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4580         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4581         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4582         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4583         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4584         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4585         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4586         /* we omit DEVCONF_STABLE_SECRET for now */
4587 }
4588
4589 static inline size_t inet6_ifla6_size(void)
4590 {
4591         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4592              + nla_total_size(sizeof(struct ifla_cacheinfo))
4593              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4594              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4595              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4596              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4597 }
4598
4599 static inline size_t inet6_if_nlmsg_size(void)
4600 {
4601         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4602                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4603                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4604                + nla_total_size(4) /* IFLA_MTU */
4605                + nla_total_size(4) /* IFLA_LINK */
4606                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4607 }
4608
4609 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4610                                       int items, int bytes)
4611 {
4612         int i;
4613         int pad = bytes - sizeof(u64) * items;
4614         BUG_ON(pad < 0);
4615
4616         /* Use put_unaligned() because stats may not be aligned for u64. */
4617         put_unaligned(items, &stats[0]);
4618         for (i = 1; i < items; i++)
4619                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4620
4621         memset(&stats[items], 0, pad);
4622 }
4623
4624 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4625                                       int items, int bytes, size_t syncpoff)
4626 {
4627         int i;
4628         int pad = bytes - sizeof(u64) * items;
4629         BUG_ON(pad < 0);
4630
4631         /* Use put_unaligned() because stats may not be aligned for u64. */
4632         put_unaligned(items, &stats[0]);
4633         for (i = 1; i < items; i++)
4634                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4635
4636         memset(&stats[items], 0, pad);
4637 }
4638
4639 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4640                              int bytes)
4641 {
4642         switch (attrtype) {
4643         case IFLA_INET6_STATS:
4644                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4645                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4646                 break;
4647         case IFLA_INET6_ICMP6STATS:
4648                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4649                 break;
4650         }
4651 }
4652
4653 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4654 {
4655         struct nlattr *nla;
4656         struct ifla_cacheinfo ci;
4657
4658         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4659                 goto nla_put_failure;
4660         ci.max_reasm_len = IPV6_MAXPLEN;
4661         ci.tstamp = cstamp_delta(idev->tstamp);
4662         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4663         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4664         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4665                 goto nla_put_failure;
4666         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4667         if (!nla)
4668                 goto nla_put_failure;
4669         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4670
4671         /* XXX - MC not implemented */
4672
4673         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4674         if (!nla)
4675                 goto nla_put_failure;
4676         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4677
4678         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4679         if (!nla)
4680                 goto nla_put_failure;
4681         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4682
4683         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4684         if (!nla)
4685                 goto nla_put_failure;
4686
4687         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4688                 goto nla_put_failure;
4689
4690         read_lock_bh(&idev->lock);
4691         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4692         read_unlock_bh(&idev->lock);
4693
4694         return 0;
4695
4696 nla_put_failure:
4697         return -EMSGSIZE;
4698 }
4699
4700 static size_t inet6_get_link_af_size(const struct net_device *dev)
4701 {
4702         if (!__in6_dev_get(dev))
4703                 return 0;
4704
4705         return inet6_ifla6_size();
4706 }
4707
4708 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4709 {
4710         struct inet6_dev *idev = __in6_dev_get(dev);
4711
4712         if (!idev)
4713                 return -ENODATA;
4714
4715         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4716                 return -EMSGSIZE;
4717
4718         return 0;
4719 }
4720
4721 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4722 {
4723         struct inet6_ifaddr *ifp;
4724         struct net_device *dev = idev->dev;
4725         bool update_rs = false;
4726         struct in6_addr ll_addr;
4727
4728         ASSERT_RTNL();
4729
4730         if (!token)
4731                 return -EINVAL;
4732         if (ipv6_addr_any(token))
4733                 return -EINVAL;
4734         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4735                 return -EINVAL;
4736         if (!ipv6_accept_ra(idev))
4737                 return -EINVAL;
4738         if (idev->cnf.rtr_solicits <= 0)
4739                 return -EINVAL;
4740
4741         write_lock_bh(&idev->lock);
4742
4743         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4744         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4745
4746         write_unlock_bh(&idev->lock);
4747
4748         if (!idev->dead && (idev->if_flags & IF_READY) &&
4749             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4750                              IFA_F_OPTIMISTIC)) {
4751
4752                 /* If we're not ready, then normal ifup will take care
4753                  * of this. Otherwise, we need to request our rs here.
