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/sched/signal.h>
  47 #include <linux/socket.h>
  48 #include <linux/sockios.h>
  49 #include <linux/net.h>
  50 #include <linux/inet.h>
  51 #include <linux/in6.h>
  52 #include <linux/netdevice.h>
  53 #include <linux/if_addr.h>
  54 #include <linux/if_arp.h>
  55 #include <linux/if_arcnet.h>
  56 #include <linux/if_infiniband.h>
  57 #include <linux/route.h>
  58 #include <linux/inetdevice.h>
  59 #include <linux/init.h>
  60 #include <linux/slab.h>
  61 #ifdef CONFIG_SYSCTL
  62 #include <linux/sysctl.h>
  63 #endif
  64 #include <linux/capability.h>
  65 #include <linux/delay.h>
  66 #include <linux/notifier.h>
  67 #include <linux/string.h>
  68 #include <linux/hash.h>
  69
  70 #include <net/net_namespace.h>
  71 #include <net/sock.h>
  72 #include <net/snmp.h>
  73
  74 #include <net/6lowpan.h>
  75 #include <net/firewire.h>
  76 #include <net/ipv6.h>
  77 #include <net/protocol.h>
  78 #include <net/ndisc.h>
  79 #include <net/ip6_route.h>
  80 #include <net/addrconf.h>
  81 #include <net/tcp.h>
  82 #include <net/ip.h>
  83 #include <net/netlink.h>
  84 #include <net/pkt_sched.h>
  85 #include <net/l3mdev.h>
  86 #include <linux/if_tunnel.h>
  87 #include <linux/rtnetlink.h>
  88 #include <linux/netconf.h>
  89 #include <linux/random.h>
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  98
  99 #define IPV6_MAX_STRLEN \
 100         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 108 {
 109         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 110         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 111         do_div(tmp, 1000000);
 112         return (s32)tmp;
 113 }
 114
 115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 116 {
 117         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 118         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 119         do_div(tmp, 1000000);
 120         if ((s32)tmp > mrt) {
 121                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 122                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 123                 do_div(tmp, 1000000);
 124         }
 125         return (s32)tmp;
 126 }
 127
 128 #ifdef CONFIG_SYSCTL
 129 static int addrconf_sysctl_register(struct inet6_dev *idev);
 130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 131 #else
 132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 133 {
 134         return 0;
 135 }
 136
 137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 138 {
 139 }
 140 #endif
 141
 142 static void ipv6_regen_rndid(struct inet6_dev *idev);
 143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 144
 145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 146 static int ipv6_count_addresses(const struct inet6_dev *idev);
 147 static int ipv6_generate_stable_address(struct in6_addr *addr,
 148                                         u8 dad_count,
 149                                         const struct inet6_dev *idev);
 150
 151 #define IN6_ADDR_HSIZE_SHIFT    8
 152 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 153 /*
 154  *      Configured unicast address hash table
 155  */
 156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 157 static DEFINE_SPINLOCK(addrconf_hash_lock);
 158
 159 static void addrconf_verify(void);
 160 static void addrconf_verify_rtnl(void);
 161 static void addrconf_verify_work(struct work_struct *);
 162
 163 static struct workqueue_struct *addrconf_wq;
 164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 165
 166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 168
 169 static void addrconf_type_change(struct net_device *dev,
 170                                  unsigned long event);
 171 static int addrconf_ifdown(struct net_device *dev, int how);
 172
 173 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 174                                                   int plen,
 175                                                   const struct net_device *dev,
 176                                                   u32 flags, u32 noflags,
 177                                                   bool no_gw);
 178
 179 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 180 static void addrconf_dad_work(struct work_struct *w);
 181 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 182                                    bool send_na);
 183 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 184 static void addrconf_rs_timer(struct timer_list *t);
 185 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 186 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 187
 188 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 189                                 struct prefix_info *pinfo);
 190
 191 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 192         .forwarding             = 0,
 193         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 194         .mtu6                   = IPV6_MIN_MTU,
 195         .accept_ra              = 1,
 196         .accept_redirects       = 1,
 197         .autoconf               = 1,
 198         .force_mld_version      = 0,
 199         .mldv1_unsolicited_report_interval = 10 * HZ,
 200         .mldv2_unsolicited_report_interval = HZ,
 201         .dad_transmits          = 1,
 202         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 203         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 204         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 205         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 206         .use_tempaddr           = 0,
 207         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 208         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 209         .regen_max_retry        = REGEN_MAX_RETRY,
 210         .max_desync_factor      = MAX_DESYNC_FACTOR,
 211         .max_addresses          = IPV6_MAX_ADDRESSES,
 212         .accept_ra_defrtr       = 1,
 213         .accept_ra_from_local   = 0,
 214         .accept_ra_min_hop_limit= 1,
 215         .accept_ra_pinfo        = 1,
 216 #ifdef CONFIG_IPV6_ROUTER_PREF
 217         .accept_ra_rtr_pref     = 1,
 218         .rtr_probe_interval     = 60 * HZ,
 219 #ifdef CONFIG_IPV6_ROUTE_INFO
 220         .accept_ra_rt_info_min_plen = 0,
 221         .accept_ra_rt_info_max_plen = 0,
 222 #endif
 223 #endif
 224         .proxy_ndp              = 0,
 225         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 226         .disable_ipv6           = 0,
 227         .accept_dad             = 0,
 228         .suppress_frag_ndisc    = 1,
 229         .accept_ra_mtu          = 1,
 230         .stable_secret          = {
 231                 .initialized = false,
 232         },
 233         .use_oif_addrs_only     = 0,
 234         .ignore_routes_with_linkdown = 0,
 235         .keep_addr_on_down      = 0,
 236         .seg6_enabled           = 0,
 237 #ifdef CONFIG_IPV6_SEG6_HMAC
 238         .seg6_require_hmac      = 0,
 239 #endif
 240         .enhanced_dad           = 1,
 241         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 242         .disable_policy         = 0,
 243 };
 244
 245 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 246         .forwarding             = 0,
 247         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 248         .mtu6                   = IPV6_MIN_MTU,
 249         .accept_ra              = 1,
 250         .accept_redirects       = 1,
 251         .autoconf               = 1,
 252         .force_mld_version      = 0,
 253         .mldv1_unsolicited_report_interval = 10 * HZ,
 254         .mldv2_unsolicited_report_interval = HZ,
 255         .dad_transmits          = 1,
 256         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 257         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 258         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 259         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 260         .use_tempaddr           = 0,
 261         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 262         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 263         .regen_max_retry        = REGEN_MAX_RETRY,
 264         .max_desync_factor      = MAX_DESYNC_FACTOR,
 265         .max_addresses          = IPV6_MAX_ADDRESSES,
 266         .accept_ra_defrtr       = 1,
 267         .accept_ra_from_local   = 0,
 268         .accept_ra_min_hop_limit= 1,
 269         .accept_ra_pinfo        = 1,
 270 #ifdef CONFIG_IPV6_ROUTER_PREF
 271         .accept_ra_rtr_pref     = 1,
 272         .rtr_probe_interval     = 60 * HZ,
 273 #ifdef CONFIG_IPV6_ROUTE_INFO
 274         .accept_ra_rt_info_min_plen = 0,
 275         .accept_ra_rt_info_max_plen = 0,
 276 #endif
 277 #endif
 278         .proxy_ndp              = 0,
 279         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 280         .disable_ipv6           = 0,
 281         .accept_dad             = 1,
 282         .suppress_frag_ndisc    = 1,
 283         .accept_ra_mtu          = 1,
 284         .stable_secret          = {
 285                 .initialized = false,
 286         },
 287         .use_oif_addrs_only     = 0,
 288         .ignore_routes_with_linkdown = 0,
 289         .keep_addr_on_down      = 0,
 290         .seg6_enabled           = 0,
 291 #ifdef CONFIG_IPV6_SEG6_HMAC
 292         .seg6_require_hmac      = 0,
 293 #endif
 294         .enhanced_dad           = 1,
 295         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 296         .disable_policy         = 0,
 297 };
 298
 299 /* Check if link is ready: is it up and is a valid qdisc available */
 300 static inline bool addrconf_link_ready(const struct net_device *dev)
 301 {
 302         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 303 }
 304
 305 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 306 {
 307         if (del_timer(&idev->rs_timer))
 308                 __in6_dev_put(idev);
 309 }
 310
 311 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 312 {
 313         if (cancel_delayed_work(&ifp->dad_work))
 314                 __in6_ifa_put(ifp);
 315 }
 316
 317 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 318                                   unsigned long when)
 319 {
 320         if (!timer_pending(&idev->rs_timer))
 321                 in6_dev_hold(idev);
 322         mod_timer(&idev->rs_timer, jiffies + when);
 323 }
 324
 325 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 326                                    unsigned long delay)
 327 {
 328         in6_ifa_hold(ifp);
 329         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 330                 in6_ifa_put(ifp);
 331 }
 332
 333 static int snmp6_alloc_dev(struct inet6_dev *idev)
 334 {
 335         int i;
 336
 337         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 338         if (!idev->stats.ipv6)
 339                 goto err_ip;
 340
 341         for_each_possible_cpu(i) {
 342                 struct ipstats_mib *addrconf_stats;
 343                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 344                 u64_stats_init(&addrconf_stats->syncp);
 345         }
 346
 347
 348         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 349                                         GFP_KERNEL);
 350         if (!idev->stats.icmpv6dev)
 351                 goto err_icmp;
 352         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 353                                            GFP_KERNEL);
 354         if (!idev->stats.icmpv6msgdev)
 355                 goto err_icmpmsg;
 356
 357         return 0;
 358
 359 err_icmpmsg:
 360         kfree(idev->stats.icmpv6dev);
 361 err_icmp:
 362         free_percpu(idev->stats.ipv6);
 363 err_ip:
 364         return -ENOMEM;
 365 }
 366
 367 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 368 {
 369         struct inet6_dev *ndev;
 370         int err = -ENOMEM;
 371
 372         ASSERT_RTNL();
 373
 374         if (dev->mtu < IPV6_MIN_MTU)
 375                 return ERR_PTR(-EINVAL);
 376
 377         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 378         if (!ndev)
 379                 return ERR_PTR(err);
 380
 381         rwlock_init(&ndev->lock);
 382         ndev->dev = dev;
 383         INIT_LIST_HEAD(&ndev->addr_list);
 384         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 385         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 386
 387         if (ndev->cnf.stable_secret.initialized)
 388                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 389
 390         ndev->cnf.mtu6 = dev->mtu;
 391         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 392         if (!ndev->nd_parms) {
 393                 kfree(ndev);
 394                 return ERR_PTR(err);
 395         }
 396         if (ndev->cnf.forwarding)
 397                 dev_disable_lro(dev);
 398         /* We refer to the device */
 399         dev_hold(dev);
 400
 401         if (snmp6_alloc_dev(ndev) < 0) {
 402                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 403                            __func__);
 404                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 405                 dev_put(dev);
 406                 kfree(ndev);
 407                 return ERR_PTR(err);
 408         }
 409
 410         if (snmp6_register_dev(ndev) < 0) {
 411                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 412                            __func__, dev->name);
 413                 goto err_release;
 414         }
 415
 416         /* One reference from device. */
 417         refcount_set(&ndev->refcnt, 1);
 418
 419         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 420                 ndev->cnf.accept_dad = -1;
 421
 422 #if IS_ENABLED(CONFIG_IPV6_SIT)
 423         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 424                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 425                 ndev->cnf.rtr_solicits = 0;
 426         }
 427 #endif
 428
 429         INIT_LIST_HEAD(&ndev->tempaddr_list);
 430         ndev->desync_factor = U32_MAX;
 431         if ((dev->flags&IFF_LOOPBACK) ||
 432             dev->type == ARPHRD_TUNNEL ||
 433             dev->type == ARPHRD_TUNNEL6 ||
 434             dev->type == ARPHRD_SIT ||
 435             dev->type == ARPHRD_NONE) {
 436                 ndev->cnf.use_tempaddr = -1;
 437         } else
 438                 ipv6_regen_rndid(ndev);
 439
 440         ndev->token = in6addr_any;
 441
 442         if (netif_running(dev) && addrconf_link_ready(dev))
 443                 ndev->if_flags |= IF_READY;
 444
 445         ipv6_mc_init_dev(ndev);
 446         ndev->tstamp = jiffies;
 447         err = addrconf_sysctl_register(ndev);
 448         if (err) {
 449                 ipv6_mc_destroy_dev(ndev);
 450                 snmp6_unregister_dev(ndev);
 451                 goto err_release;
 452         }
 453         /* protected by rtnl_lock */
 454         rcu_assign_pointer(dev->ip6_ptr, ndev);
 455
 456         /* Join interface-local all-node multicast group */
 457         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 458
 459         /* Join all-node multicast group */
 460         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 461
 462         /* Join all-router multicast group if forwarding is set */
 463         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 464                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 465
 466         return ndev;
 467
 468 err_release:
 469         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 470         ndev->dead = 1;
 471         in6_dev_finish_destroy(ndev);
 472         return ERR_PTR(err);
 473 }
 474
 475 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 476 {
 477         struct inet6_dev *idev;
 478
 479         ASSERT_RTNL();
 480
 481         idev = __in6_dev_get(dev);
 482         if (!idev) {
 483                 idev = ipv6_add_dev(dev);
 484                 if (IS_ERR(idev))
 485                         return NULL;
 486         }
 487
 488         if (dev->flags&IFF_UP)
 489                 ipv6_mc_up(idev);
 490         return idev;
 491 }
 492
 493 static int inet6_netconf_msgsize_devconf(int type)
 494 {
 495         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 496                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 497         bool all = false;
 498
 499         if (type == NETCONFA_ALL)
 500                 all = true;
 501
 502         if (all || type == NETCONFA_FORWARDING)
 503                 size += nla_total_size(4);
 504 #ifdef CONFIG_IPV6_MROUTE
 505         if (all || type == NETCONFA_MC_FORWARDING)
 506                 size += nla_total_size(4);
 507 #endif
 508         if (all || type == NETCONFA_PROXY_NEIGH)
 509                 size += nla_total_size(4);
 510
 511         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 512                 size += nla_total_size(4);
 513
 514         return size;
 515 }
 516
 517 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 518                                       struct ipv6_devconf *devconf, u32 portid,
 519                                       u32 seq, int event, unsigned int flags,
 520                                       int type)
 521 {
 522         struct nlmsghdr  *nlh;
 523         struct netconfmsg *ncm;
 524         bool all = false;
 525
 526         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 527                         flags);
 528         if (!nlh)
 529                 return -EMSGSIZE;
 530
 531         if (type == NETCONFA_ALL)
 532                 all = true;
 533
 534         ncm = nlmsg_data(nlh);
 535         ncm->ncm_family = AF_INET6;
 536
 537         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 538                 goto nla_put_failure;
 539
 540         if (!devconf)
 541                 goto out;
 542
 543         if ((all || type == NETCONFA_FORWARDING) &&
 544             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 545                 goto nla_put_failure;
 546 #ifdef CONFIG_IPV6_MROUTE
 547         if ((all || type == NETCONFA_MC_FORWARDING) &&
 548             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 549                         devconf->mc_forwarding) < 0)
 550                 goto nla_put_failure;
 551 #endif
 552         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 553             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 554                 goto nla_put_failure;
 555
 556         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 557             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 558                         devconf->ignore_routes_with_linkdown) < 0)
 559                 goto nla_put_failure;
 560
 561 out:
 562         nlmsg_end(skb, nlh);
 563         return 0;
 564
 565 nla_put_failure:
 566         nlmsg_cancel(skb, nlh);
 567         return -EMSGSIZE;
 568 }
 569
 570 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 571                                   int ifindex, struct ipv6_devconf *devconf)
 572 {
 573         struct sk_buff *skb;
 574         int err = -ENOBUFS;
 575
 576         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 577         if (!skb)
 578                 goto errout;
 579
 580         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 581                                          event, 0, type);
 582         if (err < 0) {
 583                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 584                 WARN_ON(err == -EMSGSIZE);
 585                 kfree_skb(skb);
 586                 goto errout;
 587         }
 588         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 589         return;
 590 errout:
 591         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 592 }
 593
 594 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 595         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 596         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 597         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 598         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 599 };
 600
 601 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 602                                        const struct nlmsghdr *nlh,
 603                                        struct nlattr **tb,
 604                                        struct netlink_ext_ack *extack)
 605 {
 606         int i, err;
 607
 608         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 609                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 610                 return -EINVAL;
 611         }
 612
 613         if (!netlink_strict_get_check(skb))
 614                 return nlmsg_parse(nlh, sizeof(struct netconfmsg), tb,
 615                                    NETCONFA_MAX, devconf_ipv6_policy, extack);
 616
 617         err = nlmsg_parse_strict(nlh, sizeof(struct netconfmsg), tb,
 618                                  NETCONFA_MAX, devconf_ipv6_policy, extack);
 619         if (err)
 620                 return err;
 621
 622         for (i = 0; i <= NETCONFA_MAX; i++) {
 623                 if (!tb[i])
 624                         continue;
 625
 626                 switch (i) {
 627                 case NETCONFA_IFINDEX:
 628                         break;
 629                 default:
 630                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 631                         return -EINVAL;
 632                 }
 633         }
 634
 635         return 0;
 636 }
 637
 638 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 639                                      struct nlmsghdr *nlh,
 640                                      struct netlink_ext_ack *extack)
 641 {
 642         struct net *net = sock_net(in_skb->sk);
 643         struct nlattr *tb[NETCONFA_MAX+1];
 644         struct inet6_dev *in6_dev = NULL;
 645         struct net_device *dev = NULL;
 646         struct sk_buff *skb;
 647         struct ipv6_devconf *devconf;
 648         int ifindex;
 649         int err;
 650
 651         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 652         if (err < 0)
 653                 return err;
 654
 655         if (!tb[NETCONFA_IFINDEX])
 656                 return -EINVAL;
 657
 658         err = -EINVAL;
 659         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 660         switch (ifindex) {
 661         case NETCONFA_IFINDEX_ALL:
 662                 devconf = net->ipv6.devconf_all;
 663                 break;
 664         case NETCONFA_IFINDEX_DEFAULT:
 665                 devconf = net->ipv6.devconf_dflt;
 666                 break;
 667         default:
 668                 dev = dev_get_by_index(net, ifindex);
 669                 if (!dev)
 670                         return -EINVAL;
 671                 in6_dev = in6_dev_get(dev);
 672                 if (!in6_dev)
 673                         goto errout;
 674                 devconf = &in6_dev->cnf;
 675                 break;
 676         }
 677
 678         err = -ENOBUFS;
 679         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 680         if (!skb)
 681                 goto errout;
 682
 683         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 684                                          NETLINK_CB(in_skb).portid,
 685                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 686                                          NETCONFA_ALL);
 687         if (err < 0) {
 688                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 689                 WARN_ON(err == -EMSGSIZE);
 690                 kfree_skb(skb);
 691                 goto errout;
 692         }
 693         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 694 errout:
 695         if (in6_dev)
 696                 in6_dev_put(in6_dev);
 697         if (dev)
 698                 dev_put(dev);
 699         return err;
 700 }
 701
 702 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 703                                       struct netlink_callback *cb)
 704 {
 705         const struct nlmsghdr *nlh = cb->nlh;
 706         struct net *net = sock_net(skb->sk);
 707         int h, s_h;
 708         int idx, s_idx;
 709         struct net_device *dev;
 710         struct inet6_dev *idev;
 711         struct hlist_head *head;
 712
 713         if (cb->strict_check) {
 714                 struct netlink_ext_ack *extack = cb->extack;
 715                 struct netconfmsg *ncm;
 716
 717                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 718                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 719                         return -EINVAL;
 720                 }
 721
 722                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 723                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 724                         return -EINVAL;
 725                 }
 726         }
 727
 728         s_h = cb->args[0];
 729         s_idx = idx = cb->args[1];
 730
 731         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 732                 idx = 0;
 733                 head = &net->dev_index_head[h];
 734                 rcu_read_lock();
 735                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 736                           net->dev_base_seq;
 737                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 738                         if (idx < s_idx)
 739                                 goto cont;
 740                         idev = __in6_dev_get(dev);
 741                         if (!idev)
 742                                 goto cont;
 743
 744                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 745                                                        &idev->cnf,
 746                                                        NETLINK_CB(cb->skb).portid,
 747                                                        nlh->nlmsg_seq,
 748                                                        RTM_NEWNETCONF,
 749                                                        NLM_F_MULTI,
 750                                                        NETCONFA_ALL) < 0) {
 751                                 rcu_read_unlock();
 752                                 goto done;
 753                         }
 754                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 755 cont:
 756                         idx++;
 757                 }
 758                 rcu_read_unlock();
 759         }
 760         if (h == NETDEV_HASHENTRIES) {
 761                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 762                                                net->ipv6.devconf_all,
 763                                                NETLINK_CB(cb->skb).portid,
 764                                                nlh->nlmsg_seq,
 765                                                RTM_NEWNETCONF, NLM_F_MULTI,
 766                                                NETCONFA_ALL) < 0)
 767                         goto done;
 768                 else
 769                         h++;
 770         }
 771         if (h == NETDEV_HASHENTRIES + 1) {
 772                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 773                                                net->ipv6.devconf_dflt,
 774                                                NETLINK_CB(cb->skb).portid,
 775                                                nlh->nlmsg_seq,
 776                                                RTM_NEWNETCONF, NLM_F_MULTI,
 777                                                NETCONFA_ALL) < 0)
 778                         goto done;
 779                 else
 780                         h++;
 781         }
 782 done:
 783         cb->args[0] = h;
 784         cb->args[1] = idx;
 785
 786         return skb->len;
 787 }
 788
 789 #ifdef CONFIG_SYSCTL
 790 static void dev_forward_change(struct inet6_dev *idev)
 791 {
 792         struct net_device *dev;
 793         struct inet6_ifaddr *ifa;
 794
 795         if (!idev)
 796                 return;
 797         dev = idev->dev;
 798         if (idev->cnf.forwarding)
 799                 dev_disable_lro(dev);
 800         if (dev->flags & IFF_MULTICAST) {
 801                 if (idev->cnf.forwarding) {
 802                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 803                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 804                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 805                 } else {
 806                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 807                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 808                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 809                 }
 810         }
 811
 812         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 813                 if (ifa->flags&IFA_F_TENTATIVE)
 814                         continue;
 815                 if (idev->cnf.forwarding)
 816                         addrconf_join_anycast(ifa);
 817                 else
 818                         addrconf_leave_anycast(ifa);
 819         }
 820         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 821                                      NETCONFA_FORWARDING,
 822                                      dev->ifindex, &idev->cnf);
 823 }
 824
 825
 826 static void addrconf_forward_change(struct net *net, __s32 newf)
 827 {
 828         struct net_device *dev;
 829         struct inet6_dev *idev;
 830
 831         for_each_netdev(net, dev) {
 832                 idev = __in6_dev_get(dev);
 833                 if (idev) {
 834                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 835                         idev->cnf.forwarding = newf;
 836                         if (changed)
 837                                 dev_forward_change(idev);
 838                 }
 839         }
 840 }
 841
 842 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 843 {
 844         struct net *net;
 845         int old;
 846
 847         if (!rtnl_trylock())
 848                 return restart_syscall();
 849
 850         net = (struct net *)table->extra2;
 851         old = *p;
 852         *p = newf;
 853
 854         if (p == &net->ipv6.devconf_dflt->forwarding) {
 855                 if ((!newf) ^ (!old))
 856                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 857                                                      NETCONFA_FORWARDING,
 858                                                      NETCONFA_IFINDEX_DEFAULT,
 859                                                      net->ipv6.devconf_dflt);
 860                 rtnl_unlock();
 861                 return 0;
 862         }
 863
 864         if (p == &net->ipv6.devconf_all->forwarding) {
 865                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 866
 867                 net->ipv6.devconf_dflt->forwarding = newf;
 868                 if ((!newf) ^ (!old_dflt))
 869                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 870                                                      NETCONFA_FORWARDING,
 871                                                      NETCONFA_IFINDEX_DEFAULT,
 872                                                      net->ipv6.devconf_dflt);
 873
 874                 addrconf_forward_change(net, newf);
 875                 if ((!newf) ^ (!old))
 876                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 877                                                      NETCONFA_FORWARDING,
 878                                                      NETCONFA_IFINDEX_ALL,
 879                                                      net->ipv6.devconf_all);
 880         } else if ((!newf) ^ (!old))
 881                 dev_forward_change((struct inet6_dev *)table->extra1);
 882         rtnl_unlock();
 883
 884         if (newf)
 885                 rt6_purge_dflt_routers(net);
 886         return 1;
 887 }
 888
 889 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 890 {
 891         struct net_device *dev;
 892         struct inet6_dev *idev;
 893
 894         for_each_netdev(net, dev) {
 895                 idev = __in6_dev_get(dev);
 896                 if (idev) {
 897                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 898
 899                         idev->cnf.ignore_routes_with_linkdown = newf;
 900                         if (changed)
 901                                 inet6_netconf_notify_devconf(dev_net(dev),
 902                                                              RTM_NEWNETCONF,
 903                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 904                                                              dev->ifindex,
 905                                                              &idev->cnf);
 906                 }
 907         }
 908 }
 909
 910 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 911 {
 912         struct net *net;
 913         int old;
 914
 915         if (!rtnl_trylock())
 916                 return restart_syscall();
 917
 918         net = (struct net *)table->extra2;
 919         old = *p;
 920         *p = newf;
 921
 922         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 923                 if ((!newf) ^ (!old))
 924                         inet6_netconf_notify_devconf(net,
 925                                                      RTM_NEWNETCONF,
 926                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 927                                                      NETCONFA_IFINDEX_DEFAULT,
 928                                                      net->ipv6.devconf_dflt);
 929                 rtnl_unlock();
 930                 return 0;
 931         }
 932
 933         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 934                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 935                 addrconf_linkdown_change(net, newf);
 936                 if ((!newf) ^ (!old))
 937                         inet6_netconf_notify_devconf(net,
 938                                                      RTM_NEWNETCONF,
 939                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 940                                                      NETCONFA_IFINDEX_ALL,
 941                                                      net->ipv6.devconf_all);
 942         }
 943         rtnl_unlock();
 944
 945         return 1;
 946 }
 947
 948 #endif
 949
 950 /* Nobody refers to this ifaddr, destroy it */
 951 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 952 {
 953         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 954
 955 #ifdef NET_REFCNT_DEBUG
 956         pr_debug("%s\n", __func__);
 957 #endif
 958
 959         in6_dev_put(ifp->idev);
 960
 961         if (cancel_delayed_work(&ifp->dad_work))
 962                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 963                           ifp);
 964
 965         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 966                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 967                 return;
 968         }
 969
 970         kfree_rcu(ifp, rcu);
 971 }
 972
 973 static void
 974 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 975 {
 976         struct list_head *p;
 977         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 978
 979         /*
 980          * Each device address list is sorted in order of scope -
 981          * global before linklocal.
