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 static int __ipv6_dev_get_saddr(struct net *net,
1362                                 struct ipv6_saddr_dst *dst,
1363                                 struct inet6_dev *idev,
1364                                 struct ipv6_saddr_score *scores,
1365                                 int hiscore_idx)
1366 {
1367         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1368
1369         read_lock_bh(&idev->lock);
1370         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1371                 int i;
1372
1373                 /*
1374                  * - Tentative Address (RFC2462 section 5.4)
1375                  *  - A tentative address is not considered
1376                  *    "assigned to an interface" in the traditional
1377                  *    sense, unless it is also flagged as optimistic.
1378                  * - Candidate Source Address (section 4)
1379                  *  - In any case, anycast addresses, multicast
1380                  *    addresses, and the unspecified address MUST
1381                  *    NOT be included in a candidate set.
1382                  */
1383                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1384                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1385                         continue;
1386
1387                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1388
1389                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1390                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1391                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1392                                             idev->dev->name);
1393                         continue;
1394                 }
1395
1396                 score->rule = -1;
1397                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1398
1399                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1400                         int minihiscore, miniscore;
1401
1402                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1403                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1404
1405                         if (minihiscore > miniscore) {
1406                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1407                                     score->scopedist > 0) {
1408                                         /*
1409                                          * special case:
1410                                          * each remaining entry
1411                                          * has too small (not enough)
1412                                          * scope, because ifa entries
1413                                          * are sorted by their scope
1414                                          * values.
1415                                          */
1416                                         goto out;
1417                                 }
1418                                 break;
1419                         } else if (minihiscore < miniscore) {
1420                                 if (hiscore->ifa)
1421                                         in6_ifa_put(hiscore->ifa);
1422
1423                                 in6_ifa_hold(score->ifa);
1424
1425                                 swap(hiscore, score);
1426                                 hiscore_idx = 1 - hiscore_idx;
1427
1428                                 /* restore our iterator */
1429                                 score->ifa = hiscore->ifa;
1430
1431                                 break;
1432                         }
1433                 }
1434         }
1435 out:
1436         read_unlock_bh(&idev->lock);
1437         return hiscore_idx;
1438 }
1439
1440 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1441                        const struct in6_addr *daddr, unsigned int prefs,
1442                        struct in6_addr *saddr)
1443 {
1444         struct ipv6_saddr_score scores[2], *hiscore;
1445         struct ipv6_saddr_dst dst;
1446         struct inet6_dev *idev;
1447         struct net_device *dev;
1448         int dst_type;
1449         bool use_oif_addr = false;
1450         int hiscore_idx = 0;
1451
1452         dst_type = __ipv6_addr_type(daddr);
1453         dst.addr = daddr;
1454         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1455         dst.scope = __ipv6_addr_src_scope(dst_type);
1456         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1457         dst.prefs = prefs;
1458
1459         scores[hiscore_idx].rule = -1;
1460         scores[hiscore_idx].ifa = NULL;
1461
1462         rcu_read_lock();
1463
1464         /* Candidate Source Address (section 4)
1465          *  - multicast and link-local destination address,
1466          *    the set of candidate source address MUST only
1467          *    include addresses assigned to interfaces
1468          *    belonging to the same link as the outgoing
1469          *    interface.
1470          * (- For site-local destination addresses, the
1471          *    set of candidate source addresses MUST only
1472          *    include addresses assigned to interfaces
1473          *    belonging to the same site as the outgoing
1474          *    interface.)
1475          */
1476         if (dst_dev) {
1477                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1478                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) {
1479                         idev = __in6_dev_get(dst_dev);
1480                         use_oif_addr = true;
1481                 }
1482         }
1483
1484         if (use_oif_addr) {
1485                 if (idev)
1486                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1487         } else {
1488                 for_each_netdev_rcu(net, dev) {
1489                         idev = __in6_dev_get(dev);
1490                         if (!idev)
1491                                 continue;
1492                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1493                 }
1494         }
1495         rcu_read_unlock();
1496
1497         hiscore = &scores[hiscore_idx];
1498         if (!hiscore->ifa)
1499                 return -EADDRNOTAVAIL;
1500
1501         *saddr = hiscore->ifa->addr;
1502         in6_ifa_put(hiscore->ifa);
1503         return 0;
1504 }
1505 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1506
1507 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1508                       u32 banned_flags)
1509 {
1510         struct inet6_ifaddr *ifp;
1511         int err = -EADDRNOTAVAIL;
1512
1513         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1514                 if (ifp->scope > IFA_LINK)
1515                         break;
1516                 if (ifp->scope == IFA_LINK &&
1517                     !(ifp->flags & banned_flags)) {
1518                         *addr = ifp->addr;
1519                         err = 0;
1520                         break;
1521                 }
1522         }
1523         return err;
1524 }
1525
1526 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1527                     u32 banned_flags)
1528 {
1529         struct inet6_dev *idev;
1530         int err = -EADDRNOTAVAIL;
1531
1532         rcu_read_lock();
1533         idev = __in6_dev_get(dev);
1534         if (idev) {
1535                 read_lock_bh(&idev->lock);
1536                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1537                 read_unlock_bh(&idev->lock);
1538         }
1539         rcu_read_unlock();
1540         return err;
1541 }
1542
1543 static int ipv6_count_addresses(struct inet6_dev *idev)
1544 {
1545         int cnt = 0;
1546         struct inet6_ifaddr *ifp;
1547
1548         read_lock_bh(&idev->lock);
1549         list_for_each_entry(ifp, &idev->addr_list, if_list)
1550                 cnt++;
1551         read_unlock_bh(&idev->lock);
1552         return cnt;
1553 }
1554
1555 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1556                   const struct net_device *dev, int strict)
1557 {
1558         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1559 }
1560 EXPORT_SYMBOL(ipv6_chk_addr);
1561
1562 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1563                             const struct net_device *dev, int strict,
1564                             u32 banned_flags)
1565 {
1566         struct inet6_ifaddr *ifp;
1567         unsigned int hash = inet6_addr_hash(addr);
1568         u32 ifp_flags;
1569
1570         rcu_read_lock_bh();
1571         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1572                 if (!net_eq(dev_net(ifp->idev->dev), net))
1573                         continue;
1574                 /* Decouple optimistic from tentative for evaluation here.
1575                  * Ban optimistic addresses explicitly, when required.
1576                  */
1577                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1578                             ? (ifp->flags&~IFA_F_TENTATIVE)
1579                             : ifp->flags;
1580                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1581                     !(ifp_flags&banned_flags) &&
1582                     (!dev || ifp->idev->dev == dev ||
1583                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1584                         rcu_read_unlock_bh();
1585                         return 1;
1586                 }
1587         }
1588
1589         rcu_read_unlock_bh();
1590         return 0;
1591 }
1592 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1593
1594 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1595                                struct net_device *dev)
1596 {
1597         unsigned int hash = inet6_addr_hash(addr);
1598         struct inet6_ifaddr *ifp;
1599
1600         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1601                 if (!net_eq(dev_net(ifp->idev->dev), net))
1602                         continue;
1603                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1604                         if (!dev || ifp->idev->dev == dev)
1605                                 return true;
1606                 }
1607         }
1608         return false;
1609 }
1610
1611 /* Compares an address/prefix_len with addresses on device @dev.
1612  * If one is found it returns true.
1613  */
1614 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1615         const unsigned int prefix_len, struct net_device *dev)
1616 {
1617         struct inet6_dev *idev;
1618         struct inet6_ifaddr *ifa;
1619         bool ret = false;
1620
1621         rcu_read_lock();
1622         idev = __in6_dev_get(dev);
1623         if (idev) {
1624                 read_lock_bh(&idev->lock);
1625                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1626                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1627                         if (ret)
1628                                 break;
1629                 }
1630                 read_unlock_bh(&idev->lock);
1631         }
1632         rcu_read_unlock();
1633
1634         return ret;
1635 }
1636 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1637
1638 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1639 {
1640         struct inet6_dev *idev;
1641         struct inet6_ifaddr *ifa;
1642         int     onlink;
1643
1644         onlink = 0;
1645         rcu_read_lock();
1646         idev = __in6_dev_get(dev);
1647         if (idev) {
1648                 read_lock_bh(&idev->lock);
1649                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1650                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1651                                                    ifa->prefix_len);
1652                         if (onlink)
1653                                 break;
1654                 }
1655                 read_unlock_bh(&idev->lock);
1656         }
1657         rcu_read_unlock();
1658         return onlink;
1659 }
1660 EXPORT_SYMBOL(ipv6_chk_prefix);
1661
1662 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1663                                      struct net_device *dev, int strict)
1664 {
1665         struct inet6_ifaddr *ifp, *result = NULL;
1666         unsigned int hash = inet6_addr_hash(addr);
1667
1668         rcu_read_lock_bh();
1669         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1670                 if (!net_eq(dev_net(ifp->idev->dev), net))
1671                         continue;
1672                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1673                         if (!dev || ifp->idev->dev == dev ||
1674                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1675                                 result = ifp;
1676                                 in6_ifa_hold(ifp);
1677                                 break;
1678                         }
1679                 }
1680         }
1681         rcu_read_unlock_bh();
1682
1683         return result;
1684 }
1685
1686 /* Gets referenced address, destroys ifaddr */
1687
1688 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1689 {
1690         if (ifp->flags&IFA_F_PERMANENT) {
1691                 spin_lock_bh(&ifp->lock);
1692                 addrconf_del_dad_work(ifp);
1693                 ifp->flags |= IFA_F_TENTATIVE;
1694                 if (dad_failed)
1695                         ifp->flags |= IFA_F_DADFAILED;
1696                 spin_unlock_bh(&ifp->lock);
1697                 if (dad_failed)
1698                         ipv6_ifa_notify(0, ifp);
1699                 in6_ifa_put(ifp);
1700         } else if (ifp->flags&IFA_F_TEMPORARY) {
1701                 struct inet6_ifaddr *ifpub;
1702                 spin_lock_bh(&ifp->lock);
1703                 ifpub = ifp->ifpub;
1704                 if (ifpub) {
1705                         in6_ifa_hold(ifpub);
1706                         spin_unlock_bh(&ifp->lock);
1707                         ipv6_create_tempaddr(ifpub, ifp);
1708                         in6_ifa_put(ifpub);
1709                 } else {
1710                         spin_unlock_bh(&ifp->lock);
1711                 }
1712                 ipv6_del_addr(ifp);
1713         } else {
1714                 ipv6_del_addr(ifp);
1715         }
1716 }
1717
1718 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1719 {
1720         int err = -ENOENT;
1721
1722         spin_lock_bh(&ifp->lock);
1723         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1724                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1725                 err = 0;
1726         }
1727         spin_unlock_bh(&ifp->lock);
1728
1729         return err;
1730 }
1731
1732 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1733 {
1734         struct in6_addr addr;
1735         struct inet6_dev *idev = ifp->idev;
1736         struct net *net = dev_net(ifp->idev->dev);
1737
1738         if (addrconf_dad_end(ifp)) {
1739                 in6_ifa_put(ifp);
1740                 return;
1741         }
1742
1743         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1744                              ifp->idev->dev->name, &ifp->addr);
1745
1746         spin_lock_bh(&ifp->lock);
1747
1748         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1749                 int scope = ifp->scope;
1750                 u32 flags = ifp->flags;
1751                 struct in6_addr new_addr;
1752                 struct inet6_ifaddr *ifp2;
1753                 u32 valid_lft, preferred_lft;
1754                 int pfxlen = ifp->prefix_len;
1755                 int retries = ifp->stable_privacy_retry + 1;
1756
1757                 if (retries > net->ipv6.sysctl.idgen_retries) {
1758                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1759                                              ifp->idev->dev->name);
1760                         goto errdad;
1761                 }
1762
1763                 new_addr = ifp->addr;
1764                 if (ipv6_generate_stable_address(&new_addr, retries,
1765                                                  idev))
1766                         goto errdad;
1767
1768                 valid_lft = ifp->valid_lft;
1769                 preferred_lft = ifp->prefered_lft;
1770
1771                 spin_unlock_bh(&ifp->lock);
1772
1773                 if (idev->cnf.max_addresses &&
1774                     ipv6_count_addresses(idev) >=
1775                     idev->cnf.max_addresses)
1776                         goto lock_errdad;
1777
1778                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1779                                      ifp->idev->dev->name);
1780
1781                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1782                                      scope, flags, valid_lft,
1783                                      preferred_lft);
1784                 if (IS_ERR(ifp2))
1785                         goto lock_errdad;
1786
1787                 spin_lock_bh(&ifp2->lock);
1788                 ifp2->stable_privacy_retry = retries;
1789                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1790                 spin_unlock_bh(&ifp2->lock);
1791
1792                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1793                 in6_ifa_put(ifp2);
1794 lock_errdad:
1795                 spin_lock_bh(&ifp->lock);
1796         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1797                 addr.s6_addr32[0] = htonl(0xfe800000);
1798                 addr.s6_addr32[1] = 0;
1799
1800                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1801                     ipv6_addr_equal(&ifp->addr, &addr)) {
1802                         /* DAD failed for link-local based on MAC address */
1803                         idev->cnf.disable_ipv6 = 1;
1804
1805                         pr_info("%s: IPv6 being disabled!\n",
1806                                 ifp->idev->dev->name);
1807                 }
1808         }
1809
1810 errdad:
1811         /* transition from _POSTDAD to _ERRDAD */
1812         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1813         spin_unlock_bh(&ifp->lock);
1814
1815         addrconf_mod_dad_work(ifp, 0);
1816 }
1817
1818 /* Join to solicited addr multicast group.
1819  * caller must hold RTNL */
1820 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1821 {
1822         struct in6_addr maddr;
1823
1824         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1825                 return;
1826
1827         addrconf_addr_solict_mult(addr, &maddr);
1828         ipv6_dev_mc_inc(dev, &maddr);
1829 }
1830
1831 /* caller must hold RTNL */
1832 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1833 {
1834         struct in6_addr maddr;
1835
1836         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1837                 return;
1838
1839         addrconf_addr_solict_mult(addr, &maddr);
1840         __ipv6_dev_mc_dec(idev, &maddr);
1841 }
1842
1843 /* caller must hold RTNL */
1844 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1845 {
1846         struct in6_addr addr;
1847
1848         if (ifp->prefix_len >= 127) /* RFC 6164 */
1849                 return;
1850         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1851         if (ipv6_addr_any(&addr))
1852                 return;
1853         __ipv6_dev_ac_inc(ifp->idev, &addr);
1854 }
1855
1856 /* caller must hold RTNL */
1857 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1858 {
1859         struct in6_addr addr;
1860
1861         if (ifp->prefix_len >= 127) /* RFC 6164 */
1862                 return;
1863         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1864         if (ipv6_addr_any(&addr))
1865                 return;
1866         __ipv6_dev_ac_dec(ifp->idev, &addr);
1867 }
1868
1869 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1870 {
1871         if (dev->addr_len != ETH_ALEN)
1872                 return -1;
1873         memcpy(eui, dev->dev_addr, 3);
1874         memcpy(eui + 5, dev->dev_addr + 3, 3);
1875
1876         /*
1877          * The zSeries OSA network cards can be shared among various
1878          * OS instances, but the OSA cards have only one MAC address.
1879          * This leads to duplicate address conflicts in conjunction
1880          * with IPv6 if more than one instance uses the same card.
1881          *
1882          * The driver for these cards can deliver a unique 16-bit
1883          * identifier for each instance sharing the same card.  It is
1884          * placed instead of 0xFFFE in the interface identifier.  The
1885          * "u" bit of the interface identifier is not inverted in this
1886          * case.  Hence the resulting interface identifier has local
1887          * scope according to RFC2373.
