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_deprecated(nlh, sizeof(struct netconfmsg),
 615                                               tb, NETCONFA_MAX,
 616                                               devconf_ipv6_policy, extack);
 617
 618         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 619                                             tb, NETCONFA_MAX,
 620                                             devconf_ipv6_policy, extack);
 621         if (err)
 622                 return err;
 623
 624         for (i = 0; i <= NETCONFA_MAX; i++) {
 625                 if (!tb[i])
 626                         continue;
 627
 628                 switch (i) {
 629                 case NETCONFA_IFINDEX:
 630                         break;
 631                 default:
 632                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 633                         return -EINVAL;
 634                 }
 635         }
 636
 637         return 0;
 638 }
 639
 640 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 641                                      struct nlmsghdr *nlh,
 642                                      struct netlink_ext_ack *extack)
 643 {
 644         struct net *net = sock_net(in_skb->sk);
 645         struct nlattr *tb[NETCONFA_MAX+1];
 646         struct inet6_dev *in6_dev = NULL;
 647         struct net_device *dev = NULL;
 648         struct sk_buff *skb;
 649         struct ipv6_devconf *devconf;
 650         int ifindex;
 651         int err;
 652
 653         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 654         if (err < 0)
 655                 return err;
 656
 657         if (!tb[NETCONFA_IFINDEX])
 658                 return -EINVAL;
 659
 660         err = -EINVAL;
 661         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 662         switch (ifindex) {
 663         case NETCONFA_IFINDEX_ALL:
 664                 devconf = net->ipv6.devconf_all;
 665                 break;
 666         case NETCONFA_IFINDEX_DEFAULT:
 667                 devconf = net->ipv6.devconf_dflt;
 668                 break;
 669         default:
 670                 dev = dev_get_by_index(net, ifindex);
 671                 if (!dev)
 672                         return -EINVAL;
 673                 in6_dev = in6_dev_get(dev);
 674                 if (!in6_dev)
 675                         goto errout;
 676                 devconf = &in6_dev->cnf;
 677                 break;
 678         }
 679
 680         err = -ENOBUFS;
 681         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 682         if (!skb)
 683                 goto errout;
 684
 685         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 686                                          NETLINK_CB(in_skb).portid,
 687                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 688                                          NETCONFA_ALL);
 689         if (err < 0) {
 690                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 691                 WARN_ON(err == -EMSGSIZE);
 692                 kfree_skb(skb);
 693                 goto errout;
 694         }
 695         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 696 errout:
 697         if (in6_dev)
 698                 in6_dev_put(in6_dev);
 699         if (dev)
 700                 dev_put(dev);
 701         return err;
 702 }
 703
 704 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 705                                       struct netlink_callback *cb)
 706 {
 707         const struct nlmsghdr *nlh = cb->nlh;
 708         struct net *net = sock_net(skb->sk);
 709         int h, s_h;
 710         int idx, s_idx;
 711         struct net_device *dev;
 712         struct inet6_dev *idev;
 713         struct hlist_head *head;
 714
 715         if (cb->strict_check) {
 716                 struct netlink_ext_ack *extack = cb->extack;
 717                 struct netconfmsg *ncm;
 718
 719                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 720                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 721                         return -EINVAL;
 722                 }
 723
 724                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 725                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 726                         return -EINVAL;
 727                 }
 728         }
 729
 730         s_h = cb->args[0];
 731         s_idx = idx = cb->args[1];
 732
 733         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 734                 idx = 0;
 735                 head = &net->dev_index_head[h];
 736                 rcu_read_lock();
 737                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 738                           net->dev_base_seq;
 739                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 740                         if (idx < s_idx)
 741                                 goto cont;
 742                         idev = __in6_dev_get(dev);
 743                         if (!idev)
 744                                 goto cont;
 745
 746                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 747                                                        &idev->cnf,
 748                                                        NETLINK_CB(cb->skb).portid,
 749                                                        nlh->nlmsg_seq,
 750                                                        RTM_NEWNETCONF,
 751                                                        NLM_F_MULTI,
 752                                                        NETCONFA_ALL) < 0) {
 753                                 rcu_read_unlock();
 754                                 goto done;
 755                         }
 756                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 757 cont:
 758                         idx++;
 759                 }
 760                 rcu_read_unlock();
 761         }
 762         if (h == NETDEV_HASHENTRIES) {
 763                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 764                                                net->ipv6.devconf_all,
 765                                                NETLINK_CB(cb->skb).portid,
 766                                                nlh->nlmsg_seq,
 767                                                RTM_NEWNETCONF, NLM_F_MULTI,
 768                                                NETCONFA_ALL) < 0)
 769                         goto done;
 770                 else
 771                         h++;
 772         }
 773         if (h == NETDEV_HASHENTRIES + 1) {
 774                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 775                                                net->ipv6.devconf_dflt,
 776                                                NETLINK_CB(cb->skb).portid,
 777                                                nlh->nlmsg_seq,
 778                                                RTM_NEWNETCONF, NLM_F_MULTI,
 779                                                NETCONFA_ALL) < 0)
 780                         goto done;
 781                 else
 782                         h++;
 783         }
 784 done:
 785         cb->args[0] = h;
 786         cb->args[1] = idx;
 787
 788         return skb->len;
 789 }
 790
 791 #ifdef CONFIG_SYSCTL
 792 static void dev_forward_change(struct inet6_dev *idev)
 793 {
 794         struct net_device *dev;
 795         struct inet6_ifaddr *ifa;
 796
 797         if (!idev)
 798                 return;
 799         dev = idev->dev;
 800         if (idev->cnf.forwarding)
 801                 dev_disable_lro(dev);
 802         if (dev->flags & IFF_MULTICAST) {
 803                 if (idev->cnf.forwarding) {
 804                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 805                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 806                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 807                 } else {
 808                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 809                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 810                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 811                 }
 812         }
 813
 814         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 815                 if (ifa->flags&IFA_F_TENTATIVE)
 816                         continue;
 817                 if (idev->cnf.forwarding)
 818                         addrconf_join_anycast(ifa);
 819                 else
 820                         addrconf_leave_anycast(ifa);
 821         }
 822         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 823                                      NETCONFA_FORWARDING,
 824                                      dev->ifindex, &idev->cnf);
 825 }
 826
 827
 828 static void addrconf_forward_change(struct net *net, __s32 newf)
 829 {
 830         struct net_device *dev;
 831         struct inet6_dev *idev;
 832
 833         for_each_netdev(net, dev) {
 834                 idev = __in6_dev_get(dev);
 835                 if (idev) {
 836                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 837                         idev->cnf.forwarding = newf;
 838                         if (changed)
 839                                 dev_forward_change(idev);
 840                 }
 841         }
 842 }
 843
 844 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 845 {
 846         struct net *net;
 847         int old;
 848
 849         if (!rtnl_trylock())
 850                 return restart_syscall();
 851
 852         net = (struct net *)table->extra2;
 853         old = *p;
 854         *p = newf;
 855
 856         if (p == &net->ipv6.devconf_dflt->forwarding) {
 857                 if ((!newf) ^ (!old))
 858                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 859                                                      NETCONFA_FORWARDING,
 860                                                      NETCONFA_IFINDEX_DEFAULT,
 861                                                      net->ipv6.devconf_dflt);
 862                 rtnl_unlock();
 863                 return 0;
 864         }
 865
 866         if (p == &net->ipv6.devconf_all->forwarding) {
 867                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 868
 869                 net->ipv6.devconf_dflt->forwarding = newf;
 870                 if ((!newf) ^ (!old_dflt))
 871                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 872                                                      NETCONFA_FORWARDING,
 873                                                      NETCONFA_IFINDEX_DEFAULT,
 874                                                      net->ipv6.devconf_dflt);
 875
 876                 addrconf_forward_change(net, newf);
 877                 if ((!newf) ^ (!old))
 878                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 879                                                      NETCONFA_FORWARDING,
 880                                                      NETCONFA_IFINDEX_ALL,
 881                                                      net->ipv6.devconf_all);
 882         } else if ((!newf) ^ (!old))
 883                 dev_forward_change((struct inet6_dev *)table->extra1);
 884         rtnl_unlock();
 885
 886         if (newf)
 887                 rt6_purge_dflt_routers(net);
 888         return 1;
 889 }
 890
 891 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 892 {
 893         struct net_device *dev;
 894         struct inet6_dev *idev;
 895
 896         for_each_netdev(net, dev) {
 897                 idev = __in6_dev_get(dev);
 898                 if (idev) {
 899                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 900
 901                         idev->cnf.ignore_routes_with_linkdown = newf;
 902                         if (changed)
 903                                 inet6_netconf_notify_devconf(dev_net(dev),
 904                                                              RTM_NEWNETCONF,
 905                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 906                                                              dev->ifindex,
 907                                                              &idev->cnf);
 908                 }
 909         }
 910 }
 911
 912 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 913 {
 914         struct net *net;
 915         int old;
 916
 917         if (!rtnl_trylock())
 918                 return restart_syscall();
 919
 920         net = (struct net *)table->extra2;
 921         old = *p;
 922         *p = newf;
 923
 924         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 925                 if ((!newf) ^ (!old))
 926                         inet6_netconf_notify_devconf(net,
 927                                                      RTM_NEWNETCONF,
 928                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 929                                                      NETCONFA_IFINDEX_DEFAULT,
 930                                                      net->ipv6.devconf_dflt);
 931                 rtnl_unlock();
 932                 return 0;
 933         }
 934
 935         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 936                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 937                 addrconf_linkdown_change(net, newf);
 938                 if ((!newf) ^ (!old))
 939                         inet6_netconf_notify_devconf(net,
 940                                                      RTM_NEWNETCONF,
 941                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 942                                                      NETCONFA_IFINDEX_ALL,
 943                                                      net->ipv6.devconf_all);
 944         }
 945         rtnl_unlock();
 946
 947         return 1;
 948 }
 949
 950 #endif
 951
 952 /* Nobody refers to this ifaddr, destroy it */
 953 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 954 {
 955         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 956
 957 #ifdef NET_REFCNT_DEBUG
 958         pr_debug("%s\n", __func__);
 959 #endif
 960
 961         in6_dev_put(ifp->idev);
 962
 963         if (cancel_delayed_work(&ifp->dad_work))
 964                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 965                           ifp);
 966
 967         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 968                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 969                 return;
 970         }
 971
 972         kfree_rcu(ifp, rcu);
 973 }
 974
 975 static void
 976 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 977 {
 978         struct list_head *p;
 979         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 980
 981         /*
 982          * Each device address list is sorted in order of scope -
 983          * global before linklocal.