 982          */
 983         list_for_each(p, &idev->addr_list) {
 984                 struct inet6_ifaddr *ifa
 985                         = list_entry(p, struct inet6_ifaddr, if_list);
 986                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 987                         break;
 988         }
 989
 990         list_add_tail_rcu(&ifp->if_list, p);
 991 }
 992
 993 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 994 {
 995         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 996
 997         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 998 }
 999
1000 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1001                                struct net_device *dev, unsigned int hash)
1002 {
1003         struct inet6_ifaddr *ifp;
1004
1005         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1006                 if (!net_eq(dev_net(ifp->idev->dev), net))
1007                         continue;
1008                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1009                         if (!dev || ifp->idev->dev == dev)
1010                                 return true;
1011                 }
1012         }
1013         return false;
1014 }
1015
1016 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1017 {
1018         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1019         int err = 0;
1020
1021         spin_lock(&addrconf_hash_lock);
1022
1023         /* Ignore adding duplicate addresses on an interface */
1024         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1025                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1026                 err = -EEXIST;
1027         } else {
1028                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1029         }
1030
1031         spin_unlock(&addrconf_hash_lock);
1032
1033         return err;
1034 }
1035
1036 /* On success it returns ifp with increased reference count */
1037
1038 static struct inet6_ifaddr *
1039 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1040               bool can_block, struct netlink_ext_ack *extack)
1041 {
1042         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1043         int addr_type = ipv6_addr_type(cfg->pfx);
1044         struct net *net = dev_net(idev->dev);
1045         struct inet6_ifaddr *ifa = NULL;
1046         struct fib6_info *f6i = NULL;
1047         int err = 0;
1048
1049         if (addr_type == IPV6_ADDR_ANY ||
1050             addr_type & IPV6_ADDR_MULTICAST ||
1051             (!(idev->dev->flags & IFF_LOOPBACK) &&
1052              !netif_is_l3_master(idev->dev) &&
1053              addr_type & IPV6_ADDR_LOOPBACK))
1054                 return ERR_PTR(-EADDRNOTAVAIL);
1055
1056         if (idev->dead) {
1057                 err = -ENODEV;                  /*XXX*/
1058                 goto out;
1059         }
1060
1061         if (idev->cnf.disable_ipv6) {
1062                 err = -EACCES;
1063                 goto out;
1064         }
1065
1066         /* validator notifier needs to be blocking;
1067          * do not call in atomic context
1068          */
1069         if (can_block) {
1070                 struct in6_validator_info i6vi = {
1071                         .i6vi_addr = *cfg->pfx,
1072                         .i6vi_dev = idev,
1073                         .extack = extack,
1074                 };
1075
1076                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1077                 err = notifier_to_errno(err);
1078                 if (err < 0)
1079                         goto out;
1080         }
1081
1082         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1083         if (!ifa) {
1084                 err = -ENOBUFS;
1085                 goto out;
1086         }
1087
1088         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1089         if (IS_ERR(f6i)) {
1090                 err = PTR_ERR(f6i);
1091                 f6i = NULL;
1092                 goto out;
1093         }
1094
1095         if (net->ipv6.devconf_all->disable_policy ||
1096             idev->cnf.disable_policy)
1097                 f6i->dst_nopolicy = true;
1098
1099         neigh_parms_data_state_setall(idev->nd_parms);
1100
1101         ifa->addr = *cfg->pfx;
1102         if (cfg->peer_pfx)
1103                 ifa->peer_addr = *cfg->peer_pfx;
1104
1105         spin_lock_init(&ifa->lock);
1106         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1107         INIT_HLIST_NODE(&ifa->addr_lst);
1108         ifa->scope = cfg->scope;
1109         ifa->prefix_len = cfg->plen;
1110         ifa->rt_priority = cfg->rt_priority;
1111         ifa->flags = cfg->ifa_flags;
1112         /* No need to add the TENTATIVE flag for addresses with NODAD */
1113         if (!(cfg->ifa_flags & IFA_F_NODAD))
1114                 ifa->flags |= IFA_F_TENTATIVE;
1115         ifa->valid_lft = cfg->valid_lft;
1116         ifa->prefered_lft = cfg->preferred_lft;
1117         ifa->cstamp = ifa->tstamp = jiffies;
1118         ifa->tokenized = false;
1119
1120         ifa->rt = f6i;
1121
1122         ifa->idev = idev;
1123         in6_dev_hold(idev);
1124
1125         /* For caller */
1126         refcount_set(&ifa->refcnt, 1);
1127
1128         rcu_read_lock_bh();
1129
1130         err = ipv6_add_addr_hash(idev->dev, ifa);
1131         if (err < 0) {
1132                 rcu_read_unlock_bh();
1133                 goto out;
1134         }
1135
1136         write_lock(&idev->lock);
1137
1138         /* Add to inet6_dev unicast addr list. */
1139         ipv6_link_dev_addr(idev, ifa);
1140
1141         if (ifa->flags&IFA_F_TEMPORARY) {
1142                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1143                 in6_ifa_hold(ifa);
1144         }
1145
1146         in6_ifa_hold(ifa);
1147         write_unlock(&idev->lock);
1148
1149         rcu_read_unlock_bh();
1150
1151         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1152 out:
1153         if (unlikely(err < 0)) {
1154                 fib6_info_release(f6i);
1155
1156                 if (ifa) {
1157                         if (ifa->idev)
1158                                 in6_dev_put(ifa->idev);
1159                         kfree(ifa);
1160                 }
1161                 ifa = ERR_PTR(err);
1162         }
1163
1164         return ifa;
1165 }
1166
1167 enum cleanup_prefix_rt_t {
1168         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1169         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1170         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1171 };
1172
1173 /*
1174  * Check, whether the prefix for ifp would still need a prefix route
1175  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1176  * constants.
1177  *
1178  * 1) we don't purge prefix if address was not permanent.
1179  *    prefix is managed by its own lifetime.
1180  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1181  * 3) if there are no addresses, delete prefix.
1182  * 4) if there are still other permanent address(es),
1183  *    corresponding prefix is still permanent.
1184  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1185  *    don't purge the prefix, assume user space is managing it.
1186  * 6) otherwise, update prefix lifetime to the
1187  *    longest valid lifetime among the corresponding
1188  *    addresses on the device.
1189  *    Note: subsequent RA will update lifetime.
1190  **/
1191 static enum cleanup_prefix_rt_t
1192 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1193 {
1194         struct inet6_ifaddr *ifa;
1195         struct inet6_dev *idev = ifp->idev;
1196         unsigned long lifetime;
1197         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1198
1199         *expires = jiffies;
1200
1201         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1202                 if (ifa == ifp)
1203                         continue;
1204                 if (ifa->prefix_len != ifp->prefix_len ||
1205                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1206                                        ifp->prefix_len))
1207                         continue;
1208                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1209                         return CLEANUP_PREFIX_RT_NOP;
1210
1211                 action = CLEANUP_PREFIX_RT_EXPIRE;
1212
1213                 spin_lock(&ifa->lock);
1214
1215                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1216                 /*
1217                  * Note: Because this address is
1218                  * not permanent, lifetime <
1219                  * LONG_MAX / HZ here.
1220                  */
1221                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1222                         *expires = ifa->tstamp + lifetime * HZ;
1223                 spin_unlock(&ifa->lock);
1224         }
1225
1226         return action;
1227 }
1228
1229 static void
1230 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1231 {
1232         struct fib6_info *f6i;
1233
1234         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
1235                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1236         if (f6i) {
1237                 if (del_rt)
1238                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1239                 else {
1240                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1241                                 fib6_set_expires(f6i, expires);
1242                         fib6_info_release(f6i);
1243                 }
1244         }
1245 }
1246
1247
1248 /* This function wants to get referenced ifp and releases it before return */
1249
1250 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1251 {
1252         int state;
1253         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1254         unsigned long expires;
1255
1256         ASSERT_RTNL();
1257
1258         spin_lock_bh(&ifp->lock);
1259         state = ifp->state;
1260         ifp->state = INET6_IFADDR_STATE_DEAD;
1261         spin_unlock_bh(&ifp->lock);
1262
1263         if (state == INET6_IFADDR_STATE_DEAD)
1264                 goto out;
1265
1266         spin_lock_bh(&addrconf_hash_lock);
1267         hlist_del_init_rcu(&ifp->addr_lst);
1268         spin_unlock_bh(&addrconf_hash_lock);
1269
1270         write_lock_bh(&ifp->idev->lock);
1271
1272         if (ifp->flags&IFA_F_TEMPORARY) {
1273                 list_del(&ifp->tmp_list);
1274                 if (ifp->ifpub) {
1275                         in6_ifa_put(ifp->ifpub);
1276                         ifp->ifpub = NULL;
1277                 }
1278                 __in6_ifa_put(ifp);
1279         }
1280
1281         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1282                 action = check_cleanup_prefix_route(ifp, &expires);
1283
1284         list_del_rcu(&ifp->if_list);
1285         __in6_ifa_put(ifp);
1286
1287         write_unlock_bh(&ifp->idev->lock);
1288
1289         addrconf_del_dad_work(ifp);
1290
1291         ipv6_ifa_notify(RTM_DELADDR, ifp);
1292
1293         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1294
1295         if (action != CLEANUP_PREFIX_RT_NOP) {
1296                 cleanup_prefix_route(ifp, expires,
1297                         action == CLEANUP_PREFIX_RT_DEL);
1298         }
1299
1300         /* clean up prefsrc entries */
1301         rt6_remove_prefsrc(ifp);
1302 out:
1303         in6_ifa_put(ifp);
1304 }
1305
1306 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1307                                 struct inet6_ifaddr *ift,
1308                                 bool block)
1309 {
1310         struct inet6_dev *idev = ifp->idev;
1311         struct in6_addr addr, *tmpaddr;
1312         unsigned long tmp_tstamp, age;
1313         unsigned long regen_advance;
1314         struct ifa6_config cfg;
1315         int ret = 0;
1316         unsigned long now = jiffies;
1317         long max_desync_factor;
1318         s32 cnf_temp_preferred_lft;
1319
1320         write_lock_bh(&idev->lock);
1321         if (ift) {
1322                 spin_lock_bh(&ift->lock);
1323                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1324                 spin_unlock_bh(&ift->lock);
1325                 tmpaddr = &addr;
1326         } else {
1327                 tmpaddr = NULL;
1328         }
1329 retry:
1330         in6_dev_hold(idev);
1331         if (idev->cnf.use_tempaddr <= 0) {
1332                 write_unlock_bh(&idev->lock);
1333                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1334                 in6_dev_put(idev);
1335                 ret = -1;
1336                 goto out;
1337         }
1338         spin_lock_bh(&ifp->lock);
1339         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1340                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1341                 spin_unlock_bh(&ifp->lock);
1342                 write_unlock_bh(&idev->lock);
1343                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1344                         __func__);
1345                 in6_dev_put(idev);
1346                 ret = -1;
1347                 goto out;
1348         }
1349         in6_ifa_hold(ifp);
1350         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1351         ipv6_try_regen_rndid(idev, tmpaddr);
1352         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1353         age = (now - ifp->tstamp) / HZ;
1354
1355         regen_advance = idev->cnf.regen_max_retry *
1356                         idev->cnf.dad_transmits *
1357                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1358
1359         /* recalculate max_desync_factor each time and update
1360          * idev->desync_factor if it's larger
1361          */
1362         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1363         max_desync_factor = min_t(__u32,
1364                                   idev->cnf.max_desync_factor,
1365                                   cnf_temp_preferred_lft - regen_advance);
1366
1367         if (unlikely(idev->desync_factor > max_desync_factor)) {
1368                 if (max_desync_factor > 0) {
1369                         get_random_bytes(&idev->desync_factor,
1370                                          sizeof(idev->desync_factor));
1371                         idev->desync_factor %= max_desync_factor;
1372                 } else {
1373                         idev->desync_factor = 0;
1374                 }
1375         }
1376
1377         memset(&cfg, 0, sizeof(cfg));
1378         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1379                               idev->cnf.temp_valid_lft + age);
1380         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1381         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1382
1383         cfg.plen = ifp->prefix_len;
1384         tmp_tstamp = ifp->tstamp;
1385         spin_unlock_bh(&ifp->lock);
1386
1387         write_unlock_bh(&idev->lock);
1388
1389         /* A temporary address is created only if this calculated Preferred
1390          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1391          * an implementation must not create a temporary address with a zero
1392          * Preferred Lifetime.
1393          * Use age calculation as in addrconf_verify to avoid unnecessary
1394          * temporary addresses being generated.
1395          */
1396         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1397         if (cfg.preferred_lft <= regen_advance + age) {
1398                 in6_ifa_put(ifp);
1399                 in6_dev_put(idev);
1400                 ret = -1;
1401                 goto out;
1402         }
1403
1404         cfg.ifa_flags = IFA_F_TEMPORARY;
1405         /* set in addrconf_prefix_rcv() */
1406         if (ifp->flags & IFA_F_OPTIMISTIC)
1407                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1408
1409         cfg.pfx = &addr;
1410         cfg.scope = ipv6_addr_scope(cfg.pfx);
1411
1412         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1413         if (IS_ERR(ift)) {
1414                 in6_ifa_put(ifp);
1415                 in6_dev_put(idev);
1416                 pr_info("%s: retry temporary address regeneration\n", __func__);
1417                 tmpaddr = &addr;
1418                 write_lock_bh(&idev->lock);
1419                 goto retry;
1420         }
1421
1422         spin_lock_bh(&ift->lock);
1423         ift->ifpub = ifp;
1424         ift->cstamp = now;
1425         ift->tstamp = tmp_tstamp;
1426         spin_unlock_bh(&ift->lock);
1427
1428         addrconf_dad_start(ift);
1429         in6_ifa_put(ift);
1430         in6_dev_put(idev);
1431 out:
1432         return ret;
1433 }
1434
1435 /*
1436  *      Choose an appropriate source address (RFC3484)
1437  */
1438 enum {
1439         IPV6_SADDR_RULE_INIT = 0,
1440         IPV6_SADDR_RULE_LOCAL,
1441         IPV6_SADDR_RULE_SCOPE,
1442         IPV6_SADDR_RULE_PREFERRED,
1443 #ifdef CONFIG_IPV6_MIP6
1444         IPV6_SADDR_RULE_HOA,
1445 #endif
1446         IPV6_SADDR_RULE_OIF,
1447         IPV6_SADDR_RULE_LABEL,
1448         IPV6_SADDR_RULE_PRIVACY,
1449         IPV6_SADDR_RULE_ORCHID,
1450         IPV6_SADDR_RULE_PREFIX,
1451 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1452         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1453 #endif
1454         IPV6_SADDR_RULE_MAX
1455 };
1456
1457 struct ipv6_saddr_score {
1458         int                     rule;
1459         int                     addr_type;
1460         struct inet6_ifaddr     *ifa;
1461         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1462         int                     scopedist;
1463         int                     matchlen;
1464 };
1465
1466 struct ipv6_saddr_dst {
1467         const struct in6_addr *addr;
1468         int ifindex;
1469         int scope;
1470         int label;
1471         unsigned int prefs;
1472 };
1473
1474 static inline int ipv6_saddr_preferred(int type)
1475 {
1476         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1477                 return 1;
1478         return 0;
1479 }
1480
1481 static bool ipv6_use_optimistic_addr(struct net *net,
1482                                      struct inet6_dev *idev)
1483 {
1484 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1485         if (!idev)
1486                 return false;
1487         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1488                 return false;
1489         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1490                 return false;
1491
1492         return true;
1493 #else
1494         return false;
1495 #endif
1496 }
1497
1498 static bool ipv6_allow_optimistic_dad(struct net *net,
1499                                       struct inet6_dev *idev)
1500 {
1501 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1502         if (!idev)
1503                 return false;
1504         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1505                 return false;
1506
1507         return true;
1508 #else
1509         return false;
1510 #endif
1511 }
1512
1513 static int ipv6_get_saddr_eval(struct net *net,
1514                                struct ipv6_saddr_score *score,
1515                                struct ipv6_saddr_dst *dst,
1516                                int i)
1517 {
1518         int ret;
1519
1520         if (i <= score->rule) {
1521                 switch (i) {
1522                 case IPV6_SADDR_RULE_SCOPE:
1523                         ret = score->scopedist;
1524                         break;
1525                 case IPV6_SADDR_RULE_PREFIX:
1526                         ret = score->matchlen;
1527                         break;
1528                 default:
1529                         ret = !!test_bit(i, score->scorebits);
1530                 }
1531                 goto out;
1532         }
1533
1534         switch (i) {
1535         case IPV6_SADDR_RULE_INIT:
1536                 /* Rule 0: remember if hiscore is not ready yet */
1537                 ret = !!score->ifa;
1538                 break;
1539         case IPV6_SADDR_RULE_LOCAL:
1540                 /* Rule 1: Prefer same address */
1541                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1542                 break;
1543         case IPV6_SADDR_RULE_SCOPE:
1544                 /* Rule 2: Prefer appropriate scope
1545                  *
1546                  *      ret
1547                  *       ^
1548                  *    -1 |  d 15
1549                  *    ---+--+-+---> scope
1550                  *       |
1551                  *       |             d is scope of the destination.
1552                  *  B-d  |  \
1553                  *       |   \      <- smaller scope is better if
1554                  *  B-15 |    \        if scope is enough for destination.
1555                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1556                  * d-C-1 | /
1557                  *       |/         <- greater is better
1558                  *   -C  /             if scope is not enough for destination.
1559                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1560                  *
1561                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1562                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1563                  * Assume B = 0 and we get C > 29.
1564                  */
1565                 ret = __ipv6_addr_src_scope(score->addr_type);
1566                 if (ret >= dst->scope)
1567                         ret = -ret;
1568                 else
1569                         ret -= 128;     /* 30 is enough */
1570                 score->scopedist = ret;
1571                 break;
1572         case IPV6_SADDR_RULE_PREFERRED:
1573             {
1574                 /* Rule 3: Avoid deprecated and optimistic addresses */
1575                 u8 avoid = IFA_F_DEPRECATED;
1576
1577                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1578                         avoid |= IFA_F_OPTIMISTIC;
1579                 ret = ipv6_saddr_preferred(score->addr_type) ||
1580                       !(score->ifa->flags & avoid);
1581                 break;
1582             }
1583 #ifdef CONFIG_IPV6_MIP6
1584         case IPV6_SADDR_RULE_HOA:
1585             {
1586                 /* Rule 4: Prefer home address */
1587                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1588                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1589                 break;
1590             }
1591 #endif
1592         case IPV6_SADDR_RULE_OIF:
1593                 /* Rule 5: Prefer outgoing interface */
1594                 ret = (!dst->ifindex ||
1595                        dst->ifindex == score->ifa->idev->dev->ifindex);
1596                 break;
1597         case IPV6_SADDR_RULE_LABEL:
1598                 /* Rule 6: Prefer matching label */
1599                 ret = ipv6_addr_label(net,
1600                                       &score->ifa->addr, score->addr_type,
1601                                       score->ifa->idev->dev->ifindex) == dst->label;
1602                 break;
1603         case IPV6_SADDR_RULE_PRIVACY:
1604             {
1605                 /* Rule 7: Prefer public address
1606                  * Note: prefer temporary address if use_tempaddr >= 2
1607                  */
1608                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1609                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1610                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1611                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1612                 break;
1613             }
1614         case IPV6_SADDR_RULE_ORCHID:
1615                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1616                  *          non-ORCHID vs non-ORCHID
1617                  */
1618                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1619                         ipv6_addr_orchid(dst->addr));
1620                 break;
1621         case IPV6_SADDR_RULE_PREFIX:
1622                 /* Rule 8: Use longest matching prefix */
1623                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1624                 if (ret > score->ifa->prefix_len)
1625                         ret = score->ifa->prefix_len;
1626                 score->matchlen = ret;
1627                 break;
1628 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1629         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1630                 /* Optimistic addresses still have lower precedence than other
1631                  * preferred addresses.
1632                  */
1633                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1634                 break;
1635 #endif
1636         default:
1637                 ret = 0;
1638         }
1639
1640         if (ret)
1641                 __set_bit(i, score->scorebits);
1642         score->rule = i;
1643 out:
1644         return ret;
1645 }
1646
1647 static int __ipv6_dev_get_saddr(struct net *net,
1648                                 struct ipv6_saddr_dst *dst,
1649                                 struct inet6_dev *idev,
1650                                 struct ipv6_saddr_score *scores,
1651                                 int hiscore_idx)
1652 {
1653         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1654
1655         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1656                 int i;
1657
1658                 /*
1659                  * - Tentative Address (RFC2462 section 5.4)
1660                  *  - A tentative address is not considered
1661                  *    "assigned to an interface" in the traditional
1662                  *    sense, unless it is also flagged as optimistic.
1663                  * - Candidate Source Address (section 4)
1664                  *  - In any case, anycast addresses, multicast
1665                  *    addresses, and the unspecified address MUST
1666                  *    NOT be included in a candidate set.
1667                  */
1668                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1669                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1670                         continue;
1671
1672                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1673
1674                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1675                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1676                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1677                                             idev->dev->name);
1678                         continue;
1679                 }
1680
1681                 score->rule = -1;
1682                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1683
1684                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1685                         int minihiscore, miniscore;
1686
1687                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1688                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1689
1690                         if (minihiscore > miniscore) {
1691                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1692                                     score->scopedist > 0) {
1693                                         /*
1694                                          * special case:
1695                                          * each remaining entry
1696                                          * has too small (not enough)
1697                                          * scope, because ifa entries
1698                                          * are sorted by their scope
1699                                          * values.
1700                                          */
1701                                         goto out;
1702                                 }
1703                                 break;
1704                         } else if (minihiscore < miniscore) {
1705                                 swap(hiscore, score);
1706                                 hiscore_idx = 1 - hiscore_idx;
1707
1708                                 /* restore our iterator */
1709                                 score->ifa = hiscore->ifa;
1710
1711                                 break;
1712                         }
1713                 }
1714         }
1715 out:
1716         return hiscore_idx;
1717 }
1718
1719 static int ipv6_get_saddr_master(struct net *net,
1720                                  const struct net_device *dst_dev,
1721                                  const struct net_device *master,
1722                                  struct ipv6_saddr_dst *dst,
1723                                  struct ipv6_saddr_score *scores,
1724                                  int hiscore_idx)
1725 {
1726         struct inet6_dev *idev;
1727
1728         idev = __in6_dev_get(dst_dev);
1729         if (idev)
1730                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1731                                                    scores, hiscore_idx);
1732
1733         idev = __in6_dev_get(master);
1734         if (idev)
1735                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1736                                                    scores, hiscore_idx);
1737
1738         return hiscore_idx;
1739 }
1740
1741 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1742                        const struct in6_addr *daddr, unsigned int prefs,
1743                        struct in6_addr *saddr)
1744 {
1745         struct ipv6_saddr_score scores[2], *hiscore;
1746         struct ipv6_saddr_dst dst;
1747         struct inet6_dev *idev;
1748         struct net_device *dev;
1749         int dst_type;
1750         bool use_oif_addr = false;
1751         int hiscore_idx = 0;
1752         int ret = 0;
1753
1754         dst_type = __ipv6_addr_type(daddr);
1755         dst.addr = daddr;
1756         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1757         dst.scope = __ipv6_addr_src_scope(dst_type);
1758         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1759         dst.prefs = prefs;
1760
1761         scores[hiscore_idx].rule = -1;
1762         scores[hiscore_idx].ifa = NULL;
1763
1764         rcu_read_lock();
1765
1766         /* Candidate Source Address (section 4)
1767          *  - multicast and link-local destination address,
1768          *    the set of candidate source address MUST only
1769          *    include addresses assigned to interfaces
1770          *    belonging to the same link as the outgoing
1771          *    interface.
1772          * (- For site-local destination addresses, the
1773          *    set of candidate source addresses MUST only
1774          *    include addresses assigned to interfaces
1775          *    belonging to the same site as the outgoing
1776          *    interface.)
1777          *  - "It is RECOMMENDED that the candidate source addresses
1778          *    be the set of unicast addresses assigned to the
1779          *    interface that will be used to send to the destination
1780          *    (the 'outgoing' interface)." (RFC 6724)
1781          */
1782         if (dst_dev) {
1783                 idev = __in6_dev_get(dst_dev);
1784                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1785                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1786                     (idev && idev->cnf.use_oif_addrs_only)) {
1787                         use_oif_addr = true;
1788                 }
1789         }
1790
1791         if (use_oif_addr) {
1792                 if (idev)
1793                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1794         } else {
1795                 const struct net_device *master;
1796                 int master_idx = 0;
1797
1798                 /* if dst_dev exists and is enslaved to an L3 device, then
1799                  * prefer addresses from dst_dev and then the master over
1800                  * any other enslaved devices in the L3 domain.
