net/ipv6/addrconf.c

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