 984          */
 985         list_for_each(p, &idev->addr_list) {
 986                 struct inet6_ifaddr *ifa
 987                         = list_entry(p, struct inet6_ifaddr, if_list);
 988                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 989                         break;
 990         }
 991
 992         list_add_tail_rcu(&ifp->if_list, p);
 993 }
 994
 995 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 996 {
 997         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 998
 999         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1000 }
1001
1002 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1003                                struct net_device *dev, unsigned int hash)
1004 {
1005         struct inet6_ifaddr *ifp;
1006
1007         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1008                 if (!net_eq(dev_net(ifp->idev->dev), net))
1009                         continue;
1010                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1011                         if (!dev || ifp->idev->dev == dev)
1012                                 return true;
1013                 }
1014         }
1015         return false;
1016 }
1017
1018 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1019 {
1020         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1021         int err = 0;
1022
1023         spin_lock(&addrconf_hash_lock);
1024
1025         /* Ignore adding duplicate addresses on an interface */
1026         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1027                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1028                 err = -EEXIST;
1029         } else {
1030                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1031         }
1032
1033         spin_unlock(&addrconf_hash_lock);
1034
1035         return err;
1036 }
1037
1038 /* On success it returns ifp with increased reference count */
1039
1040 static struct inet6_ifaddr *
1041 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1042               bool can_block, struct netlink_ext_ack *extack)
1043 {
1044         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1045         int addr_type = ipv6_addr_type(cfg->pfx);
1046         struct net *net = dev_net(idev->dev);
1047         struct inet6_ifaddr *ifa = NULL;
1048         struct fib6_info *f6i = NULL;
1049         int err = 0;
1050
1051         if (addr_type == IPV6_ADDR_ANY ||
1052             addr_type & IPV6_ADDR_MULTICAST ||
1053             (!(idev->dev->flags & IFF_LOOPBACK) &&
1054              !netif_is_l3_master(idev->dev) &&
1055              addr_type & IPV6_ADDR_LOOPBACK))
1056                 return ERR_PTR(-EADDRNOTAVAIL);
1057
1058         if (idev->dead) {
1059                 err = -ENODEV;                  /*XXX*/
1060                 goto out;
1061         }
1062
1063         if (idev->cnf.disable_ipv6) {
1064                 err = -EACCES;
1065                 goto out;
1066         }
1067
1068         /* validator notifier needs to be blocking;
1069          * do not call in atomic context
1070          */
1071         if (can_block) {
1072                 struct in6_validator_info i6vi = {
1073                         .i6vi_addr = *cfg->pfx,
1074                         .i6vi_dev = idev,
1075                         .extack = extack,
1076                 };
1077
1078                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1079                 err = notifier_to_errno(err);
1080                 if (err < 0)
1081                         goto out;
1082         }
1083
1084         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1085         if (!ifa) {
1086                 err = -ENOBUFS;
1087                 goto out;
1088         }
1089
1090         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1091         if (IS_ERR(f6i)) {
1092                 err = PTR_ERR(f6i);
1093                 f6i = NULL;
1094                 goto out;
1095         }
1096
1097         if (net->ipv6.devconf_all->disable_policy ||
1098             idev->cnf.disable_policy)
1099                 f6i->dst_nopolicy = true;
1100
1101         neigh_parms_data_state_setall(idev->nd_parms);
1102
1103         ifa->addr = *cfg->pfx;
1104         if (cfg->peer_pfx)
1105                 ifa->peer_addr = *cfg->peer_pfx;
1106
1107         spin_lock_init(&ifa->lock);
1108         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1109         INIT_HLIST_NODE(&ifa->addr_lst);
1110         ifa->scope = cfg->scope;
1111         ifa->prefix_len = cfg->plen;
1112         ifa->rt_priority = cfg->rt_priority;
1113         ifa->flags = cfg->ifa_flags;
1114         /* No need to add the TENTATIVE flag for addresses with NODAD */
1115         if (!(cfg->ifa_flags & IFA_F_NODAD))
1116                 ifa->flags |= IFA_F_TENTATIVE;
1117         ifa->valid_lft = cfg->valid_lft;
1118         ifa->prefered_lft = cfg->preferred_lft;
1119         ifa->cstamp = ifa->tstamp = jiffies;
1120         ifa->tokenized = false;
1121
1122         ifa->rt = f6i;
1123
1124         ifa->idev = idev;
1125         in6_dev_hold(idev);
1126
1127         /* For caller */
1128         refcount_set(&ifa->refcnt, 1);
1129
1130         rcu_read_lock_bh();
1131
1132         err = ipv6_add_addr_hash(idev->dev, ifa);
1133         if (err < 0) {
1134                 rcu_read_unlock_bh();
1135                 goto out;
1136         }
1137
1138         write_lock(&idev->lock);
1139
1140         /* Add to inet6_dev unicast addr list. */
1141         ipv6_link_dev_addr(idev, ifa);
1142
1143         if (ifa->flags&IFA_F_TEMPORARY) {
1144                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1145                 in6_ifa_hold(ifa);
1146         }
1147
1148         in6_ifa_hold(ifa);
1149         write_unlock(&idev->lock);
1150
1151         rcu_read_unlock_bh();
1152
1153         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1154 out:
1155         if (unlikely(err < 0)) {
1156                 fib6_info_release(f6i);
1157
1158                 if (ifa) {
1159                         if (ifa->idev)
1160                                 in6_dev_put(ifa->idev);
1161                         kfree(ifa);
1162                 }
1163                 ifa = ERR_PTR(err);
1164         }
1165
1166         return ifa;
1167 }
1168
1169 enum cleanup_prefix_rt_t {
1170         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1171         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1172         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1173 };
1174
1175 /*
1176  * Check, whether the prefix for ifp would still need a prefix route
1177  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1178  * constants.
1179  *
1180  * 1) we don't purge prefix if address was not permanent.
1181  *    prefix is managed by its own lifetime.
1182  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1183  * 3) if there are no addresses, delete prefix.
1184  * 4) if there are still other permanent address(es),
1185  *    corresponding prefix is still permanent.
1186  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1187  *    don't purge the prefix, assume user space is managing it.
1188  * 6) otherwise, update prefix lifetime to the
1189  *    longest valid lifetime among the corresponding
1190  *    addresses on the device.
1191  *    Note: subsequent RA will update lifetime.
1192  **/
1193 static enum cleanup_prefix_rt_t
1194 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1195 {
1196         struct inet6_ifaddr *ifa;
1197         struct inet6_dev *idev = ifp->idev;
1198         unsigned long lifetime;
1199         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1200
1201         *expires = jiffies;
1202
1203         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1204                 if (ifa == ifp)
1205                         continue;
1206                 if (ifa->prefix_len != ifp->prefix_len ||
1207                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1208                                        ifp->prefix_len))
1209                         continue;
1210                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1211                         return CLEANUP_PREFIX_RT_NOP;
1212
1213                 action = CLEANUP_PREFIX_RT_EXPIRE;
1214
1215                 spin_lock(&ifa->lock);
1216
1217                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1218                 /*
1219                  * Note: Because this address is
1220                  * not permanent, lifetime <
1221                  * LONG_MAX / HZ here.