1801                  */
1802                 master = l3mdev_master_dev_rcu(dst_dev);
1803                 if (master) {
1804                         master_idx = master->ifindex;
1805
1806                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1807                                                             master, &dst,
1808                                                             scores, hiscore_idx);
1809
1810                         if (scores[hiscore_idx].ifa)
1811                                 goto out;
1812                 }
1813
1814                 for_each_netdev_rcu(net, dev) {
1815                         /* only consider addresses on devices in the
1816                          * same L3 domain
1817                          */
1818                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1819                                 continue;
1820                         idev = __in6_dev_get(dev);
1821                         if (!idev)
1822                                 continue;
1823                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1824                 }
1825         }
1826
1827 out:
1828         hiscore = &scores[hiscore_idx];
1829         if (!hiscore->ifa)
1830                 ret = -EADDRNOTAVAIL;
1831         else
1832                 *saddr = hiscore->ifa->addr;
1833
1834         rcu_read_unlock();
1835         return ret;
1836 }
1837 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1838
1839 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1840                       u32 banned_flags)
1841 {
1842         struct inet6_ifaddr *ifp;
1843         int err = -EADDRNOTAVAIL;
1844
1845         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1846                 if (ifp->scope > IFA_LINK)
1847                         break;
1848                 if (ifp->scope == IFA_LINK &&
1849                     !(ifp->flags & banned_flags)) {
1850                         *addr = ifp->addr;
1851                         err = 0;
1852                         break;
1853                 }
1854         }
1855         return err;
1856 }
1857
1858 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1859                     u32 banned_flags)
1860 {
1861         struct inet6_dev *idev;
1862         int err = -EADDRNOTAVAIL;
1863
1864         rcu_read_lock();
1865         idev = __in6_dev_get(dev);
1866         if (idev) {
1867                 read_lock_bh(&idev->lock);
1868                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1869                 read_unlock_bh(&idev->lock);
1870         }
1871         rcu_read_unlock();
1872         return err;
1873 }
1874
1875 static int ipv6_count_addresses(const struct inet6_dev *idev)
1876 {
1877         const struct inet6_ifaddr *ifp;
1878         int cnt = 0;
1879
1880         rcu_read_lock();
1881         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1882                 cnt++;
1883         rcu_read_unlock();
1884         return cnt;
1885 }
1886
1887 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1888                   const struct net_device *dev, int strict)
1889 {
1890         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1891                                        strict, IFA_F_TENTATIVE);
1892 }
1893 EXPORT_SYMBOL(ipv6_chk_addr);
1894
1895 /* device argument is used to find the L3 domain of interest. If
1896  * skip_dev_check is set, then the ifp device is not checked against
1897  * the passed in dev argument. So the 2 cases for addresses checks are:
1898  *   1. does the address exist in the L3 domain that dev is part of
1899  *      (skip_dev_check = true), or
1900  *
1901  *   2. does the address exist on the specific device
1902  *      (skip_dev_check = false)
1903  */
1904 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1905                             const struct net_device *dev, bool skip_dev_check,
1906                             int strict, u32 banned_flags)
1907 {
1908         unsigned int hash = inet6_addr_hash(net, addr);
1909         const struct net_device *l3mdev;
1910         struct inet6_ifaddr *ifp;
1911         u32 ifp_flags;
1912
1913         rcu_read_lock();
1914
1915         l3mdev = l3mdev_master_dev_rcu(dev);
1916         if (skip_dev_check)
1917                 dev = NULL;
1918
1919         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1920                 if (!net_eq(dev_net(ifp->idev->dev), net))
1921                         continue;
1922
1923                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1924                         continue;
1925
1926                 /* Decouple optimistic from tentative for evaluation here.
1927                  * Ban optimistic addresses explicitly, when required.
1928                  */
1929                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1930                             ? (ifp->flags&~IFA_F_TENTATIVE)
1931                             : ifp->flags;
1932                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1933                     !(ifp_flags&banned_flags) &&
1934                     (!dev || ifp->idev->dev == dev ||
1935                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1936                         rcu_read_unlock();
1937                         return 1;
1938                 }
1939         }
1940
1941         rcu_read_unlock();
1942         return 0;
1943 }
1944 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1945
1946
1947 /* Compares an address/prefix_len with addresses on device @dev.
1948  * If one is found it returns true.
1949  */
1950 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1951         const unsigned int prefix_len, struct net_device *dev)
1952 {
1953         const struct inet6_ifaddr *ifa;
1954         const struct inet6_dev *idev;
1955         bool ret = false;
1956
1957         rcu_read_lock();
1958         idev = __in6_dev_get(dev);
1959         if (idev) {
1960                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1961                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1962                         if (ret)
1963                                 break;
1964                 }
1965         }
1966         rcu_read_unlock();
1967
1968         return ret;
1969 }
1970 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1971
1972 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1973 {
1974         const struct inet6_ifaddr *ifa;
1975         const struct inet6_dev *idev;
1976         int     onlink;
1977
1978         onlink = 0;
1979         rcu_read_lock();
1980         idev = __in6_dev_get(dev);
1981         if (idev) {
1982                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1983                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1984                                                    ifa->prefix_len);
1985                         if (onlink)
1986                                 break;
1987                 }
1988         }
1989         rcu_read_unlock();
1990         return onlink;
1991 }
1992 EXPORT_SYMBOL(ipv6_chk_prefix);
1993
1994 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1995                                      struct net_device *dev, int strict)
1996 {
1997         unsigned int hash = inet6_addr_hash(net, addr);
1998         struct inet6_ifaddr *ifp, *result = NULL;
1999
2000         rcu_read_lock();
2001         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2002                 if (!net_eq(dev_net(ifp->idev->dev), net))
2003                         continue;
2004                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2005                         if (!dev || ifp->idev->dev == dev ||
2006                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2007                                 result = ifp;
2008                                 in6_ifa_hold(ifp);
2009                                 break;
2010                         }
2011                 }
2012         }
2013         rcu_read_unlock();
2014
2015         return result;
2016 }
2017
2018 /* Gets referenced address, destroys ifaddr */
2019
2020 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2021 {
2022         if (dad_failed)
2023                 ifp->flags |= IFA_F_DADFAILED;
2024
2025         if (ifp->flags&IFA_F_TEMPORARY) {
2026                 struct inet6_ifaddr *ifpub;
2027                 spin_lock_bh(&ifp->lock);
2028                 ifpub = ifp->ifpub;
2029                 if (ifpub) {
2030                         in6_ifa_hold(ifpub);
2031                         spin_unlock_bh(&ifp->lock);
2032                         ipv6_create_tempaddr(ifpub, ifp, true);
2033                         in6_ifa_put(ifpub);
2034                 } else {
2035                         spin_unlock_bh(&ifp->lock);
2036                 }
2037                 ipv6_del_addr(ifp);
2038         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2039                 spin_lock_bh(&ifp->lock);
2040                 addrconf_del_dad_work(ifp);
2041                 ifp->flags |= IFA_F_TENTATIVE;
2042                 if (dad_failed)
2043                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2044                 spin_unlock_bh(&ifp->lock);
2045                 if (dad_failed)
2046                         ipv6_ifa_notify(0, ifp);
2047                 in6_ifa_put(ifp);
2048         } else {
2049                 ipv6_del_addr(ifp);
2050         }
2051 }
2052
2053 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2054 {
2055         int err = -ENOENT;
2056
2057         spin_lock_bh(&ifp->lock);
2058         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2059                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2060                 err = 0;
2061         }
2062         spin_unlock_bh(&ifp->lock);
2063
2064         return err;
2065 }
2066
2067 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2068 {
2069         struct inet6_dev *idev = ifp->idev;
2070         struct net *net = dev_net(ifp->idev->dev);
2071
2072         if (addrconf_dad_end(ifp)) {
2073                 in6_ifa_put(ifp);
2074                 return;
2075         }
2076
2077         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2078                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2079
2080         spin_lock_bh(&ifp->lock);
2081
2082         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2083                 struct in6_addr new_addr;
2084                 struct inet6_ifaddr *ifp2;
2085                 int retries = ifp->stable_privacy_retry + 1;
2086                 struct ifa6_config cfg = {
2087                         .pfx = &new_addr,
2088                         .plen = ifp->prefix_len,
2089                         .ifa_flags = ifp->flags,
2090                         .valid_lft = ifp->valid_lft,
2091                         .preferred_lft = ifp->prefered_lft,
2092                         .scope = ifp->scope,
2093                 };
2094
2095                 if (retries > net->ipv6.sysctl.idgen_retries) {
2096                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2097                                              ifp->idev->dev->name);
2098                         goto errdad;
2099                 }
2100
2101                 new_addr = ifp->addr;
2102                 if (ipv6_generate_stable_address(&new_addr, retries,
2103                                                  idev))
2104                         goto errdad;
2105
2106                 spin_unlock_bh(&ifp->lock);
2107
2108                 if (idev->cnf.max_addresses &&
2109                     ipv6_count_addresses(idev) >=
2110                     idev->cnf.max_addresses)
2111                         goto lock_errdad;
2112
2113                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2114                                      ifp->idev->dev->name);
2115
2116                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2117                 if (IS_ERR(ifp2))
2118                         goto lock_errdad;
2119
2120                 spin_lock_bh(&ifp2->lock);
2121                 ifp2->stable_privacy_retry = retries;
2122                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2123                 spin_unlock_bh(&ifp2->lock);
2124
2125                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2126                 in6_ifa_put(ifp2);
2127 lock_errdad:
2128                 spin_lock_bh(&ifp->lock);
2129         }
2130
2131 errdad:
2132         /* transition from _POSTDAD to _ERRDAD */
2133         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2134         spin_unlock_bh(&ifp->lock);
2135
2136         addrconf_mod_dad_work(ifp, 0);
2137         in6_ifa_put(ifp);
2138 }
2139
2140 /* Join to solicited addr multicast group.
2141  * caller must hold RTNL */
2142 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2143 {
2144         struct in6_addr maddr;
2145
2146         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2147                 return;
2148
2149         addrconf_addr_solict_mult(addr, &maddr);
2150         ipv6_dev_mc_inc(dev, &maddr);
2151 }
2152
2153 /* caller must hold RTNL */
2154 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2155 {
2156         struct in6_addr maddr;
2157
2158         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2159                 return;
2160
2161         addrconf_addr_solict_mult(addr, &maddr);
2162         __ipv6_dev_mc_dec(idev, &maddr);
2163 }
2164
2165 /* caller must hold RTNL */
2166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2167 {
2168         struct in6_addr addr;
2169
2170         if (ifp->prefix_len >= 127) /* RFC 6164 */
2171                 return;
2172         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2173         if (ipv6_addr_any(&addr))
2174                 return;
2175         __ipv6_dev_ac_inc(ifp->idev, &addr);
2176 }
2177
2178 /* caller must hold RTNL */
2179 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2180 {
2181         struct in6_addr addr;
2182
2183         if (ifp->prefix_len >= 127) /* RFC 6164 */
2184                 return;
2185         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2186         if (ipv6_addr_any(&addr))
2187                 return;
2188         __ipv6_dev_ac_dec(ifp->idev, &addr);
2189 }
2190
2191 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2192 {
2193         switch (dev->addr_len) {
2194         case ETH_ALEN:
2195                 memcpy(eui, dev->dev_addr, 3);
2196                 eui[3] = 0xFF;
2197                 eui[4] = 0xFE;
2198                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2199                 break;
2200         case EUI64_ADDR_LEN:
2201                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2202                 eui[0] ^= 2;
2203                 break;
2204         default:
2205                 return -1;
2206         }
2207
2208         return 0;
2209 }
2210
2211 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2212 {
2213         union fwnet_hwaddr *ha;
2214
2215         if (dev->addr_len != FWNET_ALEN)
2216                 return -1;
2217
2218         ha = (union fwnet_hwaddr *)dev->dev_addr;
2219
2220         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2221         eui[0] ^= 2;
2222         return 0;
2223 }
2224
2225 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2226 {
2227         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2228         if (dev->addr_len != ARCNET_ALEN)
2229                 return -1;
2230         memset(eui, 0, 7);
2231         eui[7] = *(u8 *)dev->dev_addr;
2232         return 0;
2233 }
2234
2235 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2236 {
2237         if (dev->addr_len != INFINIBAND_ALEN)
2238                 return -1;
2239         memcpy(eui, dev->dev_addr + 12, 8);
2240         eui[0] |= 2;
2241         return 0;
2242 }
2243
2244 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2245 {
2246         if (addr == 0)
2247                 return -1;
2248         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2249                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2250                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2251                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2252                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2253                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2254         eui[1] = 0;
2255         eui[2] = 0x5E;
2256         eui[3] = 0xFE;
2257         memcpy(eui + 4, &addr, 4);
2258         return 0;
2259 }
2260
2261 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2262 {
2263         if (dev->priv_flags & IFF_ISATAP)
2264                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2265         return -1;
2266 }
2267
2268 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2269 {
2270         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2271 }
2272
2273 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2274 {
2275         memcpy(eui, dev->perm_addr, 3);
2276         memcpy(eui + 5, dev->perm_addr + 3, 3);
2277         eui[3] = 0xFF;
2278         eui[4] = 0xFE;
2279         eui[0] ^= 2;
2280         return 0;
2281 }
2282
2283 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2284 {
2285         switch (dev->type) {
2286         case ARPHRD_ETHER:
2287         case ARPHRD_FDDI:
2288                 return addrconf_ifid_eui48(eui, dev);
2289         case ARPHRD_ARCNET:
2290                 return addrconf_ifid_arcnet(eui, dev);
2291         case ARPHRD_INFINIBAND:
2292                 return addrconf_ifid_infiniband(eui, dev);
2293         case ARPHRD_SIT:
2294                 return addrconf_ifid_sit(eui, dev);
2295         case ARPHRD_IPGRE:
2296         case ARPHRD_TUNNEL:
2297                 return addrconf_ifid_gre(eui, dev);
2298         case ARPHRD_6LOWPAN:
2299                 return addrconf_ifid_6lowpan(eui, dev);
2300         case ARPHRD_IEEE1394:
2301                 return addrconf_ifid_ieee1394(eui, dev);
2302         case ARPHRD_TUNNEL6:
2303         case ARPHRD_IP6GRE:
2304         case ARPHRD_RAWIP:
2305                 return addrconf_ifid_ip6tnl(eui, dev);
2306         }
2307         return -1;
2308 }
2309
2310 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2311 {
2312         int err = -1;
2313         struct inet6_ifaddr *ifp;
2314
2315         read_lock_bh(&idev->lock);
2316         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2317                 if (ifp->scope > IFA_LINK)
2318                         break;
2319                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2320                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2321                         err = 0;
2322                         break;
2323                 }
2324         }
2325         read_unlock_bh(&idev->lock);
2326         return err;
2327 }
2328
2329 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2330 static void ipv6_regen_rndid(struct inet6_dev *idev)
2331 {
2332 regen:
2333         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2334         idev->rndid[0] &= ~0x02;
2335
2336         /*
2337          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2338          * check if generated address is not inappropriate
2339          *
2340          *  - Reserved subnet anycast (RFC 2526)
2341          *      11111101 11....11 1xxxxxxx
2342          *  - ISATAP (RFC4214) 6.1
2343          *      00-00-5E-FE-xx-xx-xx-xx
2344          *  - value 0
2345          *  - XXX: already assigned to an address on the device
2346          */
2347         if (idev->rndid[0] == 0xfd &&
2348             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2349             (idev->rndid[7]&0x80))
2350                 goto regen;
2351         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2352                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2353                         goto regen;
2354                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2355                         goto regen;
2356         }
2357 }
2358
2359 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2360 {
2361         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2362                 ipv6_regen_rndid(idev);
2363 }
2364
2365 /*
2366  *      Add prefix route.
2367  */
2368
2369 static void
2370 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2371                       struct net_device *dev, unsigned long expires,
2372                       u32 flags, gfp_t gfp_flags)
2373 {
2374         struct fib6_config cfg = {
2375                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2376                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2377                 .fc_ifindex = dev->ifindex,
2378                 .fc_expires = expires,
2379                 .fc_dst_len = plen,
2380                 .fc_flags = RTF_UP | flags,
2381                 .fc_nlinfo.nl_net = dev_net(dev),
2382                 .fc_protocol = RTPROT_KERNEL,
2383                 .fc_type = RTN_UNICAST,
2384         };
2385
2386         cfg.fc_dst = *pfx;
2387
2388         /* Prevent useless cloning on PtP SIT.
2389            This thing is done here expecting that the whole
2390            class of non-broadcast devices need not cloning.
2391          */
2392 #if IS_ENABLED(CONFIG_IPV6_SIT)
2393         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2394                 cfg.fc_flags |= RTF_NONEXTHOP;
2395 #endif
2396
2397         ip6_route_add(&cfg, gfp_flags, NULL);
2398 }
2399
2400
2401 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2402                                                   int plen,
2403                                                   const struct net_device *dev,
2404                                                   u32 flags, u32 noflags,
2405                                                   bool no_gw)
2406 {
2407         struct fib6_node *fn;
2408         struct fib6_info *rt = NULL;
2409         struct fib6_table *table;
2410         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2411
2412         table = fib6_get_table(dev_net(dev), tb_id);
2413         if (!table)
2414                 return NULL;
2415
2416         rcu_read_lock();
2417         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2418         if (!fn)
2419                 goto out;
2420
2421         for_each_fib6_node_rt_rcu(fn) {
2422                 if (rt->fib6_nh.fib_nh_dev->ifindex != dev->ifindex)
2423                         continue;
2424                 if (no_gw && rt->fib6_nh.fib_nh_has_gw)
2425                         continue;
2426                 if ((rt->fib6_flags & flags) != flags)
2427                         continue;
2428                 if ((rt->fib6_flags & noflags) != 0)
2429                         continue;
2430                 if (!fib6_info_hold_safe(rt))
2431                         continue;
2432                 break;
2433         }
2434 out:
2435         rcu_read_unlock();
2436         return rt;
2437 }
2438
2439
2440 /* Create "default" multicast route to the interface */
2441
2442 static void addrconf_add_mroute(struct net_device *dev)
2443 {
2444         struct fib6_config cfg = {
2445                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2446                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2447                 .fc_ifindex = dev->ifindex,
2448                 .fc_dst_len = 8,
2449                 .fc_flags = RTF_UP,
2450                 .fc_type = RTN_UNICAST,
2451                 .fc_nlinfo.nl_net = dev_net(dev),
2452         };
2453
2454         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2455
2456         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2457 }
2458
2459 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2460 {
2461         struct inet6_dev *idev;
2462
2463         ASSERT_RTNL();
2464
2465         idev = ipv6_find_idev(dev);
2466         if (!idev)
2467                 return ERR_PTR(-ENOBUFS);
2468
2469         if (idev->cnf.disable_ipv6)
2470                 return ERR_PTR(-EACCES);
2471
2472         /* Add default multicast route */
2473         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2474                 addrconf_add_mroute(dev);
2475
2476         return idev;
2477 }
2478
2479 static void manage_tempaddrs(struct inet6_dev *idev,
2480                              struct inet6_ifaddr *ifp,
2481                              __u32 valid_lft, __u32 prefered_lft,
2482                              bool create, unsigned long now)
2483 {
2484         u32 flags;
2485         struct inet6_ifaddr *ift;
2486
2487         read_lock_bh(&idev->lock);
2488         /* update all temporary addresses in the list */
2489         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2490                 int age, max_valid, max_prefered;
2491
2492                 if (ifp != ift->ifpub)
2493                         continue;
2494
2495                 /* RFC 4941 section 3.3:
2496                  * If a received option will extend the lifetime of a public
2497                  * address, the lifetimes of temporary addresses should
2498                  * be extended, subject to the overall constraint that no
2499                  * temporary addresses should ever remain "valid" or "preferred"
2500                  * for a time longer than (TEMP_VALID_LIFETIME) or
2501                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2502                  */
2503                 age = (now - ift->cstamp) / HZ;
2504                 max_valid = idev->cnf.temp_valid_lft - age;
2505                 if (max_valid < 0)
2506                         max_valid = 0;
2507
2508                 max_prefered = idev->cnf.temp_prefered_lft -
2509                                idev->desync_factor - age;
2510                 if (max_prefered < 0)
2511                         max_prefered = 0;
2512
2513                 if (valid_lft > max_valid)
2514                         valid_lft = max_valid;
2515
2516                 if (prefered_lft > max_prefered)
2517                         prefered_lft = max_prefered;
2518
2519                 spin_lock(&ift->lock);
2520                 flags = ift->flags;
2521                 ift->valid_lft = valid_lft;
2522                 ift->prefered_lft = prefered_lft;
2523                 ift->tstamp = now;
2524                 if (prefered_lft > 0)
2525                         ift->flags &= ~IFA_F_DEPRECATED;
2526
2527                 spin_unlock(&ift->lock);
2528                 if (!(flags&IFA_F_TENTATIVE))
2529                         ipv6_ifa_notify(0, ift);
2530         }
2531
2532         if ((create || list_empty(&idev->tempaddr_list)) &&
2533             idev->cnf.use_tempaddr > 0) {
2534                 /* When a new public address is created as described
2535                  * in [ADDRCONF], also create a new temporary address.
2536                  * Also create a temporary address if it's enabled but
2537                  * no temporary address currently exists.
2538                  */
2539                 read_unlock_bh(&idev->lock);
2540                 ipv6_create_tempaddr(ifp, NULL, false);
2541         } else {
2542                 read_unlock_bh(&idev->lock);
2543         }
2544 }
2545
2546 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2547 {
2548         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2549                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2550 }
2551
2552 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2553                                  const struct prefix_info *pinfo,
2554                                  struct inet6_dev *in6_dev,
2555                                  const struct in6_addr *addr, int addr_type,
2556                                  u32 addr_flags, bool sllao, bool tokenized,
2557                                  __u32 valid_lft, u32 prefered_lft)
2558 {
2559         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2560         int create = 0, update_lft = 0;
2561
2562         if (!ifp && valid_lft) {
2563                 int max_addresses = in6_dev->cnf.max_addresses;
2564                 struct ifa6_config cfg = {
2565                         .pfx = addr,
2566                         .plen = pinfo->prefix_len,
2567                         .ifa_flags = addr_flags,
2568                         .valid_lft = valid_lft,
2569                         .preferred_lft = prefered_lft,
2570                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2571                 };
2572
2573 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2574                 if ((net->ipv6.devconf_all->optimistic_dad ||
2575                      in6_dev->cnf.optimistic_dad) &&
2576                     !net->ipv6.devconf_all->forwarding && sllao)
2577                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2578 #endif
2579
2580                 /* Do not allow to create too much of autoconfigured
2581                  * addresses; this would be too easy way to crash kernel.
2582                  */
2583                 if (!max_addresses ||
2584                     ipv6_count_addresses(in6_dev) < max_addresses)
2585                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2586
2587                 if (IS_ERR_OR_NULL(ifp))
2588                         return -1;
2589
2590                 create = 1;
2591                 spin_lock_bh(&ifp->lock);
2592                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2593                 ifp->cstamp = jiffies;
2594                 ifp->tokenized = tokenized;
2595                 spin_unlock_bh(&ifp->lock);
2596                 addrconf_dad_start(ifp);
2597         }
2598
2599         if (ifp) {
2600                 u32 flags;
2601                 unsigned long now;
2602                 u32 stored_lft;
2603
2604                 /* update lifetime (RFC2462 5.5.3 e) */
2605                 spin_lock_bh(&ifp->lock);
2606                 now = jiffies;
2607                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2608                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2609                 else
2610                         stored_lft = 0;
2611                 if (!create && stored_lft) {
2612                         const u32 minimum_lft = min_t(u32,
2613                                 stored_lft, MIN_VALID_LIFETIME);
2614                         valid_lft = max(valid_lft, minimum_lft);
2615
2616                         /* RFC4862 Section 5.5.3e:
2617                          * "Note that the preferred lifetime of the
2618                          *  corresponding address is always reset to
2619                          *  the Preferred Lifetime in the received
2620                          *  Prefix Information option, regardless of
2621                          *  whether the valid lifetime is also reset or
2622                          *  ignored."
2623                          *
2624                          * So we should always update prefered_lft here.