1888          */
1889         if (dev->dev_id) {
1890                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1891                 eui[4] = dev->dev_id & 0xFF;
1892         } else {
1893                 eui[3] = 0xFF;
1894                 eui[4] = 0xFE;
1895                 eui[0] ^= 2;
1896         }
1897         return 0;
1898 }
1899
1900 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1901 {
1902         if (dev->addr_len != IEEE802154_ADDR_LEN)
1903                 return -1;
1904         memcpy(eui, dev->dev_addr, 8);
1905         eui[0] ^= 2;
1906         return 0;
1907 }
1908
1909 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1910 {
1911         union fwnet_hwaddr *ha;
1912
1913         if (dev->addr_len != FWNET_ALEN)
1914                 return -1;
1915
1916         ha = (union fwnet_hwaddr *)dev->dev_addr;
1917
1918         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1919         eui[0] ^= 2;
1920         return 0;
1921 }
1922
1923 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1924 {
1925         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1926         if (dev->addr_len != ARCNET_ALEN)
1927                 return -1;
1928         memset(eui, 0, 7);
1929         eui[7] = *(u8 *)dev->dev_addr;
1930         return 0;
1931 }
1932
1933 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1934 {
1935         if (dev->addr_len != INFINIBAND_ALEN)
1936                 return -1;
1937         memcpy(eui, dev->dev_addr + 12, 8);
1938         eui[0] |= 2;
1939         return 0;
1940 }
1941
1942 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1943 {
1944         if (addr == 0)
1945                 return -1;
1946         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1947                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1948                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1949                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1950                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1951                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1952         eui[1] = 0;
1953         eui[2] = 0x5E;
1954         eui[3] = 0xFE;
1955         memcpy(eui + 4, &addr, 4);
1956         return 0;
1957 }
1958
1959 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1960 {
1961         if (dev->priv_flags & IFF_ISATAP)
1962                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1963         return -1;
1964 }
1965
1966 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1967 {
1968         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1969 }
1970
1971 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1972 {
1973         memcpy(eui, dev->perm_addr, 3);
1974         memcpy(eui + 5, dev->perm_addr + 3, 3);
1975         eui[3] = 0xFF;
1976         eui[4] = 0xFE;
1977         eui[0] ^= 2;
1978         return 0;
1979 }
1980
1981 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1982 {
1983         switch (dev->type) {
1984         case ARPHRD_ETHER:
1985         case ARPHRD_FDDI:
1986                 return addrconf_ifid_eui48(eui, dev);
1987         case ARPHRD_ARCNET:
1988                 return addrconf_ifid_arcnet(eui, dev);
1989         case ARPHRD_INFINIBAND:
1990                 return addrconf_ifid_infiniband(eui, dev);
1991         case ARPHRD_SIT:
1992                 return addrconf_ifid_sit(eui, dev);
1993         case ARPHRD_IPGRE:
1994                 return addrconf_ifid_gre(eui, dev);
1995         case ARPHRD_6LOWPAN:
1996         case ARPHRD_IEEE802154:
1997                 return addrconf_ifid_eui64(eui, dev);
1998         case ARPHRD_IEEE1394:
1999                 return addrconf_ifid_ieee1394(eui, dev);
2000         case ARPHRD_TUNNEL6:
2001                 return addrconf_ifid_ip6tnl(eui, dev);
2002         }
2003         return -1;
2004 }
2005
2006 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2007 {
2008         int err = -1;
2009         struct inet6_ifaddr *ifp;
2010
2011         read_lock_bh(&idev->lock);
2012         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2013                 if (ifp->scope > IFA_LINK)
2014                         break;
2015                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2016                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2017                         err = 0;
2018                         break;
2019                 }
2020         }
2021         read_unlock_bh(&idev->lock);
2022         return err;
2023 }
2024
2025 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2026 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2027 {
2028 regen:
2029         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2030         idev->rndid[0] &= ~0x02;
2031
2032         /*
2033          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2034          * check if generated address is not inappropriate
2035          *
2036          *  - Reserved subnet anycast (RFC 2526)
2037          *      11111101 11....11 1xxxxxxx
2038          *  - ISATAP (RFC4214) 6.1
2039          *      00-00-5E-FE-xx-xx-xx-xx
2040          *  - value 0
2041          *  - XXX: already assigned to an address on the device
2042          */
2043         if (idev->rndid[0] == 0xfd &&
2044             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2045             (idev->rndid[7]&0x80))
2046                 goto regen;
2047         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2048                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2049                         goto regen;
2050                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2051                         goto regen;
2052         }
2053 }
2054
2055 static void ipv6_regen_rndid(unsigned long data)
2056 {
2057         struct inet6_dev *idev = (struct inet6_dev *) data;
2058         unsigned long expires;
2059
2060         rcu_read_lock_bh();
2061         write_lock_bh(&idev->lock);
2062
2063         if (idev->dead)
2064                 goto out;
2065
2066         __ipv6_regen_rndid(idev);
2067
2068         expires = jiffies +
2069                 idev->cnf.temp_prefered_lft * HZ -
2070                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2071                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2072                 idev->cnf.max_desync_factor * HZ;
2073         if (time_before(expires, jiffies)) {
2074                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2075                         __func__, idev->dev->name);
2076                 goto out;
2077         }
2078
2079         if (!mod_timer(&idev->regen_timer, expires))
2080                 in6_dev_hold(idev);
2081
2082 out:
2083         write_unlock_bh(&idev->lock);
2084         rcu_read_unlock_bh();
2085         in6_dev_put(idev);
2086 }
2087
2088 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2089 {
2090         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2091                 __ipv6_regen_rndid(idev);
2092 }
2093
2094 /*
2095  *      Add prefix route.
2096  */
2097
2098 static void
2099 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2100                       unsigned long expires, u32 flags)
2101 {
2102         struct fib6_config cfg = {
2103                 .fc_table = RT6_TABLE_PREFIX,
2104                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2105                 .fc_ifindex = dev->ifindex,
2106                 .fc_expires = expires,
2107                 .fc_dst_len = plen,
2108                 .fc_flags = RTF_UP | flags,
2109                 .fc_nlinfo.nl_net = dev_net(dev),
2110                 .fc_protocol = RTPROT_KERNEL,
2111         };
2112
2113         cfg.fc_dst = *pfx;
2114
2115         /* Prevent useless cloning on PtP SIT.
2116            This thing is done here expecting that the whole
2117            class of non-broadcast devices need not cloning.
2118          */
2119 #if IS_ENABLED(CONFIG_IPV6_SIT)
2120         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2121                 cfg.fc_flags |= RTF_NONEXTHOP;
2122 #endif
2123
2124         ip6_route_add(&cfg);
2125 }
2126
2127
2128 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2129                                                   int plen,
2130                                                   const struct net_device *dev,
2131                                                   u32 flags, u32 noflags)
2132 {
2133         struct fib6_node *fn;
2134         struct rt6_info *rt = NULL;
2135         struct fib6_table *table;
2136
2137         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2138         if (!table)
2139                 return NULL;
2140
2141         read_lock_bh(&table->tb6_lock);
2142         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2143         if (!fn)
2144                 goto out;
2145
2146         noflags |= RTF_CACHE;
2147         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2148                 if (rt->dst.dev->ifindex != dev->ifindex)
2149                         continue;
2150                 if ((rt->rt6i_flags & flags) != flags)
2151                         continue;
2152                 if ((rt->rt6i_flags & noflags) != 0)
2153                         continue;
2154                 dst_hold(&rt->dst);
2155                 break;
2156         }
2157 out:
2158         read_unlock_bh(&table->tb6_lock);
2159         return rt;
2160 }
2161
2162
2163 /* Create "default" multicast route to the interface */
2164
2165 static void addrconf_add_mroute(struct net_device *dev)
2166 {
2167         struct fib6_config cfg = {
2168                 .fc_table = RT6_TABLE_LOCAL,
2169                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2170                 .fc_ifindex = dev->ifindex,
2171                 .fc_dst_len = 8,
2172                 .fc_flags = RTF_UP,
2173                 .fc_nlinfo.nl_net = dev_net(dev),
2174         };
2175
2176         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2177
2178         ip6_route_add(&cfg);
2179 }
2180
2181 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2182 {
2183         struct inet6_dev *idev;
2184
2185         ASSERT_RTNL();
2186
2187         idev = ipv6_find_idev(dev);
2188         if (!idev)
2189                 return ERR_PTR(-ENOBUFS);
2190
2191         if (idev->cnf.disable_ipv6)
2192                 return ERR_PTR(-EACCES);
2193
2194         /* Add default multicast route */
2195         if (!(dev->flags & IFF_LOOPBACK))
2196                 addrconf_add_mroute(dev);
2197
2198         return idev;
2199 }
2200
2201 static void manage_tempaddrs(struct inet6_dev *idev,
2202                              struct inet6_ifaddr *ifp,
2203                              __u32 valid_lft, __u32 prefered_lft,
2204                              bool create, unsigned long now)
2205 {
2206         u32 flags;
2207         struct inet6_ifaddr *ift;
2208
2209         read_lock_bh(&idev->lock);
2210         /* update all temporary addresses in the list */
2211         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2212                 int age, max_valid, max_prefered;
2213
2214                 if (ifp != ift->ifpub)
2215                         continue;
2216
2217                 /* RFC 4941 section 3.3:
2218                  * If a received option will extend the lifetime of a public
2219                  * address, the lifetimes of temporary addresses should
2220                  * be extended, subject to the overall constraint that no
2221                  * temporary addresses should ever remain "valid" or "preferred"
2222                  * for a time longer than (TEMP_VALID_LIFETIME) or
2223                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2224                  */
2225                 age = (now - ift->cstamp) / HZ;
2226                 max_valid = idev->cnf.temp_valid_lft - age;
2227                 if (max_valid < 0)
2228                         max_valid = 0;
2229
2230                 max_prefered = idev->cnf.temp_prefered_lft -
2231                                idev->cnf.max_desync_factor - age;
2232                 if (max_prefered < 0)
2233                         max_prefered = 0;
2234
2235                 if (valid_lft > max_valid)
2236                         valid_lft = max_valid;
2237
2238                 if (prefered_lft > max_prefered)
2239                         prefered_lft = max_prefered;
2240
2241                 spin_lock(&ift->lock);
2242                 flags = ift->flags;
2243                 ift->valid_lft = valid_lft;
2244                 ift->prefered_lft = prefered_lft;
2245                 ift->tstamp = now;
2246                 if (prefered_lft > 0)
2247                         ift->flags &= ~IFA_F_DEPRECATED;
2248
2249                 spin_unlock(&ift->lock);
2250                 if (!(flags&IFA_F_TENTATIVE))
2251                         ipv6_ifa_notify(0, ift);
2252         }
2253
2254         if ((create || list_empty(&idev->tempaddr_list)) &&
2255             idev->cnf.use_tempaddr > 0) {
2256                 /* When a new public address is created as described
2257                  * in [ADDRCONF], also create a new temporary address.
2258                  * Also create a temporary address if it's enabled but
2259                  * no temporary address currently exists.
2260                  */
2261                 read_unlock_bh(&idev->lock);
2262                 ipv6_create_tempaddr(ifp, NULL);
2263         } else {
2264                 read_unlock_bh(&idev->lock);
2265         }
2266 }
2267
2268 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2269 {
2270         struct prefix_info *pinfo;
2271         __u32 valid_lft;
2272         __u32 prefered_lft;
2273         int addr_type;
2274         u32 addr_flags = 0;
2275         struct inet6_dev *in6_dev;
2276         struct net *net = dev_net(dev);
2277
2278         pinfo = (struct prefix_info *) opt;
2279
2280         if (len < sizeof(struct prefix_info)) {
2281                 ADBG("addrconf: prefix option too short\n");
2282                 return;
2283         }
2284
2285         /*
2286          *      Validation checks ([ADDRCONF], page 19)
2287          */
2288
2289         addr_type = ipv6_addr_type(&pinfo->prefix);
2290
2291         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2292                 return;
2293
2294         valid_lft = ntohl(pinfo->valid);
2295         prefered_lft = ntohl(pinfo->prefered);
2296
2297         if (prefered_lft > valid_lft) {
2298                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2299                 return;
2300         }
2301
2302         in6_dev = in6_dev_get(dev);
2303
2304         if (!in6_dev) {
2305                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2306                                     dev->name);
2307                 return;
2308         }
2309
2310         /*
2311          *      Two things going on here:
2312          *      1) Add routes for on-link prefixes
2313          *      2) Configure prefixes with the auto flag set
2314          */
2315
2316         if (pinfo->onlink) {
2317                 struct rt6_info *rt;
2318                 unsigned long rt_expires;
2319
2320                 /* Avoid arithmetic overflow. Really, we could
2321                  * save rt_expires in seconds, likely valid_lft,
2322                  * but it would require division in fib gc, that it
2323                  * not good.
2324                  */
2325                 if (HZ > USER_HZ)
2326                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2327                 else
2328                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2329
2330                 if (addrconf_finite_timeout(rt_expires))
2331                         rt_expires *= HZ;
2332
2333                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2334                                                pinfo->prefix_len,
2335                                                dev,
2336                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2337                                                RTF_GATEWAY | RTF_DEFAULT);
2338
2339                 if (rt) {
2340                         /* Autoconf prefix route */
2341                         if (valid_lft == 0) {
2342                                 ip6_del_rt(rt);
2343                                 rt = NULL;
2344                         } else if (addrconf_finite_timeout(rt_expires)) {
2345                                 /* not infinity */
2346                                 rt6_set_expires(rt, jiffies + rt_expires);
2347                         } else {
2348                                 rt6_clean_expires(rt);
2349                         }
2350                 } else if (valid_lft) {
2351                         clock_t expires = 0;
2352                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2353                         if (addrconf_finite_timeout(rt_expires)) {
2354                                 /* not infinity */
2355                                 flags |= RTF_EXPIRES;
2356                                 expires = jiffies_to_clock_t(rt_expires);
2357                         }
2358                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2359                                               dev, expires, flags);
2360                 }
2361                 ip6_rt_put(rt);
2362         }
2363
2364         /* Try to figure out our local address for this prefix */
2365
2366         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2367                 struct inet6_ifaddr *ifp;
2368                 struct in6_addr addr;
2369                 int create = 0, update_lft = 0;
2370                 bool tokenized = false;
2371
2372                 if (pinfo->prefix_len == 64) {
2373                         memcpy(&addr, &pinfo->prefix, 8);
2374
2375                         if (!ipv6_addr_any(&in6_dev->token)) {
2376                                 read_lock_bh(&in6_dev->lock);
2377                                 memcpy(addr.s6_addr + 8,
2378                                        in6_dev->token.s6_addr + 8, 8);
2379                                 read_unlock_bh(&in6_dev->lock);
2380                                 tokenized = true;
2381                         } else if (in6_dev->addr_gen_mode ==
2382                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2383                                    !ipv6_generate_stable_address(&addr, 0,
2384                                                                  in6_dev)) {
2385                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2386                                 goto ok;
2387                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2388                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2389                                 in6_dev_put(in6_dev);
2390                                 return;
2391                         }
2392                         goto ok;
2393                 }
2394                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2395                                     pinfo->prefix_len);
2396                 in6_dev_put(in6_dev);
2397                 return;
2398
2399 ok:
2400
2401                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2402
2403                 if (!ifp && valid_lft) {
2404                         int max_addresses = in6_dev->cnf.max_addresses;
2405
2406 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2407                         if (in6_dev->cnf.optimistic_dad &&
2408                             !net->ipv6.devconf_all->forwarding && sllao)
2409                                 addr_flags = IFA_F_OPTIMISTIC;
2410 #endif
2411
2412                         /* Do not allow to create too much of autoconfigured
2413                          * addresses; this would be too easy way to crash kernel.
2414                          */
2415                         if (!max_addresses ||
2416                             ipv6_count_addresses(in6_dev) < max_addresses)
2417                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2418                                                     pinfo->prefix_len,
2419                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2420                                                     addr_flags, valid_lft,
2421                                                     prefered_lft);
2422
2423                         if (IS_ERR_OR_NULL(ifp)) {
2424                                 in6_dev_put(in6_dev);
2425                                 return;
2426                         }
2427
2428                         update_lft = 0;
2429                         create = 1;
2430                         spin_lock_bh(&ifp->lock);
2431                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2432                         ifp->cstamp = jiffies;
2433                         ifp->tokenized = tokenized;
2434                         spin_unlock_bh(&ifp->lock);
2435                         addrconf_dad_start(ifp);
2436                 }
2437
2438                 if (ifp) {
2439                         u32 flags;
2440                         unsigned long now;
2441                         u32 stored_lft;
2442
2443                         /* update lifetime (RFC2462 5.5.3 e) */
2444                         spin_lock_bh(&ifp->lock);
2445                         now = jiffies;
2446                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2447                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2448                         else
2449                                 stored_lft = 0;
2450                         if (!update_lft && !create && stored_lft) {
2451                                 const u32 minimum_lft = min_t(u32,
2452                                         stored_lft, MIN_VALID_LIFETIME);
2453                                 valid_lft = max(valid_lft, minimum_lft);
2454
2455                                 /* RFC4862 Section 5.5.3e:
2456                                  * "Note that the preferred lifetime of the
2457                                  *  corresponding address is always reset to
2458                                  *  the Preferred Lifetime in the received
2459                                  *  Prefix Information option, regardless of
2460                                  *  whether the valid lifetime is also reset or
2461                                  *  ignored."