1222                  */
1223                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1224                         *expires = ifa->tstamp + lifetime * HZ;
1225                 spin_unlock(&ifa->lock);
1226         }
1227
1228         return action;
1229 }
1230
1231 static void
1232 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1233 {
1234         struct fib6_info *f6i;
1235
1236         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
1237                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1238         if (f6i) {
1239                 if (del_rt)
1240                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1241                 else {
1242                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1243                                 fib6_set_expires(f6i, expires);
1244                         fib6_info_release(f6i);
1245                 }
1246         }
1247 }
1248
1249
1250 /* This function wants to get referenced ifp and releases it before return */
1251
1252 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1253 {
1254         int state;
1255         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1256         unsigned long expires;
1257
1258         ASSERT_RTNL();
1259
1260         spin_lock_bh(&ifp->lock);
1261         state = ifp->state;
1262         ifp->state = INET6_IFADDR_STATE_DEAD;
1263         spin_unlock_bh(&ifp->lock);
1264
1265         if (state == INET6_IFADDR_STATE_DEAD)
1266                 goto out;
1267
1268         spin_lock_bh(&addrconf_hash_lock);
1269         hlist_del_init_rcu(&ifp->addr_lst);
1270         spin_unlock_bh(&addrconf_hash_lock);
1271
1272         write_lock_bh(&ifp->idev->lock);
1273
1274         if (ifp->flags&IFA_F_TEMPORARY) {
1275                 list_del(&ifp->tmp_list);
1276                 if (ifp->ifpub) {
1277                         in6_ifa_put(ifp->ifpub);
1278                         ifp->ifpub = NULL;
1279                 }
1280                 __in6_ifa_put(ifp);
1281         }
1282
1283         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1284                 action = check_cleanup_prefix_route(ifp, &expires);
1285
1286         list_del_rcu(&ifp->if_list);
1287         __in6_ifa_put(ifp);
1288
1289         write_unlock_bh(&ifp->idev->lock);
1290
1291         addrconf_del_dad_work(ifp);
1292
1293         ipv6_ifa_notify(RTM_DELADDR, ifp);
1294
1295         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1296
1297         if (action != CLEANUP_PREFIX_RT_NOP) {
1298                 cleanup_prefix_route(ifp, expires,
1299                         action == CLEANUP_PREFIX_RT_DEL);
1300         }
1301
1302         /* clean up prefsrc entries */
1303         rt6_remove_prefsrc(ifp);
1304 out:
1305         in6_ifa_put(ifp);
1306 }
1307
1308 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1309                                 struct inet6_ifaddr *ift,
1310                                 bool block)
1311 {
1312         struct inet6_dev *idev = ifp->idev;
1313         struct in6_addr addr, *tmpaddr;
1314         unsigned long tmp_tstamp, age;
1315         unsigned long regen_advance;
1316         struct ifa6_config cfg;
1317         int ret = 0;
1318         unsigned long now = jiffies;
1319         long max_desync_factor;
1320         s32 cnf_temp_preferred_lft;
1321
1322         write_lock_bh(&idev->lock);
1323         if (ift) {
1324                 spin_lock_bh(&ift->lock);
1325                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1326                 spin_unlock_bh(&ift->lock);
1327                 tmpaddr = &addr;
1328         } else {
1329                 tmpaddr = NULL;
1330         }
1331 retry:
1332         in6_dev_hold(idev);
1333         if (idev->cnf.use_tempaddr <= 0) {
1334                 write_unlock_bh(&idev->lock);
1335                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1336                 in6_dev_put(idev);
1337                 ret = -1;
1338                 goto out;
1339         }
1340         spin_lock_bh(&ifp->lock);
1341         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1342                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1343                 spin_unlock_bh(&ifp->lock);
1344                 write_unlock_bh(&idev->lock);
1345                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1346                         __func__);
1347                 in6_dev_put(idev);
1348                 ret = -1;
1349                 goto out;
1350         }
1351         in6_ifa_hold(ifp);
1352         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1353         ipv6_try_regen_rndid(idev, tmpaddr);
1354         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1355         age = (now - ifp->tstamp) / HZ;
1356
1357         regen_advance = idev->cnf.regen_max_retry *
1358                         idev->cnf.dad_transmits *
1359                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1360
1361         /* recalculate max_desync_factor each time and update
1362          * idev->desync_factor if it's larger
1363          */
1364         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1365         max_desync_factor = min_t(__u32,
1366                                   idev->cnf.max_desync_factor,
1367                                   cnf_temp_preferred_lft - regen_advance);
1368
1369         if (unlikely(idev->desync_factor > max_desync_factor)) {
1370                 if (max_desync_factor > 0) {
1371                         get_random_bytes(&idev->desync_factor,
1372                                          sizeof(idev->desync_factor));
1373                         idev->desync_factor %= max_desync_factor;
1374                 } else {
1375                         idev->desync_factor = 0;
1376                 }
1377         }
1378
1379         memset(&cfg, 0, sizeof(cfg));
1380         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1381                               idev->cnf.temp_valid_lft + age);
1382         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1383         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1384
1385         cfg.plen = ifp->prefix_len;
1386         tmp_tstamp = ifp->tstamp;
1387         spin_unlock_bh(&ifp->lock);
1388
1389         write_unlock_bh(&idev->lock);
1390
1391         /* A temporary address is created only if this calculated Preferred
1392          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1393          * an implementation must not create a temporary address with a zero
1394          * Preferred Lifetime.
1395          * Use age calculation as in addrconf_verify to avoid unnecessary
1396          * temporary addresses being generated.
1397          */
1398         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1399         if (cfg.preferred_lft <= regen_advance + age) {
1400                 in6_ifa_put(ifp);
1401                 in6_dev_put(idev);
1402                 ret = -1;
1403                 goto out;
1404         }
1405
1406         cfg.ifa_flags = IFA_F_TEMPORARY;
1407         /* set in addrconf_prefix_rcv() */
1408         if (ifp->flags & IFA_F_OPTIMISTIC)
1409                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1410
1411         cfg.pfx = &addr;
1412         cfg.scope = ipv6_addr_scope(cfg.pfx);
1413
1414         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1415         if (IS_ERR(ift)) {
1416                 in6_ifa_put(ifp);
1417                 in6_dev_put(idev);
1418                 pr_info("%s: retry temporary address regeneration\n", __func__);
1419                 tmpaddr = &addr;
1420                 write_lock_bh(&idev->lock);
1421                 goto retry;
1422         }
1423
1424         spin_lock_bh(&ift->lock);
1425         ift->ifpub = ifp;
1426         ift->cstamp = now;
1427         ift->tstamp = tmp_tstamp;
1428         spin_unlock_bh(&ift->lock);
1429
1430         addrconf_dad_start(ift);
1431         in6_ifa_put(ift);
1432         in6_dev_put(idev);
1433 out:
1434         return ret;
1435 }
1436
1437 /*
1438  *      Choose an appropriate source address (RFC3484)
1439  */
1440 enum {
1441         IPV6_SADDR_RULE_INIT = 0,
1442         IPV6_SADDR_RULE_LOCAL,
1443         IPV6_SADDR_RULE_SCOPE,
1444         IPV6_SADDR_RULE_PREFERRED,
1445 #ifdef CONFIG_IPV6_MIP6
1446         IPV6_SADDR_RULE_HOA,
1447 #endif
1448         IPV6_SADDR_RULE_OIF,
1449         IPV6_SADDR_RULE_LABEL,
1450         IPV6_SADDR_RULE_PRIVACY,
1451         IPV6_SADDR_RULE_ORCHID,
1452         IPV6_SADDR_RULE_PREFIX,
1453 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1454         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1455 #endif
1456         IPV6_SADDR_RULE_MAX
1457 };
1458
1459 struct ipv6_saddr_score {
1460         int                     rule;
1461         int                     addr_type;
1462         struct inet6_ifaddr     *ifa;
1463         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1464         int                     scopedist;
1465         int                     matchlen;
1466 };
1467
1468 struct ipv6_saddr_dst {
1469         const struct in6_addr *addr;
1470         int ifindex;
1471         int scope;
1472         int label;
1473         unsigned int prefs;
1474 };
1475
1476 static inline int ipv6_saddr_preferred(int type)
1477 {
1478         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1479                 return 1;
1480         return 0;
1481 }
1482
1483 static bool ipv6_use_optimistic_addr(struct net *net,
1484                                      struct inet6_dev *idev)
1485 {
1486 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1487         if (!idev)
1488                 return false;
1489         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1490                 return false;
1491         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1492                 return false;
1493
1494         return true;
1495 #else
1496         return false;
1497 #endif
1498 }
1499
1500 static bool ipv6_allow_optimistic_dad(struct net *net,
1501                                       struct inet6_dev *idev)
1502 {
1503 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1504         if (!idev)
1505                 return false;
1506         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1507                 return false;
1508
1509         return true;
1510 #else
1511         return false;
1512 #endif
1513 }
1514
1515 static int ipv6_get_saddr_eval(struct net *net,
1516                                struct ipv6_saddr_score *score,
1517                                struct ipv6_saddr_dst *dst,
1518                                int i)
1519 {
1520         int ret;
1521
1522         if (i <= score->rule) {
1523                 switch (i) {
1524                 case IPV6_SADDR_RULE_SCOPE:
1525                         ret = score->scopedist;
1526                         break;
1527                 case IPV6_SADDR_RULE_PREFIX:
1528                         ret = score->matchlen;
1529                         break;
1530                 default:
1531                         ret = !!test_bit(i, score->scorebits);
1532                 }
1533                 goto out;
1534         }
1535
1536         switch (i) {
1537         case IPV6_SADDR_RULE_INIT:
1538                 /* Rule 0: remember if hiscore is not ready yet */
1539                 ret = !!score->ifa;
1540                 break;
1541         case IPV6_SADDR_RULE_LOCAL:
1542                 /* Rule 1: Prefer same address */
1543                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1544                 break;
1545         case IPV6_SADDR_RULE_SCOPE:
1546                 /* Rule 2: Prefer appropriate scope
1547                  *
1548                  *      ret
1549                  *       ^
1550                  *    -1 |  d 15
1551                  *    ---+--+-+---> scope
1552                  *       |
1553                  *       |             d is scope of the destination.
1554                  *  B-d  |  \
1555                  *       |   \      <- smaller scope is better if
1556                  *  B-15 |    \        if scope is enough for destination.
1557                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1558                  * d-C-1 | /
1559                  *       |/         <- greater is better
1560                  *   -C  /             if scope is not enough for destination.
1561                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1562                  *
1563                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1564                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1565                  * Assume B = 0 and we get C > 29.