2625                          */
2626                         update_lft = 1;
2627                 }
2628
2629                 if (update_lft) {
2630                         ifp->valid_lft = valid_lft;
2631                         ifp->prefered_lft = prefered_lft;
2632                         ifp->tstamp = now;
2633                         flags = ifp->flags;
2634                         ifp->flags &= ~IFA_F_DEPRECATED;
2635                         spin_unlock_bh(&ifp->lock);
2636
2637                         if (!(flags&IFA_F_TENTATIVE))
2638                                 ipv6_ifa_notify(0, ifp);
2639                 } else
2640                         spin_unlock_bh(&ifp->lock);
2641
2642                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2643                                  create, now);
2644
2645                 in6_ifa_put(ifp);
2646                 addrconf_verify();
2647         }
2648
2649         return 0;
2650 }
2651 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2652
2653 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2654 {
2655         struct prefix_info *pinfo;
2656         __u32 valid_lft;
2657         __u32 prefered_lft;
2658         int addr_type, err;
2659         u32 addr_flags = 0;
2660         struct inet6_dev *in6_dev;
2661         struct net *net = dev_net(dev);
2662
2663         pinfo = (struct prefix_info *) opt;
2664
2665         if (len < sizeof(struct prefix_info)) {
2666                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2667                 return;
2668         }
2669
2670         /*
2671          *      Validation checks ([ADDRCONF], page 19)
2672          */
2673
2674         addr_type = ipv6_addr_type(&pinfo->prefix);
2675
2676         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2677                 return;
2678
2679         valid_lft = ntohl(pinfo->valid);
2680         prefered_lft = ntohl(pinfo->prefered);
2681
2682         if (prefered_lft > valid_lft) {
2683                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2684                 return;
2685         }
2686
2687         in6_dev = in6_dev_get(dev);
2688
2689         if (!in6_dev) {
2690                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2691                                     dev->name);
2692                 return;
2693         }
2694
2695         /*
2696          *      Two things going on here:
2697          *      1) Add routes for on-link prefixes
2698          *      2) Configure prefixes with the auto flag set
2699          */
2700
2701         if (pinfo->onlink) {
2702                 struct fib6_info *rt;
2703                 unsigned long rt_expires;
2704
2705                 /* Avoid arithmetic overflow. Really, we could
2706                  * save rt_expires in seconds, likely valid_lft,
2707                  * but it would require division in fib gc, that it
2708                  * not good.
2709                  */
2710                 if (HZ > USER_HZ)
2711                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2712                 else
2713                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2714
2715                 if (addrconf_finite_timeout(rt_expires))
2716                         rt_expires *= HZ;
2717
2718                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2719                                                pinfo->prefix_len,
2720                                                dev,
2721                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2722                                                RTF_DEFAULT, true);
2723
2724                 if (rt) {
2725                         /* Autoconf prefix route */
2726                         if (valid_lft == 0) {
2727                                 ip6_del_rt(net, rt);
2728                                 rt = NULL;
2729                         } else if (addrconf_finite_timeout(rt_expires)) {
2730                                 /* not infinity */
2731                                 fib6_set_expires(rt, jiffies + rt_expires);
2732                         } else {
2733                                 fib6_clean_expires(rt);
2734                         }
2735                 } else if (valid_lft) {
2736                         clock_t expires = 0;
2737                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2738                         if (addrconf_finite_timeout(rt_expires)) {
2739                                 /* not infinity */
2740                                 flags |= RTF_EXPIRES;
2741                                 expires = jiffies_to_clock_t(rt_expires);
2742                         }
2743                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2744                                               0, dev, expires, flags,
2745                                               GFP_ATOMIC);
2746                 }
2747                 fib6_info_release(rt);
2748         }
2749
2750         /* Try to figure out our local address for this prefix */
2751
2752         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2753                 struct in6_addr addr;
2754                 bool tokenized = false, dev_addr_generated = false;
2755
2756                 if (pinfo->prefix_len == 64) {
2757                         memcpy(&addr, &pinfo->prefix, 8);
2758
2759                         if (!ipv6_addr_any(&in6_dev->token)) {
2760                                 read_lock_bh(&in6_dev->lock);
2761                                 memcpy(addr.s6_addr + 8,
2762                                        in6_dev->token.s6_addr + 8, 8);
2763                                 read_unlock_bh(&in6_dev->lock);
2764                                 tokenized = true;
2765                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2766                                    !ipv6_generate_stable_address(&addr, 0,
2767                                                                  in6_dev)) {
2768                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2769                                 goto ok;
2770                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2771                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2772                                 goto put;
2773                         } else {
2774                                 dev_addr_generated = true;
2775                         }
2776                         goto ok;
2777                 }
2778                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2779                                     pinfo->prefix_len);
2780                 goto put;
2781
2782 ok:
2783                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2784                                                    &addr, addr_type,
2785                                                    addr_flags, sllao,
2786                                                    tokenized, valid_lft,
2787                                                    prefered_lft);
2788                 if (err)
2789                         goto put;
2790
2791                 /* Ignore error case here because previous prefix add addr was
2792                  * successful which will be notified.
2793                  */
2794                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2795                                               addr_type, addr_flags, sllao,
2796                                               tokenized, valid_lft,
2797                                               prefered_lft,
2798                                               dev_addr_generated);
2799         }
2800         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2801 put:
2802         in6_dev_put(in6_dev);
2803 }
2804
2805 /*
2806  *      Set destination address.
2807  *      Special case for SIT interfaces where we create a new "virtual"
2808  *      device.
2809  */
2810 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2811 {
2812         struct in6_ifreq ireq;
2813         struct net_device *dev;
2814         int err = -EINVAL;
2815
2816         rtnl_lock();
2817
2818         err = -EFAULT;
2819         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2820                 goto err_exit;
2821
2822         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2823
2824         err = -ENODEV;
2825         if (!dev)
2826                 goto err_exit;
2827
2828 #if IS_ENABLED(CONFIG_IPV6_SIT)
2829         if (dev->type == ARPHRD_SIT) {
2830                 const struct net_device_ops *ops = dev->netdev_ops;
2831                 struct ifreq ifr;
2832                 struct ip_tunnel_parm p;
2833
2834                 err = -EADDRNOTAVAIL;
2835                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2836                         goto err_exit;
2837
2838                 memset(&p, 0, sizeof(p));
2839                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2840                 p.iph.saddr = 0;
2841                 p.iph.version = 4;
2842                 p.iph.ihl = 5;
2843                 p.iph.protocol = IPPROTO_IPV6;
2844                 p.iph.ttl = 64;
2845                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2846
2847                 if (ops->ndo_do_ioctl) {
2848                         mm_segment_t oldfs = get_fs();
2849
2850                         set_fs(KERNEL_DS);
2851                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2852                         set_fs(oldfs);
2853                 } else
2854                         err = -EOPNOTSUPP;
2855
2856                 if (err == 0) {
2857                         err = -ENOBUFS;
2858                         dev = __dev_get_by_name(net, p.name);
2859                         if (!dev)
2860                                 goto err_exit;
2861                         err = dev_open(dev, NULL);
2862                 }
2863         }
2864 #endif
2865
2866 err_exit:
2867         rtnl_unlock();
2868         return err;
2869 }
2870
2871 static int ipv6_mc_config(struct sock *sk, bool join,
2872                           const struct in6_addr *addr, int ifindex)
2873 {
2874         int ret;
2875
2876         ASSERT_RTNL();
2877
2878         lock_sock(sk);
2879         if (join)
2880                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2881         else
2882                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2883         release_sock(sk);
2884
2885         return ret;
2886 }
2887
2888 /*
2889  *      Manual configuration of address on an interface
2890  */
2891 static int inet6_addr_add(struct net *net, int ifindex,
2892                           struct ifa6_config *cfg,
2893                           struct netlink_ext_ack *extack)
2894 {
2895         struct inet6_ifaddr *ifp;
2896         struct inet6_dev *idev;
2897         struct net_device *dev;
2898         unsigned long timeout;
2899         clock_t expires;
2900         u32 flags;
2901
2902         ASSERT_RTNL();
2903
2904         if (cfg->plen > 128)
2905                 return -EINVAL;
2906
2907         /* check the lifetime */
2908         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2909                 return -EINVAL;
2910
2911         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2912                 return -EINVAL;
2913
2914         dev = __dev_get_by_index(net, ifindex);
2915         if (!dev)
2916                 return -ENODEV;
2917
2918         idev = addrconf_add_dev(dev);
2919         if (IS_ERR(idev))
2920                 return PTR_ERR(idev);
2921
2922         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2923                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2924                                          true, cfg->pfx, ifindex);
2925
2926                 if (ret < 0)
2927                         return ret;
2928         }
2929
2930         cfg->scope = ipv6_addr_scope(cfg->pfx);
2931
2932         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2933         if (addrconf_finite_timeout(timeout)) {
2934                 expires = jiffies_to_clock_t(timeout * HZ);
2935                 cfg->valid_lft = timeout;
2936                 flags = RTF_EXPIRES;
2937         } else {
2938                 expires = 0;
2939                 flags = 0;
2940                 cfg->ifa_flags |= IFA_F_PERMANENT;
2941         }
2942
2943         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2944         if (addrconf_finite_timeout(timeout)) {
2945                 if (timeout == 0)
2946                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2947                 cfg->preferred_lft = timeout;
2948         }
2949
2950         ifp = ipv6_add_addr(idev, cfg, true, extack);
2951         if (!IS_ERR(ifp)) {
2952                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2953                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2954                                               ifp->rt_priority, dev, expires,
2955                                               flags, GFP_KERNEL);
2956                 }
2957
2958                 /* Send a netlink notification if DAD is enabled and
2959                  * optimistic flag is not set
2960                  */
2961                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2962                         ipv6_ifa_notify(0, ifp);
2963                 /*
2964                  * Note that section 3.1 of RFC 4429 indicates
2965                  * that the Optimistic flag should not be set for
2966                  * manually configured addresses
2967                  */
2968                 addrconf_dad_start(ifp);
2969                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2970                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2971                                          cfg->preferred_lft, true, jiffies);
2972                 in6_ifa_put(ifp);
2973                 addrconf_verify_rtnl();
2974                 return 0;
2975         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2976                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2977                                cfg->pfx, ifindex);
2978         }
2979
2980         return PTR_ERR(ifp);
2981 }
2982
2983 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2984                           const struct in6_addr *pfx, unsigned int plen)
2985 {
2986         struct inet6_ifaddr *ifp;
2987         struct inet6_dev *idev;
2988         struct net_device *dev;
2989
2990         if (plen > 128)
2991                 return -EINVAL;
2992
2993         dev = __dev_get_by_index(net, ifindex);
2994         if (!dev)
2995                 return -ENODEV;
2996
2997         idev = __in6_dev_get(dev);
2998         if (!idev)
2999                 return -ENXIO;
3000
3001         read_lock_bh(&idev->lock);
3002         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3003                 if (ifp->prefix_len == plen &&
3004                     ipv6_addr_equal(pfx, &ifp->addr)) {
3005                         in6_ifa_hold(ifp);
3006                         read_unlock_bh(&idev->lock);
3007
3008                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3009                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3010                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3011                                                  jiffies);
3012                         ipv6_del_addr(ifp);
3013                         addrconf_verify_rtnl();
3014                         if (ipv6_addr_is_multicast(pfx)) {
3015                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3016                                                false, pfx, dev->ifindex);
3017                         }
3018                         return 0;
3019                 }
3020         }
3021         read_unlock_bh(&idev->lock);
3022         return -EADDRNOTAVAIL;
3023 }
3024
3025
3026 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3027 {
3028         struct ifa6_config cfg = {
3029                 .ifa_flags = IFA_F_PERMANENT,
3030                 .preferred_lft = INFINITY_LIFE_TIME,
3031                 .valid_lft = INFINITY_LIFE_TIME,
3032         };
3033         struct in6_ifreq ireq;
3034         int err;
3035
3036         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3037                 return -EPERM;
3038
3039         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3040                 return -EFAULT;
3041
3042         cfg.pfx = &ireq.ifr6_addr;
3043         cfg.plen = ireq.ifr6_prefixlen;
3044
3045         rtnl_lock();
3046         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3047         rtnl_unlock();
3048         return err;
3049 }
3050
3051 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3052 {
3053         struct in6_ifreq ireq;
3054         int err;
3055
3056         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3057                 return -EPERM;
3058
3059         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3060                 return -EFAULT;
3061
3062         rtnl_lock();
3063         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3064                              ireq.ifr6_prefixlen);
3065         rtnl_unlock();
3066         return err;
3067 }
3068
3069 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3070                      int plen, int scope)
3071 {
3072         struct inet6_ifaddr *ifp;
3073         struct ifa6_config cfg = {
3074                 .pfx = addr,
3075                 .plen = plen,
3076                 .ifa_flags = IFA_F_PERMANENT,
3077                 .valid_lft = INFINITY_LIFE_TIME,
3078                 .preferred_lft = INFINITY_LIFE_TIME,
3079                 .scope = scope
3080         };
3081
3082         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3083         if (!IS_ERR(ifp)) {
3084                 spin_lock_bh(&ifp->lock);
3085                 ifp->flags &= ~IFA_F_TENTATIVE;
3086                 spin_unlock_bh(&ifp->lock);
3087                 rt_genid_bump_ipv6(dev_net(idev->dev));
3088                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3089                 in6_ifa_put(ifp);
3090         }
3091 }
3092
3093 #if IS_ENABLED(CONFIG_IPV6_SIT)
3094 static void sit_add_v4_addrs(struct inet6_dev *idev)
3095 {
3096         struct in6_addr addr;
3097         struct net_device *dev;
3098         struct net *net = dev_net(idev->dev);
3099         int scope, plen;
3100         u32 pflags = 0;
3101
3102         ASSERT_RTNL();
3103
3104         memset(&addr, 0, sizeof(struct in6_addr));
3105         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3106
3107         if (idev->dev->flags&IFF_POINTOPOINT) {
3108                 addr.s6_addr32[0] = htonl(0xfe800000);
3109                 scope = IFA_LINK;
3110                 plen = 64;
3111         } else {
3112                 scope = IPV6_ADDR_COMPATv4;
3113                 plen = 96;
3114                 pflags |= RTF_NONEXTHOP;
3115         }
3116
3117         if (addr.s6_addr32[3]) {
3118                 add_addr(idev, &addr, plen, scope);
3119                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3120                                       GFP_KERNEL);
3121                 return;
3122         }
3123
3124         for_each_netdev(net, dev) {
3125                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3126                 if (in_dev && (dev->flags & IFF_UP)) {
3127                         struct in_ifaddr *ifa;
3128
3129                         int flag = scope;
3130
3131                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3132
3133                                 addr.s6_addr32[3] = ifa->ifa_local;
3134
3135                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3136                                         continue;
3137                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3138                                         if (idev->dev->flags&IFF_POINTOPOINT)
3139                                                 continue;
3140                                         flag |= IFA_HOST;
3141                                 }
3142
3143                                 add_addr(idev, &addr, plen, flag);
3144                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3145                                                       0, pflags, GFP_KERNEL);
3146                         }
3147                 }
3148         }
3149 }
3150 #endif
3151
3152 static void init_loopback(struct net_device *dev)
3153 {
3154         struct inet6_dev  *idev;
3155
3156         /* ::1 */
3157
3158         ASSERT_RTNL();
3159
3160         idev = ipv6_find_idev(dev);
3161         if (!idev) {
3162                 pr_debug("%s: add_dev failed\n", __func__);
3163                 return;
3164         }
3165
3166         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3167 }
3168
3169 void addrconf_add_linklocal(struct inet6_dev *idev,
3170                             const struct in6_addr *addr, u32 flags)
3171 {
3172         struct ifa6_config cfg = {
3173                 .pfx = addr,
3174                 .plen = 64,
3175                 .ifa_flags = flags | IFA_F_PERMANENT,
3176                 .valid_lft = INFINITY_LIFE_TIME,
3177                 .preferred_lft = INFINITY_LIFE_TIME,
3178                 .scope = IFA_LINK
3179         };
3180         struct inet6_ifaddr *ifp;
3181
3182 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3183         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3184              idev->cnf.optimistic_dad) &&
3185             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3186                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3187 #endif
3188
3189         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3190         if (!IS_ERR(ifp)) {
3191                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3192                                       0, 0, GFP_ATOMIC);
3193                 addrconf_dad_start(ifp);
3194                 in6_ifa_put(ifp);
3195         }
3196 }
3197 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3198
3199 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3200 {
3201         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3202                 return true;
3203
3204         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3205             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3206                 return true;
3207
3208         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3209             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3210                 return true;
3211
3212         return false;
3213 }
3214
3215 static int ipv6_generate_stable_address(struct in6_addr *address,
3216                                         u8 dad_count,
3217                                         const struct inet6_dev *idev)
3218 {
3219         static DEFINE_SPINLOCK(lock);
3220         static __u32 digest[SHA_DIGEST_WORDS];
3221         static __u32 workspace[SHA_WORKSPACE_WORDS];
3222
3223         static union {
3224                 char __data[SHA_MESSAGE_BYTES];
3225                 struct {
3226                         struct in6_addr secret;
3227                         __be32 prefix[2];
3228                         unsigned char hwaddr[MAX_ADDR_LEN];
3229                         u8 dad_count;
3230                 } __packed;
3231         } data;
3232
3233         struct in6_addr secret;
3234         struct in6_addr temp;
3235         struct net *net = dev_net(idev->dev);
3236
3237         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3238
3239         if (idev->cnf.stable_secret.initialized)
3240                 secret = idev->cnf.stable_secret.secret;
3241         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3242                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3243         else
3244                 return -1;
3245
3246 retry:
3247         spin_lock_bh(&lock);
3248
3249         sha_init(digest);
3250         memset(&data, 0, sizeof(data));
3251         memset(workspace, 0, sizeof(workspace));
3252         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3253         data.prefix[0] = address->s6_addr32[0];
3254         data.prefix[1] = address->s6_addr32[1];
3255         data.secret = secret;
3256         data.dad_count = dad_count;
3257
3258         sha_transform(digest, data.__data, workspace);
3259
3260         temp = *address;
3261         temp.s6_addr32[2] = (__force __be32)digest[0];
3262         temp.s6_addr32[3] = (__force __be32)digest[1];
3263
3264         spin_unlock_bh(&lock);
3265
3266         if (ipv6_reserved_interfaceid(temp)) {
3267                 dad_count++;
3268                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3269                         return -1;
3270                 goto retry;
3271         }
3272
3273         *address = temp;
3274         return 0;
3275 }
3276
3277 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3278 {
3279         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3280
3281         if (s->initialized)
3282                 return;
3283         s = &idev->cnf.stable_secret;
3284         get_random_bytes(&s->secret, sizeof(s->secret));
3285         s->initialized = true;
3286 }
3287
3288 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3289 {
3290         struct in6_addr addr;
3291
3292         /* no link local addresses on L3 master devices */
3293         if (netif_is_l3_master(idev->dev))
3294                 return;
3295
3296         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3297
3298         switch (idev->cnf.addr_gen_mode) {
3299         case IN6_ADDR_GEN_MODE_RANDOM:
3300                 ipv6_gen_mode_random_init(idev);
3301                 /* fallthrough */
3302         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3303                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3304                         addrconf_add_linklocal(idev, &addr,
3305                                                IFA_F_STABLE_PRIVACY);
3306                 else if (prefix_route)
3307                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3308                                               0, 0, GFP_KERNEL);
3309                 break;
3310         case IN6_ADDR_GEN_MODE_EUI64:
3311                 /* addrconf_add_linklocal also adds a prefix_route and we
3312                  * only need to care about prefix routes if ipv6_generate_eui64
3313                  * couldn't generate one.
3314                  */
3315                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3316                         addrconf_add_linklocal(idev, &addr, 0);
3317                 else if (prefix_route)
3318                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3319                                               0, 0, GFP_KERNEL);
3320                 break;
3321         case IN6_ADDR_GEN_MODE_NONE:
3322         default:
3323                 /* will not add any link local address */
3324                 break;
3325         }
3326 }
3327
3328 static void addrconf_dev_config(struct net_device *dev)
3329 {
3330         struct inet6_dev *idev;
3331
3332         ASSERT_RTNL();
3333
3334         if ((dev->type != ARPHRD_ETHER) &&
3335             (dev->type != ARPHRD_FDDI) &&
3336             (dev->type != ARPHRD_ARCNET) &&
3337             (dev->type != ARPHRD_INFINIBAND) &&
3338             (dev->type != ARPHRD_IEEE1394) &&
3339             (dev->type != ARPHRD_TUNNEL6) &&
3340             (dev->type != ARPHRD_6LOWPAN) &&
3341             (dev->type != ARPHRD_IP6GRE) &&
3342             (dev->type != ARPHRD_IPGRE) &&
3343             (dev->type != ARPHRD_TUNNEL) &&
3344             (dev->type != ARPHRD_NONE) &&
3345             (dev->type != ARPHRD_RAWIP)) {
3346                 /* Alas, we support only Ethernet autoconfiguration. */
3347                 return;
3348         }
3349
3350         idev = addrconf_add_dev(dev);
3351         if (IS_ERR(idev))
3352                 return;
3353
3354         /* this device type has no EUI support */
3355         if (dev->type == ARPHRD_NONE &&
3356             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3357                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3358
3359         addrconf_addr_gen(idev, false);
3360 }
3361
3362 #if IS_ENABLED(CONFIG_IPV6_SIT)
3363 static void addrconf_sit_config(struct net_device *dev)
3364 {
3365         struct inet6_dev *idev;
3366
3367         ASSERT_RTNL();
3368
3369         /*
3370          * Configure the tunnel with one of our IPv4
3371          * addresses... we should configure all of
3372          * our v4 addrs in the tunnel
3373          */
3374
3375         idev = ipv6_find_idev(dev);
3376         if (!idev) {
3377                 pr_debug("%s: add_dev failed\n", __func__);
3378                 return;
3379         }
3380
3381         if (dev->priv_flags & IFF_ISATAP) {
3382                 addrconf_addr_gen(idev, false);
3383                 return;
3384         }
3385
3386         sit_add_v4_addrs(idev);
3387
3388         if (dev->flags&IFF_POINTOPOINT)
3389                 addrconf_add_mroute(dev);
3390 }
3391 #endif
3392
3393 #if IS_ENABLED(CONFIG_NET_IPGRE)
3394 static void addrconf_gre_config(struct net_device *dev)
3395 {
3396         struct inet6_dev *idev;
3397
3398         ASSERT_RTNL();
3399
3400         idev = ipv6_find_idev(dev);
3401         if (!idev) {
3402                 pr_debug("%s: add_dev failed\n", __func__);
3403                 return;
3404         }
3405
3406         addrconf_addr_gen(idev, true);
3407         if (dev->flags & IFF_POINTOPOINT)
3408                 addrconf_add_mroute(dev);
3409 }
3410 #endif
3411
3412 static int fixup_permanent_addr(struct net *net,
3413                                 struct inet6_dev *idev,
3414                                 struct inet6_ifaddr *ifp)
3415 {
3416         /* !fib6_node means the host route was removed from the
3417          * FIB, for example, if 'lo' device is taken down. In that
3418          * case regenerate the host route.
3419          */
3420         if (!ifp->rt || !ifp->rt->fib6_node) {
3421                 struct fib6_info *f6i, *prev;
3422
3423                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3424                                          GFP_ATOMIC);
3425                 if (IS_ERR(f6i))
3426                         return PTR_ERR(f6i);
3427
3428                 /* ifp->rt can be accessed outside of rtnl */
3429                 spin_lock(&ifp->lock);
3430                 prev = ifp->rt;
3431                 ifp->rt = f6i;
3432                 spin_unlock(&ifp->lock);
3433
3434                 fib6_info_release(prev);
3435         }
3436
3437         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3438                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3439                                       ifp->rt_priority, idev->dev, 0, 0,
3440                                       GFP_ATOMIC);
3441         }
3442
3443         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3444                 addrconf_dad_start(ifp);
3445
3446         return 0;
3447 }
3448
3449 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3450 {
3451         struct inet6_ifaddr *ifp, *tmp;
3452         struct inet6_dev *idev;
3453
3454         idev = __in6_dev_get(dev);
3455         if (!idev)
3456                 return;
3457
3458         write_lock_bh(&idev->lock);
3459
3460         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3461                 if ((ifp->flags & IFA_F_PERMANENT) &&
3462                     fixup_permanent_addr(net, idev, ifp) < 0) {
3463                         write_unlock_bh(&idev->lock);
3464                         in6_ifa_hold(ifp);
3465                         ipv6_del_addr(ifp);
3466                         write_lock_bh(&idev->lock);
3467
3468                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3469                                              idev->dev->name, &ifp->addr);
3470                 }
3471         }
3472
3473         write_unlock_bh(&idev->lock);
3474 }
3475
3476 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3477                            void *ptr)
3478 {
3479         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3480         struct netdev_notifier_change_info *change_info;
3481         struct netdev_notifier_changeupper_info *info;
3482         struct inet6_dev *idev = __in6_dev_get(dev);
3483         struct net *net = dev_net(dev);
3484         int run_pending = 0;
3485         int err;
3486
3487         switch (event) {
3488         case NETDEV_REGISTER:
3489                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3490                         idev = ipv6_add_dev(dev);
3491                         if (IS_ERR(idev))
3492                                 return notifier_from_errno(PTR_ERR(idev));
3493                 }
3494                 break;
3495
3496         case NETDEV_CHANGEMTU:
3497                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3498                 if (dev->mtu < IPV6_MIN_MTU) {
3499                         addrconf_ifdown(dev, dev != net->loopback_dev);
3500                         break;
3501                 }
3502
3503                 if (idev) {
3504                         rt6_mtu_change(dev, dev->mtu);
3505                         idev->cnf.mtu6 = dev->mtu;
3506                         break;
3507                 }
3508
3509                 /* allocate new idev */
3510                 idev = ipv6_add_dev(dev);
3511                 if (IS_ERR(idev))
3512                         break;
3513
3514                 /* device is still not ready */
3515                 if (!(idev->if_flags & IF_READY))
3516                         break;
3517
3518                 run_pending = 1;
3519
3520                 /* fall through */
3521
3522         case NETDEV_UP:
3523         case NETDEV_CHANGE:
3524                 if (dev->flags & IFF_SLAVE)
3525                         break;
3526
3527                 if (idev && idev->cnf.disable_ipv6)
3528                         break;
3529
3530                 if (event == NETDEV_UP) {
3531                         /* restore routes for permanent addresses */
3532                         addrconf_permanent_addr(net, dev);
3533
3534                         if (!addrconf_link_ready(dev)) {
3535                                 /* device is not ready yet. */
3536                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3537                                          dev->name);
3538                                 break;
3539                         }
3540
3541                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3542                                 idev = ipv6_add_dev(dev);
3543
3544                         if (!IS_ERR_OR_NULL(idev)) {
3545                                 idev->if_flags |= IF_READY;
3546                                 run_pending = 1;
3547                         }
3548                 } else if (event == NETDEV_CHANGE) {
3549                         if (!addrconf_link_ready(dev)) {
3550                                 /* device is still not ready. */
3551                                 rt6_sync_down_dev(dev, event);
3552                                 break;
3553                         }
3554
3555                         if (!IS_ERR_OR_NULL(idev)) {
3556                                 if (idev->if_flags & IF_READY) {
3557                                         /* device is already configured -
3558                                          * but resend MLD reports, we might
3559                                          * have roamed and need to update
3560                                          * multicast snooping switches
3561                                          */
3562                                         ipv6_mc_up(idev);
3563                                         change_info = ptr;
3564                                         if (change_info->flags_changed & IFF_NOARP)
3565                                                 addrconf_dad_run(idev, true);
3566                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3567                                         break;
3568                                 }
3569                                 idev->if_flags |= IF_READY;
3570                         }
3571
3572                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3573                                 dev->name);
3574
3575                         run_pending = 1;
3576                 }
3577
3578                 switch (dev->type) {
3579 #if IS_ENABLED(CONFIG_IPV6_SIT)
3580                 case ARPHRD_SIT:
3581                         addrconf_sit_config(dev);
3582                         break;
3583 #endif
3584 #if IS_ENABLED(CONFIG_NET_IPGRE)
3585                 case ARPHRD_IPGRE:
3586                         addrconf_gre_config(dev);
3587                         break;
3588 #endif
3589                 case ARPHRD_LOOPBACK:
3590                         init_loopback(dev);
3591                         break;
3592
3593                 default:
3594                         addrconf_dev_config(dev);
3595                         break;
3596                 }
3597
3598                 if (!IS_ERR_OR_NULL(idev)) {
3599                         if (run_pending)
3600                                 addrconf_dad_run(idev, false);
3601
3602                         /* Device has an address by now */
3603                         rt6_sync_up(dev, RTNH_F_DEAD);
3604
3605                         /*
3606                          * If the MTU changed during the interface down,
3607                          * when the interface up, the changed MTU must be
3608                          * reflected in the idev as well as routers.