4754                  */
4755                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4756                 update_rs = true;
4757         }
4758
4759         write_lock_bh(&idev->lock);
4760
4761         if (update_rs) {
4762                 idev->if_flags |= IF_RS_SENT;
4763                 idev->rs_probes = 1;
4764                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4765         }
4766
4767         /* Well, that's kinda nasty ... */
4768         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4769                 spin_lock(&ifp->lock);
4770                 if (ifp->tokenized) {
4771                         ifp->valid_lft = 0;
4772                         ifp->prefered_lft = 0;
4773                 }
4774                 spin_unlock(&ifp->lock);
4775         }
4776
4777         write_unlock_bh(&idev->lock);
4778         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4779         addrconf_verify_rtnl();
4780         return 0;
4781 }
4782
4783 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4784         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4785         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4786 };
4787
4788 static int inet6_validate_link_af(const struct net_device *dev,
4789                                   const struct nlattr *nla)
4790 {
4791         struct nlattr *tb[IFLA_INET6_MAX + 1];
4792
4793         if (dev && !__in6_dev_get(dev))
4794                 return -EAFNOSUPPORT;
4795
4796         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4797 }
4798
4799 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4800 {
4801         int err = -EINVAL;
4802         struct inet6_dev *idev = __in6_dev_get(dev);
4803         struct nlattr *tb[IFLA_INET6_MAX + 1];
4804
4805         if (!idev)
4806                 return -EAFNOSUPPORT;
4807
4808         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4809                 BUG();
4810
4811         if (tb[IFLA_INET6_TOKEN]) {
4812                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4813                 if (err)
4814                         return err;
4815         }
4816
4817         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4818                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4819
4820                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4821                     mode != IN6_ADDR_GEN_MODE_NONE &&
4822                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4823                         return -EINVAL;
4824
4825                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4826                     !idev->cnf.stable_secret.initialized &&
4827                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4828                         return -EINVAL;
4829
4830                 idev->addr_gen_mode = mode;
4831                 err = 0;
4832         }
4833
4834         return err;
4835 }
4836
4837 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4838                              u32 portid, u32 seq, int event, unsigned int flags)
4839 {
4840         struct net_device *dev = idev->dev;
4841         struct ifinfomsg *hdr;
4842         struct nlmsghdr *nlh;
4843         void *protoinfo;
4844
4845         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4846         if (!nlh)
4847                 return -EMSGSIZE;
4848
4849         hdr = nlmsg_data(nlh);
4850         hdr->ifi_family = AF_INET6;
4851         hdr->__ifi_pad = 0;
4852         hdr->ifi_type = dev->type;
4853         hdr->ifi_index = dev->ifindex;
4854         hdr->ifi_flags = dev_get_flags(dev);
4855         hdr->ifi_change = 0;
4856
4857         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4858             (dev->addr_len &&
4859              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4860             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4861             (dev->ifindex != dev_get_iflink(dev) &&
4862              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4863                 goto nla_put_failure;
4864         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4865         if (!protoinfo)
4866                 goto nla_put_failure;
4867
4868         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4869                 goto nla_put_failure;
4870
4871         nla_nest_end(skb, protoinfo);
4872         nlmsg_end(skb, nlh);
4873         return 0;
4874
4875 nla_put_failure:
4876         nlmsg_cancel(skb, nlh);
4877         return -EMSGSIZE;
4878 }
4879
4880 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4881 {
4882         struct net *net = sock_net(skb->sk);
4883         int h, s_h;
4884         int idx = 0, s_idx;
4885         struct net_device *dev;
4886         struct inet6_dev *idev;
4887         struct hlist_head *head;
4888
4889         s_h = cb->args[0];
4890         s_idx = cb->args[1];
4891
4892         rcu_read_lock();
4893         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4894                 idx = 0;
4895                 head = &net->dev_index_head[h];
4896                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4897                         if (idx < s_idx)
4898                                 goto cont;
4899                         idev = __in6_dev_get(dev);
4900                         if (!idev)
4901                                 goto cont;
4902                         if (inet6_fill_ifinfo(skb, idev,
4903                                               NETLINK_CB(cb->skb).portid,
4904                                               cb->nlh->nlmsg_seq,
4905                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
4906                                 goto out;
4907 cont:
4908                         idx++;
4909                 }
4910         }
4911 out:
4912         rcu_read_unlock();
4913         cb->args[1] = idx;
4914         cb->args[0] = h;
4915
4916         return skb->len;
4917 }
4918
4919 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4920 {
4921         struct sk_buff *skb;
4922         struct net *net = dev_net(idev->dev);