2462                                  *
2463                                  * So we should always update prefered_lft here.
2464                                  */
2465                                 update_lft = 1;
2466                         }
2467
2468                         if (update_lft) {
2469                                 ifp->valid_lft = valid_lft;
2470                                 ifp->prefered_lft = prefered_lft;
2471                                 ifp->tstamp = now;
2472                                 flags = ifp->flags;
2473                                 ifp->flags &= ~IFA_F_DEPRECATED;
2474                                 spin_unlock_bh(&ifp->lock);
2475
2476                                 if (!(flags&IFA_F_TENTATIVE))
2477                                         ipv6_ifa_notify(0, ifp);
2478                         } else
2479                                 spin_unlock_bh(&ifp->lock);
2480
2481                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2482                                          create, now);
2483
2484                         in6_ifa_put(ifp);
2485                         addrconf_verify();
2486                 }
2487         }
2488         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2489         in6_dev_put(in6_dev);
2490 }
2491
2492 /*
2493  *      Set destination address.
2494  *      Special case for SIT interfaces where we create a new "virtual"
2495  *      device.
2496  */
2497 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2498 {
2499         struct in6_ifreq ireq;
2500         struct net_device *dev;
2501         int err = -EINVAL;
2502
2503         rtnl_lock();
2504
2505         err = -EFAULT;
2506         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2507                 goto err_exit;
2508
2509         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2510
2511         err = -ENODEV;
2512         if (!dev)
2513                 goto err_exit;
2514
2515 #if IS_ENABLED(CONFIG_IPV6_SIT)
2516         if (dev->type == ARPHRD_SIT) {
2517                 const struct net_device_ops *ops = dev->netdev_ops;
2518                 struct ifreq ifr;
2519                 struct ip_tunnel_parm p;
2520
2521                 err = -EADDRNOTAVAIL;
2522                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2523                         goto err_exit;
2524
2525                 memset(&p, 0, sizeof(p));
2526                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2527                 p.iph.saddr = 0;
2528                 p.iph.version = 4;
2529                 p.iph.ihl = 5;
2530                 p.iph.protocol = IPPROTO_IPV6;
2531                 p.iph.ttl = 64;
2532                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2533
2534                 if (ops->ndo_do_ioctl) {
2535                         mm_segment_t oldfs = get_fs();
2536
2537                         set_fs(KERNEL_DS);
2538                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2539                         set_fs(oldfs);
2540                 } else
2541                         err = -EOPNOTSUPP;
2542
2543                 if (err == 0) {
2544                         err = -ENOBUFS;
2545                         dev = __dev_get_by_name(net, p.name);
2546                         if (!dev)
2547                                 goto err_exit;
2548                         err = dev_open(dev);
2549                 }
2550         }
2551 #endif
2552
2553 err_exit:
2554         rtnl_unlock();
2555         return err;
2556 }
2557
2558 static int ipv6_mc_config(struct sock *sk, bool join,
2559                           const struct in6_addr *addr, int ifindex)
2560 {
2561         int ret;
2562
2563         ASSERT_RTNL();
2564
2565         lock_sock(sk);
2566         if (join)
2567                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2568         else
2569                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2570         release_sock(sk);
2571
2572         return ret;
2573 }
2574
2575 /*
2576  *      Manual configuration of address on an interface
2577  */
2578 static int inet6_addr_add(struct net *net, int ifindex,
2579                           const struct in6_addr *pfx,
2580                           const struct in6_addr *peer_pfx,
2581                           unsigned int plen, __u32 ifa_flags,
2582                           __u32 prefered_lft, __u32 valid_lft)
2583 {
2584         struct inet6_ifaddr *ifp;
2585         struct inet6_dev *idev;
2586         struct net_device *dev;
2587         unsigned long timeout;
2588         clock_t expires;
2589         int scope;
2590         u32 flags;
2591
2592         ASSERT_RTNL();
2593
2594         if (plen > 128)
2595                 return -EINVAL;
2596
2597         /* check the lifetime */
2598         if (!valid_lft || prefered_lft > valid_lft)
2599                 return -EINVAL;
2600
2601         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2602                 return -EINVAL;
2603
2604         dev = __dev_get_by_index(net, ifindex);
2605         if (!dev)
2606                 return -ENODEV;
2607
2608         idev = addrconf_add_dev(dev);
2609         if (IS_ERR(idev))
2610                 return PTR_ERR(idev);
2611
2612         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2613                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2614                                          true, pfx, ifindex);
2615
2616                 if (ret < 0)
2617                         return ret;
2618         }
2619
2620         scope = ipv6_addr_scope(pfx);
2621
2622         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2623         if (addrconf_finite_timeout(timeout)) {
2624                 expires = jiffies_to_clock_t(timeout * HZ);
2625                 valid_lft = timeout;
2626                 flags = RTF_EXPIRES;
2627         } else {
2628                 expires = 0;
2629                 flags = 0;
2630                 ifa_flags |= IFA_F_PERMANENT;
2631         }
2632
2633         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2634         if (addrconf_finite_timeout(timeout)) {
2635                 if (timeout == 0)
2636                         ifa_flags |= IFA_F_DEPRECATED;
2637                 prefered_lft = timeout;
2638         }
2639
2640         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2641                             valid_lft, prefered_lft);
2642
2643         if (!IS_ERR(ifp)) {
2644                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2645                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2646                                               expires, flags);
2647                 }
2648
2649                 /*
2650                  * Note that section 3.1 of RFC 4429 indicates
2651                  * that the Optimistic flag should not be set for
2652                  * manually configured addresses
2653                  */
2654                 addrconf_dad_start(ifp);
2655                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2656                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2657                                          true, jiffies);
2658                 in6_ifa_put(ifp);
2659                 addrconf_verify_rtnl();
2660                 return 0;
2661         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2662                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2663                                false, pfx, ifindex);
2664         }
2665
2666         return PTR_ERR(ifp);
2667 }
2668
2669 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2670                           const struct in6_addr *pfx, unsigned int plen)
2671 {
2672         struct inet6_ifaddr *ifp;
2673         struct inet6_dev *idev;
2674         struct net_device *dev;
2675
2676         if (plen > 128)
2677                 return -EINVAL;
2678
2679         dev = __dev_get_by_index(net, ifindex);
2680         if (!dev)
2681                 return -ENODEV;
2682
2683         idev = __in6_dev_get(dev);
2684         if (!idev)
2685                 return -ENXIO;
2686
2687         read_lock_bh(&idev->lock);
2688         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2689                 if (ifp->prefix_len == plen &&
2690                     ipv6_addr_equal(pfx, &ifp->addr)) {
2691                         in6_ifa_hold(ifp);
2692                         read_unlock_bh(&idev->lock);
2693
2694                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2695                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2696                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2697                                                  jiffies);
2698                         ipv6_del_addr(ifp);
2699                         addrconf_verify_rtnl();
2700                         if (ipv6_addr_is_multicast(pfx)) {
2701                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2702                                                false, pfx, dev->ifindex);
2703                         }
2704                         return 0;
2705                 }
2706         }
2707         read_unlock_bh(&idev->lock);
2708         return -EADDRNOTAVAIL;
2709 }
2710
2711
2712 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2713 {
2714         struct in6_ifreq ireq;
2715         int err;
2716
2717         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2718                 return -EPERM;
2719
2720         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2721                 return -EFAULT;
2722
2723         rtnl_lock();
2724         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2725                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2726                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2727         rtnl_unlock();
2728         return err;
2729 }
2730
2731 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2732 {
2733         struct in6_ifreq ireq;
2734         int err;
2735
2736         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2737                 return -EPERM;
2738
2739         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2740                 return -EFAULT;
2741
2742         rtnl_lock();
2743         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2744                              ireq.ifr6_prefixlen);
2745         rtnl_unlock();
2746         return err;
2747 }
2748
2749 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2750                      int plen, int scope)
2751 {
2752         struct inet6_ifaddr *ifp;
2753
2754         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2755                             scope, IFA_F_PERMANENT,
2756                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2757         if (!IS_ERR(ifp)) {
2758                 spin_lock_bh(&ifp->lock);
2759                 ifp->flags &= ~IFA_F_TENTATIVE;
2760                 spin_unlock_bh(&ifp->lock);
2761                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2762                 in6_ifa_put(ifp);
2763         }
2764 }
2765
2766 #if IS_ENABLED(CONFIG_IPV6_SIT)
2767 static void sit_add_v4_addrs(struct inet6_dev *idev)
2768 {
2769         struct in6_addr addr;
2770         struct net_device *dev;
2771         struct net *net = dev_net(idev->dev);
2772         int scope, plen;
2773         u32 pflags = 0;
2774
2775         ASSERT_RTNL();
2776
2777         memset(&addr, 0, sizeof(struct in6_addr));
2778         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2779
2780         if (idev->dev->flags&IFF_POINTOPOINT) {
2781                 addr.s6_addr32[0] = htonl(0xfe800000);
2782                 scope = IFA_LINK;
2783                 plen = 64;
2784         } else {
2785                 scope = IPV6_ADDR_COMPATv4;
2786                 plen = 96;
2787                 pflags |= RTF_NONEXTHOP;
2788         }
2789
2790         if (addr.s6_addr32[3]) {
2791                 add_addr(idev, &addr, plen, scope);
2792                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2793                 return;
2794         }
2795
2796         for_each_netdev(net, dev) {
2797                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2798                 if (in_dev && (dev->flags & IFF_UP)) {
2799                         struct in_ifaddr *ifa;
2800
2801                         int flag = scope;
2802
2803                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2804
2805                                 addr.s6_addr32[3] = ifa->ifa_local;
2806
2807                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2808                                         continue;
2809                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2810                                         if (idev->dev->flags&IFF_POINTOPOINT)
2811                                                 continue;
2812                                         flag |= IFA_HOST;
2813                                 }
2814
2815                                 add_addr(idev, &addr, plen, flag);
2816                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2817                                                       pflags);
2818                         }
2819                 }
2820         }
2821 }
2822 #endif
2823
2824 static void init_loopback(struct net_device *dev)
2825 {
2826         struct inet6_dev  *idev;
2827         struct net_device *sp_dev;
2828         struct inet6_ifaddr *sp_ifa;
2829         struct rt6_info *sp_rt;
2830
2831         /* ::1 */
2832
2833         ASSERT_RTNL();
2834
2835         idev = ipv6_find_idev(dev);
2836         if (!idev) {
2837                 pr_debug("%s: add_dev failed\n", __func__);
2838                 return;
2839         }
2840
2841         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2842
2843         /* Add routes to other interface's IPv6 addresses */
2844         for_each_netdev(dev_net(dev), sp_dev) {
2845                 if (!strcmp(sp_dev->name, dev->name))
2846                         continue;
2847
2848                 idev = __in6_dev_get(sp_dev);
2849                 if (!idev)
2850                         continue;
2851
2852                 read_lock_bh(&idev->lock);
2853                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2854
2855                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2856                                 continue;
2857
2858                         if (sp_ifa->rt) {
2859                                 /* This dst has been added to garbage list when
2860                                  * lo device down, release this obsolete dst and
2861                                  * reallocate a new router for ifa.
2862                                  */
2863                                 if (sp_ifa->rt->dst.obsolete > 0) {
2864                                         ip6_rt_put(sp_ifa->rt);
2865                                         sp_ifa->rt = NULL;
2866                                 } else {
2867                                         continue;
2868                                 }
2869                         }
2870
2871                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2872
2873                         /* Failure cases are ignored */
2874                         if (!IS_ERR(sp_rt)) {
2875                                 sp_ifa->rt = sp_rt;
2876                                 ip6_ins_rt(sp_rt);
2877                         }
2878                 }
2879                 read_unlock_bh(&idev->lock);
2880         }
2881 }
2882
2883 static void addrconf_add_linklocal(struct inet6_dev *idev,
2884                                    const struct in6_addr *addr, u32 flags)
2885 {
2886         struct inet6_ifaddr *ifp;
2887         u32 addr_flags = flags | IFA_F_PERMANENT;
2888
2889 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2890         if (idev->cnf.optimistic_dad &&
2891             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2892                 addr_flags |= IFA_F_OPTIMISTIC;
2893 #endif
2894
2895         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2896                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2897         if (!IS_ERR(ifp)) {
2898                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2899                 addrconf_dad_start(ifp);
2900                 in6_ifa_put(ifp);
2901         }
2902 }
2903
2904 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2905 {
2906         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2907                 return true;
2908
2909         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2910             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2911                 return true;
2912
2913         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2914             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2915                 return true;
2916
2917         return false;
2918 }
2919
2920 static int ipv6_generate_stable_address(struct in6_addr *address,
2921                                         u8 dad_count,
2922                                         const struct inet6_dev *idev)
2923 {
2924         static DEFINE_SPINLOCK(lock);
2925         static __u32 digest[SHA_DIGEST_WORDS];
2926         static __u32 workspace[SHA_WORKSPACE_WORDS];
2927
2928         static union {
2929                 char __data[SHA_MESSAGE_BYTES];
2930                 struct {
2931                         struct in6_addr secret;
2932                         __be32 prefix[2];
2933                         unsigned char hwaddr[MAX_ADDR_LEN];
2934                         u8 dad_count;
2935                 } __packed;
2936         } data;
2937
2938         struct in6_addr secret;
2939         struct in6_addr temp;
2940         struct net *net = dev_net(idev->dev);
2941
2942         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2943
2944         if (idev->cnf.stable_secret.initialized)
2945                 secret = idev->cnf.stable_secret.secret;
2946         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
2947                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
2948         else
2949                 return -1;
2950
2951 retry:
2952         spin_lock_bh(&lock);
2953
2954         sha_init(digest);
2955         memset(&data, 0, sizeof(data));
2956         memset(workspace, 0, sizeof(workspace));
2957         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
2958         data.prefix[0] = address->s6_addr32[0];
2959         data.prefix[1] = address->s6_addr32[1];
2960         data.secret = secret;
2961         data.dad_count = dad_count;
2962
2963         sha_transform(digest, data.__data, workspace);
2964
2965         temp = *address;
2966         temp.s6_addr32[2] = (__force __be32)digest[0];
2967         temp.s6_addr32[3] = (__force __be32)digest[1];
2968
2969         spin_unlock_bh(&lock);
2970
2971         if (ipv6_reserved_interfaceid(temp)) {
2972                 dad_count++;
2973                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
2974                         return -1;
2975                 goto retry;
2976         }
2977
2978         *address = temp;
2979         return 0;
2980 }
2981
2982 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2983 {
2984         struct in6_addr addr;
2985
2986         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2987
2988         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
2989                 if (!ipv6_generate_stable_address(&addr, 0, idev))
2990                         addrconf_add_linklocal(idev, &addr,
2991                                                IFA_F_STABLE_PRIVACY);
2992                 else if (prefix_route)
2993                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2994         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2995                 /* addrconf_add_linklocal also adds a prefix_route and we
2996                  * only need to care about prefix routes if ipv6_generate_eui64
2997                  * couldn't generate one.