1566                  */
1567                 ret = __ipv6_addr_src_scope(score->addr_type);
1568                 if (ret >= dst->scope)
1569                         ret = -ret;
1570                 else
1571                         ret -= 128;     /* 30 is enough */
1572                 score->scopedist = ret;
1573                 break;
1574         case IPV6_SADDR_RULE_PREFERRED:
1575             {
1576                 /* Rule 3: Avoid deprecated and optimistic addresses */
1577                 u8 avoid = IFA_F_DEPRECATED;
1578
1579                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1580                         avoid |= IFA_F_OPTIMISTIC;
1581                 ret = ipv6_saddr_preferred(score->addr_type) ||
1582                       !(score->ifa->flags & avoid);
1583                 break;
1584             }
1585 #ifdef CONFIG_IPV6_MIP6
1586         case IPV6_SADDR_RULE_HOA:
1587             {
1588                 /* Rule 4: Prefer home address */
1589                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1590                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1591                 break;
1592             }
1593 #endif
1594         case IPV6_SADDR_RULE_OIF:
1595                 /* Rule 5: Prefer outgoing interface */
1596                 ret = (!dst->ifindex ||
1597                        dst->ifindex == score->ifa->idev->dev->ifindex);
1598                 break;
1599         case IPV6_SADDR_RULE_LABEL:
1600                 /* Rule 6: Prefer matching label */
1601                 ret = ipv6_addr_label(net,
1602                                       &score->ifa->addr, score->addr_type,
1603                                       score->ifa->idev->dev->ifindex) == dst->label;
1604                 break;
1605         case IPV6_SADDR_RULE_PRIVACY:
1606             {
1607                 /* Rule 7: Prefer public address
1608                  * Note: prefer temporary address if use_tempaddr >= 2
1609                  */
1610                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1611                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1612                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1613                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1614                 break;
1615             }
1616         case IPV6_SADDR_RULE_ORCHID:
1617                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1618                  *          non-ORCHID vs non-ORCHID
1619                  */
1620                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1621                         ipv6_addr_orchid(dst->addr));
1622                 break;
1623         case IPV6_SADDR_RULE_PREFIX:
1624                 /* Rule 8: Use longest matching prefix */
1625                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1626                 if (ret > score->ifa->prefix_len)
1627                         ret = score->ifa->prefix_len;
1628                 score->matchlen = ret;
1629                 break;
1630 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1631         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1632                 /* Optimistic addresses still have lower precedence than other
1633                  * preferred addresses.
1634                  */
1635                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1636                 break;
1637 #endif
1638         default:
1639                 ret = 0;
1640         }
1641
1642         if (ret)
1643                 __set_bit(i, score->scorebits);
1644         score->rule = i;
1645 out:
1646         return ret;
1647 }
1648
1649 static int __ipv6_dev_get_saddr(struct net *net,
1650                                 struct ipv6_saddr_dst *dst,
1651                                 struct inet6_dev *idev,
1652                                 struct ipv6_saddr_score *scores,
1653                                 int hiscore_idx)
1654 {
1655         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1656
1657         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1658                 int i;
1659
1660                 /*
1661                  * - Tentative Address (RFC2462 section 5.4)
1662                  *  - A tentative address is not considered
1663                  *    "assigned to an interface" in the traditional
1664                  *    sense, unless it is also flagged as optimistic.
1665                  * - Candidate Source Address (section 4)
1666                  *  - In any case, anycast addresses, multicast
1667                  *    addresses, and the unspecified address MUST
1668                  *    NOT be included in a candidate set.
1669                  */
1670                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1671                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1672                         continue;
1673
1674                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1675
1676                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1677                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1678                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1679                                             idev->dev->name);
1680                         continue;
1681                 }
1682
1683                 score->rule = -1;
1684                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1685
1686                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1687                         int minihiscore, miniscore;
1688
1689                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1690                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1691
1692                         if (minihiscore > miniscore) {
1693                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1694                                     score->scopedist > 0) {
1695                                         /*
1696                                          * special case:
1697                                          * each remaining entry
1698                                          * has too small (not enough)
1699                                          * scope, because ifa entries
1700                                          * are sorted by their scope
1701                                          * values.
1702                                          */
1703                                         goto out;
1704                                 }
1705                                 break;
1706                         } else if (minihiscore < miniscore) {
1707                                 swap(hiscore, score);
1708                                 hiscore_idx = 1 - hiscore_idx;
1709
1710                                 /* restore our iterator */
1711                                 score->ifa = hiscore->ifa;
1712
1713                                 break;
1714                         }
1715                 }
1716         }
1717 out:
1718         return hiscore_idx;
1719 }
1720
1721 static int ipv6_get_saddr_master(struct net *net,
1722                                  const struct net_device *dst_dev,
1723                                  const struct net_device *master,
1724                                  struct ipv6_saddr_dst *dst,
1725                                  struct ipv6_saddr_score *scores,
1726                                  int hiscore_idx)
1727 {
1728         struct inet6_dev *idev;
1729
1730         idev = __in6_dev_get(dst_dev);
1731         if (idev)
1732                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1733                                                    scores, hiscore_idx);
1734
1735         idev = __in6_dev_get(master);
1736         if (idev)
1737                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1738                                                    scores, hiscore_idx);
1739
1740         return hiscore_idx;
1741 }
1742
1743 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1744                        const struct in6_addr *daddr, unsigned int prefs,
1745                        struct in6_addr *saddr)
1746 {
1747         struct ipv6_saddr_score scores[2], *hiscore;
1748         struct ipv6_saddr_dst dst;
1749         struct inet6_dev *idev;
1750         struct net_device *dev;
1751         int dst_type;
1752         bool use_oif_addr = false;
1753         int hiscore_idx = 0;
1754         int ret = 0;
1755
1756         dst_type = __ipv6_addr_type(daddr);
1757         dst.addr = daddr;
1758         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1759         dst.scope = __ipv6_addr_src_scope(dst_type);
1760         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1761         dst.prefs = prefs;
1762
1763         scores[hiscore_idx].rule = -1;
1764         scores[hiscore_idx].ifa = NULL;
1765
1766         rcu_read_lock();
1767
1768         /* Candidate Source Address (section 4)
1769          *  - multicast and link-local destination address,
1770          *    the set of candidate source address MUST only
1771          *    include addresses assigned to interfaces
1772          *    belonging to the same link as the outgoing
1773          *    interface.
1774          * (- For site-local destination addresses, the
1775          *    set of candidate source addresses MUST only
1776          *    include addresses assigned to interfaces
1777          *    belonging to the same site as the outgoing
1778          *    interface.)
1779          *  - "It is RECOMMENDED that the candidate source addresses
1780          *    be the set of unicast addresses assigned to the
1781          *    interface that will be used to send to the destination
1782          *    (the 'outgoing' interface)." (RFC 6724)
1783          */
1784         if (dst_dev) {
1785                 idev = __in6_dev_get(dst_dev);
1786                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1787                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1788                     (idev && idev->cnf.use_oif_addrs_only)) {
1789                         use_oif_addr = true;
1790                 }
1791         }
1792
1793         if (use_oif_addr) {
1794                 if (idev)
1795                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1796         } else {
1797                 const struct net_device *master;
1798                 int master_idx = 0;
1799
1800                 /* if dst_dev exists and is enslaved to an L3 device, then
1801                  * prefer addresses from dst_dev and then the master over
1802                  * any other enslaved devices in the L3 domain.
1803                  */
1804                 master = l3mdev_master_dev_rcu(dst_dev);
1805                 if (master) {
1806                         master_idx = master->ifindex;
1807
1808                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1809                                                             master, &dst,
1810                                                             scores, hiscore_idx);
1811
1812                         if (scores[hiscore_idx].ifa)
1813                                 goto out;
1814                 }
1815
1816                 for_each_netdev_rcu(net, dev) {
1817                         /* only consider addresses on devices in the
1818                          * same L3 domain
1819                          */
1820                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1821                                 continue;
1822                         idev = __in6_dev_get(dev);
1823                         if (!idev)
1824                                 continue;
1825                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1826                 }
1827         }
1828
1829 out:
1830         hiscore = &scores[hiscore_idx];
1831         if (!hiscore->ifa)
1832                 ret = -EADDRNOTAVAIL;
1833         else
1834                 *saddr = hiscore->ifa->addr;
1835
1836         rcu_read_unlock();
1837         return ret;
1838 }
1839 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1840
1841 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1842                       u32 banned_flags)
1843 {
1844         struct inet6_ifaddr *ifp;
1845         int err = -EADDRNOTAVAIL;
1846
1847         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1848                 if (ifp->scope > IFA_LINK)
1849                         break;
1850                 if (ifp->scope == IFA_LINK &&
1851                     !(ifp->flags & banned_flags)) {
1852                         *addr = ifp->addr;
1853                         err = 0;
1854                         break;
1855                 }
1856         }
1857         return err;
1858 }
1859
1860 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1861                     u32 banned_flags)
1862 {
1863         struct inet6_dev *idev;
1864         int err = -EADDRNOTAVAIL;
1865
1866         rcu_read_lock();
1867         idev = __in6_dev_get(dev);
1868         if (idev) {
1869                 read_lock_bh(&idev->lock);
1870                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1871                 read_unlock_bh(&idev->lock);
1872         }
1873         rcu_read_unlock();
1874         return err;
1875 }
1876
1877 static int ipv6_count_addresses(const struct inet6_dev *idev)
1878 {
1879         const struct inet6_ifaddr *ifp;
1880         int cnt = 0;
1881
1882         rcu_read_lock();
1883         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1884                 cnt++;
1885         rcu_read_unlock();
1886         return cnt;
1887 }
1888
1889 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1890                   const struct net_device *dev, int strict)
1891 {
1892         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1893                                        strict, IFA_F_TENTATIVE);
1894 }
1895 EXPORT_SYMBOL(ipv6_chk_addr);
1896
1897 /* device argument is used to find the L3 domain of interest. If
1898  * skip_dev_check is set, then the ifp device is not checked against
1899  * the passed in dev argument. So the 2 cases for addresses checks are:
1900  *   1. does the address exist in the L3 domain that dev is part of
1901  *      (skip_dev_check = true), or
1902  *
1903  *   2. does the address exist on the specific device
1904  *      (skip_dev_check = false)
1905  */
1906 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1907                             const struct net_device *dev, bool skip_dev_check,
1908                             int strict, u32 banned_flags)
1909 {
1910         unsigned int hash = inet6_addr_hash(net, addr);
1911         const struct net_device *l3mdev;
1912         struct inet6_ifaddr *ifp;
1913         u32 ifp_flags;
1914
1915         rcu_read_lock();
1916
1917         l3mdev = l3mdev_master_dev_rcu(dev);
1918         if (skip_dev_check)
1919                 dev = NULL;
1920
1921         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1922                 if (!net_eq(dev_net(ifp->idev->dev), net))
1923                         continue;
1924
1925                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1926                         continue;
1927
1928                 /* Decouple optimistic from tentative for evaluation here.