4923         int err = -ENOBUFS;
4924
4925         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4926         if (!skb)
4927                 goto errout;
4928
4929         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4930         if (err < 0) {
4931                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4932                 WARN_ON(err == -EMSGSIZE);
4933                 kfree_skb(skb);
4934                 goto errout;
4935         }
4936         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4937         return;
4938 errout:
4939         if (err < 0)
4940                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4941 }
4942
4943 static inline size_t inet6_prefix_nlmsg_size(void)
4944 {
4945         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4946                + nla_total_size(sizeof(struct in6_addr))
4947                + nla_total_size(sizeof(struct prefix_cacheinfo));
4948 }
4949
4950 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4951                              struct prefix_info *pinfo, u32 portid, u32 seq,
4952                              int event, unsigned int flags)
4953 {
4954         struct prefixmsg *pmsg;
4955         struct nlmsghdr *nlh;
4956         struct prefix_cacheinfo ci;
4957
4958         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4959         if (!nlh)
4960                 return -EMSGSIZE;
4961
4962         pmsg = nlmsg_data(nlh);
4963         pmsg->prefix_family = AF_INET6;
4964         pmsg->prefix_pad1 = 0;
4965         pmsg->prefix_pad2 = 0;
4966         pmsg->prefix_ifindex = idev->dev->ifindex;
4967         pmsg->prefix_len = pinfo->prefix_len;
4968         pmsg->prefix_type = pinfo->type;
4969         pmsg->prefix_pad3 = 0;
4970         pmsg->prefix_flags = 0;
4971         if (pinfo->onlink)
4972                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4973         if (pinfo->autoconf)
4974                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4975
4976         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4977                 goto nla_put_failure;
4978         ci.preferred_time = ntohl(pinfo->prefered);
4979         ci.valid_time = ntohl(pinfo->valid);
4980         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4981                 goto nla_put_failure;
4982         nlmsg_end(skb, nlh);
4983         return 0;
4984
4985 nla_put_failure:
4986         nlmsg_cancel(skb, nlh);
4987         return -EMSGSIZE;
4988 }
4989
4990 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4991                          struct prefix_info *pinfo)
4992 {
4993         struct sk_buff *skb;
4994         struct net *net = dev_net(idev->dev);
4995         int err = -ENOBUFS;
4996
4997         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4998         if (!skb)
4999                 goto errout;
5000
5001         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5002         if (err < 0) {
5003                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5004                 WARN_ON(err == -EMSGSIZE);
5005                 kfree_skb(skb);
5006                 goto errout;
5007         }
5008         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5009         return;
5010 errout:
5011         if (err < 0)
5012                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5013 }
5014
5015 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5016 {
5017         struct net *net = dev_net(ifp->idev->dev);
5018
5019         if (event)
5020                 ASSERT_RTNL();
5021
5022         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5023
5024         switch (event) {
5025         case RTM_NEWADDR:
5026                 /*
5027                  * If the address was optimistic
5028                  * we inserted the route at the start of
5029                  * our DAD process, so we don't need
5030                  * to do it again
5031                  */
5032                 if (!(ifp->rt->rt6i_node))
5033                         ip6_ins_rt(ifp->rt);
5034                 if (ifp->idev->cnf.forwarding)
5035                         addrconf_join_anycast(ifp);
5036                 if (!ipv6_addr_any(&ifp->peer_addr))
5037                         addrconf_prefix_route(&ifp->peer_addr, 128,
5038                                               ifp->idev->dev, 0, 0);
5039                 break;
5040         case RTM_DELADDR:
5041                 if (ifp->idev->cnf.forwarding)
5042                         addrconf_leave_anycast(ifp);
5043                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5044                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5045                         struct rt6_info *rt;
5046
5047                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5048                                                        ifp->idev->dev, 0, 0);
5049                         if (rt && ip6_del_rt(rt))
5050                                 dst_free(&rt->dst);
5051                 }
5052                 dst_hold(&ifp->rt->dst);
5053
5054                 if (ip6_del_rt(ifp->rt))
5055                         dst_free(&ifp->rt->dst);
5056
5057                 rt_genid_bump_ipv6(net);
5058                 break;
5059         }
5060         atomic_inc(&net->ipv6.dev_addr_genid);
5061 }
5062
5063 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5064 {
5065         rcu_read_lock_bh();
5066         if (likely(ifp->idev->dead == 0))
5067                 __ipv6_ifa_notify(event, ifp);
5068         rcu_read_unlock_bh();
5069 }
5070
5071 #ifdef CONFIG_SYSCTL
5072
5073 static
5074 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5075                            void __user *buffer, size_t *lenp, loff_t *ppos)
5076 {
5077         int *valp = ctl->data;
5078         int val = *valp;
5079         loff_t pos = *ppos;
5080         struct ctl_table lctl;
5081         int ret;
5082
5083         /*
5084          * ctl->data points to idev->cnf.forwarding, we should
5085          * not modify it until we get the rtnl lock.