3609                          */
3610                         if (idev->cnf.mtu6 != dev->mtu &&
3611                             dev->mtu >= IPV6_MIN_MTU) {
3612                                 rt6_mtu_change(dev, dev->mtu);
3613                                 idev->cnf.mtu6 = dev->mtu;
3614                         }
3615                         idev->tstamp = jiffies;
3616                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3617
3618                         /*
3619                          * If the changed mtu during down is lower than
3620                          * IPV6_MIN_MTU stop IPv6 on this interface.
3621                          */
3622                         if (dev->mtu < IPV6_MIN_MTU)
3623                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3624                 }
3625                 break;
3626
3627         case NETDEV_DOWN:
3628         case NETDEV_UNREGISTER:
3629                 /*
3630                  *      Remove all addresses from this interface.
3631                  */
3632                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3633                 break;
3634
3635         case NETDEV_CHANGENAME:
3636                 if (idev) {
3637                         snmp6_unregister_dev(idev);
3638                         addrconf_sysctl_unregister(idev);
3639                         err = addrconf_sysctl_register(idev);
3640                         if (err)
3641                                 return notifier_from_errno(err);
3642                         err = snmp6_register_dev(idev);
3643                         if (err) {
3644                                 addrconf_sysctl_unregister(idev);
3645                                 return notifier_from_errno(err);
3646                         }
3647                 }
3648                 break;
3649
3650         case NETDEV_PRE_TYPE_CHANGE:
3651         case NETDEV_POST_TYPE_CHANGE:
3652                 if (idev)
3653                         addrconf_type_change(dev, event);
3654                 break;
3655
3656         case NETDEV_CHANGEUPPER:
3657                 info = ptr;
3658
3659                 /* flush all routes if dev is linked to or unlinked from
3660                  * an L3 master device (e.g., VRF)
3661                  */
3662                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3663                         addrconf_ifdown(dev, 0);
3664         }
3665
3666         return NOTIFY_OK;
3667 }
3668
3669 /*
3670  *      addrconf module should be notified of a device going up
3671  */
3672 static struct notifier_block ipv6_dev_notf = {
3673         .notifier_call = addrconf_notify,
3674         .priority = ADDRCONF_NOTIFY_PRIORITY,
3675 };
3676
3677 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3678 {
3679         struct inet6_dev *idev;
3680         ASSERT_RTNL();
3681
3682         idev = __in6_dev_get(dev);
3683
3684         if (event == NETDEV_POST_TYPE_CHANGE)
3685                 ipv6_mc_remap(idev);
3686         else if (event == NETDEV_PRE_TYPE_CHANGE)
3687                 ipv6_mc_unmap(idev);
3688 }
3689
3690 static bool addr_is_local(const struct in6_addr *addr)
3691 {
3692         return ipv6_addr_type(addr) &
3693                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3694 }
3695
3696 static int addrconf_ifdown(struct net_device *dev, int how)
3697 {
3698         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3699         struct net *net = dev_net(dev);
3700         struct inet6_dev *idev;
3701         struct inet6_ifaddr *ifa, *tmp;
3702         bool keep_addr = false;
3703         int state, i;
3704
3705         ASSERT_RTNL();
3706
3707         rt6_disable_ip(dev, event);
3708
3709         idev = __in6_dev_get(dev);
3710         if (!idev)
3711                 return -ENODEV;
3712
3713         /*
3714          * Step 1: remove reference to ipv6 device from parent device.
3715          *         Do not dev_put!
3716          */
3717         if (how) {
3718                 idev->dead = 1;
3719
3720                 /* protected by rtnl_lock */
3721                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3722
3723                 /* Step 1.5: remove snmp6 entry */
3724                 snmp6_unregister_dev(idev);
3725
3726         }
3727
3728         /* combine the user config with event to determine if permanent
3729          * addresses are to be removed from address hash table
3730          */
3731         if (!how && !idev->cnf.disable_ipv6) {
3732                 /* aggregate the system setting and interface setting */
3733                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3734
3735                 if (!_keep_addr)
3736                         _keep_addr = idev->cnf.keep_addr_on_down;
3737
3738                 keep_addr = (_keep_addr > 0);
3739         }
3740
3741         /* Step 2: clear hash table */
3742         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3743                 struct hlist_head *h = &inet6_addr_lst[i];
3744
3745                 spin_lock_bh(&addrconf_hash_lock);
3746 restart:
3747                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3748                         if (ifa->idev == idev) {
3749                                 addrconf_del_dad_work(ifa);
3750                                 /* combined flag + permanent flag decide if
3751                                  * address is retained on a down event
3752                                  */
3753                                 if (!keep_addr ||
3754                                     !(ifa->flags & IFA_F_PERMANENT) ||
3755                                     addr_is_local(&ifa->addr)) {
3756                                         hlist_del_init_rcu(&ifa->addr_lst);
3757                                         goto restart;
3758                                 }
3759                         }
3760                 }
3761                 spin_unlock_bh(&addrconf_hash_lock);
3762         }
3763
3764         write_lock_bh(&idev->lock);
3765
3766         addrconf_del_rs_timer(idev);
3767
3768         /* Step 2: clear flags for stateless addrconf */
3769         if (!how)
3770                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3771
3772         /* Step 3: clear tempaddr list */
3773         while (!list_empty(&idev->tempaddr_list)) {
3774                 ifa = list_first_entry(&idev->tempaddr_list,
3775                                        struct inet6_ifaddr, tmp_list);
3776                 list_del(&ifa->tmp_list);
3777                 write_unlock_bh(&idev->lock);
3778                 spin_lock_bh(&ifa->lock);
3779
3780                 if (ifa->ifpub) {
3781                         in6_ifa_put(ifa->ifpub);
3782                         ifa->ifpub = NULL;
3783                 }
3784                 spin_unlock_bh(&ifa->lock);
3785                 in6_ifa_put(ifa);
3786                 write_lock_bh(&idev->lock);
3787         }
3788
3789         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3790                 struct fib6_info *rt = NULL;
3791                 bool keep;
3792
3793                 addrconf_del_dad_work(ifa);
3794
3795                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3796                         !addr_is_local(&ifa->addr);
3797
3798                 write_unlock_bh(&idev->lock);
3799                 spin_lock_bh(&ifa->lock);
3800
3801                 if (keep) {
3802                         /* set state to skip the notifier below */
3803                         state = INET6_IFADDR_STATE_DEAD;
3804                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3805                         if (!(ifa->flags & IFA_F_NODAD))
3806                                 ifa->flags |= IFA_F_TENTATIVE;
3807
3808                         rt = ifa->rt;
3809                         ifa->rt = NULL;
3810                 } else {
3811                         state = ifa->state;
3812                         ifa->state = INET6_IFADDR_STATE_DEAD;
3813                 }
3814
3815                 spin_unlock_bh(&ifa->lock);
3816
3817                 if (rt)
3818                         ip6_del_rt(net, rt);
3819
3820                 if (state != INET6_IFADDR_STATE_DEAD) {
3821                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3822                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3823                 } else {
3824                         if (idev->cnf.forwarding)
3825                                 addrconf_leave_anycast(ifa);
3826                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3827                 }
3828
3829                 write_lock_bh(&idev->lock);
3830                 if (!keep) {
3831                         list_del_rcu(&ifa->if_list);
3832                         in6_ifa_put(ifa);
3833                 }
3834         }
3835
3836         write_unlock_bh(&idev->lock);
3837
3838         /* Step 5: Discard anycast and multicast list */
3839         if (how) {
3840                 ipv6_ac_destroy_dev(idev);
3841                 ipv6_mc_destroy_dev(idev);
3842         } else {
3843                 ipv6_mc_down(idev);
3844         }
3845
3846         idev->tstamp = jiffies;
3847
3848         /* Last: Shot the device (if unregistered) */
3849         if (how) {
3850                 addrconf_sysctl_unregister(idev);
3851                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3852                 neigh_ifdown(&nd_tbl, dev);
3853                 in6_dev_put(idev);
3854         }
3855         return 0;
3856 }
3857
3858 static void addrconf_rs_timer(struct timer_list *t)
3859 {
3860         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3861         struct net_device *dev = idev->dev;
3862         struct in6_addr lladdr;
3863
3864         write_lock(&idev->lock);
3865         if (idev->dead || !(idev->if_flags & IF_READY))
3866                 goto out;
3867
3868         if (!ipv6_accept_ra(idev))
3869                 goto out;
3870
3871         /* Announcement received after solicitation was sent */
3872         if (idev->if_flags & IF_RA_RCVD)
3873                 goto out;
3874
3875         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3876                 write_unlock(&idev->lock);
3877                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3878                         ndisc_send_rs(dev, &lladdr,
3879                                       &in6addr_linklocal_allrouters);
3880                 else
3881                         goto put;
3882
3883                 write_lock(&idev->lock);
3884                 idev->rs_interval = rfc3315_s14_backoff_update(
3885                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3886                 /* The wait after the last probe can be shorter */
3887                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3888                                              idev->cnf.rtr_solicits) ?
3889                                       idev->cnf.rtr_solicit_delay :
3890                                       idev->rs_interval);
3891         } else {
3892                 /*
3893                  * Note: we do not support deprecated "all on-link"
3894                  * assumption any longer.
3895                  */
3896                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3897         }
3898
3899 out:
3900         write_unlock(&idev->lock);
3901 put:
3902         in6_dev_put(idev);
3903 }
3904
3905 /*
3906  *      Duplicate Address Detection
3907  */
3908 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3909 {
3910         unsigned long rand_num;
3911         struct inet6_dev *idev = ifp->idev;
3912         u64 nonce;
3913
3914         if (ifp->flags & IFA_F_OPTIMISTIC)
3915                 rand_num = 0;
3916         else
3917                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3918
3919         nonce = 0;
3920         if (idev->cnf.enhanced_dad ||
3921             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3922                 do
3923                         get_random_bytes(&nonce, 6);
3924                 while (nonce == 0);
3925         }
3926         ifp->dad_nonce = nonce;
3927         ifp->dad_probes = idev->cnf.dad_transmits;
3928         addrconf_mod_dad_work(ifp, rand_num);
3929 }
3930
3931 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3932 {
3933         struct inet6_dev *idev = ifp->idev;
3934         struct net_device *dev = idev->dev;
3935         bool bump_id, notify = false;
3936         struct net *net;
3937
3938         addrconf_join_solict(dev, &ifp->addr);
3939
3940         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3941
3942         read_lock_bh(&idev->lock);
3943         spin_lock(&ifp->lock);
3944         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3945                 goto out;
3946
3947         net = dev_net(dev);
3948         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3949             (net->ipv6.devconf_all->accept_dad < 1 &&
3950              idev->cnf.accept_dad < 1) ||
3951             !(ifp->flags&IFA_F_TENTATIVE) ||
3952             ifp->flags & IFA_F_NODAD) {
3953                 bool send_na = false;
3954
3955                 if (ifp->flags & IFA_F_TENTATIVE &&
3956                     !(ifp->flags & IFA_F_OPTIMISTIC))
3957                         send_na = true;
3958                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3959                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3960                 spin_unlock(&ifp->lock);
3961                 read_unlock_bh(&idev->lock);
3962
3963                 addrconf_dad_completed(ifp, bump_id, send_na);
3964                 return;
3965         }
3966
3967         if (!(idev->if_flags & IF_READY)) {
3968                 spin_unlock(&ifp->lock);
3969                 read_unlock_bh(&idev->lock);
3970                 /*
3971                  * If the device is not ready:
3972                  * - keep it tentative if it is a permanent address.
3973                  * - otherwise, kill it.
3974                  */
3975                 in6_ifa_hold(ifp);
3976                 addrconf_dad_stop(ifp, 0);
3977                 return;
3978         }
3979
3980         /*
3981          * Optimistic nodes can start receiving
3982          * Frames right away
3983          */
3984         if (ifp->flags & IFA_F_OPTIMISTIC) {
3985                 ip6_ins_rt(net, ifp->rt);
3986                 if (ipv6_use_optimistic_addr(net, idev)) {
3987                         /* Because optimistic nodes can use this address,
3988                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3989                          */
3990                         notify = true;
3991                 }
3992         }
3993
3994         addrconf_dad_kick(ifp);
3995 out:
3996         spin_unlock(&ifp->lock);
3997         read_unlock_bh(&idev->lock);
3998         if (notify)
3999                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4000 }
4001
4002 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4003 {
4004         bool begin_dad = false;
4005
4006         spin_lock_bh(&ifp->lock);
4007         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4008                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4009                 begin_dad = true;
4010         }
4011         spin_unlock_bh(&ifp->lock);
4012
4013         if (begin_dad)
4014                 addrconf_mod_dad_work(ifp, 0);
4015 }
4016
4017 static void addrconf_dad_work(struct work_struct *w)
4018 {
4019         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4020                                                 struct inet6_ifaddr,
4021                                                 dad_work);
4022         struct inet6_dev *idev = ifp->idev;
4023         bool bump_id, disable_ipv6 = false;
4024         struct in6_addr mcaddr;
4025
4026         enum {
4027                 DAD_PROCESS,
4028                 DAD_BEGIN,
4029                 DAD_ABORT,
4030         } action = DAD_PROCESS;
4031
4032         rtnl_lock();
4033
4034         spin_lock_bh(&ifp->lock);
4035         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4036                 action = DAD_BEGIN;
4037                 ifp->state = INET6_IFADDR_STATE_DAD;
4038         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4039                 action = DAD_ABORT;
4040                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4041
4042                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4043                      idev->cnf.accept_dad > 1) &&
4044                     !idev->cnf.disable_ipv6 &&
4045                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4046                         struct in6_addr addr;
4047
4048                         addr.s6_addr32[0] = htonl(0xfe800000);
4049                         addr.s6_addr32[1] = 0;
4050
4051                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4052                             ipv6_addr_equal(&ifp->addr, &addr)) {
4053                                 /* DAD failed for link-local based on MAC */
4054                                 idev->cnf.disable_ipv6 = 1;
4055
4056                                 pr_info("%s: IPv6 being disabled!\n",
4057                                         ifp->idev->dev->name);
4058                                 disable_ipv6 = true;
4059                         }
4060                 }
4061         }
4062         spin_unlock_bh(&ifp->lock);
4063
4064         if (action == DAD_BEGIN) {
4065                 addrconf_dad_begin(ifp);
4066                 goto out;
4067         } else if (action == DAD_ABORT) {
4068                 in6_ifa_hold(ifp);
4069                 addrconf_dad_stop(ifp, 1);
4070                 if (disable_ipv6)
4071                         addrconf_ifdown(idev->dev, 0);
4072                 goto out;
4073         }
4074
4075         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4076                 goto out;
4077
4078         write_lock_bh(&idev->lock);
4079         if (idev->dead || !(idev->if_flags & IF_READY)) {
4080                 write_unlock_bh(&idev->lock);
4081                 goto out;
4082         }
4083
4084         spin_lock(&ifp->lock);
4085         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4086                 spin_unlock(&ifp->lock);
4087                 write_unlock_bh(&idev->lock);
4088                 goto out;
4089         }
4090
4091         if (ifp->dad_probes == 0) {
4092                 bool send_na = false;
4093
4094                 /*
4095                  * DAD was successful
4096                  */
4097
4098                 if (ifp->flags & IFA_F_TENTATIVE &&
4099                     !(ifp->flags & IFA_F_OPTIMISTIC))
4100                         send_na = true;
4101                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4102                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4103                 spin_unlock(&ifp->lock);
4104                 write_unlock_bh(&idev->lock);
4105
4106                 addrconf_dad_completed(ifp, bump_id, send_na);
4107
4108                 goto out;
4109         }
4110
4111         ifp->dad_probes--;
4112         addrconf_mod_dad_work(ifp,
4113                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4114         spin_unlock(&ifp->lock);
4115         write_unlock_bh(&idev->lock);
4116
4117         /* send a neighbour solicitation for our addr */
4118         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4119         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4120                       ifp->dad_nonce);
4121 out:
4122         in6_ifa_put(ifp);
4123         rtnl_unlock();
4124 }
4125
4126 /* ifp->idev must be at least read locked */
4127 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4128 {
4129         struct inet6_ifaddr *ifpiter;
4130         struct inet6_dev *idev = ifp->idev;
4131
4132         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4133                 if (ifpiter->scope > IFA_LINK)
4134                         break;
4135                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4136                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4137                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4138                     IFA_F_PERMANENT)
4139                         return false;
4140         }
4141         return true;
4142 }
4143
4144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4145                                    bool send_na)
4146 {
4147         struct net_device *dev = ifp->idev->dev;
4148         struct in6_addr lladdr;
4149         bool send_rs, send_mld;
4150
4151         addrconf_del_dad_work(ifp);
4152
4153         /*
4154          *      Configure the address for reception. Now it is valid.
4155          */
4156
4157         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4158
4159         /* If added prefix is link local and we are prepared to process
4160            router advertisements, start sending router solicitations.
4161          */
4162
4163         read_lock_bh(&ifp->idev->lock);
4164         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4165         send_rs = send_mld &&
4166                   ipv6_accept_ra(ifp->idev) &&
4167                   ifp->idev->cnf.rtr_solicits != 0 &&
4168                   (dev->flags&IFF_LOOPBACK) == 0;
4169         read_unlock_bh(&ifp->idev->lock);
4170
4171         /* While dad is in progress mld report's source address is in6_addrany.
4172          * Resend with proper ll now.
4173          */
4174         if (send_mld)
4175                 ipv6_mc_dad_complete(ifp->idev);
4176
4177         /* send unsolicited NA if enabled */
4178         if (send_na &&
4179             (ifp->idev->cnf.ndisc_notify ||
4180              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4181                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4182                               /*router=*/ !!ifp->idev->cnf.forwarding,
4183                               /*solicited=*/ false, /*override=*/ true,
4184                               /*inc_opt=*/ true);
4185         }
4186
4187         if (send_rs) {
4188                 /*
4189                  *      If a host as already performed a random delay
4190                  *      [...] as part of DAD [...] there is no need
4191                  *      to delay again before sending the first RS
4192                  */
4193                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4194                         return;
4195                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4196
4197                 write_lock_bh(&ifp->idev->lock);
4198                 spin_lock(&ifp->lock);
4199                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4200                         ifp->idev->cnf.rtr_solicit_interval);
4201                 ifp->idev->rs_probes = 1;
4202                 ifp->idev->if_flags |= IF_RS_SENT;
4203                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4204                 spin_unlock(&ifp->lock);
4205                 write_unlock_bh(&ifp->idev->lock);
4206         }
4207
4208         if (bump_id)
4209                 rt_genid_bump_ipv6(dev_net(dev));
4210
4211         /* Make sure that a new temporary address will be created
4212          * before this temporary address becomes deprecated.
4213          */
4214         if (ifp->flags & IFA_F_TEMPORARY)
4215                 addrconf_verify_rtnl();
4216 }
4217
4218 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4219 {
4220         struct inet6_ifaddr *ifp;
4221
4222         read_lock_bh(&idev->lock);
4223         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4224                 spin_lock(&ifp->lock);
4225                 if ((ifp->flags & IFA_F_TENTATIVE &&
4226                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4227                         if (restart)
4228                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4229                         addrconf_dad_kick(ifp);
4230                 }
4231                 spin_unlock(&ifp->lock);
4232         }
4233         read_unlock_bh(&idev->lock);
4234 }
4235
4236 #ifdef CONFIG_PROC_FS
4237 struct if6_iter_state {
4238         struct seq_net_private p;
4239         int bucket;
4240         int offset;
4241 };
4242
4243 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4244 {
4245         struct if6_iter_state *state = seq->private;
4246         struct net *net = seq_file_net(seq);
4247         struct inet6_ifaddr *ifa = NULL;
4248         int p = 0;
4249
4250         /* initial bucket if pos is 0 */
4251         if (pos == 0) {
4252                 state->bucket = 0;
4253                 state->offset = 0;
4254         }
4255
4256         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4257                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4258                                          addr_lst) {
4259                         if (!net_eq(dev_net(ifa->idev->dev), net))
4260                                 continue;
4261                         /* sync with offset */
4262                         if (p < state->offset) {
4263                                 p++;
4264                                 continue;
4265                         }
4266                         return ifa;
4267                 }
4268
4269                 /* prepare for next bucket */
4270                 state->offset = 0;
4271                 p = 0;
4272         }
4273         return NULL;
4274 }
4275
4276 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4277                                          struct inet6_ifaddr *ifa)
4278 {
4279         struct if6_iter_state *state = seq->private;
4280         struct net *net = seq_file_net(seq);
4281
4282         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4283                 if (!net_eq(dev_net(ifa->idev->dev), net))
4284                         continue;
4285                 state->offset++;
4286                 return ifa;
4287         }
4288
4289         state->offset = 0;
4290         while (++state->bucket < IN6_ADDR_HSIZE) {
4291                 hlist_for_each_entry_rcu(ifa,
4292                                      &inet6_addr_lst[state->bucket], addr_lst) {
4293                         if (!net_eq(dev_net(ifa->idev->dev), net))
4294                                 continue;
4295                         return ifa;
4296                 }
4297         }
4298
4299         return NULL;
4300 }
4301
4302 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4303         __acquires(rcu)
4304 {
4305         rcu_read_lock();
4306         return if6_get_first(seq, *pos);
4307 }
4308
4309 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4310 {
4311         struct inet6_ifaddr *ifa;
4312
4313         ifa = if6_get_next(seq, v);
4314         ++*pos;
4315         return ifa;
4316 }
4317
4318 static void if6_seq_stop(struct seq_file *seq, void *v)
4319         __releases(rcu)
4320 {
4321         rcu_read_unlock();
4322 }
4323
4324 static int if6_seq_show(struct seq_file *seq, void *v)
4325 {
4326         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4327         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4328                    &ifp->addr,
4329                    ifp->idev->dev->ifindex,
4330                    ifp->prefix_len,
4331                    ifp->scope,
4332                    (u8) ifp->flags,
4333                    ifp->idev->dev->name);
4334         return 0;
4335 }
4336
4337 static const struct seq_operations if6_seq_ops = {
4338         .start  = if6_seq_start,
4339         .next   = if6_seq_next,
4340         .show   = if6_seq_show,
4341         .stop   = if6_seq_stop,
4342 };
4343
4344 static int __net_init if6_proc_net_init(struct net *net)
4345 {
4346         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4347                         sizeof(struct if6_iter_state)))
4348                 return -ENOMEM;
4349         return 0;
4350 }
4351
4352 static void __net_exit if6_proc_net_exit(struct net *net)
4353 {
4354         remove_proc_entry("if_inet6", net->proc_net);
4355 }
4356
4357 static struct pernet_operations if6_proc_net_ops = {
4358         .init = if6_proc_net_init,
4359         .exit = if6_proc_net_exit,
4360 };
4361
4362 int __init if6_proc_init(void)
4363 {
4364         return register_pernet_subsys(&if6_proc_net_ops);
4365 }
4366
4367 void if6_proc_exit(void)
4368 {
4369         unregister_pernet_subsys(&if6_proc_net_ops);
4370 }
4371 #endif  /* CONFIG_PROC_FS */
4372
4373 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4374 /* Check if address is a home address configured on any interface. */
4375 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4376 {
4377         unsigned int hash = inet6_addr_hash(net, addr);
4378         struct inet6_ifaddr *ifp = NULL;
4379         int ret = 0;
4380
4381         rcu_read_lock();
4382         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4383                 if (!net_eq(dev_net(ifp->idev->dev), net))
4384                         continue;
4385                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4386                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4387                         ret = 1;
4388                         break;
4389                 }
4390         }
4391         rcu_read_unlock();
4392         return ret;
4393 }
4394 #endif
4395
4396 /*
4397  *      Periodic address status verification
4398  */
4399
4400 static void addrconf_verify_rtnl(void)
4401 {
4402         unsigned long now, next, next_sec, next_sched;
4403         struct inet6_ifaddr *ifp;
4404         int i;
4405
4406         ASSERT_RTNL();
4407
4408         rcu_read_lock_bh();
4409         now = jiffies;
4410         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4411
4412         cancel_delayed_work(&addr_chk_work);
4413
4414         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4415 restart:
4416                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4417                         unsigned long age;
4418
4419                         /* When setting preferred_lft to a value not zero or
4420                          * infinity, while valid_lft is infinity
4421                          * IFA_F_PERMANENT has a non-infinity life time.