1929                  * Ban optimistic addresses explicitly, when required.
1930                  */
1931                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1932                             ? (ifp->flags&~IFA_F_TENTATIVE)
1933                             : ifp->flags;
1934                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1935                     !(ifp_flags&banned_flags) &&
1936                     (!dev || ifp->idev->dev == dev ||
1937                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1938                         rcu_read_unlock();
1939                         return 1;
1940                 }
1941         }
1942
1943         rcu_read_unlock();
1944         return 0;
1945 }
1946 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1947
1948
1949 /* Compares an address/prefix_len with addresses on device @dev.
1950  * If one is found it returns true.
1951  */
1952 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1953         const unsigned int prefix_len, struct net_device *dev)
1954 {
1955         const struct inet6_ifaddr *ifa;
1956         const struct inet6_dev *idev;
1957         bool ret = false;
1958
1959         rcu_read_lock();
1960         idev = __in6_dev_get(dev);
1961         if (idev) {
1962                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1963                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1964                         if (ret)
1965                                 break;
1966                 }
1967         }
1968         rcu_read_unlock();
1969
1970         return ret;
1971 }
1972 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1973
1974 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1975 {
1976         const struct inet6_ifaddr *ifa;
1977         const struct inet6_dev *idev;
1978         int     onlink;
1979
1980         onlink = 0;
1981         rcu_read_lock();
1982         idev = __in6_dev_get(dev);
1983         if (idev) {
1984                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1985                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1986                                                    ifa->prefix_len);
1987                         if (onlink)
1988                                 break;
1989                 }
1990         }
1991         rcu_read_unlock();
1992         return onlink;
1993 }
1994 EXPORT_SYMBOL(ipv6_chk_prefix);
1995
1996 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1997                                      struct net_device *dev, int strict)
1998 {
1999         unsigned int hash = inet6_addr_hash(net, addr);
2000         struct inet6_ifaddr *ifp, *result = NULL;
2001
2002         rcu_read_lock();
2003         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2004                 if (!net_eq(dev_net(ifp->idev->dev), net))
2005                         continue;
2006                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2007                         if (!dev || ifp->idev->dev == dev ||
2008                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2009                                 result = ifp;
2010                                 in6_ifa_hold(ifp);
2011                                 break;
2012                         }
2013                 }
2014         }
2015         rcu_read_unlock();
2016
2017         return result;
2018 }
2019
2020 /* Gets referenced address, destroys ifaddr */
2021
2022 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2023 {
2024         if (dad_failed)
2025                 ifp->flags |= IFA_F_DADFAILED;
2026
2027         if (ifp->flags&IFA_F_TEMPORARY) {
2028                 struct inet6_ifaddr *ifpub;
2029                 spin_lock_bh(&ifp->lock);
2030                 ifpub = ifp->ifpub;
2031                 if (ifpub) {
2032                         in6_ifa_hold(ifpub);
2033                         spin_unlock_bh(&ifp->lock);
2034                         ipv6_create_tempaddr(ifpub, ifp, true);
2035                         in6_ifa_put(ifpub);
2036                 } else {
2037                         spin_unlock_bh(&ifp->lock);
2038                 }
2039                 ipv6_del_addr(ifp);
2040         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2041                 spin_lock_bh(&ifp->lock);
2042                 addrconf_del_dad_work(ifp);
2043                 ifp->flags |= IFA_F_TENTATIVE;
2044                 if (dad_failed)
2045                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2046                 spin_unlock_bh(&ifp->lock);
2047                 if (dad_failed)
2048                         ipv6_ifa_notify(0, ifp);
2049                 in6_ifa_put(ifp);
2050         } else {
2051                 ipv6_del_addr(ifp);
2052         }
2053 }
2054
2055 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2056 {
2057         int err = -ENOENT;
2058
2059         spin_lock_bh(&ifp->lock);
2060         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2061                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2062                 err = 0;
2063         }
2064         spin_unlock_bh(&ifp->lock);
2065
2066         return err;
2067 }
2068
2069 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2070 {
2071         struct inet6_dev *idev = ifp->idev;
2072         struct net *net = dev_net(ifp->idev->dev);
2073
2074         if (addrconf_dad_end(ifp)) {
2075                 in6_ifa_put(ifp);
2076                 return;
2077         }
2078
2079         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2080                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2081
2082         spin_lock_bh(&ifp->lock);
2083
2084         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2085                 struct in6_addr new_addr;
2086                 struct inet6_ifaddr *ifp2;
2087                 int retries = ifp->stable_privacy_retry + 1;
2088                 struct ifa6_config cfg = {
2089                         .pfx = &new_addr,
2090                         .plen = ifp->prefix_len,
2091                         .ifa_flags = ifp->flags,
2092                         .valid_lft = ifp->valid_lft,
2093                         .preferred_lft = ifp->prefered_lft,
2094                         .scope = ifp->scope,
2095                 };
2096
2097                 if (retries > net->ipv6.sysctl.idgen_retries) {
2098                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2099                                              ifp->idev->dev->name);
2100                         goto errdad;
2101                 }
2102
2103                 new_addr = ifp->addr;
2104                 if (ipv6_generate_stable_address(&new_addr, retries,
2105                                                  idev))
2106                         goto errdad;
2107
2108                 spin_unlock_bh(&ifp->lock);
2109
2110                 if (idev->cnf.max_addresses &&
2111                     ipv6_count_addresses(idev) >=
2112                     idev->cnf.max_addresses)
2113                         goto lock_errdad;
2114
2115                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2116                                      ifp->idev->dev->name);
2117
2118                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2119                 if (IS_ERR(ifp2))
2120                         goto lock_errdad;
2121
2122                 spin_lock_bh(&ifp2->lock);
2123                 ifp2->stable_privacy_retry = retries;
2124                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2125                 spin_unlock_bh(&ifp2->lock);
2126
2127                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2128                 in6_ifa_put(ifp2);
2129 lock_errdad:
2130                 spin_lock_bh(&ifp->lock);
2131         }
2132
2133 errdad:
2134         /* transition from _POSTDAD to _ERRDAD */
2135         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2136         spin_unlock_bh(&ifp->lock);
2137
2138         addrconf_mod_dad_work(ifp, 0);
2139         in6_ifa_put(ifp);
2140 }
2141
2142 /* Join to solicited addr multicast group.