5086          */
5087         lctl = *ctl;
5088         lctl.data = &val;
5089
5090         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5091
5092         if (write)
5093                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5094         if (ret)
5095                 *ppos = pos;
5096         return ret;
5097 }
5098
5099 static
5100 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5101                         void __user *buffer, size_t *lenp, loff_t *ppos)
5102 {
5103         struct inet6_dev *idev = ctl->extra1;
5104         int min_mtu = IPV6_MIN_MTU;
5105         struct ctl_table lctl;
5106
5107         lctl = *ctl;
5108         lctl.extra1 = &min_mtu;
5109         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5110
5111         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5112 }
5113
5114 static void dev_disable_change(struct inet6_dev *idev)
5115 {
5116         struct netdev_notifier_info info;
5117
5118         if (!idev || !idev->dev)
5119                 return;
5120
5121         netdev_notifier_info_init(&info, idev->dev);
5122         if (idev->cnf.disable_ipv6)
5123                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5124         else
5125                 addrconf_notify(NULL, NETDEV_UP, &info);
5126 }
5127
5128 static void addrconf_disable_change(struct net *net, __s32 newf)
5129 {
5130         struct net_device *dev;
5131         struct inet6_dev *idev;
5132
5133         rcu_read_lock();
5134         for_each_netdev_rcu(net, dev) {
5135                 idev = __in6_dev_get(dev);
5136                 if (idev) {
5137                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5138                         idev->cnf.disable_ipv6 = newf;
5139                         if (changed)
5140                                 dev_disable_change(idev);
5141                 }
5142         }
5143         rcu_read_unlock();
5144 }
5145
5146 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5147 {
5148         struct net *net;
5149         int old;
5150
5151         if (!rtnl_trylock())
5152                 return restart_syscall();
5153
5154         net = (struct net *)table->extra2;
5155         old = *p;
5156         *p = newf;
5157
5158         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5159                 rtnl_unlock();
5160                 return 0;
5161         }
5162
5163         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5164                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5165                 addrconf_disable_change(net, newf);
5166         } else if ((!newf) ^ (!old))
5167                 dev_disable_change((struct inet6_dev *)table->extra1);
5168
5169         rtnl_unlock();
5170         return 0;
5171 }
5172
5173 static
5174 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5175                             void __user *buffer, size_t *lenp, loff_t *ppos)
5176 {
5177         int *valp = ctl->data;
5178         int val = *valp;
5179         loff_t pos = *ppos;
5180         struct ctl_table lctl;
5181         int ret;
5182
5183         /*
5184          * ctl->data points to idev->cnf.disable_ipv6, we should
5185          * not modify it until we get the rtnl lock.
5186          */
5187         lctl = *ctl;
5188         lctl.data = &val;
5189
5190         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5191
5192         if (write)
5193                 ret = addrconf_disable_ipv6(ctl, valp, val);
5194         if (ret)
5195                 *ppos = pos;
5196         return ret;
5197 }
5198
5199 static
5200 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5201                               void __user *buffer, size_t *lenp, loff_t *ppos)
5202 {
5203         int *valp = ctl->data;
5204         int ret;
5205         int old, new;
5206
5207         old = *valp;
5208         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5209         new = *valp;
5210
5211         if (write && old != new) {
5212                 struct net *net = ctl->extra2;
5213
5214                 if (!rtnl_trylock())
5215                         return restart_syscall();
5216
5217                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5218                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5219                                                      NETCONFA_IFINDEX_DEFAULT,
5220                                                      net->ipv6.devconf_dflt);
5221                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5222                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5223                                                      NETCONFA_IFINDEX_ALL,
5224                                                      net->ipv6.devconf_all);
5225                 else {
5226                         struct inet6_dev *idev = ctl->extra1;
5227
5228                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5229                                                      idev->dev->ifindex,
5230                                                      &idev->cnf);
5231                 }
5232                 rtnl_unlock();
5233         }
5234
5235         return ret;
5236 }
5237
5238 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5239                                          void __user *buffer, size_t *lenp,
5240                                          loff_t *ppos)
5241 {
5242         int err;
5243         struct in6_addr addr;
5244         char str[IPV6_MAX_STRLEN];
5245         struct ctl_table lctl = *ctl;
5246         struct net *net = ctl->extra2;
5247         struct ipv6_stable_secret *secret = ctl->data;
5248
5249         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5250                 return -EIO;
5251
5252         lctl.maxlen = IPV6_MAX_STRLEN;
5253         lctl.data = str;
5254
5255         if (!rtnl_trylock())
5256                 return restart_syscall();
5257
5258         if (!write && !secret->initialized) {
5259                 err = -EIO;
5260                 goto out;
5261         }
5262
5263         if (!write) {
5264                 err = snprintf(str, sizeof(str), "%pI6",
5265                                &secret->secret);