4422                          */
4423                         if ((ifp->flags & IFA_F_PERMANENT) &&
4424                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4425                                 continue;
4426
4427                         spin_lock(&ifp->lock);
4428                         /* We try to batch several events at once. */
4429                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4430
4431                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4432                             age >= ifp->valid_lft) {
4433                                 spin_unlock(&ifp->lock);
4434                                 in6_ifa_hold(ifp);
4435                                 ipv6_del_addr(ifp);
4436                                 goto restart;
4437                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4438                                 spin_unlock(&ifp->lock);
4439                                 continue;
4440                         } else if (age >= ifp->prefered_lft) {
4441                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4442                                 int deprecate = 0;
4443
4444                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4445                                         deprecate = 1;
4446                                         ifp->flags |= IFA_F_DEPRECATED;
4447                                 }
4448
4449                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4450                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4451                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4452
4453                                 spin_unlock(&ifp->lock);
4454
4455                                 if (deprecate) {
4456                                         in6_ifa_hold(ifp);
4457
4458                                         ipv6_ifa_notify(0, ifp);
4459                                         in6_ifa_put(ifp);
4460                                         goto restart;
4461                                 }
4462                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4463                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4464                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4465                                         ifp->idev->cnf.dad_transmits *
4466                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4467
4468                                 if (age >= ifp->prefered_lft - regen_advance) {
4469                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4470                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4471                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4472                                         if (!ifp->regen_count && ifpub) {
4473                                                 ifp->regen_count++;
4474                                                 in6_ifa_hold(ifp);
4475                                                 in6_ifa_hold(ifpub);
4476                                                 spin_unlock(&ifp->lock);
4477
4478                                                 spin_lock(&ifpub->lock);
4479                                                 ifpub->regen_count = 0;
4480                                                 spin_unlock(&ifpub->lock);
4481                                                 rcu_read_unlock_bh();
4482                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4483                                                 in6_ifa_put(ifpub);
4484                                                 in6_ifa_put(ifp);
4485                                                 rcu_read_lock_bh();
4486                                                 goto restart;
4487                                         }
4488                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4489                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4490                                 spin_unlock(&ifp->lock);
4491                         } else {
4492                                 /* ifp->prefered_lft <= ifp->valid_lft */
4493                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4494                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4495                                 spin_unlock(&ifp->lock);
4496                         }
4497                 }
4498         }
4499
4500         next_sec = round_jiffies_up(next);
4501         next_sched = next;
4502
4503         /* If rounded timeout is accurate enough, accept it. */
4504         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4505                 next_sched = next_sec;
4506
4507         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4508         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4509                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4510
4511         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4512                  now, next, next_sec, next_sched);
4513         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4514         rcu_read_unlock_bh();
4515 }
4516
4517 static void addrconf_verify_work(struct work_struct *w)
4518 {
4519         rtnl_lock();
4520         addrconf_verify_rtnl();
4521         rtnl_unlock();
4522 }
4523
4524 static void addrconf_verify(void)
4525 {
4526         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4527 }
4528
4529 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4530                                      struct in6_addr **peer_pfx)
4531 {
4532         struct in6_addr *pfx = NULL;
4533
4534         *peer_pfx = NULL;
4535
4536         if (addr)
4537                 pfx = nla_data(addr);
4538
4539         if (local) {
4540                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4541                         *peer_pfx = pfx;
4542                 pfx = nla_data(local);
4543         }
4544
4545         return pfx;
4546 }
4547
4548 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4549         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4550         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4551         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4552         [IFA_FLAGS]             = { .len = sizeof(u32) },
4553         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4554         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4555 };
4556
4557 static int
4558 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4559                   struct netlink_ext_ack *extack)
4560 {
4561         struct net *net = sock_net(skb->sk);
4562         struct ifaddrmsg *ifm;
4563         struct nlattr *tb[IFA_MAX+1];
4564         struct in6_addr *pfx, *peer_pfx;
4565         u32 ifa_flags;
4566         int err;
4567
4568         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4569                           extack);
4570         if (err < 0)
4571                 return err;
4572
4573         ifm = nlmsg_data(nlh);
4574         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4575         if (!pfx)
4576                 return -EINVAL;
4577
4578         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4579
4580         /* We ignore other flags so far. */
4581         ifa_flags &= IFA_F_MANAGETEMPADDR;
4582
4583         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4584                               ifm->ifa_prefixlen);
4585 }
4586
4587 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4588                                unsigned long expires, u32 flags)
4589 {
4590         struct fib6_info *f6i;
4591         u32 prio;
4592
4593         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
4594                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4595         if (!f6i)
4596                 return -ENOENT;
4597
4598         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4599         if (f6i->fib6_metric != prio) {
4600                 /* delete old one */
4601                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4602
4603                 /* add new one */
4604                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4605                                       ifp->rt_priority, ifp->idev->dev,
4606                                       expires, flags, GFP_KERNEL);
4607         } else {
4608                 if (!expires)
4609                         fib6_clean_expires(f6i);
4610                 else
4611                         fib6_set_expires(f6i, expires);
4612
4613                 fib6_info_release(f6i);
4614         }
4615
4616         return 0;
4617 }
4618
4619 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4620 {
4621         u32 flags;
4622         clock_t expires;
4623         unsigned long timeout;
4624         bool was_managetempaddr;
4625         bool had_prefixroute;
4626
4627         ASSERT_RTNL();
4628
4629         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4630                 return -EINVAL;
4631
4632         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4633             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4634                 return -EINVAL;
4635
4636         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4637                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4638
4639         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4640         if (addrconf_finite_timeout(timeout)) {
4641                 expires = jiffies_to_clock_t(timeout * HZ);
4642                 cfg->valid_lft = timeout;
4643                 flags = RTF_EXPIRES;
4644         } else {
4645                 expires = 0;
4646                 flags = 0;
4647                 cfg->ifa_flags |= IFA_F_PERMANENT;
4648         }
4649
4650         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4651         if (addrconf_finite_timeout(timeout)) {
4652                 if (timeout == 0)
4653                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4654                 cfg->preferred_lft = timeout;
4655         }
4656
4657         spin_lock_bh(&ifp->lock);
4658         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4659         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4660                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4661         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4662                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4663                         IFA_F_NOPREFIXROUTE);
4664         ifp->flags |= cfg->ifa_flags;
4665         ifp->tstamp = jiffies;
4666         ifp->valid_lft = cfg->valid_lft;
4667         ifp->prefered_lft = cfg->preferred_lft;
4668
4669         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4670                 ifp->rt_priority = cfg->rt_priority;
4671
4672         spin_unlock_bh(&ifp->lock);
4673         if (!(ifp->flags&IFA_F_TENTATIVE))
4674                 ipv6_ifa_notify(0, ifp);
4675
4676         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4677                 int rc = -ENOENT;
4678
4679                 if (had_prefixroute)
4680                         rc = modify_prefix_route(ifp, expires, flags);
4681
4682                 /* prefix route could have been deleted; if so restore it */
4683                 if (rc == -ENOENT) {
4684                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4685                                               ifp->rt_priority, ifp->idev->dev,
4686                                               expires, flags, GFP_KERNEL);
4687                 }
4688         } else if (had_prefixroute) {
4689                 enum cleanup_prefix_rt_t action;
4690                 unsigned long rt_expires;
4691
4692                 write_lock_bh(&ifp->idev->lock);
4693                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4694                 write_unlock_bh(&ifp->idev->lock);
4695
4696                 if (action != CLEANUP_PREFIX_RT_NOP) {
4697                         cleanup_prefix_route(ifp, rt_expires,
4698                                 action == CLEANUP_PREFIX_RT_DEL);
4699                 }
4700         }
4701
4702         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4703                 if (was_managetempaddr &&
4704                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4705                         cfg->valid_lft = 0;
4706                         cfg->preferred_lft = 0;
4707                 }
4708                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4709                                  cfg->preferred_lft, !was_managetempaddr,
4710                                  jiffies);
4711         }
4712
4713         addrconf_verify_rtnl();
4714
4715         return 0;
4716 }
4717
4718 static int
4719 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4720                   struct netlink_ext_ack *extack)
4721 {
4722         struct net *net = sock_net(skb->sk);
4723         struct ifaddrmsg *ifm;
4724         struct nlattr *tb[IFA_MAX+1];
4725         struct in6_addr *peer_pfx;
4726         struct inet6_ifaddr *ifa;
4727         struct net_device *dev;
4728         struct inet6_dev *idev;
4729         struct ifa6_config cfg;
4730         int err;
4731
4732         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4733                           extack);
4734         if (err < 0)
4735                 return err;
4736
4737         memset(&cfg, 0, sizeof(cfg));
4738
4739         ifm = nlmsg_data(nlh);
4740         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4741         if (!cfg.pfx)
4742                 return -EINVAL;
4743
4744         cfg.peer_pfx = peer_pfx;
4745         cfg.plen = ifm->ifa_prefixlen;
4746         if (tb[IFA_RT_PRIORITY])
4747                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4748
4749         cfg.valid_lft = INFINITY_LIFE_TIME;
4750         cfg.preferred_lft = INFINITY_LIFE_TIME;
4751
4752         if (tb[IFA_CACHEINFO]) {
4753                 struct ifa_cacheinfo *ci;
4754
4755                 ci = nla_data(tb[IFA_CACHEINFO]);
4756                 cfg.valid_lft = ci->ifa_valid;
4757                 cfg.preferred_lft = ci->ifa_prefered;
4758         }
4759
4760         dev =  __dev_get_by_index(net, ifm->ifa_index);
4761         if (!dev)
4762                 return -ENODEV;
4763
4764         if (tb[IFA_FLAGS])
4765                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4766         else
4767                 cfg.ifa_flags = ifm->ifa_flags;
4768
4769         /* We ignore other flags so far. */
4770         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4771                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4772                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4773
4774         idev = ipv6_find_idev(dev);
4775         if (!idev)
4776                 return -ENOBUFS;
4777
4778         if (!ipv6_allow_optimistic_dad(net, idev))
4779                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4780
4781         if (cfg.ifa_flags & IFA_F_NODAD &&
4782             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4783                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4784                 return -EINVAL;
4785         }
4786
4787         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4788         if (!ifa) {
4789                 /*
4790                  * It would be best to check for !NLM_F_CREATE here but
4791                  * userspace already relies on not having to provide this.
4792                  */
4793                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4794         }
4795
4796         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4797             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4798                 err = -EEXIST;
4799         else
4800                 err = inet6_addr_modify(ifa, &cfg);
4801
4802         in6_ifa_put(ifa);
4803
4804         return err;
4805 }
4806
4807 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4808                           u8 scope, int ifindex)
4809 {
4810         struct ifaddrmsg *ifm;
4811
4812         ifm = nlmsg_data(nlh);
4813         ifm->ifa_family = AF_INET6;
4814         ifm->ifa_prefixlen = prefixlen;
4815         ifm->ifa_flags = flags;
4816         ifm->ifa_scope = scope;
4817         ifm->ifa_index = ifindex;
4818 }
4819
4820 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4821                          unsigned long tstamp, u32 preferred, u32 valid)
4822 {
4823         struct ifa_cacheinfo ci;
4824
4825         ci.cstamp = cstamp_delta(cstamp);
4826         ci.tstamp = cstamp_delta(tstamp);
4827         ci.ifa_prefered = preferred;
4828         ci.ifa_valid = valid;
4829
4830         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4831 }
4832
4833 static inline int rt_scope(int ifa_scope)
4834 {
4835         if (ifa_scope & IFA_HOST)
4836                 return RT_SCOPE_HOST;
4837         else if (ifa_scope & IFA_LINK)
4838                 return RT_SCOPE_LINK;
4839         else if (ifa_scope & IFA_SITE)
4840                 return RT_SCOPE_SITE;
4841         else
4842                 return RT_SCOPE_UNIVERSE;
4843 }
4844
4845 static inline int inet6_ifaddr_msgsize(void)
4846 {
4847         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4848                + nla_total_size(16) /* IFA_LOCAL */
4849                + nla_total_size(16) /* IFA_ADDRESS */
4850                + nla_total_size(sizeof(struct ifa_cacheinfo))
4851                + nla_total_size(4)  /* IFA_FLAGS */
4852                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4853 }
4854
4855 enum addr_type_t {
4856         UNICAST_ADDR,
4857         MULTICAST_ADDR,
4858         ANYCAST_ADDR,
4859 };
4860
4861 struct inet6_fill_args {
4862         u32 portid;
4863         u32 seq;
4864         int event;
4865         unsigned int flags;
4866         int netnsid;
4867         int ifindex;
4868         enum addr_type_t type;
4869 };
4870
4871 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4872                              struct inet6_fill_args *args)
4873 {
4874         struct nlmsghdr  *nlh;
4875         u32 preferred, valid;
4876
4877         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4878                         sizeof(struct ifaddrmsg), args->flags);
4879         if (!nlh)
4880                 return -EMSGSIZE;
4881
4882         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4883                       ifa->idev->dev->ifindex);
4884
4885         if (args->netnsid >= 0 &&
4886             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4887                 goto error;
4888
4889         if (!((ifa->flags&IFA_F_PERMANENT) &&
4890               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4891                 preferred = ifa->prefered_lft;
4892                 valid = ifa->valid_lft;
4893                 if (preferred != INFINITY_LIFE_TIME) {
4894                         long tval = (jiffies - ifa->tstamp)/HZ;
4895                         if (preferred > tval)
4896                                 preferred -= tval;
4897                         else
4898                                 preferred = 0;
4899                         if (valid != INFINITY_LIFE_TIME) {
4900                                 if (valid > tval)
4901                                         valid -= tval;
4902                                 else
4903                                         valid = 0;
4904                         }
4905                 }
4906         } else {
4907                 preferred = INFINITY_LIFE_TIME;
4908                 valid = INFINITY_LIFE_TIME;
4909         }
4910
4911         if (!ipv6_addr_any(&ifa->peer_addr)) {
4912                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4913                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4914                         goto error;
4915         } else
4916                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4917                         goto error;
4918
4919         if (ifa->rt_priority &&
4920             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4921                 goto error;
4922
4923         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4924                 goto error;
4925
4926         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4927                 goto error;
4928
4929         nlmsg_end(skb, nlh);
4930         return 0;
4931
4932 error:
4933         nlmsg_cancel(skb, nlh);
4934         return -EMSGSIZE;
4935 }
4936
4937 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4938                                struct inet6_fill_args *args)
4939 {
4940         struct nlmsghdr  *nlh;
4941         u8 scope = RT_SCOPE_UNIVERSE;
4942         int ifindex = ifmca->idev->dev->ifindex;
4943
4944         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4945                 scope = RT_SCOPE_SITE;
4946
4947         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4948                         sizeof(struct ifaddrmsg), args->flags);
4949         if (!nlh)
4950                 return -EMSGSIZE;
4951
4952         if (args->netnsid >= 0 &&
4953             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4954                 return -EMSGSIZE;
4955
4956         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4957         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4958             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4959                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4960                 nlmsg_cancel(skb, nlh);
4961                 return -EMSGSIZE;
4962         }
4963
4964         nlmsg_end(skb, nlh);
4965         return 0;
4966 }
4967
4968 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4969                                struct inet6_fill_args *args)
4970 {
4971         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4972         int ifindex = dev ? dev->ifindex : 1;
4973         struct nlmsghdr  *nlh;
4974         u8 scope = RT_SCOPE_UNIVERSE;
4975
4976         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4977                 scope = RT_SCOPE_SITE;
4978
4979         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4980                         sizeof(struct ifaddrmsg), args->flags);
4981         if (!nlh)
4982                 return -EMSGSIZE;
4983
4984         if (args->netnsid >= 0 &&
4985             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4986                 return -EMSGSIZE;
4987
4988         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4989         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4990             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4991                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4992                 nlmsg_cancel(skb, nlh);
4993                 return -EMSGSIZE;
4994         }
4995
4996         nlmsg_end(skb, nlh);
4997         return 0;
4998 }
4999
5000 /* called with rcu_read_lock() */
5001 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5002                           struct netlink_callback *cb, int s_ip_idx,
5003                           struct inet6_fill_args *fillargs)
5004 {
5005         struct ifmcaddr6 *ifmca;
5006         struct ifacaddr6 *ifaca;
5007         int ip_idx = 0;
5008         int err = 1;
5009
5010         read_lock_bh(&idev->lock);
5011         switch (fillargs->type) {
5012         case UNICAST_ADDR: {
5013                 struct inet6_ifaddr *ifa;
5014                 fillargs->event = RTM_NEWADDR;
5015
5016                 /* unicast address incl. temp addr */
5017                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5018                         if (ip_idx < s_ip_idx)
5019                                 goto next;
5020                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5021                         if (err < 0)
5022                                 break;
5023                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5024 next:
5025                         ip_idx++;
5026                 }
5027                 break;
5028         }
5029         case MULTICAST_ADDR:
5030                 fillargs->event = RTM_GETMULTICAST;
5031
5032                 /* multicast address */
5033                 for (ifmca = idev->mc_list; ifmca;
5034                      ifmca = ifmca->next, ip_idx++) {
5035                         if (ip_idx < s_ip_idx)
5036                                 continue;
5037                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5038                         if (err < 0)
5039                                 break;
5040                 }
5041                 break;
5042         case ANYCAST_ADDR:
5043                 fillargs->event = RTM_GETANYCAST;
5044                 /* anycast address */
5045                 for (ifaca = idev->ac_list; ifaca;
5046                      ifaca = ifaca->aca_next, ip_idx++) {
5047                         if (ip_idx < s_ip_idx)
5048                                 continue;
5049                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5050                         if (err < 0)
5051                                 break;
5052                 }
5053                 break;
5054         default:
5055                 break;
5056         }
5057         read_unlock_bh(&idev->lock);
5058         cb->args[2] = ip_idx;
5059         return err;
5060 }
5061
5062 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5063                                        struct inet6_fill_args *fillargs,
5064                                        struct net **tgt_net, struct sock *sk,
5065                                        struct netlink_callback *cb)
5066 {
5067         struct netlink_ext_ack *extack = cb->extack;
5068         struct nlattr *tb[IFA_MAX+1];
5069         struct ifaddrmsg *ifm;
5070         int err, i;
5071
5072         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5073                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5074                 return -EINVAL;
5075         }
5076
5077         ifm = nlmsg_data(nlh);
5078         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5079                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5080                 return -EINVAL;
5081         }
5082
5083         fillargs->ifindex = ifm->ifa_index;
5084         if (fillargs->ifindex) {
5085                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5086                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5087         }
5088
5089         err = nlmsg_parse_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5090                                  ifa_ipv6_policy, extack);
5091         if (err < 0)
5092                 return err;
5093
5094         for (i = 0; i <= IFA_MAX; ++i) {
5095                 if (!tb[i])
5096                         continue;
5097
5098                 if (i == IFA_TARGET_NETNSID) {
5099                         struct net *net;
5100
5101                         fillargs->netnsid = nla_get_s32(tb[i]);
5102                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5103                         if (IS_ERR(net)) {
5104                                 fillargs->netnsid = -1;
5105                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5106                                 return PTR_ERR(net);
5107                         }
5108                         *tgt_net = net;
5109                 } else {
5110                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5111                         return -EINVAL;
5112                 }
5113         }
5114
5115         return 0;
5116 }
5117
5118 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5119                            enum addr_type_t type)
5120 {
5121         const struct nlmsghdr *nlh = cb->nlh;
5122         struct inet6_fill_args fillargs = {
5123                 .portid = NETLINK_CB(cb->skb).portid,
5124                 .seq = cb->nlh->nlmsg_seq,
5125                 .flags = NLM_F_MULTI,
5126                 .netnsid = -1,
5127                 .type = type,
5128         };
5129         struct net *net = sock_net(skb->sk);
5130         struct net *tgt_net = net;
5131         int idx, s_idx, s_ip_idx;
5132         int h, s_h;
5133         struct net_device *dev;
5134         struct inet6_dev *idev;
5135         struct hlist_head *head;
5136         int err = 0;
5137
5138         s_h = cb->args[0];
5139         s_idx = idx = cb->args[1];
5140         s_ip_idx = cb->args[2];
5141
5142         if (cb->strict_check) {
5143                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5144                                                   skb->sk, cb);
5145                 if (err < 0)
5146                         goto put_tgt_net;
5147
5148                 err = 0;
5149                 if (fillargs.ifindex) {
5150                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5151                         if (!dev) {
5152                                 err = -ENODEV;
5153                                 goto put_tgt_net;
5154                         }
5155                         idev = __in6_dev_get(dev);
5156                         if (idev) {
5157                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5158                                                      &fillargs);
5159                                 if (err > 0)
5160                                         err = 0;
5161                         }
5162                         goto put_tgt_net;
5163                 }
5164         }
5165
5166         rcu_read_lock();
5167         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5168         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5169                 idx = 0;
5170                 head = &tgt_net->dev_index_head[h];
5171                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5172                         if (idx < s_idx)
5173                                 goto cont;
5174                         if (h > s_h || idx > s_idx)
5175                                 s_ip_idx = 0;
5176                         idev = __in6_dev_get(dev);
5177                         if (!idev)
5178                                 goto cont;
5179
5180                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5181                                            &fillargs) < 0)
5182                                 goto done;
5183 cont:
5184                         idx++;
5185                 }
5186         }
5187 done:
5188         rcu_read_unlock();
5189         cb->args[0] = h;
5190         cb->args[1] = idx;
5191 put_tgt_net:
5192         if (fillargs.netnsid >= 0)
5193                 put_net(tgt_net);
5194
5195         return skb->len ? : err;
5196 }
5197
5198 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5199 {
5200         enum addr_type_t type = UNICAST_ADDR;
5201
5202         return inet6_dump_addr(skb, cb, type);
5203 }
5204
5205 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5206 {
5207         enum addr_type_t type = MULTICAST_ADDR;
5208
5209         return inet6_dump_addr(skb, cb, type);
5210 }
5211
5212
5213 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5214 {
5215         enum addr_type_t type = ANYCAST_ADDR;
5216
5217         return inet6_dump_addr(skb, cb, type);
5218 }
5219
5220 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5221                                        const struct nlmsghdr *nlh,
5222                                        struct nlattr **tb,
5223                                        struct netlink_ext_ack *extack)
5224 {
5225         struct ifaddrmsg *ifm;
5226         int i, err;
5227
5228         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5229                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5230                 return -EINVAL;
5231         }
5232
5233         ifm = nlmsg_data(nlh);
5234         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5235                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5236                 return -EINVAL;
5237         }
5238
5239         if (!netlink_strict_get_check(skb))
5240                 return nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX,
5241                                    ifa_ipv6_policy, extack);
5242
5243         err = nlmsg_parse_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5244                                  ifa_ipv6_policy, extack);
5245         if (err)
5246                 return err;
5247
5248         for (i = 0; i <= IFA_MAX; i++) {
5249                 if (!tb[i])
5250                         continue;
5251
5252                 switch (i) {
5253                 case IFA_TARGET_NETNSID:
5254                 case IFA_ADDRESS:
5255                 case IFA_LOCAL:
5256                         break;
5257                 default:
5258                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5259                         return -EINVAL;
5260                 }
5261         }
5262
5263         return 0;
5264 }
5265
5266 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5267                              struct netlink_ext_ack *extack)
5268 {
5269         struct net *net = sock_net(in_skb->sk);
5270         struct inet6_fill_args fillargs = {
5271                 .portid = NETLINK_CB(in_skb).portid,
5272                 .seq = nlh->nlmsg_seq,
5273                 .event = RTM_NEWADDR,
5274                 .flags = 0,
5275                 .netnsid = -1,
5276         };
5277         struct net *tgt_net = net;
5278         struct ifaddrmsg *ifm;
5279         struct nlattr *tb[IFA_MAX+1];
5280         struct in6_addr *addr = NULL, *peer;
5281         struct net_device *dev = NULL;
5282         struct inet6_ifaddr *ifa;
5283         struct sk_buff *skb;
5284         int err;
5285
5286         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5287         if (err < 0)
5288                 return err;
5289
5290         if (tb[IFA_TARGET_NETNSID]) {
5291                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5292
5293                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5294                                                   fillargs.netnsid);
5295                 if (IS_ERR(tgt_net))
5296                         return PTR_ERR(tgt_net);
5297         }
5298
5299         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5300         if (!addr)
5301                 return -EINVAL;
5302
5303         ifm = nlmsg_data(nlh);
5304         if (ifm->ifa_index)
5305                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5306
5307         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5308         if (!ifa) {
5309                 err = -EADDRNOTAVAIL;
5310                 goto errout;
5311         }
5312
5313         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5314         if (!skb) {
5315                 err = -ENOBUFS;
5316                 goto errout_ifa;
5317         }
5318
5319         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5320         if (err < 0) {
5321                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5322                 WARN_ON(err == -EMSGSIZE);
5323                 kfree_skb(skb);
5324                 goto errout_ifa;
5325         }
5326         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5327 errout_ifa:
5328         in6_ifa_put(ifa);
5329 errout:
5330         if (dev)
5331                 dev_put(dev);
5332         if (fillargs.netnsid >= 0)
5333                 put_net(tgt_net);
5334
5335         return err;
5336 }
5337
5338 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5339 {
5340         struct sk_buff *skb;
5341         struct net *net = dev_net(ifa->idev->dev);
5342         struct inet6_fill_args fillargs = {
5343                 .portid = 0,
5344                 .seq = 0,
5345                 .event = event,
5346                 .flags = 0,
5347                 .netnsid = -1,
5348         };
5349         int err = -ENOBUFS;
5350
5351         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5352         if (!skb)
5353                 goto errout;
5354
5355         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5356         if (err < 0) {
5357                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5358                 WARN_ON(err == -EMSGSIZE);
5359                 kfree_skb(skb);
5360                 goto errout;
5361         }
5362         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5363         return;
5364 errout:
5365         if (err < 0)
5366                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5367 }
5368
5369 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5370                                 __s32 *array, int bytes)
5371 {
5372         BUG_ON(bytes < (DEVCONF_MAX * 4));
5373
5374         memset(array, 0, bytes);
5375         array[DEVCONF_FORWARDING] = cnf->forwarding;
5376         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5377         array[DEVCONF_MTU6] = cnf->mtu6;
5378         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5379         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5380         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5381         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5382         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5383         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5384                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5385         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5386                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5387         array[DEVCONF_RTR_SOLICIT_DELAY] =
5388                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5389         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5390         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5391                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5392         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5393                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5394         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5395         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5396         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5397         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5398         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5399         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5400         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5401         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5402         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5403 #ifdef CONFIG_IPV6_ROUTER_PREF
5404         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5405         array[DEVCONF_RTR_PROBE_INTERVAL] =
5406                 jiffies_to_msecs(cnf->rtr_probe_interval);
5407 #ifdef CONFIG_IPV6_ROUTE_INFO
5408         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5409         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5410 #endif
5411 #endif
5412         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5413         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5415         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5416         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5417 #endif
5418 #ifdef CONFIG_IPV6_MROUTE
5419         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5420 #endif
5421         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5422         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5423         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5424         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5425         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5426         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5427         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5428         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5429         /* we omit DEVCONF_STABLE_SECRET for now */
5430         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5431         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5432         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5433         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5434         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5435 #ifdef CONFIG_IPV6_SEG6_HMAC
5436         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5437 #endif
5438         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5439         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5440         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5441         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5442 }
5443
5444 static inline size_t inet6_ifla6_size(void)
5445 {
5446         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5447              + nla_total_size(sizeof(struct ifla_cacheinfo))
5448              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5449              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5450              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5451              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5452              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5453              + 0;
5454 }
5455
5456 static inline size_t inet6_if_nlmsg_size(void)
5457 {
5458         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5459                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5460                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5461                + nla_total_size(4) /* IFLA_MTU */
5462                + nla_total_size(4) /* IFLA_LINK */
5463                + nla_total_size(1) /* IFLA_OPERSTATE */
5464                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5465 }
5466
5467 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5468                                         int bytes)
5469 {
5470         int i;
5471         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5472         BUG_ON(pad < 0);
5473
5474         /* Use put_unaligned() because stats may not be aligned for u64. */
5475         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5476         for (i = 1; i < ICMP6_MIB_MAX; i++)
5477                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5478
5479         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5480 }
5481
5482 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5483                                         int bytes, size_t syncpoff)
5484 {
5485         int i, c;
5486         u64 buff[IPSTATS_MIB_MAX];
5487         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5488
5489         BUG_ON(pad < 0);
5490
5491         memset(buff, 0, sizeof(buff));
5492         buff[0] = IPSTATS_MIB_MAX;
5493
5494         for_each_possible_cpu(c) {
5495                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5496                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5497         }
5498
5499         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5500         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5501 }
5502
5503 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5504                              int bytes)
5505 {
5506         switch (attrtype) {
5507         case IFLA_INET6_STATS:
5508                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5509                                      offsetof(struct ipstats_mib, syncp));
5510                 break;
5511         case IFLA_INET6_ICMP6STATS:
5512                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5513                 break;
5514         }
5515 }
5516
5517 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5518                                   u32 ext_filter_mask)
5519 {
5520         struct nlattr *nla;
5521         struct ifla_cacheinfo ci;
5522
5523         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5524                 goto nla_put_failure;
5525         ci.max_reasm_len = IPV6_MAXPLEN;
5526         ci.tstamp = cstamp_delta(idev->tstamp);
5527         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5528         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5529         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5530                 goto nla_put_failure;
5531         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5532         if (!nla)
5533                 goto nla_put_failure;
5534         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5535
5536         /* XXX - MC not implemented */
5537
5538         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5539                 return 0;
5540
5541         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5542         if (!nla)
5543                 goto nla_put_failure;
5544         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5545
5546         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5547         if (!nla)
5548                 goto nla_put_failure;
5549         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5550
5551         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5552         if (!nla)
5553                 goto nla_put_failure;
5554
5555         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5556                 goto nla_put_failure;
5557
5558         read_lock_bh(&idev->lock);
5559         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5560         read_unlock_bh(&idev->lock);
5561
5562         return 0;
5563
5564 nla_put_failure:
5565         return -EMSGSIZE;
5566 }
5567
5568 static size_t inet6_get_link_af_size(const struct net_device *dev,
5569                                      u32 ext_filter_mask)
5570 {
5571         if (!__in6_dev_get(dev))
5572                 return 0;
5573
5574         return inet6_ifla6_size();
5575 }
5576
5577 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5578                               u32 ext_filter_mask)
5579 {
5580         struct inet6_dev *idev = __in6_dev_get(dev);
5581
5582         if (!idev)
5583                 return -ENODATA;
5584
5585         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5586                 return -EMSGSIZE;
5587
5588         return 0;
5589 }
5590
5591 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5592 {
5593         struct inet6_ifaddr *ifp;
5594         struct net_device *dev = idev->dev;
5595         bool clear_token, update_rs = false;
5596         struct in6_addr ll_addr;
5597
5598         ASSERT_RTNL();
5599
5600         if (!token)
5601                 return -EINVAL;
5602         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5603                 return -EINVAL;
5604         if (!ipv6_accept_ra(idev))
5605                 return -EINVAL;
5606         if (idev->cnf.rtr_solicits == 0)
5607                 return -EINVAL;
5608
5609         write_lock_bh(&idev->lock);
5610
5611         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5612         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5613
5614         write_unlock_bh(&idev->lock);
5615
5616         clear_token = ipv6_addr_any(token);
5617         if (clear_token)
5618                 goto update_lft;
5619
5620         if (!idev->dead && (idev->if_flags & IF_READY) &&
5621             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5622                              IFA_F_OPTIMISTIC)) {
5623                 /* If we're not ready, then normal ifup will take care
5624                  * of this. Otherwise, we need to request our rs here.