2143  * caller must hold RTNL */
2144 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2145 {
2146         struct in6_addr maddr;
2147
2148         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2149                 return;
2150
2151         addrconf_addr_solict_mult(addr, &maddr);
2152         ipv6_dev_mc_inc(dev, &maddr);
2153 }
2154
2155 /* caller must hold RTNL */
2156 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2157 {
2158         struct in6_addr maddr;
2159
2160         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2161                 return;
2162
2163         addrconf_addr_solict_mult(addr, &maddr);
2164         __ipv6_dev_mc_dec(idev, &maddr);
2165 }
2166
2167 /* caller must hold RTNL */
2168 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2169 {
2170         struct in6_addr addr;
2171
2172         if (ifp->prefix_len >= 127) /* RFC 6164 */
2173                 return;
2174         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2175         if (ipv6_addr_any(&addr))
2176                 return;
2177         __ipv6_dev_ac_inc(ifp->idev, &addr);
2178 }
2179
2180 /* caller must hold RTNL */
2181 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2182 {
2183         struct in6_addr addr;
2184
2185         if (ifp->prefix_len >= 127) /* RFC 6164 */
2186                 return;
2187         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2188         if (ipv6_addr_any(&addr))
2189                 return;
2190         __ipv6_dev_ac_dec(ifp->idev, &addr);
2191 }
2192
2193 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2194 {
2195         switch (dev->addr_len) {
2196         case ETH_ALEN:
2197                 memcpy(eui, dev->dev_addr, 3);
2198                 eui[3] = 0xFF;
2199                 eui[4] = 0xFE;
2200                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2201                 break;
2202         case EUI64_ADDR_LEN:
2203                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2204                 eui[0] ^= 2;
2205                 break;
2206         default:
2207                 return -1;
2208         }
2209
2210         return 0;
2211 }
2212
2213 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2214 {
2215         union fwnet_hwaddr *ha;
2216
2217         if (dev->addr_len != FWNET_ALEN)
2218                 return -1;
2219
2220         ha = (union fwnet_hwaddr *)dev->dev_addr;
2221
2222         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2223         eui[0] ^= 2;
2224         return 0;
2225 }
2226
2227 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2228 {
2229         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2230         if (dev->addr_len != ARCNET_ALEN)
2231                 return -1;
2232         memset(eui, 0, 7);
2233         eui[7] = *(u8 *)dev->dev_addr;
2234         return 0;
2235 }
2236
2237 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2238 {
2239         if (dev->addr_len != INFINIBAND_ALEN)
2240                 return -1;
2241         memcpy(eui, dev->dev_addr + 12, 8);
2242         eui[0] |= 2;
2243         return 0;
2244 }
2245
2246 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2247 {
2248         if (addr == 0)
2249                 return -1;
2250         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2251                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2252                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2253                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2254                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2255                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2256         eui[1] = 0;
2257         eui[2] = 0x5E;
2258         eui[3] = 0xFE;
2259         memcpy(eui + 4, &addr, 4);
2260         return 0;
2261 }
2262
2263 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2264 {
2265         if (dev->priv_flags & IFF_ISATAP)
2266                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2267         return -1;
2268 }
2269
2270 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2271 {
2272         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2273 }
2274
2275 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2276 {
2277         memcpy(eui, dev->perm_addr, 3);
2278         memcpy(eui + 5, dev->perm_addr + 3, 3);
2279         eui[3] = 0xFF;
2280         eui[4] = 0xFE;
2281         eui[0] ^= 2;
2282         return 0;
2283 }
2284
2285 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2286 {
2287         switch (dev->type) {
2288         case ARPHRD_ETHER:
2289         case ARPHRD_FDDI:
2290                 return addrconf_ifid_eui48(eui, dev);
2291         case ARPHRD_ARCNET:
2292                 return addrconf_ifid_arcnet(eui, dev);
2293         case ARPHRD_INFINIBAND:
2294                 return addrconf_ifid_infiniband(eui, dev);
2295         case ARPHRD_SIT:
2296                 return addrconf_ifid_sit(eui, dev);
2297         case ARPHRD_IPGRE:
2298         case ARPHRD_TUNNEL:
2299                 return addrconf_ifid_gre(eui, dev);
2300         case ARPHRD_6LOWPAN:
2301                 return addrconf_ifid_6lowpan(eui, dev);
2302         case ARPHRD_IEEE1394:
2303                 return addrconf_ifid_ieee1394(eui, dev);
2304         case ARPHRD_TUNNEL6:
2305         case ARPHRD_IP6GRE:
2306         case ARPHRD_RAWIP:
2307                 return addrconf_ifid_ip6tnl(eui, dev);
2308         }
2309         return -1;
2310 }
2311
2312 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2313 {
2314         int err = -1;
2315         struct inet6_ifaddr *ifp;
2316
2317         read_lock_bh(&idev->lock);
2318         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2319                 if (ifp->scope > IFA_LINK)
2320                         break;
2321                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2322                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2323                         err = 0;
2324                         break;
2325                 }
2326         }
2327         read_unlock_bh(&idev->lock);
2328         return err;
2329 }
2330
2331 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2332 static void ipv6_regen_rndid(struct inet6_dev *idev)
2333 {
2334 regen:
2335         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2336         idev->rndid[0] &= ~0x02;
2337
2338         /*
2339          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2340          * check if generated address is not inappropriate
2341          *
2342          *  - Reserved subnet anycast (RFC 2526)
2343          *      11111101 11....11 1xxxxxxx
2344          *  - ISATAP (RFC4214) 6.1
2345          *      00-00-5E-FE-xx-xx-xx-xx
2346          *  - value 0
2347          *  - XXX: already assigned to an address on the device
2348          */
2349         if (idev->rndid[0] == 0xfd &&
2350             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2351             (idev->rndid[7]&0x80))
2352                 goto regen;
2353         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2354                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2355                         goto regen;
2356                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2357                         goto regen;
2358         }
2359 }
2360
2361 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2362 {
2363         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2364                 ipv6_regen_rndid(idev);
2365 }
2366
2367 /*
2368  *      Add prefix route.
2369  */
2370
2371 static void
2372 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2373                       struct net_device *dev, unsigned long expires,
2374                       u32 flags, gfp_t gfp_flags)
2375 {
2376         struct fib6_config cfg = {
2377                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2378                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2379                 .fc_ifindex = dev->ifindex,
2380                 .fc_expires = expires,
2381                 .fc_dst_len = plen,
2382                 .fc_flags = RTF_UP | flags,
2383                 .fc_nlinfo.nl_net = dev_net(dev),
2384                 .fc_protocol = RTPROT_KERNEL,
2385                 .fc_type = RTN_UNICAST,
2386         };
2387
2388         cfg.fc_dst = *pfx;
2389
2390         /* Prevent useless cloning on PtP SIT.
2391            This thing is done here expecting that the whole
2392            class of non-broadcast devices need not cloning.
2393          */
2394 #if IS_ENABLED(CONFIG_IPV6_SIT)
2395         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2396                 cfg.fc_flags |= RTF_NONEXTHOP;
2397 #endif
2398
2399         ip6_route_add(&cfg, gfp_flags, NULL);
2400 }
2401
2402
2403 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2404                                                   int plen,
2405                                                   const struct net_device *dev,
2406                                                   u32 flags, u32 noflags,
2407                                                   bool no_gw)
2408 {
2409         struct fib6_node *fn;
2410         struct fib6_info *rt = NULL;
2411         struct fib6_table *table;
2412         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2413
2414         table = fib6_get_table(dev_net(dev), tb_id);
2415         if (!table)
2416                 return NULL;
2417
2418         rcu_read_lock();
2419         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2420         if (!fn)
2421                 goto out;
2422
2423         for_each_fib6_node_rt_rcu(fn) {
2424                 if (rt->fib6_nh.fib_nh_dev->ifindex != dev->ifindex)
2425                         continue;
2426                 if (no_gw && rt->fib6_nh.fib_nh_gw_family)
2427                         continue;
2428                 if ((rt->fib6_flags & flags) != flags)
2429                         continue;
2430                 if ((rt->fib6_flags & noflags) != 0)
2431                         continue;
2432                 if (!fib6_info_hold_safe(rt))
2433                         continue;
2434                 break;
2435         }
2436 out:
2437         rcu_read_unlock();
2438         return rt;
2439 }
2440
2441
2442 /* Create "default" multicast route to the interface */
2443
2444 static void addrconf_add_mroute(struct net_device *dev)
2445 {
2446         struct fib6_config cfg = {
2447                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2448                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2449                 .fc_ifindex = dev->ifindex,
2450                 .fc_dst_len = 8,
2451                 .fc_flags = RTF_UP,
2452                 .fc_type = RTN_UNICAST,
2453                 .fc_nlinfo.nl_net = dev_net(dev),
2454         };
2455
2456         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2457
2458         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2459 }
2460
2461 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2462 {
2463         struct inet6_dev *idev;
2464
2465         ASSERT_RTNL();
2466
2467         idev = ipv6_find_idev(dev);
2468         if (!idev)
2469                 return ERR_PTR(-ENOBUFS);
2470
2471         if (idev->cnf.disable_ipv6)
2472                 return ERR_PTR(-EACCES);
2473
2474         /* Add default multicast route */
2475         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2476                 addrconf_add_mroute(dev);
2477
2478         return idev;
2479 }
2480
2481 static void manage_tempaddrs(struct inet6_dev *idev,
2482                              struct inet6_ifaddr *ifp,
2483                              __u32 valid_lft, __u32 prefered_lft,
2484                              bool create, unsigned long now)
2485 {
2486         u32 flags;
2487         struct inet6_ifaddr *ift;
2488
2489         read_lock_bh(&idev->lock);
2490         /* update all temporary addresses in the list */
2491         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2492                 int age, max_valid, max_prefered;
2493
2494                 if (ifp != ift->ifpub)
2495                         continue;
2496
2497                 /* RFC 4941 section 3.3:
2498                  * If a received option will extend the lifetime of a public
2499                  * address, the lifetimes of temporary addresses should
2500                  * be extended, subject to the overall constraint that no
2501                  * temporary addresses should ever remain "valid" or "preferred"
2502                  * for a time longer than (TEMP_VALID_LIFETIME) or
2503                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2504                  */
2505                 age = (now - ift->cstamp) / HZ;
2506                 max_valid = idev->cnf.temp_valid_lft - age;
2507                 if (max_valid < 0)
2508                         max_valid = 0;
2509
2510                 max_prefered = idev->cnf.temp_prefered_lft -
2511                                idev->desync_factor - age;
2512                 if (max_prefered < 0)
2513                         max_prefered = 0;
2514
2515                 if (valid_lft > max_valid)
2516                         valid_lft = max_valid;
2517
2518                 if (prefered_lft > max_prefered)
2519                         prefered_lft = max_prefered;
2520
2521                 spin_lock(&ift->lock);
2522                 flags = ift->flags;
2523                 ift->valid_lft = valid_lft;
2524                 ift->prefered_lft = prefered_lft;
2525                 ift->tstamp = now;
2526                 if (prefered_lft > 0)
2527                         ift->flags &= ~IFA_F_DEPRECATED;
2528
2529                 spin_unlock(&ift->lock);
2530                 if (!(flags&IFA_F_TENTATIVE))
2531                         ipv6_ifa_notify(0, ift);
2532         }
2533
2534         if ((create || list_empty(&idev->tempaddr_list)) &&
2535             idev->cnf.use_tempaddr > 0) {
2536                 /* When a new public address is created as described
2537                  * in [ADDRCONF], also create a new temporary address.
2538                  * Also create a temporary address if it's enabled but
2539                  * no temporary address currently exists.