5266                 if (err >= sizeof(str)) {
5267                         err = -EIO;
5268                         goto out;
5269                 }
5270         }
5271
5272         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5273         if (err || !write)
5274                 goto out;
5275
5276         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5277                 err = -EIO;
5278                 goto out;
5279         }
5280
5281         secret->initialized = true;
5282         secret->secret = addr;
5283
5284         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5285                 struct net_device *dev;
5286
5287                 for_each_netdev(net, dev) {
5288                         struct inet6_dev *idev = __in6_dev_get(dev);
5289
5290                         if (idev) {
5291                                 idev->addr_gen_mode =
5292                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5293                         }
5294                 }
5295         } else {
5296                 struct inet6_dev *idev = ctl->extra1;
5297
5298                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5299         }
5300
5301 out:
5302         rtnl_unlock();
5303
5304         return err;
5305 }
5306
5307 static struct addrconf_sysctl_table
5308 {
5309         struct ctl_table_header *sysctl_header;
5310         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5311 } addrconf_sysctl __read_mostly = {
5312         .sysctl_header = NULL,
5313         .addrconf_vars = {
5314                 {
5315                         .procname       = "forwarding",
5316                         .data           = &ipv6_devconf.forwarding,
5317                         .maxlen         = sizeof(int),
5318                         .mode           = 0644,
5319                         .proc_handler   = addrconf_sysctl_forward,
5320                 },
5321                 {
5322                         .procname       = "hop_limit",
5323                         .data           = &ipv6_devconf.hop_limit,
5324                         .maxlen         = sizeof(int),
5325                         .mode           = 0644,
5326                         .proc_handler   = proc_dointvec,
5327                 },
5328                 {
5329                         .procname       = "mtu",
5330                         .data           = &ipv6_devconf.mtu6,
5331                         .maxlen         = sizeof(int),
5332                         .mode           = 0644,
5333                         .proc_handler   = addrconf_sysctl_mtu,
5334                 },
5335                 {
5336                         .procname       = "accept_ra",
5337                         .data           = &ipv6_devconf.accept_ra,
5338                         .maxlen         = sizeof(int),
5339                         .mode           = 0644,
5340                         .proc_handler   = proc_dointvec,
5341                 },
5342                 {
5343                         .procname       = "accept_redirects",
5344                         .data           = &ipv6_devconf.accept_redirects,
5345                         .maxlen         = sizeof(int),
5346                         .mode           = 0644,
5347                         .proc_handler   = proc_dointvec,
5348                 },
5349                 {
5350                         .procname       = "autoconf",
5351                         .data           = &ipv6_devconf.autoconf,
5352                         .maxlen         = sizeof(int),
5353                         .mode           = 0644,
5354                         .proc_handler   = proc_dointvec,
5355                 },
5356                 {
5357                         .procname       = "dad_transmits",
5358                         .data           = &ipv6_devconf.dad_transmits,
5359                         .maxlen         = sizeof(int),
5360                         .mode           = 0644,
5361                         .proc_handler   = proc_dointvec,
5362                 },
5363                 {
5364                         .procname       = "router_solicitations",
5365                         .data           = &ipv6_devconf.rtr_solicits,
5366                         .maxlen         = sizeof(int),
5367                         .mode           = 0644,
5368                         .proc_handler   = proc_dointvec,
5369                 },
5370                 {
5371                         .procname       = "router_solicitation_interval",
5372                         .data           = &ipv6_devconf.rtr_solicit_interval,
5373                         .maxlen         = sizeof(int),
5374                         .mode           = 0644,
5375                         .proc_handler   = proc_dointvec_jiffies,
5376                 },
5377                 {
5378                         .procname       = "router_solicitation_delay",
5379                         .data           = &ipv6_devconf.rtr_solicit_delay,
5380                         .maxlen         = sizeof(int),
5381                         .mode           = 0644,
5382                         .proc_handler   = proc_dointvec_jiffies,
5383                 },
5384                 {
5385                         .procname       = "force_mld_version",
5386                         .data           = &ipv6_devconf.force_mld_version,
5387                         .maxlen         = sizeof(int),
5388                         .mode           = 0644,
5389                         .proc_handler   = proc_dointvec,
5390                 },
5391                 {
5392                         .procname       = "mldv1_unsolicited_report_interval",
5393                         .data           =
5394                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5395                         .maxlen         = sizeof(int),
5396                         .mode           = 0644,
5397                         .proc_handler   = proc_dointvec_ms_jiffies,
5398                 },
5399                 {
5400                         .procname       = "mldv2_unsolicited_report_interval",
5401                         .data           =