5625                  */
5626                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5627                 update_rs = true;
5628         }
5629
5630 update_lft:
5631         write_lock_bh(&idev->lock);
5632
5633         if (update_rs) {
5634                 idev->if_flags |= IF_RS_SENT;
5635                 idev->rs_interval = rfc3315_s14_backoff_init(
5636                         idev->cnf.rtr_solicit_interval);
5637                 idev->rs_probes = 1;
5638                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5639         }
5640
5641         /* Well, that's kinda nasty ... */
5642         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5643                 spin_lock(&ifp->lock);
5644                 if (ifp->tokenized) {
5645                         ifp->valid_lft = 0;
5646                         ifp->prefered_lft = 0;
5647                 }
5648                 spin_unlock(&ifp->lock);
5649         }
5650
5651         write_unlock_bh(&idev->lock);
5652         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5653         addrconf_verify_rtnl();
5654         return 0;
5655 }
5656
5657 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5658         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5659         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5660 };
5661
5662 static int inet6_validate_link_af(const struct net_device *dev,
5663                                   const struct nlattr *nla)
5664 {
5665         struct nlattr *tb[IFLA_INET6_MAX + 1];
5666
5667         if (dev && !__in6_dev_get(dev))
5668                 return -EAFNOSUPPORT;
5669
5670         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5671                                 NULL);
5672 }
5673
5674 static int check_addr_gen_mode(int mode)
5675 {
5676         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5677             mode != IN6_ADDR_GEN_MODE_NONE &&
5678             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5679             mode != IN6_ADDR_GEN_MODE_RANDOM)
5680                 return -EINVAL;
5681         return 1;
5682 }
5683
5684 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5685                                 int mode)
5686 {
5687         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5688             !idev->cnf.stable_secret.initialized &&
5689             !net->ipv6.devconf_dflt->stable_secret.initialized)
5690                 return -EINVAL;
5691         return 1;
5692 }
5693
5694 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5695 {
5696         int err = -EINVAL;
5697         struct inet6_dev *idev = __in6_dev_get(dev);
5698         struct nlattr *tb[IFLA_INET6_MAX + 1];
5699
5700         if (!idev)
5701                 return -EAFNOSUPPORT;
5702
5703         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5704                 BUG();
5705
5706         if (tb[IFLA_INET6_TOKEN]) {
5707                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5708                 if (err)
5709                         return err;
5710         }
5711
5712         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5713                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5714
5715                 if (check_addr_gen_mode(mode) < 0 ||
5716                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5717                         return -EINVAL;
5718
5719                 idev->cnf.addr_gen_mode = mode;
5720                 err = 0;
5721         }
5722
5723         return err;
5724 }
5725
5726 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5727                              u32 portid, u32 seq, int event, unsigned int flags)
5728 {
5729         struct net_device *dev = idev->dev;
5730         struct ifinfomsg *hdr;
5731         struct nlmsghdr *nlh;
5732         void *protoinfo;
5733
5734         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5735         if (!nlh)
5736                 return -EMSGSIZE;
5737
5738         hdr = nlmsg_data(nlh);
5739         hdr->ifi_family = AF_INET6;
5740         hdr->__ifi_pad = 0;
5741         hdr->ifi_type = dev->type;
5742         hdr->ifi_index = dev->ifindex;
5743         hdr->ifi_flags = dev_get_flags(dev);
5744         hdr->ifi_change = 0;
5745
5746         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5747             (dev->addr_len &&
5748              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5749             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5750             (dev->ifindex != dev_get_iflink(dev) &&
5751              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5752             nla_put_u8(skb, IFLA_OPERSTATE,
5753                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5754                 goto nla_put_failure;
5755         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5756         if (!protoinfo)
5757                 goto nla_put_failure;
5758
5759         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5760                 goto nla_put_failure;
5761
5762         nla_nest_end(skb, protoinfo);
5763         nlmsg_end(skb, nlh);
5764         return 0;
5765
5766 nla_put_failure:
5767         nlmsg_cancel(skb, nlh);
5768         return -EMSGSIZE;
5769 }
5770
5771 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5772                                    struct netlink_ext_ack *extack)
5773 {
5774         struct ifinfomsg *ifm;
5775
5776         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5777                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5778                 return -EINVAL;
5779         }
5780
5781         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5782                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5783                 return -EINVAL;
5784         }
5785
5786         ifm = nlmsg_data(nlh);
5787         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5788             ifm->ifi_change || ifm->ifi_index) {
5789                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5790                 return -EINVAL;
5791         }
5792
5793         return 0;
5794 }
5795
5796 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5797 {
5798         struct net *net = sock_net(skb->sk);
5799         int h, s_h;
5800         int idx = 0, s_idx;
5801         struct net_device *dev;
5802         struct inet6_dev *idev;
5803         struct hlist_head *head;
5804
5805         /* only requests using strict checking can pass data to
5806          * influence the dump
5807          */
5808         if (cb->strict_check) {
5809                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5810
5811                 if (err < 0)
5812                         return err;
5813         }
5814
5815         s_h = cb->args[0];
5816         s_idx = cb->args[1];
5817
5818         rcu_read_lock();
5819         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5820                 idx = 0;
5821                 head = &net->dev_index_head[h];
5822                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5823                         if (idx < s_idx)
5824                                 goto cont;
5825                         idev = __in6_dev_get(dev);
5826                         if (!idev)
5827                                 goto cont;
5828                         if (inet6_fill_ifinfo(skb, idev,
5829                                               NETLINK_CB(cb->skb).portid,
5830                                               cb->nlh->nlmsg_seq,
5831                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5832                                 goto out;
5833 cont:
5834                         idx++;
5835                 }
5836         }
5837 out:
5838         rcu_read_unlock();
5839         cb->args[1] = idx;
5840         cb->args[0] = h;
5841
5842         return skb->len;
5843 }
5844
5845 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5846 {
5847         struct sk_buff *skb;
5848         struct net *net = dev_net(idev->dev);
5849         int err = -ENOBUFS;
5850
5851         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5852         if (!skb)
5853                 goto errout;
5854
5855         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5856         if (err < 0) {
5857                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5858                 WARN_ON(err == -EMSGSIZE);
5859                 kfree_skb(skb);
5860                 goto errout;
5861         }
5862         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5863         return;
5864 errout:
5865         if (err < 0)
5866                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5867 }
5868
5869 static inline size_t inet6_prefix_nlmsg_size(void)
5870 {
5871         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5872                + nla_total_size(sizeof(struct in6_addr))
5873                + nla_total_size(sizeof(struct prefix_cacheinfo));
5874 }
5875
5876 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5877                              struct prefix_info *pinfo, u32 portid, u32 seq,
5878                              int event, unsigned int flags)
5879 {
5880         struct prefixmsg *pmsg;
5881         struct nlmsghdr *nlh;
5882         struct prefix_cacheinfo ci;
5883
5884         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5885         if (!nlh)
5886                 return -EMSGSIZE;
5887
5888         pmsg = nlmsg_data(nlh);
5889         pmsg->prefix_family = AF_INET6;
5890         pmsg->prefix_pad1 = 0;
5891         pmsg->prefix_pad2 = 0;
5892         pmsg->prefix_ifindex = idev->dev->ifindex;
5893         pmsg->prefix_len = pinfo->prefix_len;
5894         pmsg->prefix_type = pinfo->type;
5895         pmsg->prefix_pad3 = 0;
5896         pmsg->prefix_flags = 0;
5897         if (pinfo->onlink)
5898                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5899         if (pinfo->autoconf)
5900                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5901
5902         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5903                 goto nla_put_failure;
5904         ci.preferred_time = ntohl(pinfo->prefered);
5905         ci.valid_time = ntohl(pinfo->valid);
5906         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5907                 goto nla_put_failure;
5908         nlmsg_end(skb, nlh);
5909         return 0;
5910
5911 nla_put_failure:
5912         nlmsg_cancel(skb, nlh);
5913         return -EMSGSIZE;
5914 }
5915
5916 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5917                          struct prefix_info *pinfo)
5918 {
5919         struct sk_buff *skb;
5920         struct net *net = dev_net(idev->dev);
5921         int err = -ENOBUFS;
5922
5923         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5924         if (!skb)
5925                 goto errout;
5926
5927         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5928         if (err < 0) {
5929                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5930                 WARN_ON(err == -EMSGSIZE);
5931                 kfree_skb(skb);
5932                 goto errout;
5933         }
5934         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5935         return;
5936 errout:
5937         if (err < 0)
5938                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5939 }
5940
5941 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5942 {
5943         struct net *net = dev_net(ifp->idev->dev);
5944
5945         if (event)
5946                 ASSERT_RTNL();
5947
5948         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5949
5950         switch (event) {
5951         case RTM_NEWADDR:
5952                 /*
5953                  * If the address was optimistic
5954                  * we inserted the route at the start of
5955                  * our DAD process, so we don't need
5956                  * to do it again
5957                  */
5958                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5959                         ip6_ins_rt(net, ifp->rt);
5960                 if (ifp->idev->cnf.forwarding)
5961                         addrconf_join_anycast(ifp);
5962                 if (!ipv6_addr_any(&ifp->peer_addr))
5963                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5964                                               ifp->idev->dev, 0, 0,
5965                                               GFP_ATOMIC);
5966                 break;
5967         case RTM_DELADDR:
5968                 if (ifp->idev->cnf.forwarding)
5969                         addrconf_leave_anycast(ifp);
5970                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5971                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5972                         struct fib6_info *rt;
5973
5974                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5975                                                        ifp->idev->dev, 0, 0,
5976                                                        false);
5977                         if (rt)
5978                                 ip6_del_rt(net, rt);
5979                 }
5980                 if (ifp->rt) {
5981                         ip6_del_rt(net, ifp->rt);
5982                         ifp->rt = NULL;
5983                 }
5984                 rt_genid_bump_ipv6(net);
5985                 break;
5986         }
5987         atomic_inc(&net->ipv6.dev_addr_genid);
5988 }
5989
5990 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5991 {
5992         rcu_read_lock_bh();
5993         if (likely(ifp->idev->dead == 0))
5994                 __ipv6_ifa_notify(event, ifp);
5995         rcu_read_unlock_bh();
5996 }
5997
5998 #ifdef CONFIG_SYSCTL
5999
6000 static
6001 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6002                            void __user *buffer, size_t *lenp, loff_t *ppos)
6003 {
6004         int *valp = ctl->data;
6005         int val = *valp;
6006         loff_t pos = *ppos;
6007         struct ctl_table lctl;
6008         int ret;
6009
6010         /*
6011          * ctl->data points to idev->cnf.forwarding, we should
6012          * not modify it until we get the rtnl lock.
6013          */
6014         lctl = *ctl;
6015         lctl.data = &val;
6016
6017         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6018
6019         if (write)
6020                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6021         if (ret)
6022                 *ppos = pos;
6023         return ret;
6024 }
6025
6026 static
6027 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6028                         void __user *buffer, size_t *lenp, loff_t *ppos)
6029 {
6030         struct inet6_dev *idev = ctl->extra1;
6031         int min_mtu = IPV6_MIN_MTU;
6032         struct ctl_table lctl;
6033
6034         lctl = *ctl;
6035         lctl.extra1 = &min_mtu;
6036         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6037
6038         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6039 }
6040
6041 static void dev_disable_change(struct inet6_dev *idev)
6042 {
6043         struct netdev_notifier_info info;
6044
6045         if (!idev || !idev->dev)
6046                 return;
6047
6048         netdev_notifier_info_init(&info, idev->dev);
6049         if (idev->cnf.disable_ipv6)
6050                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6051         else
6052                 addrconf_notify(NULL, NETDEV_UP, &info);
6053 }
6054
6055 static void addrconf_disable_change(struct net *net, __s32 newf)
6056 {
6057         struct net_device *dev;
6058         struct inet6_dev *idev;
6059
6060         for_each_netdev(net, dev) {
6061                 idev = __in6_dev_get(dev);
6062                 if (idev) {
6063                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6064                         idev->cnf.disable_ipv6 = newf;
6065                         if (changed)
6066                                 dev_disable_change(idev);
6067                 }
6068         }
6069 }
6070
6071 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6072 {
6073         struct net *net;
6074         int old;
6075
6076         if (!rtnl_trylock())
6077                 return restart_syscall();
6078
6079         net = (struct net *)table->extra2;
6080         old = *p;
6081         *p = newf;
6082
6083         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6084                 rtnl_unlock();
6085                 return 0;
6086         }
6087
6088         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6089                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6090                 addrconf_disable_change(net, newf);
6091         } else if ((!newf) ^ (!old))
6092                 dev_disable_change((struct inet6_dev *)table->extra1);
6093
6094         rtnl_unlock();
6095         return 0;
6096 }
6097
6098 static
6099 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6100                             void __user *buffer, size_t *lenp, loff_t *ppos)
6101 {
6102         int *valp = ctl->data;
6103         int val = *valp;
6104         loff_t pos = *ppos;
6105         struct ctl_table lctl;
6106         int ret;
6107
6108         /*
6109          * ctl->data points to idev->cnf.disable_ipv6, we should
6110          * not modify it until we get the rtnl lock.
6111          */
6112         lctl = *ctl;
6113         lctl.data = &val;
6114
6115         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6116
6117         if (write)
6118                 ret = addrconf_disable_ipv6(ctl, valp, val);
6119         if (ret)
6120                 *ppos = pos;
6121         return ret;
6122 }
6123
6124 static
6125 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6126                               void __user *buffer, size_t *lenp, loff_t *ppos)
6127 {
6128         int *valp = ctl->data;
6129         int ret;
6130         int old, new;
6131
6132         old = *valp;
6133         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6134         new = *valp;
6135
6136         if (write && old != new) {
6137                 struct net *net = ctl->extra2;
6138
6139                 if (!rtnl_trylock())
6140                         return restart_syscall();
6141
6142                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6143                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6144                                                      NETCONFA_PROXY_NEIGH,
6145                                                      NETCONFA_IFINDEX_DEFAULT,
6146                                                      net->ipv6.devconf_dflt);
6147                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6148                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6149                                                      NETCONFA_PROXY_NEIGH,
6150                                                      NETCONFA_IFINDEX_ALL,
6151                                                      net->ipv6.devconf_all);
6152                 else {
6153                         struct inet6_dev *idev = ctl->extra1;
6154
6155                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6156                                                      NETCONFA_PROXY_NEIGH,
6157                                                      idev->dev->ifindex,
6158                                                      &idev->cnf);
6159                 }
6160                 rtnl_unlock();
6161         }
6162
6163         return ret;
6164 }
6165
6166 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6167                                          void __user *buffer, size_t *lenp,
6168                                          loff_t *ppos)
6169 {
6170         int ret = 0;
6171         u32 new_val;
6172         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6173         struct net *net = (struct net *)ctl->extra2;
6174         struct ctl_table tmp = {
6175                 .data = &new_val,
6176                 .maxlen = sizeof(new_val),
6177                 .mode = ctl->mode,
6178         };
6179
6180         if (!rtnl_trylock())
6181                 return restart_syscall();
6182
6183         new_val = *((u32 *)ctl->data);
6184
6185         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6186         if (ret != 0)
6187                 goto out;
6188
6189         if (write) {
6190                 if (check_addr_gen_mode(new_val) < 0) {
6191                         ret = -EINVAL;
6192                         goto out;
6193                 }
6194
6195                 if (idev) {
6196                         if (check_stable_privacy(idev, net, new_val) < 0) {
6197                                 ret = -EINVAL;
6198                                 goto out;
6199                         }
6200
6201                         if (idev->cnf.addr_gen_mode != new_val) {
6202                                 idev->cnf.addr_gen_mode = new_val;
6203                                 addrconf_dev_config(idev->dev);
6204                         }
6205                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6206                         struct net_device *dev;
6207
6208                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6209                         for_each_netdev(net, dev) {
6210                                 idev = __in6_dev_get(dev);
6211                                 if (idev &&
6212                                     idev->cnf.addr_gen_mode != new_val) {
6213                                         idev->cnf.addr_gen_mode = new_val;
6214                                         addrconf_dev_config(idev->dev);
6215                                 }
6216                         }
6217                 }
6218
6219                 *((u32 *)ctl->data) = new_val;
6220         }
6221
6222 out:
6223         rtnl_unlock();
6224
6225         return ret;
6226 }
6227
6228 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6229                                          void __user *buffer, size_t *lenp,
6230                                          loff_t *ppos)
6231 {
6232         int err;
6233         struct in6_addr addr;
6234         char str[IPV6_MAX_STRLEN];
6235         struct ctl_table lctl = *ctl;
6236         struct net *net = ctl->extra2;
6237         struct ipv6_stable_secret *secret = ctl->data;
6238
6239         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6240                 return -EIO;
6241
6242         lctl.maxlen = IPV6_MAX_STRLEN;
6243         lctl.data = str;
6244
6245         if (!rtnl_trylock())
6246                 return restart_syscall();
6247
6248         if (!write && !secret->initialized) {
6249                 err = -EIO;
6250                 goto out;
6251         }
6252
6253         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6254         if (err >= sizeof(str)) {
6255                 err = -EIO;
6256                 goto out;
6257         }
6258
6259         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6260         if (err || !write)
6261                 goto out;
6262
6263         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6264                 err = -EIO;
6265                 goto out;
6266         }
6267
6268         secret->initialized = true;
6269         secret->secret = addr;
6270
6271         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6272                 struct net_device *dev;
6273
6274                 for_each_netdev(net, dev) {
6275                         struct inet6_dev *idev = __in6_dev_get(dev);
6276
6277                         if (idev) {
6278                                 idev->cnf.addr_gen_mode =
6279                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6280                         }
6281                 }
6282         } else {
6283                 struct inet6_dev *idev = ctl->extra1;
6284
6285                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6286         }
6287
6288 out:
6289         rtnl_unlock();
6290
6291         return err;
6292 }
6293
6294 static
6295 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6296                                                 int write,
6297                                                 void __user *buffer,
6298                                                 size_t *lenp,
6299                                                 loff_t *ppos)
6300 {
6301         int *valp = ctl->data;
6302         int val = *valp;
6303         loff_t pos = *ppos;
6304         struct ctl_table lctl;
6305         int ret;
6306
6307         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6308          * we should not modify it until we get the rtnl lock.