2998                  */
2999                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3000                         addrconf_add_linklocal(idev, &addr, 0);
3001                 else if (prefix_route)
3002                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3003         }
3004 }
3005
3006 static void addrconf_dev_config(struct net_device *dev)
3007 {
3008         struct inet6_dev *idev;
3009
3010         ASSERT_RTNL();
3011
3012         if ((dev->type != ARPHRD_ETHER) &&
3013             (dev->type != ARPHRD_FDDI) &&
3014             (dev->type != ARPHRD_ARCNET) &&
3015             (dev->type != ARPHRD_INFINIBAND) &&
3016             (dev->type != ARPHRD_IEEE802154) &&
3017             (dev->type != ARPHRD_IEEE1394) &&
3018             (dev->type != ARPHRD_TUNNEL6) &&
3019             (dev->type != ARPHRD_6LOWPAN)) {
3020                 /* Alas, we support only Ethernet autoconfiguration. */
3021                 return;
3022         }
3023
3024         idev = addrconf_add_dev(dev);
3025         if (IS_ERR(idev))
3026                 return;
3027
3028         addrconf_addr_gen(idev, false);
3029 }
3030
3031 #if IS_ENABLED(CONFIG_IPV6_SIT)
3032 static void addrconf_sit_config(struct net_device *dev)
3033 {
3034         struct inet6_dev *idev;
3035
3036         ASSERT_RTNL();
3037
3038         /*
3039          * Configure the tunnel with one of our IPv4
3040          * addresses... we should configure all of
3041          * our v4 addrs in the tunnel
3042          */
3043
3044         idev = ipv6_find_idev(dev);
3045         if (!idev) {
3046                 pr_debug("%s: add_dev failed\n", __func__);
3047                 return;
3048         }
3049
3050         if (dev->priv_flags & IFF_ISATAP) {
3051                 addrconf_addr_gen(idev, false);
3052                 return;
3053         }
3054
3055         sit_add_v4_addrs(idev);
3056
3057         if (dev->flags&IFF_POINTOPOINT)
3058                 addrconf_add_mroute(dev);
3059 }
3060 #endif
3061
3062 #if IS_ENABLED(CONFIG_NET_IPGRE)
3063 static void addrconf_gre_config(struct net_device *dev)
3064 {
3065         struct inet6_dev *idev;
3066
3067         ASSERT_RTNL();
3068
3069         idev = ipv6_find_idev(dev);
3070         if (!idev) {
3071                 pr_debug("%s: add_dev failed\n", __func__);
3072                 return;
3073         }
3074
3075         addrconf_addr_gen(idev, true);
3076 }
3077 #endif
3078
3079 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3080                            void *ptr)
3081 {
3082         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3083         struct inet6_dev *idev = __in6_dev_get(dev);
3084         int run_pending = 0;
3085         int err;
3086
3087         switch (event) {
3088         case NETDEV_REGISTER:
3089                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3090                         idev = ipv6_add_dev(dev);
3091                         if (IS_ERR(idev))
3092                                 return notifier_from_errno(PTR_ERR(idev));
3093                 }
3094                 break;
3095
3096         case NETDEV_UP:
3097         case NETDEV_CHANGE:
3098                 if (dev->flags & IFF_SLAVE)
3099                         break;
3100
3101                 if (idev && idev->cnf.disable_ipv6)
3102                         break;
3103
3104                 if (event == NETDEV_UP) {
3105                         if (!addrconf_qdisc_ok(dev)) {
3106                                 /* device is not ready yet. */
3107                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3108                                         dev->name);
3109                                 break;
3110                         }
3111
3112                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3113                                 idev = ipv6_add_dev(dev);
3114
3115                         if (!IS_ERR_OR_NULL(idev)) {
3116                                 idev->if_flags |= IF_READY;
3117                                 run_pending = 1;
3118                         }
3119                 } else {
3120                         if (!addrconf_qdisc_ok(dev)) {
3121                                 /* device is still not ready. */
3122                                 break;
3123                         }
3124
3125                         if (idev) {
3126                                 if (idev->if_flags & IF_READY)
3127                                         /* device is already configured. */
3128                                         break;
3129                                 idev->if_flags |= IF_READY;
3130                         }
3131
3132                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3133                                 dev->name);
3134
3135                         run_pending = 1;
3136                 }
3137
3138                 switch (dev->type) {
3139 #if IS_ENABLED(CONFIG_IPV6_SIT)
3140                 case ARPHRD_SIT:
3141                         addrconf_sit_config(dev);
3142                         break;
3143 #endif
3144 #if IS_ENABLED(CONFIG_NET_IPGRE)
3145                 case ARPHRD_IPGRE:
3146                         addrconf_gre_config(dev);
3147                         break;
3148 #endif
3149                 case ARPHRD_LOOPBACK:
3150                         init_loopback(dev);
3151                         break;
3152
3153                 default:
3154                         addrconf_dev_config(dev);
3155                         break;
3156                 }
3157
3158                 if (!IS_ERR_OR_NULL(idev)) {
3159                         if (run_pending)
3160                                 addrconf_dad_run(idev);
3161
3162                         /*
3163                          * If the MTU changed during the interface down,
3164                          * when the interface up, the changed MTU must be
3165                          * reflected in the idev as well as routers.
3166                          */
3167                         if (idev->cnf.mtu6 != dev->mtu &&
3168                             dev->mtu >= IPV6_MIN_MTU) {
3169                                 rt6_mtu_change(dev, dev->mtu);
3170                                 idev->cnf.mtu6 = dev->mtu;
3171                         }
3172                         idev->tstamp = jiffies;
3173                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3174
3175                         /*
3176                          * If the changed mtu during down is lower than
3177                          * IPV6_MIN_MTU stop IPv6 on this interface.
3178                          */
3179                         if (dev->mtu < IPV6_MIN_MTU)
3180                                 addrconf_ifdown(dev, 1);
3181                 }
3182                 break;
3183
3184         case NETDEV_CHANGEMTU:
3185                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3186                         rt6_mtu_change(dev, dev->mtu);
3187                         idev->cnf.mtu6 = dev->mtu;
3188                         break;
3189                 }
3190
3191                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3192                         idev = ipv6_add_dev(dev);
3193                         if (!IS_ERR(idev))
3194                                 break;
3195                 }
3196
3197                 /*
3198                  * if MTU under IPV6_MIN_MTU.
3199                  * Stop IPv6 on this interface.
3200                  */
3201
3202         case NETDEV_DOWN:
3203         case NETDEV_UNREGISTER:
3204                 /*
3205                  *      Remove all addresses from this interface.
3206                  */
3207                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3208                 break;
3209
3210         case NETDEV_CHANGENAME:
3211                 if (idev) {
3212                         snmp6_unregister_dev(idev);
3213                         addrconf_sysctl_unregister(idev);
3214                         err = addrconf_sysctl_register(idev);
3215                         if (err)
3216                                 return notifier_from_errno(err);
3217                         err = snmp6_register_dev(idev);
3218                         if (err) {
3219                                 addrconf_sysctl_unregister(idev);
3220                                 return notifier_from_errno(err);
3221                         }
3222                 }
3223                 break;
3224
3225         case NETDEV_PRE_TYPE_CHANGE:
3226         case NETDEV_POST_TYPE_CHANGE:
3227                 addrconf_type_change(dev, event);
3228                 break;
3229         }
3230
3231         return NOTIFY_OK;
3232 }
3233
3234 /*
3235  *      addrconf module should be notified of a device going up
3236  */
3237 static struct notifier_block ipv6_dev_notf = {
3238         .notifier_call = addrconf_notify,
3239 };
3240
3241 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3242 {
3243         struct inet6_dev *idev;
3244         ASSERT_RTNL();
3245
3246         idev = __in6_dev_get(dev);
3247
3248         if (event == NETDEV_POST_TYPE_CHANGE)
3249                 ipv6_mc_remap(idev);
3250         else if (event == NETDEV_PRE_TYPE_CHANGE)
3251                 ipv6_mc_unmap(idev);
3252 }
3253
3254 static int addrconf_ifdown(struct net_device *dev, int how)
3255 {
3256         struct net *net = dev_net(dev);
3257         struct inet6_dev *idev;
3258         struct inet6_ifaddr *ifa;
3259         int state, i;
3260
3261         ASSERT_RTNL();
3262
3263         rt6_ifdown(net, dev);
3264         neigh_ifdown(&nd_tbl, dev);
3265
3266         idev = __in6_dev_get(dev);
3267         if (!idev)
3268                 return -ENODEV;
3269
3270         /*
3271          * Step 1: remove reference to ipv6 device from parent device.
3272          *         Do not dev_put!
3273          */
3274         if (how) {
3275                 idev->dead = 1;
3276
3277                 /* protected by rtnl_lock */
3278                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3279
3280                 /* Step 1.5: remove snmp6 entry */
3281                 snmp6_unregister_dev(idev);
3282
3283         }
3284
3285         /* Step 2: clear hash table */
3286         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3287                 struct hlist_head *h = &inet6_addr_lst[i];
3288
3289                 spin_lock_bh(&addrconf_hash_lock);
3290 restart:
3291                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3292                         if (ifa->idev == idev) {
3293                                 hlist_del_init_rcu(&ifa->addr_lst);
3294                                 addrconf_del_dad_work(ifa);
3295                                 goto restart;
3296                         }
3297                 }
3298                 spin_unlock_bh(&addrconf_hash_lock);
3299         }
3300
3301         write_lock_bh(&idev->lock);
3302
3303         addrconf_del_rs_timer(idev);
3304
3305         /* Step 2: clear flags for stateless addrconf */
3306         if (!how)
3307                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3308
3309         if (how && del_timer(&idev->regen_timer))
3310                 in6_dev_put(idev);
3311
3312         /* Step 3: clear tempaddr list */
3313         while (!list_empty(&idev->tempaddr_list)) {
3314                 ifa = list_first_entry(&idev->tempaddr_list,
3315                                        struct inet6_ifaddr, tmp_list);
3316                 list_del(&ifa->tmp_list);
3317                 write_unlock_bh(&idev->lock);
3318                 spin_lock_bh(&ifa->lock);
3319
3320                 if (ifa->ifpub) {
3321                         in6_ifa_put(ifa->ifpub);
3322                         ifa->ifpub = NULL;
3323                 }
3324                 spin_unlock_bh(&ifa->lock);
3325                 in6_ifa_put(ifa);
3326                 write_lock_bh(&idev->lock);
3327         }
3328
3329         while (!list_empty(&idev->addr_list)) {
3330                 ifa = list_first_entry(&idev->addr_list,
3331                                        struct inet6_ifaddr, if_list);
3332                 addrconf_del_dad_work(ifa);
3333
3334                 list_del(&ifa->if_list);
3335
3336                 write_unlock_bh(&idev->lock);
3337
3338                 spin_lock_bh(&ifa->lock);
3339                 state = ifa->state;
3340                 ifa->state = INET6_IFADDR_STATE_DEAD;
3341                 spin_unlock_bh(&ifa->lock);
3342
3343                 if (state != INET6_IFADDR_STATE_DEAD) {
3344                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3345                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3346                 }
3347                 in6_ifa_put(ifa);
3348
3349                 write_lock_bh(&idev->lock);
3350         }
3351
3352         write_unlock_bh(&idev->lock);
3353
3354         /* Step 5: Discard anycast and multicast list */
3355         if (how) {
3356                 ipv6_ac_destroy_dev(idev);
3357                 ipv6_mc_destroy_dev(idev);
3358         } else {
3359                 ipv6_mc_down(idev);
3360         }
3361
3362         idev->tstamp = jiffies;
3363
3364         /* Last: Shot the device (if unregistered) */
3365         if (how) {
3366                 addrconf_sysctl_unregister(idev);
3367                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3368                 neigh_ifdown(&nd_tbl, dev);
3369                 in6_dev_put(idev);
3370         }
3371         return 0;
3372 }
3373
3374 static void addrconf_rs_timer(unsigned long data)
3375 {
3376         struct inet6_dev *idev = (struct inet6_dev *)data;
3377         struct net_device *dev = idev->dev;
3378         struct in6_addr lladdr;
3379
3380         write_lock(&idev->lock);
3381         if (idev->dead || !(idev->if_flags & IF_READY))
3382                 goto out;
3383
3384         if (!ipv6_accept_ra(idev))
3385                 goto out;
3386
3387         /* Announcement received after solicitation was sent */
3388         if (idev->if_flags & IF_RA_RCVD)
3389                 goto out;
3390
3391         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3392                 write_unlock(&idev->lock);
3393                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3394                         ndisc_send_rs(dev, &lladdr,
3395                                       &in6addr_linklocal_allrouters);
3396                 else
3397                         goto put;
3398
3399                 write_lock(&idev->lock);
3400                 /* The wait after the last probe can be shorter */
3401                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3402                                              idev->cnf.rtr_solicits) ?
3403                                       idev->cnf.rtr_solicit_delay :
3404                                       idev->cnf.rtr_solicit_interval);
3405         } else {
3406                 /*
3407                  * Note: we do not support deprecated "all on-link"
3408                  * assumption any longer.
3409                  */
3410                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3411         }
3412
3413 out:
3414         write_unlock(&idev->lock);
3415 put:
3416         in6_dev_put(idev);
3417 }
3418
3419 /*
3420  *      Duplicate Address Detection
3421  */
3422 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3423 {
3424         unsigned long rand_num;
3425         struct inet6_dev *idev = ifp->idev;
3426
3427         if (ifp->flags & IFA_F_OPTIMISTIC)
3428                 rand_num = 0;
3429         else
3430                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3431
3432         ifp->dad_probes = idev->cnf.dad_transmits;
3433         addrconf_mod_dad_work(ifp, rand_num);
3434 }
3435
3436 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3437 {
3438         struct inet6_dev *idev = ifp->idev;
3439         struct net_device *dev = idev->dev;
3440
3441         addrconf_join_solict(dev, &ifp->addr);
3442
3443         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3444
3445         read_lock_bh(&idev->lock);
3446         spin_lock(&ifp->lock);
3447         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3448                 goto out;
3449
3450         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3451             idev->cnf.accept_dad < 1 ||
3452             !(ifp->flags&IFA_F_TENTATIVE) ||
3453             ifp->flags & IFA_F_NODAD) {
3454                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3455                 spin_unlock(&ifp->lock);
3456                 read_unlock_bh(&idev->lock);
3457
3458                 addrconf_dad_completed(ifp);
3459                 return;
3460         }
3461
3462         if (!(idev->if_flags & IF_READY)) {
3463                 spin_unlock(&ifp->lock);
3464                 read_unlock_bh(&idev->lock);
3465                 /*
3466                  * If the device is not ready:
3467                  * - keep it tentative if it is a permanent address.
3468                  * - otherwise, kill it.
3469                  */
3470                 in6_ifa_hold(ifp);
3471                 addrconf_dad_stop(ifp, 0);
3472                 return;
3473         }
3474
3475         /*
3476          * Optimistic nodes can start receiving
3477          * Frames right away
3478          */
3479         if (ifp->flags & IFA_F_OPTIMISTIC) {
3480                 ip6_ins_rt(ifp->rt);
3481                 if (ipv6_use_optimistic_addr(idev)) {
3482                         /* Because optimistic nodes can use this address,
3483                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3484                          */
3485                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3486                 }
3487         }
3488
3489         addrconf_dad_kick(ifp);
3490 out:
3491         spin_unlock(&ifp->lock);
3492         read_unlock_bh(&idev->lock);
3493 }
3494
3495 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3496 {
3497         bool begin_dad = false;
3498
3499         spin_lock_bh(&ifp->lock);
3500         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3501                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3502                 begin_dad = true;
3503         }
3504         spin_unlock_bh(&ifp->lock);
3505
3506         if (begin_dad)
3507                 addrconf_mod_dad_work(ifp, 0);
3508 }
3509
3510 static void addrconf_dad_work(struct work_struct *w)
3511 {
3512         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3513                                                 struct inet6_ifaddr,
3514                                                 dad_work);
3515         struct inet6_dev *idev = ifp->idev;
3516         struct in6_addr mcaddr;
3517
3518         enum {
3519                 DAD_PROCESS,
3520                 DAD_BEGIN,
3521                 DAD_ABORT,
3522         } action = DAD_PROCESS;
3523
3524         rtnl_lock();
3525
3526         spin_lock_bh(&ifp->lock);
3527         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3528                 action = DAD_BEGIN;
3529                 ifp->state = INET6_IFADDR_STATE_DAD;
3530         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3531                 action = DAD_ABORT;
3532                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3533         }
3534         spin_unlock_bh(&ifp->lock);
3535
3536         if (action == DAD_BEGIN) {
3537                 addrconf_dad_begin(ifp);
3538                 goto out;
3539         } else if (action == DAD_ABORT) {
3540                 addrconf_dad_stop(ifp, 1);
3541                 goto out;
3542         }
3543
3544         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3545                 goto out;
3546
3547         write_lock_bh(&idev->lock);
3548         if (idev->dead || !(idev->if_flags & IF_READY)) {
3549                 write_unlock_bh(&idev->lock);
3550                 goto out;
3551         }
3552
3553         spin_lock(&ifp->lock);
3554         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3555                 spin_unlock(&ifp->lock);
3556                 write_unlock_bh(&idev->lock);
3557                 goto out;
3558         }
3559
3560         if (ifp->dad_probes == 0) {
3561                 /*
3562                  * DAD was successful
3563                  */
3564
3565                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3566                 spin_unlock(&ifp->lock);
3567                 write_unlock_bh(&idev->lock);
3568
3569                 addrconf_dad_completed(ifp);
3570
3571                 goto out;
3572         }
3573
3574         ifp->dad_probes--;
3575         addrconf_mod_dad_work(ifp,
3576                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3577         spin_unlock(&ifp->lock);
3578         write_unlock_bh(&idev->lock);
3579
3580         /* send a neighbour solicitation for our addr */
3581         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3582         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3583 out:
3584         in6_ifa_put(ifp);
3585         rtnl_unlock();
3586 }
3587
3588 /* ifp->idev must be at least read locked */
3589 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3590 {
3591         struct inet6_ifaddr *ifpiter;
3592         struct inet6_dev *idev = ifp->idev;
3593
3594         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3595                 if (ifpiter->scope > IFA_LINK)
3596                         break;
3597                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3598                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3599                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3600                     IFA_F_PERMANENT)
3601                         return false;
3602         }
3603         return true;
3604 }
3605
3606 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3607 {
3608         struct net_device *dev = ifp->idev->dev;
3609         struct in6_addr lladdr;
3610         bool send_rs, send_mld;
3611
3612         addrconf_del_dad_work(ifp);
3613
3614         /*
3615          *      Configure the address for reception. Now it is valid.