5402                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5403                         .maxlen         = sizeof(int),
5404                         .mode           = 0644,
5405                         .proc_handler   = proc_dointvec_ms_jiffies,
5406                 },
5407                 {
5408                         .procname       = "use_tempaddr",
5409                         .data           = &ipv6_devconf.use_tempaddr,
5410                         .maxlen         = sizeof(int),
5411                         .mode           = 0644,
5412                         .proc_handler   = proc_dointvec,
5413                 },
5414                 {
5415                         .procname       = "temp_valid_lft",
5416                         .data           = &ipv6_devconf.temp_valid_lft,
5417                         .maxlen         = sizeof(int),
5418                         .mode           = 0644,
5419                         .proc_handler   = proc_dointvec,
5420                 },
5421                 {
5422                         .procname       = "temp_prefered_lft",
5423                         .data           = &ipv6_devconf.temp_prefered_lft,
5424                         .maxlen         = sizeof(int),
5425                         .mode           = 0644,
5426                         .proc_handler   = proc_dointvec,
5427                 },
5428                 {
5429                         .procname       = "regen_max_retry",
5430                         .data           = &ipv6_devconf.regen_max_retry,
5431                         .maxlen         = sizeof(int),
5432                         .mode           = 0644,
5433                         .proc_handler   = proc_dointvec,
5434                 },
5435                 {
5436                         .procname       = "max_desync_factor",
5437                         .data           = &ipv6_devconf.max_desync_factor,
5438                         .maxlen         = sizeof(int),
5439                         .mode           = 0644,
5440                         .proc_handler   = proc_dointvec,
5441                 },
5442                 {
5443                         .procname       = "max_addresses",
5444                         .data           = &ipv6_devconf.max_addresses,
5445                         .maxlen         = sizeof(int),
5446                         .mode           = 0644,
5447                         .proc_handler   = proc_dointvec,
5448                 },
5449                 {
5450                         .procname       = "accept_ra_defrtr",
5451                         .data           = &ipv6_devconf.accept_ra_defrtr,
5452                         .maxlen         = sizeof(int),
5453                         .mode           = 0644,
5454                         .proc_handler   = proc_dointvec,
5455                 },
5456                 {
5457                         .procname       = "accept_ra_pinfo",
5458                         .data           = &ipv6_devconf.accept_ra_pinfo,
5459                         .maxlen         = sizeof(int),
5460                         .mode           = 0644,
5461                         .proc_handler   = proc_dointvec,
5462                 },
5463 #ifdef CONFIG_IPV6_ROUTER_PREF
5464                 {
5465                         .procname       = "accept_ra_rtr_pref",
5466                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5467                         .maxlen         = sizeof(int),
5468                         .mode           = 0644,
5469                         .proc_handler   = proc_dointvec,
5470                 },
5471                 {
5472                         .procname       = "router_probe_interval",
5473                         .data           = &ipv6_devconf.rtr_probe_interval,
5474                         .maxlen         = sizeof(int),
5475                         .mode           = 0644,
5476                         .proc_handler   = proc_dointvec_jiffies,
5477                 },
5478 #ifdef CONFIG_IPV6_ROUTE_INFO
5479                 {
5480                         .procname       = "accept_ra_rt_info_max_plen",
5481                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5482                         .maxlen         = sizeof(int),
5483                         .mode           = 0644,
5484                         .proc_handler   = proc_dointvec,
5485                 },
5486 #endif
5487 #endif
5488                 {
5489                         .procname       = "proxy_ndp",
5490                         .data           = &ipv6_devconf.proxy_ndp,
5491                         .maxlen         = sizeof(int),
5492                         .mode           = 0644,
5493                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5494                 },
5495                 {
5496                         .procname       = "accept_source_route",
5497                         .data           = &ipv6_devconf.accept_source_route,
5498                         .maxlen         = sizeof(int),
5499                         .mode           = 0644,
5500                         .proc_handler   = proc_dointvec,
5501                 },
5502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5503                 {
5504                         .procname       = "optimistic_dad",
5505                         .data           = &ipv6_devconf.optimistic_dad,
5506                         .maxlen         = sizeof(int),
5507                         .mode           = 0644,
5508                         .proc_handler   = proc_dointvec,
5509
5510                 },
5511                 {
5512                         .procname       = "use_optimistic",
5513                         .data           = &ipv6_devconf.use_optimistic,
5514                         .maxlen         = sizeof(int),
5515                         .mode           = 0644,
5516                         .proc_handler   = proc_dointvec,
5517
5518                 },
5519 #endif
5520 #ifdef CONFIG_IPV6_MROUTE
5521                 {
5522                         .procname       = "mc_forwarding",
5523                         .data           = &ipv6_devconf.mc_forwarding,
5524                         .maxlen         = sizeof(int),