6309          */
6310         lctl = *ctl;
6311         lctl.data = &val;
6312
6313         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6314
6315         if (write)
6316                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6317         if (ret)
6318                 *ppos = pos;
6319         return ret;
6320 }
6321
6322 static
6323 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6324 {
6325         if (rt) {
6326                 if (action)
6327                         rt->dst.flags |= DST_NOPOLICY;
6328                 else
6329                         rt->dst.flags &= ~DST_NOPOLICY;
6330         }
6331 }
6332
6333 static
6334 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6335 {
6336         struct inet6_ifaddr *ifa;
6337
6338         read_lock_bh(&idev->lock);
6339         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6340                 spin_lock(&ifa->lock);
6341                 if (ifa->rt) {
6342                         struct fib6_info *rt = ifa->rt;
6343                         int cpu;
6344
6345                         rcu_read_lock();
6346                         ifa->rt->dst_nopolicy = val ? true : false;
6347                         if (rt->rt6i_pcpu) {
6348                                 for_each_possible_cpu(cpu) {
6349                                         struct rt6_info **rtp;
6350
6351                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6352                                         addrconf_set_nopolicy(*rtp, val);
6353                                 }
6354                         }
6355                         rcu_read_unlock();
6356                 }
6357                 spin_unlock(&ifa->lock);
6358         }
6359         read_unlock_bh(&idev->lock);
6360 }
6361
6362 static
6363 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6364 {
6365         struct inet6_dev *idev;
6366         struct net *net;
6367
6368         if (!rtnl_trylock())
6369                 return restart_syscall();
6370
6371         *valp = val;
6372
6373         net = (struct net *)ctl->extra2;
6374         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6375                 rtnl_unlock();
6376                 return 0;
6377         }
6378
6379         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6380                 struct net_device *dev;
6381
6382                 for_each_netdev(net, dev) {
6383                         idev = __in6_dev_get(dev);
6384                         if (idev)
6385                                 addrconf_disable_policy_idev(idev, val);
6386                 }
6387         } else {
6388                 idev = (struct inet6_dev *)ctl->extra1;
6389                 addrconf_disable_policy_idev(idev, val);
6390         }
6391
6392         rtnl_unlock();
6393         return 0;
6394 }
6395
6396 static
6397 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6398                                    void __user *buffer, size_t *lenp,
6399                                    loff_t *ppos)
6400 {
6401         int *valp = ctl->data;
6402         int val = *valp;
6403         loff_t pos = *ppos;
6404         struct ctl_table lctl;
6405         int ret;
6406
6407         lctl = *ctl;
6408         lctl.data = &val;
6409         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6410
6411         if (write && (*valp != val))
6412                 ret = addrconf_disable_policy(ctl, valp, val);
6413
6414         if (ret)
6415                 *ppos = pos;
6416
6417         return ret;
6418 }
6419
6420 static int minus_one = -1;
6421 static const int zero = 0;
6422 static const int one = 1;
6423 static const int two_five_five = 255;
6424
6425 static const struct ctl_table addrconf_sysctl[] = {
6426         {
6427                 .procname       = "forwarding",
6428                 .data           = &ipv6_devconf.forwarding,
6429                 .maxlen         = sizeof(int),
6430                 .mode           = 0644,
6431                 .proc_handler   = addrconf_sysctl_forward,
6432         },
6433         {
6434                 .procname       = "hop_limit",
6435                 .data           = &ipv6_devconf.hop_limit,
6436                 .maxlen         = sizeof(int),
6437                 .mode           = 0644,
6438                 .proc_handler   = proc_dointvec_minmax,
6439                 .extra1         = (void *)&one,
6440                 .extra2         = (void *)&two_five_five,
6441         },
6442         {
6443                 .procname       = "mtu",
6444                 .data           = &ipv6_devconf.mtu6,
6445                 .maxlen         = sizeof(int),
6446                 .mode           = 0644,
6447                 .proc_handler   = addrconf_sysctl_mtu,
6448         },
6449         {
6450                 .procname       = "accept_ra",
6451                 .data           = &ipv6_devconf.accept_ra,
6452                 .maxlen         = sizeof(int),
6453                 .mode           = 0644,
6454                 .proc_handler   = proc_dointvec,
6455         },
6456         {
6457                 .procname       = "accept_redirects",
6458                 .data           = &ipv6_devconf.accept_redirects,
6459                 .maxlen         = sizeof(int),
6460                 .mode           = 0644,
6461                 .proc_handler   = proc_dointvec,
6462         },
6463         {
6464                 .procname       = "autoconf",
6465                 .data           = &ipv6_devconf.autoconf,
6466                 .maxlen         = sizeof(int),
6467                 .mode           = 0644,
6468                 .proc_handler   = proc_dointvec,
6469         },
6470         {
6471                 .procname       = "dad_transmits",
6472                 .data           = &ipv6_devconf.dad_transmits,
6473                 .maxlen         = sizeof(int),
6474                 .mode           = 0644,
6475                 .proc_handler   = proc_dointvec,
6476         },
6477         {
6478                 .procname       = "router_solicitations",
6479                 .data           = &ipv6_devconf.rtr_solicits,
6480                 .maxlen         = sizeof(int),
6481                 .mode           = 0644,
6482                 .proc_handler   = proc_dointvec_minmax,
6483                 .extra1         = &minus_one,
6484         },
6485         {
6486                 .procname       = "router_solicitation_interval",
6487                 .data           = &ipv6_devconf.rtr_solicit_interval,
6488                 .maxlen         = sizeof(int),
6489                 .mode           = 0644,
6490                 .proc_handler   = proc_dointvec_jiffies,
6491         },
6492         {
6493                 .procname       = "router_solicitation_max_interval",
6494                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6495                 .maxlen         = sizeof(int),
6496                 .mode           = 0644,
6497                 .proc_handler   = proc_dointvec_jiffies,
6498         },
6499         {
6500                 .procname       = "router_solicitation_delay",
6501                 .data           = &ipv6_devconf.rtr_solicit_delay,
6502                 .maxlen         = sizeof(int),
6503                 .mode           = 0644,
6504                 .proc_handler   = proc_dointvec_jiffies,
6505         },
6506         {
6507                 .procname       = "force_mld_version",
6508                 .data           = &ipv6_devconf.force_mld_version,
6509                 .maxlen         = sizeof(int),
6510                 .mode           = 0644,
6511                 .proc_handler   = proc_dointvec,
6512         },
6513         {
6514                 .procname       = "mldv1_unsolicited_report_interval",
6515                 .data           =
6516                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6517                 .maxlen         = sizeof(int),
6518                 .mode           = 0644,
6519                 .proc_handler   = proc_dointvec_ms_jiffies,
6520         },
6521         {
6522                 .procname       = "mldv2_unsolicited_report_interval",
6523                 .data           =
6524                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6525                 .maxlen         = sizeof(int),
6526                 .mode           = 0644,
6527                 .proc_handler   = proc_dointvec_ms_jiffies,
6528         },
6529         {
6530                 .procname       = "use_tempaddr",
6531                 .data           = &ipv6_devconf.use_tempaddr,
6532                 .maxlen         = sizeof(int),
6533                 .mode           = 0644,
6534                 .proc_handler   = proc_dointvec,
6535         },
6536         {
6537                 .procname       = "temp_valid_lft",
6538                 .data           = &ipv6_devconf.temp_valid_lft,
6539                 .maxlen         = sizeof(int),
6540                 .mode           = 0644,
6541                 .proc_handler   = proc_dointvec,
6542         },
6543         {
6544                 .procname       = "temp_prefered_lft",
6545                 .data           = &ipv6_devconf.temp_prefered_lft,
6546                 .maxlen         = sizeof(int),
6547                 .mode           = 0644,
6548                 .proc_handler   = proc_dointvec,
6549         },
6550         {
6551                 .procname       = "regen_max_retry",
6552                 .data           = &ipv6_devconf.regen_max_retry,
6553                 .maxlen         = sizeof(int),
6554                 .mode           = 0644,
6555                 .proc_handler   = proc_dointvec,
6556         },
6557         {
6558                 .procname       = "max_desync_factor",
6559                 .data           = &ipv6_devconf.max_desync_factor,
6560                 .maxlen         = sizeof(int),
6561                 .mode           = 0644,
6562                 .proc_handler   = proc_dointvec,
6563         },
6564         {
6565                 .procname       = "max_addresses",
6566                 .data           = &ipv6_devconf.max_addresses,
6567                 .maxlen         = sizeof(int),
6568                 .mode           = 0644,
6569                 .proc_handler   = proc_dointvec,
6570         },
6571         {
6572                 .procname       = "accept_ra_defrtr",
6573                 .data           = &ipv6_devconf.accept_ra_defrtr,
6574                 .maxlen         = sizeof(int),
6575                 .mode           = 0644,
6576                 .proc_handler   = proc_dointvec,
6577         },
6578         {
6579                 .procname       = "accept_ra_min_hop_limit",
6580                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6581                 .maxlen         = sizeof(int),
6582                 .mode           = 0644,
6583                 .proc_handler   = proc_dointvec,
6584         },
6585         {
6586                 .procname       = "accept_ra_pinfo",
6587                 .data           = &ipv6_devconf.accept_ra_pinfo,
6588                 .maxlen         = sizeof(int),
6589                 .mode           = 0644,
6590                 .proc_handler   = proc_dointvec,
6591         },
6592 #ifdef CONFIG_IPV6_ROUTER_PREF
6593         {
6594                 .procname       = "accept_ra_rtr_pref",
6595                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6596                 .maxlen         = sizeof(int),
6597                 .mode           = 0644,
6598                 .proc_handler   = proc_dointvec,
6599         },
6600         {
6601                 .procname       = "router_probe_interval",
6602                 .data           = &ipv6_devconf.rtr_probe_interval,
6603                 .maxlen         = sizeof(int),
6604                 .mode           = 0644,
6605                 .proc_handler   = proc_dointvec_jiffies,
6606         },
6607 #ifdef CONFIG_IPV6_ROUTE_INFO
6608         {
6609                 .procname       = "accept_ra_rt_info_min_plen",
6610                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6611                 .maxlen         = sizeof(int),
6612                 .mode           = 0644,
6613                 .proc_handler   = proc_dointvec,
6614         },
6615         {
6616                 .procname       = "accept_ra_rt_info_max_plen",
6617                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6618                 .maxlen         = sizeof(int),
6619                 .mode           = 0644,
6620                 .proc_handler   = proc_dointvec,
6621         },
6622 #endif
6623 #endif
6624         {
6625                 .procname       = "proxy_ndp",
6626                 .data           = &ipv6_devconf.proxy_ndp,
6627                 .maxlen         = sizeof(int),
6628                 .mode           = 0644,
6629                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6630         },
6631         {
6632                 .procname       = "accept_source_route",
6633                 .data           = &ipv6_devconf.accept_source_route,
6634                 .maxlen         = sizeof(int),
6635                 .mode           = 0644,
6636                 .proc_handler   = proc_dointvec,
6637         },
6638 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6639         {
6640                 .procname       = "optimistic_dad",
6641                 .data           = &ipv6_devconf.optimistic_dad,
6642                 .maxlen         = sizeof(int),
6643                 .mode           = 0644,
6644                 .proc_handler   = proc_dointvec,
6645         },
6646         {
6647                 .procname       = "use_optimistic",
6648                 .data           = &ipv6_devconf.use_optimistic,
6649                 .maxlen         = sizeof(int),
6650                 .mode           = 0644,
6651                 .proc_handler   = proc_dointvec,
6652         },
6653 #endif
6654 #ifdef CONFIG_IPV6_MROUTE
6655         {
6656                 .procname       = "mc_forwarding",
6657                 .data           = &ipv6_devconf.mc_forwarding,
6658                 .maxlen         = sizeof(int),
6659                 .mode           = 0444,
6660                 .proc_handler   = proc_dointvec,
6661         },
6662 #endif
6663         {
6664                 .procname       = "disable_ipv6",
6665                 .data           = &ipv6_devconf.disable_ipv6,
6666                 .maxlen         = sizeof(int),
6667                 .mode           = 0644,
6668                 .proc_handler   = addrconf_sysctl_disable,
6669         },
6670         {
6671                 .procname       = "accept_dad",
6672                 .data           = &ipv6_devconf.accept_dad,
6673                 .maxlen         = sizeof(int),
6674                 .mode           = 0644,
6675                 .proc_handler   = proc_dointvec,
6676         },
6677         {
6678                 .procname       = "force_tllao",
6679                 .data           = &ipv6_devconf.force_tllao,
6680                 .maxlen         = sizeof(int),
6681                 .mode           = 0644,
6682                 .proc_handler   = proc_dointvec
6683         },
6684         {
6685                 .procname       = "ndisc_notify",
6686                 .data           = &ipv6_devconf.ndisc_notify,
6687                 .maxlen         = sizeof(int),
6688                 .mode           = 0644,
6689                 .proc_handler   = proc_dointvec
6690         },
6691         {
6692                 .procname       = "suppress_frag_ndisc",
6693                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6694                 .maxlen         = sizeof(int),
6695                 .mode           = 0644,
6696                 .proc_handler   = proc_dointvec
6697         },
6698         {
6699                 .procname       = "accept_ra_from_local",
6700                 .data           = &ipv6_devconf.accept_ra_from_local,
6701                 .maxlen         = sizeof(int),
6702                 .mode           = 0644,
6703                 .proc_handler   = proc_dointvec,
6704         },
6705         {
6706                 .procname       = "accept_ra_mtu",
6707                 .data           = &ipv6_devconf.accept_ra_mtu,
6708                 .maxlen         = sizeof(int),
6709                 .mode           = 0644,
6710                 .proc_handler   = proc_dointvec,
6711         },
6712         {
6713                 .procname       = "stable_secret",
6714                 .data           = &ipv6_devconf.stable_secret,
6715                 .maxlen         = IPV6_MAX_STRLEN,
6716                 .mode           = 0600,
6717                 .proc_handler   = addrconf_sysctl_stable_secret,
6718         },
6719         {
6720                 .procname       = "use_oif_addrs_only",
6721                 .data           = &ipv6_devconf.use_oif_addrs_only,
6722                 .maxlen         = sizeof(int),
6723                 .mode           = 0644,
6724                 .proc_handler   = proc_dointvec,
6725         },
6726         {
6727                 .procname       = "ignore_routes_with_linkdown",
6728                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6729                 .maxlen         = sizeof(int),
6730                 .mode           = 0644,
6731                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6732         },
6733         {
6734                 .procname       = "drop_unicast_in_l2_multicast",
6735                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6736                 .maxlen         = sizeof(int),
6737                 .mode           = 0644,
6738                 .proc_handler   = proc_dointvec,
6739         },
6740         {
6741                 .procname       = "drop_unsolicited_na",
6742                 .data           = &ipv6_devconf.drop_unsolicited_na,
6743                 .maxlen         = sizeof(int),
6744                 .mode           = 0644,
6745                 .proc_handler   = proc_dointvec,
6746         },
6747         {
6748                 .procname       = "keep_addr_on_down",
6749                 .data           = &ipv6_devconf.keep_addr_on_down,
6750                 .maxlen         = sizeof(int),
6751                 .mode           = 0644,
6752                 .proc_handler   = proc_dointvec,
6753
6754         },
6755         {
6756                 .procname       = "seg6_enabled",
6757                 .data           = &ipv6_devconf.seg6_enabled,
6758                 .maxlen         = sizeof(int),
6759                 .mode           = 0644,
6760                 .proc_handler   = proc_dointvec,
6761         },
6762 #ifdef CONFIG_IPV6_SEG6_HMAC
6763         {
6764                 .procname       = "seg6_require_hmac",
6765                 .data           = &ipv6_devconf.seg6_require_hmac,
6766                 .maxlen         = sizeof(int),
6767                 .mode           = 0644,
6768                 .proc_handler   = proc_dointvec,
6769         },
6770 #endif
6771         {
6772                 .procname       = "enhanced_dad",
6773                 .data           = &ipv6_devconf.enhanced_dad,
6774                 .maxlen         = sizeof(int),
6775                 .mode           = 0644,
6776                 .proc_handler   = proc_dointvec,
6777         },
6778         {
6779                 .procname               = "addr_gen_mode",
6780                 .data                   = &ipv6_devconf.addr_gen_mode,
6781                 .maxlen                 = sizeof(int),
6782                 .mode                   = 0644,
6783                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6784         },
6785         {
6786                 .procname       = "disable_policy",
6787                 .data           = &ipv6_devconf.disable_policy,
6788                 .maxlen         = sizeof(int),
6789                 .mode           = 0644,
6790                 .proc_handler   = addrconf_sysctl_disable_policy,
6791         },
6792         {
6793                 .procname       = "ndisc_tclass",
6794                 .data           = &ipv6_devconf.ndisc_tclass,
6795                 .maxlen         = sizeof(int),
6796                 .mode           = 0644,
6797                 .proc_handler   = proc_dointvec_minmax,
6798                 .extra1         = (void *)&zero,
6799                 .extra2         = (void *)&two_five_five,
6800         },
6801         {
6802                 /* sentinel */
6803         }
6804 };
6805
6806 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6807                 struct inet6_dev *idev, struct ipv6_devconf *p)
6808 {
6809         int i, ifindex;
6810         struct ctl_table *table;
6811         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6812
6813         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6814         if (!table)
6815                 goto out;
6816
6817         for (i = 0; table[i].data; i++) {
6818                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6819                 /* If one of these is already set, then it is not safe to
6820                  * overwrite either of them: this makes proc_dointvec_minmax
6821                  * usable.
6822                  */
6823                 if (!table[i].extra1 && !table[i].extra2) {
6824                         table[i].extra1 = idev; /* embedded; no ref */
6825                         table[i].extra2 = net;
6826                 }
6827         }
6828
6829         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6830
6831         p->sysctl_header = register_net_sysctl(net, path, table);
6832         if (!p->sysctl_header)
6833                 goto free;
6834
6835         if (!strcmp(dev_name, "all"))
6836                 ifindex = NETCONFA_IFINDEX_ALL;
6837         else if (!strcmp(dev_name, "default"))
6838                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6839         else
6840                 ifindex = idev->dev->ifindex;
6841         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6842                                      ifindex, p);
6843         return 0;
6844
6845 free:
6846         kfree(table);
6847 out:
6848         return -ENOBUFS;
6849 }
6850
6851 static void __addrconf_sysctl_unregister(struct net *net,
6852                                          struct ipv6_devconf *p, int ifindex)
6853 {
6854         struct ctl_table *table;
6855
6856         if (!p->sysctl_header)
6857                 return;
6858
6859         table = p->sysctl_header->ctl_table_arg;
6860         unregister_net_sysctl_table(p->sysctl_header);
6861         p->sysctl_header = NULL;
6862         kfree(table);
6863
6864         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6865 }
6866
6867 static int addrconf_sysctl_register(struct inet6_dev *idev)
6868 {
6869         int err;
6870
6871         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6872                 return -EINVAL;
6873
6874         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6875                                     &ndisc_ifinfo_sysctl_change);
6876         if (err)
6877                 return err;
6878         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6879                                          idev, &idev->cnf);
6880         if (err)
6881                 neigh_sysctl_unregister(idev->nd_parms);
6882
6883         return err;
6884 }
6885
6886 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6887 {
6888         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6889                                      idev->dev->ifindex);
6890         neigh_sysctl_unregister(idev->nd_parms);
6891 }
6892
6893
6894 #endif
6895
6896 static int __net_init addrconf_init_net(struct net *net)
6897 {
6898         int err = -ENOMEM;
6899         struct ipv6_devconf *all, *dflt;
6900
6901         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6902         if (!all)
6903                 goto err_alloc_all;
6904
6905         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6906         if (!dflt)
6907                 goto err_alloc_dflt;
6908
6909         if (IS_ENABLED(CONFIG_SYSCTL) &&
6910             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6911                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6912                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6913         }
6914
6915         /* these will be inherited by all namespaces */
6916         dflt->autoconf = ipv6_defaults.autoconf;
6917         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6918
6919         dflt->stable_secret.initialized = false;
6920         all->stable_secret.initialized = false;
6921
6922         net->ipv6.devconf_all = all;
6923         net->ipv6.devconf_dflt = dflt;
6924
6925 #ifdef CONFIG_SYSCTL
6926         err = __addrconf_sysctl_register(net, "all", NULL, all);
6927         if (err < 0)
6928                 goto err_reg_all;
6929
6930         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6931         if (err < 0)
6932                 goto err_reg_dflt;
6933 #endif
6934         return 0;
6935
6936 #ifdef CONFIG_SYSCTL
6937 err_reg_dflt:
6938         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6939 err_reg_all:
6940         kfree(dflt);
6941 #endif
6942 err_alloc_dflt:
6943         kfree(all);
6944 err_alloc_all:
6945         return err;
6946 }
6947
6948 static void __net_exit addrconf_exit_net(struct net *net)
6949 {
6950 #ifdef CONFIG_SYSCTL
6951         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6952                                      NETCONFA_IFINDEX_DEFAULT);
6953         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6954                                      NETCONFA_IFINDEX_ALL);
6955 #endif
6956         kfree(net->ipv6.devconf_dflt);
6957         kfree(net->ipv6.devconf_all);
6958 }
6959
6960 static struct pernet_operations addrconf_ops = {
6961         .init = addrconf_init_net,
6962         .exit = addrconf_exit_net,
6963 };
6964
6965 static struct rtnl_af_ops inet6_ops __read_mostly = {
6966         .family           = AF_INET6,
6967         .fill_link_af     = inet6_fill_link_af,
6968         .get_link_af_size = inet6_get_link_af_size,
6969         .validate_link_af = inet6_validate_link_af,
6970         .set_link_af      = inet6_set_link_af,
6971 };
6972
6973 /*
6974  *      Init / cleanup code
6975  */
6976
6977 int __init addrconf_init(void)
6978 {
6979         struct inet6_dev *idev;
6980         int i, err;
6981
6982         err = ipv6_addr_label_init();
6983         if (err < 0) {
6984                 pr_crit("%s: cannot initialize default policy table: %d\n",
6985                         __func__, err);
6986                 goto out;
6987         }
6988
6989         err = register_pernet_subsys(&addrconf_ops);
6990         if (err < 0)
6991                 goto out_addrlabel;
6992
6993         addrconf_wq = create_workqueue("ipv6_addrconf");
6994         if (!addrconf_wq) {
6995                 err = -ENOMEM;
6996                 goto out_nowq;
6997         }
6998
6999         /* The addrconf netdev notifier requires that loopback_dev
7000          * has it's ipv6 private information allocated and setup
7001          * before it can bring up and give link-local addresses
7002          * to other devices which are up.
7003          *
7004          * Unfortunately, loopback_dev is not necessarily the first
7005          * entry in the global dev_base list of net devices.  In fact,
7006          * it is likely to be the very last entry on that list.
7007          * So this causes the notifier registry below to try and
7008          * give link-local addresses to all devices besides loopback_dev
7009          * first, then loopback_dev, which cases all the non-loopback_dev
7010          * devices to fail to get a link-local address.
7011          *
7012          * So, as a temporary fix, allocate the ipv6 structure for
7013          * loopback_dev first by hand.
7014          * Longer term, all of the dependencies ipv6 has upon the loopback
7015          * device and it being up should be removed.
7016          */
7017         rtnl_lock();
7018         idev = ipv6_add_dev(init_net.loopback_dev);
7019         rtnl_unlock();
7020         if (IS_ERR(idev)) {
7021                 err = PTR_ERR(idev);
7022                 goto errlo;
7023         }
7024
7025         ip6_route_init_special_entries();
7026
7027         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7028                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7029
7030         register_netdevice_notifier(&ipv6_dev_notf);
7031
7032         addrconf_verify();
7033
7034         rtnl_af_register(&inet6_ops);
7035
7036         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7037                                    NULL, inet6_dump_ifinfo, 0);
7038         if (err < 0)
7039                 goto errout;
7040
7041         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7042                                    inet6_rtm_newaddr, NULL, 0);
7043         if (err < 0)
7044                 goto errout;
7045         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7046                                    inet6_rtm_deladdr, NULL, 0);
7047         if (err < 0)
7048                 goto errout;
7049         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7050                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7051                                    RTNL_FLAG_DOIT_UNLOCKED);
7052         if (err < 0)
7053                 goto errout;
7054         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7055                                    NULL, inet6_dump_ifmcaddr, 0);
7056         if (err < 0)
7057                 goto errout;
7058         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7059                                    NULL, inet6_dump_ifacaddr, 0);
7060         if (err < 0)
7061                 goto errout;
7062         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7063                                    inet6_netconf_get_devconf,
7064                                    inet6_netconf_dump_devconf,
7065                                    RTNL_FLAG_DOIT_UNLOCKED);
7066         if (err < 0)
7067                 goto errout;
7068         err = ipv6_addr_label_rtnl_register();
7069         if (err < 0)
7070                 goto errout;
7071
7072         return 0;
7073 errout:
7074         rtnl_unregister_all(PF_INET6);
7075         rtnl_af_unregister(&inet6_ops);
7076         unregister_netdevice_notifier(&ipv6_dev_notf);
7077 errlo:
7078         destroy_workqueue(addrconf_wq);
7079 out_nowq:
7080         unregister_pernet_subsys(&addrconf_ops);
7081 out_addrlabel:
7082         ipv6_addr_label_cleanup();
7083 out:
7084         return err;
7085 }
7086
7087 void addrconf_cleanup(void)
7088 {
7089         struct net_device *dev;
7090         int i;
7091
7092         unregister_netdevice_notifier(&ipv6_dev_notf);
7093         unregister_pernet_subsys(&addrconf_ops);
7094         ipv6_addr_label_cleanup();
7095
7096         rtnl_af_unregister(&inet6_ops);
7097
7098         rtnl_lock();
7099
7100         /* clean dev list */
7101         for_each_netdev(&init_net, dev) {
7102                 if (__in6_dev_get(dev) == NULL)
7103                         continue;
7104                 addrconf_ifdown(dev, 1);
7105         }
7106         addrconf_ifdown(init_net.loopback_dev, 2);
7107
7108         /*
7109          *      Check hash table.
7110          */
7111         spin_lock_bh(&addrconf_hash_lock);
7112         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7113                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7114         spin_unlock_bh(&addrconf_hash_lock);
7115         cancel_delayed_work(&addr_chk_work);
7116         rtnl_unlock();
7117
7118         destroy_workqueue(addrconf_wq);
7119 }