3616          */
3617
3618         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3619
3620         /* If added prefix is link local and we are prepared to process
3621            router advertisements, start sending router solicitations.
3622          */
3623
3624         read_lock_bh(&ifp->idev->lock);
3625         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3626         send_rs = send_mld &&
3627                   ipv6_accept_ra(ifp->idev) &&
3628                   ifp->idev->cnf.rtr_solicits > 0 &&
3629                   (dev->flags&IFF_LOOPBACK) == 0;
3630         read_unlock_bh(&ifp->idev->lock);
3631
3632         /* While dad is in progress mld report's source address is in6_addrany.
3633          * Resend with proper ll now.
3634          */
3635         if (send_mld)
3636                 ipv6_mc_dad_complete(ifp->idev);
3637
3638         if (send_rs) {
3639                 /*
3640                  *      If a host as already performed a random delay
3641                  *      [...] as part of DAD [...] there is no need
3642                  *      to delay again before sending the first RS
3643                  */
3644                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3645                         return;
3646                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3647
3648                 write_lock_bh(&ifp->idev->lock);
3649                 spin_lock(&ifp->lock);
3650                 ifp->idev->rs_probes = 1;
3651                 ifp->idev->if_flags |= IF_RS_SENT;
3652                 addrconf_mod_rs_timer(ifp->idev,
3653                                       ifp->idev->cnf.rtr_solicit_interval);
3654                 spin_unlock(&ifp->lock);
3655                 write_unlock_bh(&ifp->idev->lock);
3656         }
3657 }
3658
3659 static void addrconf_dad_run(struct inet6_dev *idev)
3660 {
3661         struct inet6_ifaddr *ifp;
3662
3663         read_lock_bh(&idev->lock);
3664         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3665                 spin_lock(&ifp->lock);
3666                 if (ifp->flags & IFA_F_TENTATIVE &&
3667                     ifp->state == INET6_IFADDR_STATE_DAD)
3668                         addrconf_dad_kick(ifp);
3669                 spin_unlock(&ifp->lock);
3670         }
3671         read_unlock_bh(&idev->lock);
3672 }
3673
3674 #ifdef CONFIG_PROC_FS
3675 struct if6_iter_state {
3676         struct seq_net_private p;
3677         int bucket;
3678         int offset;
3679 };
3680
3681 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3682 {
3683         struct inet6_ifaddr *ifa = NULL;
3684         struct if6_iter_state *state = seq->private;
3685         struct net *net = seq_file_net(seq);
3686         int p = 0;
3687
3688         /* initial bucket if pos is 0 */
3689         if (pos == 0) {
3690                 state->bucket = 0;
3691                 state->offset = 0;
3692         }
3693
3694         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3695                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3696                                          addr_lst) {
3697                         if (!net_eq(dev_net(ifa->idev->dev), net))
3698                                 continue;
3699                         /* sync with offset */
3700                         if (p < state->offset) {
3701                                 p++;
3702                                 continue;
3703                         }
3704                         state->offset++;
3705                         return ifa;
3706                 }
3707
3708                 /* prepare for next bucket */
3709                 state->offset = 0;
3710                 p = 0;
3711         }
3712         return NULL;
3713 }
3714
3715 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3716                                          struct inet6_ifaddr *ifa)
3717 {
3718         struct if6_iter_state *state = seq->private;
3719         struct net *net = seq_file_net(seq);
3720
3721         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3722                 if (!net_eq(dev_net(ifa->idev->dev), net))
3723                         continue;
3724                 state->offset++;
3725                 return ifa;
3726         }
3727
3728         while (++state->bucket < IN6_ADDR_HSIZE) {
3729                 state->offset = 0;
3730                 hlist_for_each_entry_rcu_bh(ifa,
3731                                      &inet6_addr_lst[state->bucket], addr_lst) {
3732                         if (!net_eq(dev_net(ifa->idev->dev), net))
3733                                 continue;
3734                         state->offset++;
3735                         return ifa;
3736                 }
3737         }
3738
3739         return NULL;
3740 }
3741
3742 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3743         __acquires(rcu_bh)
3744 {
3745         rcu_read_lock_bh();
3746         return if6_get_first(seq, *pos);
3747 }
3748
3749 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3750 {
3751         struct inet6_ifaddr *ifa;
3752
3753         ifa = if6_get_next(seq, v);
3754         ++*pos;
3755         return ifa;
3756 }
3757
3758 static void if6_seq_stop(struct seq_file *seq, void *v)
3759         __releases(rcu_bh)
3760 {
3761         rcu_read_unlock_bh();
3762 }
3763
3764 static int if6_seq_show(struct seq_file *seq, void *v)
3765 {
3766         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3767         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3768                    &ifp->addr,
3769                    ifp->idev->dev->ifindex,
3770                    ifp->prefix_len,
3771                    ifp->scope,
3772                    (u8) ifp->flags,
3773                    ifp->idev->dev->name);
3774         return 0;
3775 }
3776
3777 static const struct seq_operations if6_seq_ops = {
3778         .start  = if6_seq_start,
3779         .next   = if6_seq_next,
3780         .show   = if6_seq_show,
3781         .stop   = if6_seq_stop,
3782 };
3783
3784 static int if6_seq_open(struct inode *inode, struct file *file)
3785 {
3786         return seq_open_net(inode, file, &if6_seq_ops,
3787                             sizeof(struct if6_iter_state));
3788 }
3789
3790 static const struct file_operations if6_fops = {
3791         .owner          = THIS_MODULE,
3792         .open           = if6_seq_open,
3793         .read           = seq_read,
3794         .llseek         = seq_lseek,
3795         .release        = seq_release_net,
3796 };
3797
3798 static int __net_init if6_proc_net_init(struct net *net)
3799 {
3800         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3801                 return -ENOMEM;
3802         return 0;
3803 }
3804
3805 static void __net_exit if6_proc_net_exit(struct net *net)
3806 {
3807         remove_proc_entry("if_inet6", net->proc_net);
3808 }
3809
3810 static struct pernet_operations if6_proc_net_ops = {
3811         .init = if6_proc_net_init,
3812         .exit = if6_proc_net_exit,
3813 };
3814
3815 int __init if6_proc_init(void)
3816 {
3817         return register_pernet_subsys(&if6_proc_net_ops);
3818 }
3819
3820 void if6_proc_exit(void)
3821 {
3822         unregister_pernet_subsys(&if6_proc_net_ops);
3823 }
3824 #endif  /* CONFIG_PROC_FS */
3825
3826 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3827 /* Check if address is a home address configured on any interface. */
3828 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3829 {
3830         int ret = 0;
3831         struct inet6_ifaddr *ifp = NULL;
3832         unsigned int hash = inet6_addr_hash(addr);
3833
3834         rcu_read_lock_bh();
3835         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3836                 if (!net_eq(dev_net(ifp->idev->dev), net))
3837                         continue;
3838                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3839                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3840                         ret = 1;
3841                         break;
3842                 }
3843         }
3844         rcu_read_unlock_bh();
3845         return ret;
3846 }
3847 #endif
3848
3849 /*
3850  *      Periodic address status verification
3851  */
3852
3853 static void addrconf_verify_rtnl(void)
3854 {
3855         unsigned long now, next, next_sec, next_sched;
3856         struct inet6_ifaddr *ifp;
3857         int i;
3858
3859         ASSERT_RTNL();
3860
3861         rcu_read_lock_bh();
3862         now = jiffies;
3863         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3864
3865         cancel_delayed_work(&addr_chk_work);
3866
3867         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3868 restart:
3869                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3870                         unsigned long age;
3871
3872                         /* When setting preferred_lft to a value not zero or
3873                          * infinity, while valid_lft is infinity
3874                          * IFA_F_PERMANENT has a non-infinity life time.
3875                          */
3876                         if ((ifp->flags & IFA_F_PERMANENT) &&
3877                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3878                                 continue;
3879
3880                         spin_lock(&ifp->lock);
3881                         /* We try to batch several events at once. */
3882                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3883
3884                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3885                             age >= ifp->valid_lft) {
3886                                 spin_unlock(&ifp->lock);
3887                                 in6_ifa_hold(ifp);
3888                                 ipv6_del_addr(ifp);
3889                                 goto restart;
3890                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3891                                 spin_unlock(&ifp->lock);
3892                                 continue;
3893                         } else if (age >= ifp->prefered_lft) {
3894                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3895                                 int deprecate = 0;
3896
3897                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3898                                         deprecate = 1;
3899                                         ifp->flags |= IFA_F_DEPRECATED;
3900                                 }
3901
3902                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3903                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3904                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3905
3906                                 spin_unlock(&ifp->lock);
3907
3908                                 if (deprecate) {
3909                                         in6_ifa_hold(ifp);
3910
3911                                         ipv6_ifa_notify(0, ifp);
3912                                         in6_ifa_put(ifp);
3913                                         goto restart;
3914                                 }
3915                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3916                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3917                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3918                                         ifp->idev->cnf.dad_transmits *
3919                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3920
3921                                 if (age >= ifp->prefered_lft - regen_advance) {
3922                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3923                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3924                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3925                                         if (!ifp->regen_count && ifpub) {
3926                                                 ifp->regen_count++;
3927                                                 in6_ifa_hold(ifp);
3928                                                 in6_ifa_hold(ifpub);
3929                                                 spin_unlock(&ifp->lock);
3930
3931                                                 spin_lock(&ifpub->lock);
3932                                                 ifpub->regen_count = 0;
3933                                                 spin_unlock(&ifpub->lock);
3934                                                 ipv6_create_tempaddr(ifpub, ifp);
3935                                                 in6_ifa_put(ifpub);
3936                                                 in6_ifa_put(ifp);
3937                                                 goto restart;
3938                                         }
3939                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3940                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3941                                 spin_unlock(&ifp->lock);
3942                         } else {
3943                                 /* ifp->prefered_lft <= ifp->valid_lft */
3944                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3945                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3946                                 spin_unlock(&ifp->lock);
3947                         }
3948                 }
3949         }
3950
3951         next_sec = round_jiffies_up(next);
3952         next_sched = next;
3953
3954         /* If rounded timeout is accurate enough, accept it. */
3955         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3956                 next_sched = next_sec;
3957
3958         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3959         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3960                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3961
3962         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3963               now, next, next_sec, next_sched);
3964         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3965         rcu_read_unlock_bh();
3966 }
3967
3968 static void addrconf_verify_work(struct work_struct *w)
3969 {
3970         rtnl_lock();
3971         addrconf_verify_rtnl();
3972         rtnl_unlock();
3973 }
3974
3975 static void addrconf_verify(void)
3976 {
3977         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3978 }
3979
3980 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3981                                      struct in6_addr **peer_pfx)
3982 {
3983         struct in6_addr *pfx = NULL;
3984
3985         *peer_pfx = NULL;
3986
3987         if (addr)
3988                 pfx = nla_data(addr);
3989
3990         if (local) {
3991                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3992                         *peer_pfx = pfx;
3993                 pfx = nla_data(local);
3994         }
3995
3996         return pfx;
3997 }
3998
3999 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4000         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4001         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4002         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4003         [IFA_FLAGS]             = { .len = sizeof(u32) },
4004 };
4005
4006 static int
4007 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4008 {
4009         struct net *net = sock_net(skb->sk);
4010         struct ifaddrmsg *ifm;
4011         struct nlattr *tb[IFA_MAX+1];
4012         struct in6_addr *pfx, *peer_pfx;
4013         u32 ifa_flags;
4014         int err;
4015
4016         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4017         if (err < 0)
4018                 return err;
4019
4020         ifm = nlmsg_data(nlh);
4021         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4022         if (!pfx)
4023                 return -EINVAL;
4024
4025         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4026
4027         /* We ignore other flags so far. */
4028         ifa_flags &= IFA_F_MANAGETEMPADDR;
4029
4030         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4031                               ifm->ifa_prefixlen);
4032 }
4033
4034 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4035                              u32 prefered_lft, u32 valid_lft)
4036 {
4037         u32 flags;
4038         clock_t expires;
4039         unsigned long timeout;
4040         bool was_managetempaddr;
4041         bool had_prefixroute;
4042
4043         ASSERT_RTNL();
4044
4045         if (!valid_lft || (prefered_lft > valid_lft))
4046                 return -EINVAL;
4047
4048         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4049             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4050                 return -EINVAL;
4051
4052         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4053         if (addrconf_finite_timeout(timeout)) {
4054                 expires = jiffies_to_clock_t(timeout * HZ);
4055                 valid_lft = timeout;
4056                 flags = RTF_EXPIRES;
4057         } else {
4058                 expires = 0;
4059                 flags = 0;
4060                 ifa_flags |= IFA_F_PERMANENT;
4061         }
4062
4063         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4064         if (addrconf_finite_timeout(timeout)) {
4065                 if (timeout == 0)
4066                         ifa_flags |= IFA_F_DEPRECATED;
4067                 prefered_lft = timeout;
4068         }
4069
4070         spin_lock_bh(&ifp->lock);
4071         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4072         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4073                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4074         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4075                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4076                         IFA_F_NOPREFIXROUTE);
4077         ifp->flags |= ifa_flags;
4078         ifp->tstamp = jiffies;
4079         ifp->valid_lft = valid_lft;
4080         ifp->prefered_lft = prefered_lft;
4081
4082         spin_unlock_bh(&ifp->lock);
4083         if (!(ifp->flags&IFA_F_TENTATIVE))
4084                 ipv6_ifa_notify(0, ifp);
4085
4086         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4087                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4088                                       expires, flags);
4089         } else if (had_prefixroute) {
4090                 enum cleanup_prefix_rt_t action;
4091                 unsigned long rt_expires;
4092
4093                 write_lock_bh(&ifp->idev->lock);
4094                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4095                 write_unlock_bh(&ifp->idev->lock);
4096
4097                 if (action != CLEANUP_PREFIX_RT_NOP) {
4098                         cleanup_prefix_route(ifp, rt_expires,
4099                                 action == CLEANUP_PREFIX_RT_DEL);
4100                 }
4101         }
4102
4103         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4104                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4105                         valid_lft = prefered_lft = 0;
4106                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4107                                  !was_managetempaddr, jiffies);
4108         }
4109
4110         addrconf_verify_rtnl();
4111
4112         return 0;
4113 }
4114
4115 static int
4116 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4117 {
4118         struct net *net = sock_net(skb->sk);
4119         struct ifaddrmsg *ifm;
4120         struct nlattr *tb[IFA_MAX+1];
4121         struct in6_addr *pfx, *peer_pfx;
4122         struct inet6_ifaddr *ifa;
4123         struct net_device *dev;
4124         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4125         u32 ifa_flags;
4126         int err;
4127
4128         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4129         if (err < 0)
4130                 return err;
4131
4132         ifm = nlmsg_data(nlh);
4133         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4134         if (!pfx)
4135                 return -EINVAL;
4136
4137         if (tb[IFA_CACHEINFO]) {
4138                 struct ifa_cacheinfo *ci;
4139
4140                 ci = nla_data(tb[IFA_CACHEINFO]);
4141                 valid_lft = ci->ifa_valid;
4142                 preferred_lft = ci->ifa_prefered;
4143         } else {
4144                 preferred_lft = INFINITY_LIFE_TIME;
4145                 valid_lft = INFINITY_LIFE_TIME;
4146         }
4147
4148         dev =  __dev_get_by_index(net, ifm->ifa_index);
4149         if (!dev)
4150                 return -ENODEV;
4151
4152         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4153
4154         /* We ignore other flags so far. */
4155         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4156                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4157
4158         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4159         if (!ifa) {
4160                 /*
4161                  * It would be best to check for !NLM_F_CREATE here but
4162                  * userspace already relies on not having to provide this.