5525                         .mode           = 0444,
5526                         .proc_handler   = proc_dointvec,
5527                 },
5528 #endif
5529                 {
5530                         .procname       = "disable_ipv6",
5531                         .data           = &ipv6_devconf.disable_ipv6,
5532                         .maxlen         = sizeof(int),
5533                         .mode           = 0644,
5534                         .proc_handler   = addrconf_sysctl_disable,
5535                 },
5536                 {
5537                         .procname       = "accept_dad",
5538                         .data           = &ipv6_devconf.accept_dad,
5539                         .maxlen         = sizeof(int),
5540                         .mode           = 0644,
5541                         .proc_handler   = proc_dointvec,
5542                 },
5543                 {
5544                         .procname       = "force_tllao",
5545                         .data           = &ipv6_devconf.force_tllao,
5546                         .maxlen         = sizeof(int),
5547                         .mode           = 0644,
5548                         .proc_handler   = proc_dointvec
5549                 },
5550                 {
5551                         .procname       = "ndisc_notify",
5552                         .data           = &ipv6_devconf.ndisc_notify,
5553                         .maxlen         = sizeof(int),
5554                         .mode           = 0644,
5555                         .proc_handler   = proc_dointvec
5556                 },
5557                 {
5558                         .procname       = "suppress_frag_ndisc",
5559                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5560                         .maxlen         = sizeof(int),
5561                         .mode           = 0644,
5562                         .proc_handler   = proc_dointvec
5563                 },
5564                 {
5565                         .procname       = "accept_ra_from_local",
5566                         .data           = &ipv6_devconf.accept_ra_from_local,
5567                         .maxlen         = sizeof(int),
5568                         .mode           = 0644,
5569                         .proc_handler   = proc_dointvec,
5570                 },
5571                 {
5572                         .procname       = "accept_ra_mtu",
5573                         .data           = &ipv6_devconf.accept_ra_mtu,
5574                         .maxlen         = sizeof(int),
5575                         .mode           = 0644,
5576                         .proc_handler   = proc_dointvec,
5577                 },
5578                 {
5579                         .procname       = "stable_secret",
5580                         .data           = &ipv6_devconf.stable_secret,
5581                         .maxlen         = IPV6_MAX_STRLEN,
5582                         .mode           = 0600,
5583                         .proc_handler   = addrconf_sysctl_stable_secret,
5584                 },
5585                 {
5586                         /* sentinel */
5587                 }
5588         },
5589 };
5590
5591 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5592                 struct inet6_dev *idev, struct ipv6_devconf *p)
5593 {
5594         int i;
5595         struct addrconf_sysctl_table *t;
5596         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5597
5598         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5599         if (!t)
5600                 goto out;
5601
5602         for (i = 0; t->addrconf_vars[i].data; i++) {
5603                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5604                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5605                 t->addrconf_vars[i].extra2 = net;
5606         }
5607
5608         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5609
5610         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5611         if (!t->sysctl_header)
5612                 goto free;
5613
5614         p->sysctl = t;
5615         return 0;
5616
5617 free:
5618         kfree(t);
5619 out:
5620         return -ENOBUFS;
5621 }
5622
5623 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5624 {
5625         struct addrconf_sysctl_table *t;
5626
5627         if (!p->sysctl)
5628                 return;
5629
5630         t = p->sysctl;
5631         p->sysctl = NULL;
5632         unregister_net_sysctl_table(t->sysctl_header);
5633         kfree(t);
5634 }
5635
5636 static int addrconf_sysctl_register(struct inet6_dev *idev)
5637 {
5638         int err;
5639
5640         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5641                 return -EINVAL;
5642
5643         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5644                                     &ndisc_ifinfo_sysctl_change);
5645         if (err)
5646                 return err;
5647         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5648                                          idev, &idev->cnf);
5649         if (err)
5650                 neigh_sysctl_unregister(idev->nd_parms);
5651
5652         return err;
5653 }
5654
5655 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5656 {
5657         __addrconf_sysctl_unregister(&idev->cnf);
5658         neigh_sysctl_unregister(idev->nd_parms);
5659 }
5660
5661
5662 #endif
5663
5664 static int __net_init addrconf_init_net(struct net *net)
5665 {
5666         int err = -ENOMEM;
5667         struct ipv6_devconf *all, *dflt;
5668
5669         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5670         if (!all)
5671                 goto err_alloc_all;
5672
5673         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5674         if (!dflt)
5675                 goto err_alloc_dflt;
5676
5677         /* these will be inherited by all namespaces */
5678         dflt->autoconf = ipv6_defaults.autoconf;
5679         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5680
5681         dflt->stable_secret.initialized = false;
5682         all->stable_secret.initialized = false;
5683
5684         net->ipv6.devconf_all = all;