2540                  */
2541                 read_unlock_bh(&idev->lock);
2542                 ipv6_create_tempaddr(ifp, NULL, false);
2543         } else {
2544                 read_unlock_bh(&idev->lock);
2545         }
2546 }
2547
2548 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2549 {
2550         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2551                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2552 }
2553
2554 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2555                                  const struct prefix_info *pinfo,
2556                                  struct inet6_dev *in6_dev,
2557                                  const struct in6_addr *addr, int addr_type,
2558                                  u32 addr_flags, bool sllao, bool tokenized,
2559                                  __u32 valid_lft, u32 prefered_lft)
2560 {
2561         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2562         int create = 0, update_lft = 0;
2563
2564         if (!ifp && valid_lft) {
2565                 int max_addresses = in6_dev->cnf.max_addresses;
2566                 struct ifa6_config cfg = {
2567                         .pfx = addr,
2568                         .plen = pinfo->prefix_len,
2569                         .ifa_flags = addr_flags,
2570                         .valid_lft = valid_lft,
2571                         .preferred_lft = prefered_lft,
2572                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2573                 };
2574
2575 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2576                 if ((net->ipv6.devconf_all->optimistic_dad ||
2577                      in6_dev->cnf.optimistic_dad) &&
2578                     !net->ipv6.devconf_all->forwarding && sllao)
2579                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2580 #endif
2581
2582                 /* Do not allow to create too much of autoconfigured
2583                  * addresses; this would be too easy way to crash kernel.
2584                  */
2585                 if (!max_addresses ||
2586                     ipv6_count_addresses(in6_dev) < max_addresses)
2587                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2588
2589                 if (IS_ERR_OR_NULL(ifp))
2590                         return -1;
2591
2592                 create = 1;
2593                 spin_lock_bh(&ifp->lock);
2594                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2595                 ifp->cstamp = jiffies;
2596                 ifp->tokenized = tokenized;
2597                 spin_unlock_bh(&ifp->lock);
2598                 addrconf_dad_start(ifp);
2599         }
2600
2601         if (ifp) {
2602                 u32 flags;
2603                 unsigned long now;
2604                 u32 stored_lft;
2605
2606                 /* update lifetime (RFC2462 5.5.3 e) */
2607                 spin_lock_bh(&ifp->lock);
2608                 now = jiffies;
2609                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2610                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2611                 else
2612                         stored_lft = 0;
2613                 if (!create && stored_lft) {
2614                         const u32 minimum_lft = min_t(u32,
2615                                 stored_lft, MIN_VALID_LIFETIME);
2616                         valid_lft = max(valid_lft, minimum_lft);
2617
2618                         /* RFC4862 Section 5.5.3e:
2619                          * "Note that the preferred lifetime of the
2620                          *  corresponding address is always reset to
2621                          *  the Preferred Lifetime in the received
2622                          *  Prefix Information option, regardless of
2623                          *  whether the valid lifetime is also reset or
2624                          *  ignored."
2625                          *
2626                          * So we should always update prefered_lft here.
2627                          */
2628                         update_lft = 1;
2629                 }
2630
2631                 if (update_lft) {
2632                         ifp->valid_lft = valid_lft;
2633                         ifp->prefered_lft = prefered_lft;
2634                         ifp->tstamp = now;
2635                         flags = ifp->flags;
2636                         ifp->flags &= ~IFA_F_DEPRECATED;
2637                         spin_unlock_bh(&ifp->lock);
2638
2639                         if (!(flags&IFA_F_TENTATIVE))
2640                                 ipv6_ifa_notify(0, ifp);
2641                 } else
2642                         spin_unlock_bh(&ifp->lock);
2643
2644                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2645                                  create, now);
2646
2647                 in6_ifa_put(ifp);
2648                 addrconf_verify();
2649         }
2650
2651         return 0;
2652 }
2653 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2654
2655 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2656 {
2657         struct prefix_info *pinfo;
2658         __u32 valid_lft;
2659         __u32 prefered_lft;
2660         int addr_type, err;
2661         u32 addr_flags = 0;
2662         struct inet6_dev *in6_dev;
2663         struct net *net = dev_net(dev);
2664
2665         pinfo = (struct prefix_info *) opt;
2666
2667         if (len < sizeof(struct prefix_info)) {
2668                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2669                 return;
2670         }
2671
2672         /*
2673          *      Validation checks ([ADDRCONF], page 19)
2674          */
2675
2676         addr_type = ipv6_addr_type(&pinfo->prefix);
2677
2678         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2679                 return;
2680
2681         valid_lft = ntohl(pinfo->valid);
2682         prefered_lft = ntohl(pinfo->prefered);
2683
2684         if (prefered_lft > valid_lft) {
2685                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2686                 return;
2687         }
2688
2689         in6_dev = in6_dev_get(dev);
2690
2691         if (!in6_dev) {
2692                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2693                                     dev->name);
2694                 return;
2695         }
2696
2697         /*
2698          *      Two things going on here:
2699          *      1) Add routes for on-link prefixes
2700          *      2) Configure prefixes with the auto flag set
2701          */
2702
2703         if (pinfo->onlink) {
2704                 struct fib6_info *rt;
2705                 unsigned long rt_expires;
2706
2707                 /* Avoid arithmetic overflow. Really, we could
2708                  * save rt_expires in seconds, likely valid_lft,
2709                  * but it would require division in fib gc, that it
2710                  * not good.
2711                  */
2712                 if (HZ > USER_HZ)
2713                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2714                 else
2715                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2716
2717                 if (addrconf_finite_timeout(rt_expires))
2718                         rt_expires *= HZ;
2719
2720                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2721                                                pinfo->prefix_len,
2722                                                dev,
2723                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2724                                                RTF_DEFAULT, true);
2725
2726                 if (rt) {
2727                         /* Autoconf prefix route */
2728                         if (valid_lft == 0) {
2729                                 ip6_del_rt(net, rt);
2730                                 rt = NULL;
2731                         } else if (addrconf_finite_timeout(rt_expires)) {
2732                                 /* not infinity */
2733                                 fib6_set_expires(rt, jiffies + rt_expires);
2734                         } else {
2735                                 fib6_clean_expires(rt);
2736                         }
2737                 } else if (valid_lft) {
2738                         clock_t expires = 0;
2739                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2740                         if (addrconf_finite_timeout(rt_expires)) {
2741                                 /* not infinity */
2742                                 flags |= RTF_EXPIRES;
2743                                 expires = jiffies_to_clock_t(rt_expires);
2744                         }
2745                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2746                                               0, dev, expires, flags,
2747                                               GFP_ATOMIC);
2748                 }
2749                 fib6_info_release(rt);
2750         }
2751
2752         /* Try to figure out our local address for this prefix */
2753
2754         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2755                 struct in6_addr addr;
2756                 bool tokenized = false, dev_addr_generated = false;
2757
2758                 if (pinfo->prefix_len == 64) {
2759                         memcpy(&addr, &pinfo->prefix, 8);
2760
2761                         if (!ipv6_addr_any(&in6_dev->token)) {
2762                                 read_lock_bh(&in6_dev->lock);
2763                                 memcpy(addr.s6_addr + 8,
2764                                        in6_dev->token.s6_addr + 8, 8);
2765                                 read_unlock_bh(&in6_dev->lock);
2766                                 tokenized = true;
2767                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2768                                    !ipv6_generate_stable_address(&addr, 0,
2769                                                                  in6_dev)) {
2770                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2771                                 goto ok;
2772                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2773                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2774                                 goto put;
2775                         } else {
2776                                 dev_addr_generated = true;
2777                         }
2778                         goto ok;
2779                 }
2780                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2781                                     pinfo->prefix_len);
2782                 goto put;
2783
2784 ok:
2785                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2786                                                    &addr, addr_type,
2787                                                    addr_flags, sllao,
2788                                                    tokenized, valid_lft,
2789                                                    prefered_lft);
2790                 if (err)
2791                         goto put;
2792
2793                 /* Ignore error case here because previous prefix add addr was
2794                  * successful which will be notified.
2795                  */
2796                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2797                                               addr_type, addr_flags, sllao,
2798                                               tokenized, valid_lft,
2799                                               prefered_lft,
2800                                               dev_addr_generated);
2801         }
2802         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2803 put:
2804         in6_dev_put(in6_dev);
2805 }
2806
2807 /*
2808  *      Set destination address.
2809  *      Special case for SIT interfaces where we create a new "virtual"
2810  *      device.