4163                  */
4164                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4165                                       ifm->ifa_prefixlen, ifa_flags,
4166                                       preferred_lft, valid_lft);
4167         }
4168
4169         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4170             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4171                 err = -EEXIST;
4172         else
4173                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4174
4175         in6_ifa_put(ifa);
4176
4177         return err;
4178 }
4179
4180 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4181                           u8 scope, int ifindex)
4182 {
4183         struct ifaddrmsg *ifm;
4184
4185         ifm = nlmsg_data(nlh);
4186         ifm->ifa_family = AF_INET6;
4187         ifm->ifa_prefixlen = prefixlen;
4188         ifm->ifa_flags = flags;
4189         ifm->ifa_scope = scope;
4190         ifm->ifa_index = ifindex;
4191 }
4192
4193 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4194                          unsigned long tstamp, u32 preferred, u32 valid)
4195 {
4196         struct ifa_cacheinfo ci;
4197
4198         ci.cstamp = cstamp_delta(cstamp);
4199         ci.tstamp = cstamp_delta(tstamp);
4200         ci.ifa_prefered = preferred;
4201         ci.ifa_valid = valid;
4202
4203         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4204 }
4205
4206 static inline int rt_scope(int ifa_scope)
4207 {
4208         if (ifa_scope & IFA_HOST)
4209                 return RT_SCOPE_HOST;
4210         else if (ifa_scope & IFA_LINK)
4211                 return RT_SCOPE_LINK;
4212         else if (ifa_scope & IFA_SITE)
4213                 return RT_SCOPE_SITE;
4214         else
4215                 return RT_SCOPE_UNIVERSE;
4216 }
4217
4218 static inline int inet6_ifaddr_msgsize(void)
4219 {
4220         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4221                + nla_total_size(16) /* IFA_LOCAL */
4222                + nla_total_size(16) /* IFA_ADDRESS */
4223                + nla_total_size(sizeof(struct ifa_cacheinfo))
4224                + nla_total_size(4)  /* IFA_FLAGS */;
4225 }
4226
4227 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4228                              u32 portid, u32 seq, int event, unsigned int flags)
4229 {
4230         struct nlmsghdr  *nlh;
4231         u32 preferred, valid;
4232
4233         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4234         if (!nlh)
4235                 return -EMSGSIZE;
4236
4237         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4238                       ifa->idev->dev->ifindex);
4239
4240         if (!((ifa->flags&IFA_F_PERMANENT) &&
4241               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4242                 preferred = ifa->prefered_lft;
4243                 valid = ifa->valid_lft;
4244                 if (preferred != INFINITY_LIFE_TIME) {
4245                         long tval = (jiffies - ifa->tstamp)/HZ;
4246                         if (preferred > tval)
4247                                 preferred -= tval;
4248                         else
4249                                 preferred = 0;
4250                         if (valid != INFINITY_LIFE_TIME) {
4251                                 if (valid > tval)
4252                                         valid -= tval;
4253                                 else
4254                                         valid = 0;
4255                         }
4256                 }
4257         } else {
4258                 preferred = INFINITY_LIFE_TIME;
4259                 valid = INFINITY_LIFE_TIME;
4260         }
4261
4262         if (!ipv6_addr_any(&ifa->peer_addr)) {
4263                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4264                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4265                         goto error;
4266         } else
4267                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4268                         goto error;
4269
4270         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4271                 goto error;
4272
4273         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4274                 goto error;
4275
4276         nlmsg_end(skb, nlh);
4277         return 0;
4278
4279 error:
4280         nlmsg_cancel(skb, nlh);
4281         return -EMSGSIZE;
4282 }
4283
4284 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4285                                 u32 portid, u32 seq, int event, u16 flags)
4286 {
4287         struct nlmsghdr  *nlh;
4288         u8 scope = RT_SCOPE_UNIVERSE;
4289         int ifindex = ifmca->idev->dev->ifindex;
4290
4291         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4292                 scope = RT_SCOPE_SITE;
4293
4294         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4295         if (!nlh)
4296                 return -EMSGSIZE;
4297
4298         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4299         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4300             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4301                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4302                 nlmsg_cancel(skb, nlh);
4303                 return -EMSGSIZE;
4304         }
4305
4306         nlmsg_end(skb, nlh);
4307         return 0;
4308 }
4309
4310 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4311                                 u32 portid, u32 seq, int event, unsigned int flags)
4312 {
4313         struct nlmsghdr  *nlh;
4314         u8 scope = RT_SCOPE_UNIVERSE;
4315         int ifindex = ifaca->aca_idev->dev->ifindex;
4316
4317         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4318                 scope = RT_SCOPE_SITE;
4319
4320         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4321         if (!nlh)
4322                 return -EMSGSIZE;
4323
4324         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4325         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4326             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4327                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4328                 nlmsg_cancel(skb, nlh);
4329                 return -EMSGSIZE;
4330         }
4331
4332         nlmsg_end(skb, nlh);
4333         return 0;
4334 }
4335
4336 enum addr_type_t {
4337         UNICAST_ADDR,
4338         MULTICAST_ADDR,
4339         ANYCAST_ADDR,
4340 };
4341
4342 /* called with rcu_read_lock() */
4343 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4344                           struct netlink_callback *cb, enum addr_type_t type,
4345                           int s_ip_idx, int *p_ip_idx)
4346 {
4347         struct ifmcaddr6 *ifmca;
4348         struct ifacaddr6 *ifaca;
4349         int err = 1;
4350         int ip_idx = *p_ip_idx;
4351
4352         read_lock_bh(&idev->lock);
4353         switch (type) {
4354         case UNICAST_ADDR: {
4355                 struct inet6_ifaddr *ifa;
4356
4357                 /* unicast address incl. temp addr */
4358                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4359                         if (++ip_idx < s_ip_idx)
4360                                 continue;
4361                         err = inet6_fill_ifaddr(skb, ifa,
4362                                                 NETLINK_CB(cb->skb).portid,
4363                                                 cb->nlh->nlmsg_seq,
4364                                                 RTM_NEWADDR,
4365                                                 NLM_F_MULTI);
4366                         if (err < 0)
4367                                 break;
4368                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4369                 }
4370                 break;
4371         }
4372         case MULTICAST_ADDR:
4373                 /* multicast address */
4374                 for (ifmca = idev->mc_list; ifmca;
4375                      ifmca = ifmca->next, ip_idx++) {
4376                         if (ip_idx < s_ip_idx)
4377                                 continue;
4378                         err = inet6_fill_ifmcaddr(skb, ifmca,
4379                                                   NETLINK_CB(cb->skb).portid,
4380                                                   cb->nlh->nlmsg_seq,
4381                                                   RTM_GETMULTICAST,
4382                                                   NLM_F_MULTI);
4383                         if (err < 0)
4384                                 break;
4385                 }
4386                 break;
4387         case ANYCAST_ADDR:
4388                 /* anycast address */
4389                 for (ifaca = idev->ac_list; ifaca;
4390                      ifaca = ifaca->aca_next, ip_idx++) {
4391                         if (ip_idx < s_ip_idx)
4392                                 continue;
4393                         err = inet6_fill_ifacaddr(skb, ifaca,
4394                                                   NETLINK_CB(cb->skb).portid,
4395                                                   cb->nlh->nlmsg_seq,
4396                                                   RTM_GETANYCAST,
4397                                                   NLM_F_MULTI);
4398                         if (err < 0)
4399                                 break;
4400                 }
4401                 break;
4402         default:
4403                 break;
4404         }
4405         read_unlock_bh(&idev->lock);
4406         *p_ip_idx = ip_idx;
4407         return err;
4408 }
4409
4410 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4411                            enum addr_type_t type)
4412 {
4413         struct net *net = sock_net(skb->sk);
4414         int h, s_h;
4415         int idx, ip_idx;
4416         int s_idx, s_ip_idx;
4417         struct net_device *dev;
4418         struct inet6_dev *idev;
4419         struct hlist_head *head;
4420
4421         s_h = cb->args[0];
4422         s_idx = idx = cb->args[1];
4423         s_ip_idx = ip_idx = cb->args[2];
4424
4425         rcu_read_lock();
4426         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4427         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4428                 idx = 0;
4429                 head = &net->dev_index_head[h];
4430                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4431                         if (idx < s_idx)
4432                                 goto cont;
4433                         if (h > s_h || idx > s_idx)
4434                                 s_ip_idx = 0;
4435                         ip_idx = 0;
4436                         idev = __in6_dev_get(dev);
4437                         if (!idev)
4438                                 goto cont;
4439
4440                         if (in6_dump_addrs(idev, skb, cb, type,
4441                                            s_ip_idx, &ip_idx) < 0)
4442                                 goto done;
4443 cont:
4444                         idx++;
4445                 }
4446         }
4447 done:
4448         rcu_read_unlock();
4449         cb->args[0] = h;
4450         cb->args[1] = idx;
4451         cb->args[2] = ip_idx;
4452
4453         return skb->len;
4454 }
4455
4456 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4457 {
4458         enum addr_type_t type = UNICAST_ADDR;
4459
4460         return inet6_dump_addr(skb, cb, type);
4461 }
4462
4463 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4464 {
4465         enum addr_type_t type = MULTICAST_ADDR;
4466
4467         return inet6_dump_addr(skb, cb, type);
4468 }
4469
4470
4471 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4472 {
4473         enum addr_type_t type = ANYCAST_ADDR;
4474
4475         return inet6_dump_addr(skb, cb, type);
4476 }
4477
4478 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4479 {
4480         struct net *net = sock_net(in_skb->sk);
4481         struct ifaddrmsg *ifm;
4482         struct nlattr *tb[IFA_MAX+1];
4483         struct in6_addr *addr = NULL, *peer;
4484         struct net_device *dev = NULL;
4485         struct inet6_ifaddr *ifa;
4486         struct sk_buff *skb;
4487         int err;
4488
4489         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4490         if (err < 0)
4491                 goto errout;
4492
4493         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4494         if (!addr) {
4495                 err = -EINVAL;
4496                 goto errout;
4497         }
4498
4499         ifm = nlmsg_data(nlh);
4500         if (ifm->ifa_index)
4501                 dev = __dev_get_by_index(net, ifm->ifa_index);
4502
4503         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4504         if (!ifa) {
4505                 err = -EADDRNOTAVAIL;
4506                 goto errout;
4507         }
4508
4509         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4510         if (!skb) {
4511                 err = -ENOBUFS;
4512                 goto errout_ifa;
4513         }
4514
4515         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4516                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4517         if (err < 0) {
4518                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4519                 WARN_ON(err == -EMSGSIZE);
4520                 kfree_skb(skb);
4521                 goto errout_ifa;
4522         }
4523         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4524 errout_ifa:
4525         in6_ifa_put(ifa);
4526 errout:
4527         return err;
4528 }
4529
4530 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4531 {
4532         struct sk_buff *skb;
4533         struct net *net = dev_net(ifa->idev->dev);
4534         int err = -ENOBUFS;
4535
4536         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4537         if (!skb)
4538                 goto errout;
4539
4540         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4541         if (err < 0) {
4542                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4543                 WARN_ON(err == -EMSGSIZE);
4544                 kfree_skb(skb);
4545                 goto errout;
4546         }
4547         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4548         return;
4549 errout:
4550         if (err < 0)
4551                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4552 }
4553
4554 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4555                                 __s32 *array, int bytes)
4556 {
4557         BUG_ON(bytes < (DEVCONF_MAX * 4));
4558
4559         memset(array, 0, bytes);
4560         array[DEVCONF_FORWARDING] = cnf->forwarding;
4561         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4562         array[DEVCONF_MTU6] = cnf->mtu6;
4563         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4564         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4565         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4566         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4567         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4568         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4569                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4570         array[DEVCONF_RTR_SOLICIT_DELAY] =
4571                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4572         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4573         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4574                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4575         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4576                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4577         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4578         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4579         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4580         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4581         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4582         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4583         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4584         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4585 #ifdef CONFIG_IPV6_ROUTER_PREF
4586         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4587         array[DEVCONF_RTR_PROBE_INTERVAL] =
4588                 jiffies_to_msecs(cnf->rtr_probe_interval);
4589 #ifdef CONFIG_IPV6_ROUTE_INFO
4590         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4591 #endif
4592 #endif
4593         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4594         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4595 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4596         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4597         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4598 #endif
4599 #ifdef CONFIG_IPV6_MROUTE
4600         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4601 #endif
4602         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4603         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4604         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4605         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4606         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4607         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4608         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4609         /* we omit DEVCONF_STABLE_SECRET for now */
4610 }
4611
4612 static inline size_t inet6_ifla6_size(void)
4613 {
4614         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4615              + nla_total_size(sizeof(struct ifla_cacheinfo))
4616              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4617              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4618              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4619              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4620 }
4621
4622 static inline size_t inet6_if_nlmsg_size(void)
4623 {
4624         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4625                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4626                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4627                + nla_total_size(4) /* IFLA_MTU */
4628                + nla_total_size(4) /* IFLA_LINK */
4629                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4630 }
4631
4632 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4633                                       int items, int bytes)
4634 {
4635         int i;
4636         int pad = bytes - sizeof(u64) * items;
4637         BUG_ON(pad < 0);
4638
4639         /* Use put_unaligned() because stats may not be aligned for u64. */
4640         put_unaligned(items, &stats[0]);
4641         for (i = 1; i < items; i++)
4642                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4643
4644         memset(&stats[items], 0, pad);
4645 }
4646
4647 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4648                                       int items, int bytes, size_t syncpoff)
4649 {
4650         int i;
4651         int pad = bytes - sizeof(u64) * items;
4652         BUG_ON(pad < 0);
4653
4654         /* Use put_unaligned() because stats may not be aligned for u64. */
4655         put_unaligned(items, &stats[0]);
4656         for (i = 1; i < items; i++)
4657                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4658
4659         memset(&stats[items], 0, pad);
4660 }
4661
4662 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4663                              int bytes)
4664 {
4665         switch (attrtype) {
4666         case IFLA_INET6_STATS:
4667                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4668                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4669                 break;
4670         case IFLA_INET6_ICMP6STATS:
4671                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4672                 break;
4673         }
4674 }
4675
4676 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4677 {
4678         struct nlattr *nla;
4679         struct ifla_cacheinfo ci;
4680
4681         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4682                 goto nla_put_failure;
4683         ci.max_reasm_len = IPV6_MAXPLEN;
4684         ci.tstamp = cstamp_delta(idev->tstamp);
4685         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4686         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4687         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4688                 goto nla_put_failure;
4689         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4690         if (!nla)
4691                 goto nla_put_failure;
4692         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4693
4694         /* XXX - MC not implemented */
4695
4696         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4697         if (!nla)
4698                 goto nla_put_failure;
4699         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4700
4701         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4702         if (!nla)
4703                 goto nla_put_failure;
4704         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4705
4706         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4707         if (!nla)
4708                 goto nla_put_failure;
4709
4710         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4711                 goto nla_put_failure;
4712
4713         read_lock_bh(&idev->lock);
4714         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4715         read_unlock_bh(&idev->lock);
4716
4717         return 0;
4718
4719 nla_put_failure:
4720         return -EMSGSIZE;
4721 }
4722
4723 static size_t inet6_get_link_af_size(const struct net_device *dev)
4724 {
4725         if (!__in6_dev_get(dev))
4726                 return 0;
4727
4728         return inet6_ifla6_size();
4729 }
4730
4731 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4732 {
4733         struct inet6_dev *idev = __in6_dev_get(dev);
4734
4735         if (!idev)
4736                 return -ENODATA;
4737
4738         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4739                 return -EMSGSIZE;
4740
4741         return 0;
4742 }
4743
4744 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4745 {
4746         struct inet6_ifaddr *ifp;
4747         struct net_device *dev = idev->dev;
4748         bool update_rs = false;
4749         struct in6_addr ll_addr;
4750
4751         ASSERT_RTNL();
4752
4753         if (!token)
4754                 return -EINVAL;
4755         if (ipv6_addr_any(token))
4756                 return -EINVAL;
4757         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4758                 return -EINVAL;
4759         if (!ipv6_accept_ra(idev))
4760                 return -EINVAL;
4761         if (idev->cnf.rtr_solicits <= 0)
4762                 return -EINVAL;
4763
4764         write_lock_bh(&idev->lock);
4765
4766         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4767         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4768
4769         write_unlock_bh(&idev->lock);
4770
4771         if (!idev->dead && (idev->if_flags & IF_READY) &&
4772             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4773                              IFA_F_OPTIMISTIC)) {
4774
4775                 /* If we're not ready, then normal ifup will take care
4776                  * of this. Otherwise, we need to request our rs here.