5685         net->ipv6.devconf_dflt = dflt;
5686
5687 #ifdef CONFIG_SYSCTL
5688         err = __addrconf_sysctl_register(net, "all", NULL, all);
5689         if (err < 0)
5690                 goto err_reg_all;
5691
5692         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5693         if (err < 0)
5694                 goto err_reg_dflt;
5695 #endif
5696         return 0;
5697
5698 #ifdef CONFIG_SYSCTL
5699 err_reg_dflt:
5700         __addrconf_sysctl_unregister(all);
5701 err_reg_all:
5702         kfree(dflt);
5703 #endif
5704 err_alloc_dflt:
5705         kfree(all);
5706 err_alloc_all:
5707         return err;
5708 }
5709
5710 static void __net_exit addrconf_exit_net(struct net *net)
5711 {
5712 #ifdef CONFIG_SYSCTL
5713         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5714         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5715 #endif
5716         kfree(net->ipv6.devconf_dflt);
5717         kfree(net->ipv6.devconf_all);
5718 }
5719
5720 static struct pernet_operations addrconf_ops = {
5721         .init = addrconf_init_net,
5722         .exit = addrconf_exit_net,
5723 };
5724
5725 static struct rtnl_af_ops inet6_ops __read_mostly = {
5726         .family           = AF_INET6,
5727         .fill_link_af     = inet6_fill_link_af,
5728         .get_link_af_size = inet6_get_link_af_size,
5729         .validate_link_af = inet6_validate_link_af,
5730         .set_link_af      = inet6_set_link_af,
5731 };
5732
5733 /*
5734  *      Init / cleanup code
5735  */
5736
5737 int __init addrconf_init(void)
5738 {
5739         struct inet6_dev *idev;
5740         int i, err;
5741
5742         err = ipv6_addr_label_init();
5743         if (err < 0) {
5744                 pr_crit("%s: cannot initialize default policy table: %d\n",
5745                         __func__, err);
5746                 goto out;
5747         }
5748
5749         err = register_pernet_subsys(&addrconf_ops);
5750         if (err < 0)
5751                 goto out_addrlabel;
5752
5753         addrconf_wq = create_workqueue("ipv6_addrconf");
5754         if (!addrconf_wq) {
5755                 err = -ENOMEM;
5756                 goto out_nowq;
5757         }
5758
5759         /* The addrconf netdev notifier requires that loopback_dev
5760          * has it's ipv6 private information allocated and setup
5761          * before it can bring up and give link-local addresses
5762          * to other devices which are up.
5763          *
5764          * Unfortunately, loopback_dev is not necessarily the first
5765          * entry in the global dev_base list of net devices.  In fact,
5766          * it is likely to be the very last entry on that list.
5767          * So this causes the notifier registry below to try and
5768          * give link-local addresses to all devices besides loopback_dev
5769          * first, then loopback_dev, which cases all the non-loopback_dev
5770          * devices to fail to get a link-local address.
5771          *
5772          * So, as a temporary fix, allocate the ipv6 structure for
5773          * loopback_dev first by hand.
5774          * Longer term, all of the dependencies ipv6 has upon the loopback
5775          * device and it being up should be removed.
5776          */
5777         rtnl_lock();
5778         idev = ipv6_add_dev(init_net.loopback_dev);
5779         rtnl_unlock();
5780         if (IS_ERR(idev)) {
5781                 err = PTR_ERR(idev);
5782                 goto errlo;
5783         }
5784
5785         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5786                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5787
5788         register_netdevice_notifier(&ipv6_dev_notf);
5789
5790         addrconf_verify();
5791
5792         rtnl_af_register(&inet6_ops);
5793
5794         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5795                               NULL);
5796         if (err < 0)
5797                 goto errout;
5798
5799         /* Only the first call to __rtnl_register can fail */
5800         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5801         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5802         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5803                         inet6_dump_ifaddr, NULL);
5804         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5805                         inet6_dump_ifmcaddr, NULL);
5806         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5807                         inet6_dump_ifacaddr, NULL);
5808         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5809                         inet6_netconf_dump_devconf, NULL);
5810
5811         ipv6_addr_label_rtnl_register();
5812
5813         return 0;
5814 errout:
5815         rtnl_af_unregister(&inet6_ops);
5816         unregister_netdevice_notifier(&ipv6_dev_notf);
5817 errlo:
5818         destroy_workqueue(addrconf_wq);
5819 out_nowq:
5820         unregister_pernet_subsys(&addrconf_ops);
5821 out_addrlabel:
5822         ipv6_addr_label_cleanup();
5823 out:
5824         return err;
5825 }
5826
5827 void addrconf_cleanup(void)
5828 {
5829         struct net_device *dev;
5830         int i;
5831
5832         unregister_netdevice_notifier(&ipv6_dev_notf);
5833         unregister_pernet_subsys(&addrconf_ops);
5834         ipv6_addr_label_cleanup();
5835
5836         rtnl_lock();
5837
5838         __rtnl_af_unregister(&inet6_ops);
5839
5840         /* clean dev list */
5841         for_each_netdev(&init_net, dev) {
5842                 if (__in6_dev_get(dev) == NULL)
5843                         continue;
5844                 addrconf_ifdown(dev, 1);
5845         }
5846         addrconf_ifdown(init_net.loopback_dev, 2);
5847
5848         /*
5849          *      Check hash table.
5850          */
5851         spin_lock_bh(&addrconf_hash_lock);
5852         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5853                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5854         spin_unlock_bh(&addrconf_hash_lock);
5855         cancel_delayed_work(&addr_chk_work);
5856         rtnl_unlock();
5857
5858         destroy_workqueue(addrconf_wq);
5859 }