4777                  */
4778                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4779                 update_rs = true;
4780         }
4781
4782         write_lock_bh(&idev->lock);
4783
4784         if (update_rs) {
4785                 idev->if_flags |= IF_RS_SENT;
4786                 idev->rs_probes = 1;
4787                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4788         }
4789
4790         /* Well, that's kinda nasty ... */
4791         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4792                 spin_lock(&ifp->lock);
4793                 if (ifp->tokenized) {
4794                         ifp->valid_lft = 0;
4795                         ifp->prefered_lft = 0;
4796                 }
4797                 spin_unlock(&ifp->lock);
4798         }
4799
4800         write_unlock_bh(&idev->lock);
4801         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4802         addrconf_verify_rtnl();
4803         return 0;
4804 }
4805
4806 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4807         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4808         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4809 };
4810
4811 static int inet6_validate_link_af(const struct net_device *dev,
4812                                   const struct nlattr *nla)
4813 {
4814         struct nlattr *tb[IFLA_INET6_MAX + 1];
4815
4816         if (dev && !__in6_dev_get(dev))
4817                 return -EAFNOSUPPORT;
4818
4819         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4820 }
4821
4822 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4823 {
4824         int err = -EINVAL;
4825         struct inet6_dev *idev = __in6_dev_get(dev);
4826         struct nlattr *tb[IFLA_INET6_MAX + 1];
4827
4828         if (!idev)
4829                 return -EAFNOSUPPORT;
4830
4831         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4832                 BUG();
4833
4834         if (tb[IFLA_INET6_TOKEN]) {
4835                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4836                 if (err)
4837                         return err;
4838         }
4839
4840         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4841                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4842
4843                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4844                     mode != IN6_ADDR_GEN_MODE_NONE &&
4845                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4846                         return -EINVAL;
4847
4848                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4849                     !idev->cnf.stable_secret.initialized &&
4850                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4851                         return -EINVAL;
4852
4853                 idev->addr_gen_mode = mode;
4854                 err = 0;
4855         }
4856
4857         return err;
4858 }
4859
4860 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4861                              u32 portid, u32 seq, int event, unsigned int flags)
4862 {
4863         struct net_device *dev = idev->dev;
4864         struct ifinfomsg *hdr;
4865         struct nlmsghdr *nlh;
4866         void *protoinfo;
4867
4868         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4869         if (!nlh)
4870                 return -EMSGSIZE;
4871
4872         hdr = nlmsg_data(nlh);
4873         hdr->ifi_family = AF_INET6;
4874         hdr->__ifi_pad = 0;
4875         hdr->ifi_type = dev->type;
4876         hdr->ifi_index = dev->ifindex;
4877         hdr->ifi_flags = dev_get_flags(dev);
4878         hdr->ifi_change = 0;
4879
4880         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4881             (dev->addr_len &&
4882              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4883             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4884             (dev->ifindex != dev_get_iflink(dev) &&
4885              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4886                 goto nla_put_failure;
4887         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4888         if (!protoinfo)
4889                 goto nla_put_failure;
4890
4891         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4892                 goto nla_put_failure;
4893
4894         nla_nest_end(skb, protoinfo);
4895         nlmsg_end(skb, nlh);
4896         return 0;
4897
4898 nla_put_failure:
4899         nlmsg_cancel(skb, nlh);
4900         return -EMSGSIZE;
4901 }
4902
4903 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4904 {
4905         struct net *net = sock_net(skb->sk);
4906         int h, s_h;
4907         int idx = 0, s_idx;
4908         struct net_device *dev;
4909         struct inet6_dev *idev;
4910         struct hlist_head *head;
4911
4912         s_h = cb->args[0];
4913         s_idx = cb->args[1];
4914
4915         rcu_read_lock();
4916         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4917                 idx = 0;
4918                 head = &net->dev_index_head[h];
4919                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4920                         if (idx < s_idx)
4921                                 goto cont;
4922                         idev = __in6_dev_get(dev);
4923                         if (!idev)
4924                                 goto cont;
4925                         if (inet6_fill_ifinfo(skb, idev,
4926                                               NETLINK_CB(cb->skb).portid,
4927                                               cb->nlh->nlmsg_seq,
4928                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
4929                                 goto out;
4930 cont:
4931                         idx++;
4932                 }
4933         }
4934 out:
4935         rcu_read_unlock();
4936         cb->args[1] = idx;
4937         cb->args[0] = h;
4938
4939         return skb->len;
4940 }
4941
4942 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4943 {
4944         struct sk_buff *skb;
4945         struct net *net = dev_net(idev->dev);
4946         int err = -ENOBUFS;
4947
4948         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4949         if (!skb)
4950                 goto errout;
4951
4952         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4953         if (err < 0) {
4954                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4955                 WARN_ON(err == -EMSGSIZE);
4956                 kfree_skb(skb);
4957                 goto errout;
4958         }
4959         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4960         return;
4961 errout:
4962         if (err < 0)
4963                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4964 }
4965
4966 static inline size_t inet6_prefix_nlmsg_size(void)
4967 {
4968         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4969                + nla_total_size(sizeof(struct in6_addr))
4970                + nla_total_size(sizeof(struct prefix_cacheinfo));
4971 }
4972
4973 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4974                              struct prefix_info *pinfo, u32 portid, u32 seq,
4975                              int event, unsigned int flags)
4976 {
4977         struct prefixmsg *pmsg;
4978         struct nlmsghdr *nlh;
4979         struct prefix_cacheinfo ci;
4980
4981         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4982         if (!nlh)
4983                 return -EMSGSIZE;
4984
4985         pmsg = nlmsg_data(nlh);
4986         pmsg->prefix_family = AF_INET6;
4987         pmsg->prefix_pad1 = 0;
4988         pmsg->prefix_pad2 = 0;
4989         pmsg->prefix_ifindex = idev->dev->ifindex;
4990         pmsg->prefix_len = pinfo->prefix_len;
4991         pmsg->prefix_type = pinfo->type;
4992         pmsg->prefix_pad3 = 0;
4993         pmsg->prefix_flags = 0;
4994         if (pinfo->onlink)
4995                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4996         if (pinfo->autoconf)
4997                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4998
4999         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5000                 goto nla_put_failure;
5001         ci.preferred_time = ntohl(pinfo->prefered);
5002         ci.valid_time = ntohl(pinfo->valid);
5003         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5004                 goto nla_put_failure;
5005         nlmsg_end(skb, nlh);
5006         return 0;
5007
5008 nla_put_failure:
5009         nlmsg_cancel(skb, nlh);
5010         return -EMSGSIZE;
5011 }
5012
5013 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5014                          struct prefix_info *pinfo)
5015 {
5016         struct sk_buff *skb;
5017         struct net *net = dev_net(idev->dev);
5018         int err = -ENOBUFS;
5019
5020         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5021         if (!skb)
5022                 goto errout;
5023
5024         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5025         if (err < 0) {
5026                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5027                 WARN_ON(err == -EMSGSIZE);
5028                 kfree_skb(skb);
5029                 goto errout;
5030         }
5031         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5032         return;
5033 errout:
5034         if (err < 0)
5035                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5036 }
5037
5038 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5039 {
5040         struct net *net = dev_net(ifp->idev->dev);
5041
5042         if (event)
5043                 ASSERT_RTNL();
5044
5045         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5046
5047         switch (event) {
5048         case RTM_NEWADDR:
5049                 /*
5050                  * If the address was optimistic
5051                  * we inserted the route at the start of
5052                  * our DAD process, so we don't need
5053                  * to do it again
5054                  */
5055                 if (!(ifp->rt->rt6i_node))
5056                         ip6_ins_rt(ifp->rt);
5057                 if (ifp->idev->cnf.forwarding)
5058                         addrconf_join_anycast(ifp);
5059                 if (!ipv6_addr_any(&ifp->peer_addr))
5060                         addrconf_prefix_route(&ifp->peer_addr, 128,
5061                                               ifp->idev->dev, 0, 0);
5062                 break;
5063         case RTM_DELADDR:
5064                 if (ifp->idev->cnf.forwarding)
5065                         addrconf_leave_anycast(ifp);
5066                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5067                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5068                         struct rt6_info *rt;
5069
5070                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5071                                                        ifp->idev->dev, 0, 0);
5072                         if (rt && ip6_del_rt(rt))
5073                                 dst_free(&rt->dst);
5074                 }
5075                 dst_hold(&ifp->rt->dst);
5076
5077                 if (ip6_del_rt(ifp->rt))
5078                         dst_free(&ifp->rt->dst);
5079
5080                 rt_genid_bump_ipv6(net);
5081                 break;
5082         }
5083         atomic_inc(&net->ipv6.dev_addr_genid);
5084 }
5085
5086 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5087 {
5088         rcu_read_lock_bh();
5089         if (likely(ifp->idev->dead == 0))
5090                 __ipv6_ifa_notify(event, ifp);
5091         rcu_read_unlock_bh();
5092 }
5093
5094 #ifdef CONFIG_SYSCTL
5095
5096 static
5097 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5098                            void __user *buffer, size_t *lenp, loff_t *ppos)
5099 {
5100         int *valp = ctl->data;
5101         int val = *valp;
5102         loff_t pos = *ppos;
5103         struct ctl_table lctl;
5104         int ret;
5105
5106         /*
5107          * ctl->data points to idev->cnf.forwarding, we should
5108          * not modify it until we get the rtnl lock.
5109          */
5110         lctl = *ctl;
5111         lctl.data = &val;
5112
5113         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5114
5115         if (write)
5116                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5117         if (ret)
5118                 *ppos = pos;
5119         return ret;
5120 }
5121
5122 static
5123 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5124                         void __user *buffer, size_t *lenp, loff_t *ppos)
5125 {
5126         struct inet6_dev *idev = ctl->extra1;
5127         int min_mtu = IPV6_MIN_MTU;
5128         struct ctl_table lctl;
5129
5130         lctl = *ctl;
5131         lctl.extra1 = &min_mtu;
5132         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5133
5134         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5135 }
5136
5137 static void dev_disable_change(struct inet6_dev *idev)
5138 {
5139         struct netdev_notifier_info info;
5140
5141         if (!idev || !idev->dev)
5142                 return;
5143
5144         netdev_notifier_info_init(&info, idev->dev);
5145         if (idev->cnf.disable_ipv6)
5146                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5147         else
5148                 addrconf_notify(NULL, NETDEV_UP, &info);
5149 }
5150
5151 static void addrconf_disable_change(struct net *net, __s32 newf)
5152 {
5153         struct net_device *dev;
5154         struct inet6_dev *idev;
5155
5156         rcu_read_lock();
5157         for_each_netdev_rcu(net, dev) {
5158                 idev = __in6_dev_get(dev);
5159                 if (idev) {
5160                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5161                         idev->cnf.disable_ipv6 = newf;
5162                         if (changed)
5163                                 dev_disable_change(idev);
5164                 }
5165         }
5166         rcu_read_unlock();
5167 }
5168
5169 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5170 {
5171         struct net *net;
5172         int old;
5173
5174         if (!rtnl_trylock())
5175                 return restart_syscall();
5176
5177         net = (struct net *)table->extra2;
5178         old = *p;
5179         *p = newf;
5180
5181         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5182                 rtnl_unlock();
5183                 return 0;
5184         }
5185
5186         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5187                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5188                 addrconf_disable_change(net, newf);
5189         } else if ((!newf) ^ (!old))
5190                 dev_disable_change((struct inet6_dev *)table->extra1);
5191
5192         rtnl_unlock();
5193         return 0;
5194 }
5195
5196 static
5197 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5198                             void __user *buffer, size_t *lenp, loff_t *ppos)
5199 {
5200         int *valp = ctl->data;
5201         int val = *valp;
5202         loff_t pos = *ppos;
5203         struct ctl_table lctl;
5204         int ret;
5205
5206         /*
5207          * ctl->data points to idev->cnf.disable_ipv6, we should
5208          * not modify it until we get the rtnl lock.
5209          */
5210         lctl = *ctl;
5211         lctl.data = &val;
5212
5213         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5214
5215         if (write)
5216                 ret = addrconf_disable_ipv6(ctl, valp, val);
5217         if (ret)
5218                 *ppos = pos;
5219         return ret;
5220 }
5221
5222 static
5223 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5224                               void __user *buffer, size_t *lenp, loff_t *ppos)
5225 {
5226         int *valp = ctl->data;
5227         int ret;
5228         int old, new;
5229
5230         old = *valp;
5231         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5232         new = *valp;
5233
5234         if (write && old != new) {
5235                 struct net *net = ctl->extra2;
5236
5237                 if (!rtnl_trylock())
5238                         return restart_syscall();
5239
5240                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5241                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5242                                                      NETCONFA_IFINDEX_DEFAULT,
5243                                                      net->ipv6.devconf_dflt);
5244                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5245                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5246                                                      NETCONFA_IFINDEX_ALL,
5247                                                      net->ipv6.devconf_all);
5248                 else {
5249                         struct inet6_dev *idev = ctl->extra1;
5250
5251                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5252                                                      idev->dev->ifindex,
5253                                                      &idev->cnf);
5254                 }
5255                 rtnl_unlock();
5256         }
5257
5258         return ret;
5259 }
5260
5261 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5262                                          void __user *buffer, size_t *lenp,
5263                                          loff_t *ppos)
5264 {
5265         int err;
5266         struct in6_addr addr;
5267         char str[IPV6_MAX_STRLEN];
5268         struct ctl_table lctl = *ctl;
5269         struct net *net = ctl->extra2;
5270         struct ipv6_stable_secret *secret = ctl->data;
5271
5272         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5273                 return -EIO;
5274
5275         lctl.maxlen = IPV6_MAX_STRLEN;
5276         lctl.data = str;
5277
5278         if (!rtnl_trylock())
5279                 return restart_syscall();
5280
5281         if (!write && !secret->initialized) {
5282                 err = -EIO;
5283                 goto out;
5284         }
5285
5286         if (!write) {
5287                 err = snprintf(str, sizeof(str), "%pI6",
5288                                &secret->secret);
5289                 if (err >= sizeof(str)) {
5290                         err = -EIO;
5291                         goto out;
5292                 }
5293         }
5294
5295         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5296         if (err || !write)
5297                 goto out;
5298
5299         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5300                 err = -EIO;
5301                 goto out;
5302         }
5303
5304         secret->initialized = true;
5305         secret->secret = addr;
5306
5307         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5308                 struct net_device *dev;
5309
5310                 for_each_netdev(net, dev) {
5311                         struct inet6_dev *idev = __in6_dev_get(dev);
5312
5313                         if (idev) {
5314                                 idev->addr_gen_mode =
5315                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5316                         }
5317                 }
5318         } else {
5319                 struct inet6_dev *idev = ctl->extra1;
5320
5321                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5322         }
5323
5324 out:
5325         rtnl_unlock();
5326
5327         return err;
5328 }
5329
5330 static struct addrconf_sysctl_table
5331 {
5332         struct ctl_table_header *sysctl_header;
5333         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5334 } addrconf_sysctl __read_mostly = {
5335         .sysctl_header = NULL,
5336         .addrconf_vars = {
5337                 {
5338                         .procname       = "forwarding",
5339                         .data           = &ipv6_devconf.forwarding,
5340                         .maxlen         = sizeof(int),
5341                         .mode           = 0644,
5342                         .proc_handler   = addrconf_sysctl_forward,
5343                 },
5344                 {
5345                         .procname       = "hop_limit",
5346                         .data           = &ipv6_devconf.hop_limit,
5347                         .maxlen         = sizeof(int),
5348                         .mode           = 0644,
5349                         .proc_handler   = proc_dointvec,
5350                 },
5351                 {
5352                         .procname       = "mtu",
5353                         .data           = &ipv6_devconf.mtu6,
5354                         .maxlen         = sizeof(int),
5355                         .mode           = 0644,
5356                         .proc_handler   = addrconf_sysctl_mtu,
5357                 },
5358                 {
5359                         .procname       = "accept_ra",
5360                         .data           = &ipv6_devconf.accept_ra,
5361                         .maxlen         = sizeof(int),
5362                         .mode           = 0644,
5363                         .proc_handler   = proc_dointvec,
5364                 },
5365                 {
5366                         .procname       = "accept_redirects",
5367                         .data           = &ipv6_devconf.accept_redirects,
5368                         .maxlen         = sizeof(int),
5369                         .mode           = 0644,
5370                         .proc_handler   = proc_dointvec,
5371                 },
5372                 {
5373                         .procname       = "autoconf",
5374                         .data           = &ipv6_devconf.autoconf,
5375                         .maxlen         = sizeof(int),
5376                         .mode           = 0644,
5377                         .proc_handler   = proc_dointvec,
5378                 },
5379                 {
5380                         .procname       = "dad_transmits",
5381                         .data           = &ipv6_devconf.dad_transmits,
5382                         .maxlen         = sizeof(int),
5383                         .mode           = 0644,
5384                         .proc_handler   = proc_dointvec,
5385                 },
5386                 {
5387                         .procname       = "router_solicitations",
5388                         .data           = &ipv6_devconf.rtr_solicits,
5389                         .maxlen         = sizeof(int),
5390                         .mode           = 0644,
5391                         .proc_handler   = proc_dointvec,
5392                 },
5393                 {
5394                         .procname       = "router_solicitation_interval",
5395                         .data           = &ipv6_devconf.rtr_solicit_interval,
5396                         .maxlen         = sizeof(int),
5397                         .mode           = 0644,
5398                         .proc_handler   = proc_dointvec_jiffies,
5399                 },
5400                 {
5401                         .procname       = "router_solicitation_delay",
5402                         .data           = &ipv6_devconf.rtr_solicit_delay,
5403                         .maxlen         = sizeof(int),
5404                         .mode           = 0644,
5405                         .proc_handler   = proc_dointvec_jiffies,
5406                 },
5407                 {
5408                         .procname       = "force_mld_version",
5409                         .data           = &ipv6_devconf.force_mld_version,
5410                         .maxlen         = sizeof(int),
5411                         .mode           = 0644,
5412                         .proc_handler   = proc_dointvec,
5413                 },
5414                 {
5415                         .procname       = "mldv1_unsolicited_report_interval",
5416                         .data           =
5417                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5418                         .maxlen         = sizeof(int),
5419                         .mode           = 0644,
5420                         .proc_handler   = proc_dointvec_ms_jiffies,
5421                 },
5422                 {
5423                         .procname       = "mldv2_unsolicited_report_interval",
5424                         .data           =
5425                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5426                         .maxlen         = sizeof(int),
5427                         .mode           = 0644,
5428                         .proc_handler   = proc_dointvec_ms_jiffies,
5429                 },
5430                 {
5431                         .procname       = "use_tempaddr",
5432                         .data           = &ipv6_devconf.use_tempaddr,
5433                         .maxlen         = sizeof(int),
5434                         .mode           = 0644,
5435                         .proc_handler   = proc_dointvec,
5436                 },
5437                 {
5438                         .procname       = "temp_valid_lft",
5439                         .data           = &ipv6_devconf.temp_valid_lft,
5440                         .maxlen         = sizeof(int),
5441                         .mode           = 0644,
5442                         .proc_handler   = proc_dointvec,
5443                 },
5444                 {
5445                         .procname       = "temp_prefered_lft",
5446                         .data           = &ipv6_devconf.temp_prefered_lft,
5447                         .maxlen         = sizeof(int),
5448                         .mode           = 0644,
5449                         .proc_handler   = proc_dointvec,
5450                 },
5451                 {
5452                         .procname       = "regen_max_retry",
5453                         .data           = &ipv6_devconf.regen_max_retry,
5454                         .maxlen         = sizeof(int),
5455                         .mode           = 0644,
5456                         .proc_handler   = proc_dointvec,
5457                 },
5458                 {
5459                         .procname       = "max_desync_factor",
5460                         .data           = &ipv6_devconf.max_desync_factor,
5461                         .maxlen         = sizeof(int),
5462                         .mode           = 0644,
5463                         .proc_handler   = proc_dointvec,
5464                 },
5465                 {
5466                         .procname       = "max_addresses",
5467                         .data           = &ipv6_devconf.max_addresses,
5468                         .maxlen         = sizeof(int),
5469                         .mode           = 0644,
5470                         .proc_handler   = proc_dointvec,
5471                 },
5472                 {
5473                         .procname       = "accept_ra_defrtr",
5474                         .data           = &ipv6_devconf.accept_ra_defrtr,
5475                         .maxlen         = sizeof(int),
5476                         .mode           = 0644,
5477                         .proc_handler   = proc_dointvec,
5478                 },
5479                 {
5480                         .procname       = "accept_ra_pinfo",
5481                         .data           = &ipv6_devconf.accept_ra_pinfo,
5482                         .maxlen         = sizeof(int),
5483                         .mode           = 0644,
5484                         .proc_handler   = proc_dointvec,
5485                 },
5486 #ifdef CONFIG_IPV6_ROUTER_PREF
5487                 {
5488                         .procname       = "accept_ra_rtr_pref",
5489                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5490                         .maxlen         = sizeof(int),
5491                         .mode           = 0644,
5492                         .proc_handler   = proc_dointvec,
5493                 },
5494                 {
5495                         .procname       = "router_probe_interval",
5496                         .data           = &ipv6_devconf.rtr_probe_interval,
5497                         .maxlen         = sizeof(int),
5498                         .mode           = 0644,
5499                         .proc_handler   = proc_dointvec_jiffies,
5500                 },
5501 #ifdef CONFIG_IPV6_ROUTE_INFO
5502                 {
5503                         .procname       = "accept_ra_rt_info_max_plen",
5504                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5505                         .maxlen         = sizeof(int),
5506                         .mode           = 0644,
5507                         .proc_handler   = proc_dointvec,
5508                 },
5509 #endif
5510 #endif
5511                 {
5512                         .procname       = "proxy_ndp",
5513                         .data           = &ipv6_devconf.proxy_ndp,
5514                         .maxlen         = sizeof(int),
5515                         .mode           = 0644,
5516                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5517                 },
5518                 {
5519                         .procname       = "accept_source_route",
5520                         .data           = &ipv6_devconf.accept_source_route,
5521                         .maxlen         = sizeof(int),
5522                         .mode           = 0644,
5523                         .proc_handler   = proc_dointvec,
5524                 },
5525 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5526                 {
5527                         .procname       = "optimistic_dad",
5528                         .data           = &ipv6_devconf.optimistic_dad,
5529                         .maxlen         = sizeof(int),
5530                         .mode           = 0644,
5531                         .proc_handler   = proc_dointvec,
5532
5533                 },
5534                 {
5535                         .procname       = "use_optimistic",
5536                         .data           = &ipv6_devconf.use_optimistic,
5537                         .maxlen         = sizeof(int),
5538                         .mode           = 0644,
5539                         .proc_handler   = proc_dointvec,
5540
5541                 },
5542 #endif
5543 #ifdef CONFIG_IPV6_MROUTE
5544                 {
5545                         .procname       = "mc_forwarding",
5546                         .data           = &ipv6_devconf.mc_forwarding,
5547                         .maxlen         = sizeof(int),
5548                         .mode           = 0444,
5549                         .proc_handler   = proc_dointvec,
5550                 },
5551 #endif
5552                 {
5553                         .procname       = "disable_ipv6",
5554                         .data           = &ipv6_devconf.disable_ipv6,
5555                         .maxlen         = sizeof(int),
5556                         .mode           = 0644,
5557                         .proc_handler   = addrconf_sysctl_disable,
5558                 },
5559                 {
5560                         .procname       = "accept_dad",
5561                         .data           = &ipv6_devconf.accept_dad,
5562                         .maxlen         = sizeof(int),
5563                         .mode           = 0644,
5564                         .proc_handler   = proc_dointvec,
5565                 },
5566                 {
5567                         .procname       = "force_tllao",
5568                         .data           = &ipv6_devconf.force_tllao,
5569                         .maxlen         = sizeof(int),
5570                         .mode           = 0644,
5571                         .proc_handler   = proc_dointvec
5572                 },
5573                 {
5574                         .procname       = "ndisc_notify",
5575                         .data           = &ipv6_devconf.ndisc_notify,
5576                         .maxlen         = sizeof(int),
5577                         .mode           = 0644,
5578                         .proc_handler   = proc_dointvec
5579                 },
5580                 {
5581                         .procname       = "suppress_frag_ndisc",
5582                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5583                         .maxlen         = sizeof(int),
5584                         .mode           = 0644,
5585                         .proc_handler   = proc_dointvec
5586                 },
5587                 {
5588                         .procname       = "accept_ra_from_local",
5589                         .data           = &ipv6_devconf.accept_ra_from_local,
5590                         .maxlen         = sizeof(int),
5591                         .mode           = 0644,
5592                         .proc_handler   = proc_dointvec,
5593                 },
5594                 {
5595                         .procname       = "accept_ra_mtu",
5596                         .data           = &ipv6_devconf.accept_ra_mtu,
5597                         .maxlen         = sizeof(int),
5598                         .mode           = 0644,
5599                         .proc_handler   = proc_dointvec,
5600                 },
5601                 {
5602                         .procname       = "stable_secret",
5603                         .data           = &ipv6_devconf.stable_secret,
5604                         .maxlen         = IPV6_MAX_STRLEN,
5605                         .mode           = 0600,
5606                         .proc_handler   = addrconf_sysctl_stable_secret,
5607                 },
5608                 {
5609                         /* sentinel */
5610                 }
5611         },
5612 };
5613
5614 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5615                 struct inet6_dev *idev, struct ipv6_devconf *p)
5616 {
5617         int i;
5618         struct addrconf_sysctl_table *t;
5619         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5620
5621         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5622         if (!t)
5623                 goto out;
5624
5625         for (i = 0; t->addrconf_vars[i].data; i++) {
5626                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5627                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5628                 t->addrconf_vars[i].extra2 = net;
5629         }
5630
5631         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5632
5633         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5634         if (!t->sysctl_header)
5635                 goto free;
5636
5637         p->sysctl = t;
5638         return 0;
5639
5640 free:
5641         kfree(t);
5642 out:
5643         return -ENOBUFS;
5644 }
5645
5646 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5647 {
5648         struct addrconf_sysctl_table *t;
5649
5650         if (!p->sysctl)
5651                 return;
5652
5653         t = p->sysctl;
5654         p->sysctl = NULL;
5655         unregister_net_sysctl_table(t->sysctl_header);
5656         kfree(t);
5657 }
5658
5659 static int addrconf_sysctl_register(struct inet6_dev *idev)
5660 {
5661         int err;
5662
5663         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5664                 return -EINVAL;
5665
5666         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5667                                     &ndisc_ifinfo_sysctl_change);
5668         if (err)
5669                 return err;
5670         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5671                                          idev, &idev->cnf);
5672         if (err)
5673                 neigh_sysctl_unregister(idev->nd_parms);
5674
5675         return err;
5676 }
5677
5678 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5679 {
5680         __addrconf_sysctl_unregister(&idev->cnf);
5681         neigh_sysctl_unregister(idev->nd_parms);
5682 }
5683
5684
5685 #endif
5686
5687 static int __net_init addrconf_init_net(struct net *net)
5688 {
5689         int err = -ENOMEM;
5690         struct ipv6_devconf *all, *dflt;
5691
5692         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5693         if (!all)
5694                 goto err_alloc_all;
5695
5696         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5697         if (!dflt)
5698                 goto err_alloc_dflt;
5699
5700         /* these will be inherited by all namespaces */
5701         dflt->autoconf = ipv6_defaults.autoconf;
5702         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5703
5704         dflt->stable_secret.initialized = false;
5705         all->stable_secret.initialized = false;
5706
5707         net->ipv6.devconf_all = all;
5708         net->ipv6.devconf_dflt = dflt;
5709
5710 #ifdef CONFIG_SYSCTL
5711         err = __addrconf_sysctl_register(net, "all", NULL, all);
5712         if (err < 0)
5713                 goto err_reg_all;
5714
5715         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5716         if (err < 0)
5717                 goto err_reg_dflt;
5718 #endif
5719         return 0;
5720
5721 #ifdef CONFIG_SYSCTL
5722 err_reg_dflt:
5723         __addrconf_sysctl_unregister(all);
5724 err_reg_all:
5725         kfree(dflt);
5726 #endif
5727 err_alloc_dflt:
5728         kfree(all);
5729 err_alloc_all:
5730         return err;
5731 }
5732
5733 static void __net_exit addrconf_exit_net(struct net *net)
5734 {
5735 #ifdef CONFIG_SYSCTL
5736         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5737         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5738 #endif
5739         kfree(net->ipv6.devconf_dflt);
5740         kfree(net->ipv6.devconf_all);
5741 }
5742
5743 static struct pernet_operations addrconf_ops = {
5744         .init = addrconf_init_net,
5745         .exit = addrconf_exit_net,
5746 };
5747
5748 static struct rtnl_af_ops inet6_ops __read_mostly = {
5749         .family           = AF_INET6,
5750         .fill_link_af     = inet6_fill_link_af,
5751         .get_link_af_size = inet6_get_link_af_size,
5752         .validate_link_af = inet6_validate_link_af,
5753         .set_link_af      = inet6_set_link_af,
5754 };
5755
5756 /*
5757  *      Init / cleanup code
5758  */
5759
5760 int __init addrconf_init(void)
5761 {
5762         struct inet6_dev *idev;
5763         int i, err;
5764
5765         err = ipv6_addr_label_init();
5766         if (err < 0) {
5767                 pr_crit("%s: cannot initialize default policy table: %d\n",
5768                         __func__, err);
5769                 goto out;
5770         }
5771
5772         err = register_pernet_subsys(&addrconf_ops);
5773         if (err < 0)
5774                 goto out_addrlabel;
5775
5776         addrconf_wq = create_workqueue("ipv6_addrconf");
5777         if (!addrconf_wq) {
5778                 err = -ENOMEM;
5779                 goto out_nowq;
5780         }
5781
5782         /* The addrconf netdev notifier requires that loopback_dev
5783          * has it's ipv6 private information allocated and setup
5784          * before it can bring up and give link-local addresses
5785          * to other devices which are up.
5786          *
5787          * Unfortunately, loopback_dev is not necessarily the first
5788          * entry in the global dev_base list of net devices.  In fact,
5789          * it is likely to be the very last entry on that list.
5790          * So this causes the notifier registry below to try and
5791          * give link-local addresses to all devices besides loopback_dev
5792          * first, then loopback_dev, which cases all the non-loopback_dev
5793          * devices to fail to get a link-local address.
5794          *
5795          * So, as a temporary fix, allocate the ipv6 structure for
5796          * loopback_dev first by hand.
5797          * Longer term, all of the dependencies ipv6 has upon the loopback
5798          * device and it being up should be removed.
5799          */
5800         rtnl_lock();
5801         idev = ipv6_add_dev(init_net.loopback_dev);
5802         rtnl_unlock();
5803         if (IS_ERR(idev)) {
5804                 err = PTR_ERR(idev);
5805                 goto errlo;
5806         }
5807
5808         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5809                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5810
5811         register_netdevice_notifier(&ipv6_dev_notf);
5812
5813         addrconf_verify();
5814
5815         rtnl_af_register(&inet6_ops);
5816
5817         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5818                               NULL);
5819         if (err < 0)
5820                 goto errout;
5821
5822         /* Only the first call to __rtnl_register can fail */
5823         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5824         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5825         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5826                         inet6_dump_ifaddr, NULL);
5827         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5828                         inet6_dump_ifmcaddr, NULL);
5829         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5830                         inet6_dump_ifacaddr, NULL);
5831         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5832                         inet6_netconf_dump_devconf, NULL);
5833
5834         ipv6_addr_label_rtnl_register();
5835
5836         return 0;
5837 errout:
5838         rtnl_af_unregister(&inet6_ops);
5839         unregister_netdevice_notifier(&ipv6_dev_notf);
5840 errlo:
5841         destroy_workqueue(addrconf_wq);
5842 out_nowq:
5843         unregister_pernet_subsys(&addrconf_ops);
5844 out_addrlabel:
5845         ipv6_addr_label_cleanup();
5846 out:
5847         return err;
5848 }
5849
5850 void addrconf_cleanup(void)
5851 {
5852         struct net_device *dev;
5853         int i;
5854
5855         unregister_netdevice_notifier(&ipv6_dev_notf);
5856         unregister_pernet_subsys(&addrconf_ops);
5857         ipv6_addr_label_cleanup();
5858
5859         rtnl_lock();
5860
5861         __rtnl_af_unregister(&inet6_ops);
5862
5863         /* clean dev list */
5864         for_each_netdev(&init_net, dev) {
5865                 if (__in6_dev_get(dev) == NULL)
5866                         continue;
5867                 addrconf_ifdown(dev, 1);
5868         }
5869         addrconf_ifdown(init_net.loopback_dev, 2);
5870
5871         /*
5872          *      Check hash table.
5873          */
5874         spin_lock_bh(&addrconf_hash_lock);
5875         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5876                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5877         spin_unlock_bh(&addrconf_hash_lock);
5878         cancel_delayed_work(&addr_chk_work);
5879         rtnl_unlock();
5880
5881         destroy_workqueue(addrconf_wq);
5882 }