2811  */
2812 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2813 {
2814         struct in6_ifreq ireq;
2815         struct net_device *dev;
2816         int err = -EINVAL;
2817
2818         rtnl_lock();
2819
2820         err = -EFAULT;
2821         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2822                 goto err_exit;
2823
2824         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2825
2826         err = -ENODEV;
2827         if (!dev)
2828                 goto err_exit;
2829
2830 #if IS_ENABLED(CONFIG_IPV6_SIT)
2831         if (dev->type == ARPHRD_SIT) {
2832                 const struct net_device_ops *ops = dev->netdev_ops;
2833                 struct ifreq ifr;
2834                 struct ip_tunnel_parm p;
2835
2836                 err = -EADDRNOTAVAIL;
2837                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2838                         goto err_exit;
2839
2840                 memset(&p, 0, sizeof(p));
2841                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2842                 p.iph.saddr = 0;
2843                 p.iph.version = 4;
2844                 p.iph.ihl = 5;
2845                 p.iph.protocol = IPPROTO_IPV6;
2846                 p.iph.ttl = 64;
2847                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2848
2849                 if (ops->ndo_do_ioctl) {
2850                         mm_segment_t oldfs = get_fs();
2851
2852                         set_fs(KERNEL_DS);
2853                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2854                         set_fs(oldfs);
2855                 } else
2856                         err = -EOPNOTSUPP;
2857
2858                 if (err == 0) {
2859                         err = -ENOBUFS;
2860                         dev = __dev_get_by_name(net, p.name);
2861                         if (!dev)
2862                                 goto err_exit;
2863                         err = dev_open(dev, NULL);
2864                 }
2865         }
2866 #endif
2867
2868 err_exit:
2869         rtnl_unlock();
2870         return err;
2871 }
2872
2873 static int ipv6_mc_config(struct sock *sk, bool join,
2874                           const struct in6_addr *addr, int ifindex)
2875 {
2876         int ret;
2877
2878         ASSERT_RTNL();
2879
2880         lock_sock(sk);
2881         if (join)
2882                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2883         else
2884                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2885         release_sock(sk);
2886
2887         return ret;
2888 }
2889
2890 /*
2891  *      Manual configuration of address on an interface
2892  */
2893 static int inet6_addr_add(struct net *net, int ifindex,
2894                           struct ifa6_config *cfg,
2895                           struct netlink_ext_ack *extack)
2896 {
2897         struct inet6_ifaddr *ifp;
2898         struct inet6_dev *idev;
2899         struct net_device *dev;
2900         unsigned long timeout;
2901         clock_t expires;
2902         u32 flags;
2903
2904         ASSERT_RTNL();
2905
2906         if (cfg->plen > 128)
2907                 return -EINVAL;
2908
2909         /* check the lifetime */
2910         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2911                 return -EINVAL;
2912
2913         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2914                 return -EINVAL;
2915
2916         dev = __dev_get_by_index(net, ifindex);
2917         if (!dev)
2918                 return -ENODEV;
2919
2920         idev = addrconf_add_dev(dev);
2921         if (IS_ERR(idev))
2922                 return PTR_ERR(idev);
2923
2924         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2925                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2926                                          true, cfg->pfx, ifindex);
2927
2928                 if (ret < 0)
2929                         return ret;
2930         }
2931
2932         cfg->scope = ipv6_addr_scope(cfg->pfx);
2933
2934         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2935         if (addrconf_finite_timeout(timeout)) {
2936                 expires = jiffies_to_clock_t(timeout * HZ);
2937                 cfg->valid_lft = timeout;
2938                 flags = RTF_EXPIRES;
2939         } else {
2940                 expires = 0;
2941                 flags = 0;
2942                 cfg->ifa_flags |= IFA_F_PERMANENT;
2943         }
2944
2945         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2946         if (addrconf_finite_timeout(timeout)) {
2947                 if (timeout == 0)
2948                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2949                 cfg->preferred_lft = timeout;
2950         }
2951
2952         ifp = ipv6_add_addr(idev, cfg, true, extack);
2953         if (!IS_ERR(ifp)) {
2954                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2955                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2956                                               ifp->rt_priority, dev, expires,
2957                                               flags, GFP_KERNEL);
2958                 }
2959
2960                 /* Send a netlink notification if DAD is enabled and
2961                  * optimistic flag is not set
2962                  */
2963                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2964                         ipv6_ifa_notify(0, ifp);
2965                 /*
2966                  * Note that section 3.1 of RFC 4429 indicates
2967                  * that the Optimistic flag should not be set for
2968                  * manually configured addresses
2969                  */
2970                 addrconf_dad_start(ifp);
2971                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2972                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2973                                          cfg->preferred_lft, true, jiffies);
2974                 in6_ifa_put(ifp);
2975                 addrconf_verify_rtnl();
2976                 return 0;
2977         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2978                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2979                                cfg->pfx, ifindex);
2980         }
2981
2982         return PTR_ERR(ifp);
2983 }
2984
2985 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2986                           const struct in6_addr *pfx, unsigned int plen)
2987 {
2988         struct inet6_ifaddr *ifp;
2989         struct inet6_dev *idev;
2990         struct net_device *dev;
2991
2992         if (plen > 128)
2993                 return -EINVAL;
2994
2995         dev = __dev_get_by_index(net, ifindex);
2996         if (!dev)
2997                 return -ENODEV;
2998
2999         idev = __in6_dev_get(dev);
3000         if (!idev)
3001                 return -ENXIO;
3002
3003         read_lock_bh(&idev->lock);
3004         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3005                 if (ifp->prefix_len == plen &&
3006                     ipv6_addr_equal(pfx, &ifp->addr)) {
3007                         in6_ifa_hold(ifp);
3008                         read_unlock_bh(&idev->lock);
3009
3010                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3011                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3012                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3013                                                  jiffies);
3014                         ipv6_del_addr(ifp);
3015                         addrconf_verify_rtnl();
3016                         if (ipv6_addr_is_multicast(pfx)) {
3017                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3018                                                false, pfx, dev->ifindex);
3019                         }
3020                         return 0;
3021                 }
3022         }
3023         read_unlock_bh(&idev->lock);
3024         return -EADDRNOTAVAIL;
3025 }
3026
3027
3028 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3029 {
3030         struct ifa6_config cfg = {
3031                 .ifa_flags = IFA_F_PERMANENT,
3032                 .preferred_lft = INFINITY_LIFE_TIME,
3033                 .valid_lft = INFINITY_LIFE_TIME,
3034         };
3035         struct in6_ifreq ireq;
3036         int err;
3037
3038         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3039                 return -EPERM;
3040
3041         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3042                 return -EFAULT;
3043
3044         cfg.pfx = &ireq.ifr6_addr;
3045         cfg.plen = ireq.ifr6_prefixlen;
3046
3047         rtnl_lock();
3048         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3049         rtnl_unlock();
3050         return err;
3051 }
3052
3053 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3054 {
3055         struct in6_ifreq ireq;
3056         int err;
3057
3058         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3059                 return -EPERM;
3060
3061         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3062                 return -EFAULT;
3063
3064         rtnl_lock();
3065         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3066                              ireq.ifr6_prefixlen);
3067         rtnl_unlock();
3068         return err;
3069 }
3070
3071 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3072                      int plen, int scope)
3073 {
3074         struct inet6_ifaddr *ifp;
3075         struct ifa6_config cfg = {
3076                 .pfx = addr,
3077                 .plen = plen,
3078                 .ifa_flags = IFA_F_PERMANENT,
3079                 .valid_lft = INFINITY_LIFE_TIME,
3080                 .preferred_lft = INFINITY_LIFE_TIME,
3081                 .scope = scope
3082         };
3083
3084         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3085         if (!IS_ERR(ifp)) {
3086                 spin_lock_bh(&ifp->lock);
3087                 ifp->flags &= ~IFA_F_TENTATIVE;
3088                 spin_unlock_bh(&ifp->lock);
3089                 rt_genid_bump_ipv6(dev_net(idev->dev));
3090                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3091                 in6_ifa_put(ifp);
3092         }
3093 }
3094
3095 #if IS_ENABLED(CONFIG_IPV6_SIT)
3096 static void sit_add_v4_addrs(struct inet6_dev *idev)
3097 {
3098         struct in6_addr addr;
3099         struct net_device *dev;
3100         struct net *net = dev_net(idev->dev);
3101         int scope, plen;
3102         u32 pflags = 0;
3103
3104         ASSERT_RTNL();
3105
3106         memset(&addr, 0, sizeof(struct in6_addr));
3107         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3108
3109         if (idev->dev->flags&IFF_POINTOPOINT) {
3110                 addr.s6_addr32[0] = htonl(0xfe800000);
3111                 scope = IFA_LINK;
3112                 plen = 64;
3113         } else {
3114                 scope = IPV6_ADDR_COMPATv4;
3115                 plen = 96;
3116                 pflags |= RTF_NONEXTHOP;
3117         }
3118
3119         if (addr.s6_addr32[3]) {
3120                 add_addr(idev, &addr, plen, scope);
3121                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3122                                       GFP_KERNEL);
3123                 return;
3124         }
3125
3126         for_each_netdev(net, dev) {
3127                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3128                 if (in_dev && (dev->flags & IFF_UP)) {
3129                         struct in_ifaddr *ifa;
3130
3131                         int flag = scope;
3132
3133                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3134
3135                                 addr.s6_addr32[3] = ifa->ifa_local;
3136
3137                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3138                                         continue;
3139                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3140                                         if (idev->dev->flags&IFF_POINTOPOINT)
3141                                                 continue;
3142                                         flag |= IFA_HOST;
3143                                 }
3144
3145                                 add_addr(idev, &addr, plen, flag);
3146                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3147                                                       0, pflags, GFP_KERNEL);
3148                         }
3149                 }
3150         }
3151 }
3152 #endif
3153
3154 static void init_loopback(struct net_device *dev)
3155 {
3156         struct inet6_dev  *idev;
3157
3158         /* ::1 */
3159
3160         ASSERT_RTNL();
3161
3162         idev = ipv6_find_idev(dev);
3163         if (!idev) {
3164                 pr_debug("%s: add_dev failed\n", __func__);
3165                 return;
3166         }
3167
3168         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3169 }
3170
3171 void addrconf_add_linklocal(struct inet6_dev *idev,
3172                             const struct in6_addr *addr, u32 flags)
3173 {
3174         struct ifa6_config cfg = {
3175                 .pfx = addr,
3176                 .plen = 64,
3177                 .ifa_flags = flags | IFA_F_PERMANENT,
3178                 .valid_lft = INFINITY_LIFE_TIME,
3179                 .preferred_lft = INFINITY_LIFE_TIME,
3180                 .scope = IFA_LINK
3181         };
3182         struct inet6_ifaddr *ifp;
3183
3184 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3185         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3186              idev->cnf.optimistic_dad) &&
3187             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3188                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3189 #endif
3190
3191         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3192         if (!IS_ERR(ifp)) {
3193                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3194                                       0, 0, GFP_ATOMIC);
3195                 addrconf_dad_start(ifp);
3196                 in6_ifa_put(ifp);
3197         }
3198 }
3199 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3200
3201 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3202 {
3203         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3204                 return true;
3205
3206         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3207             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3208                 return true;
3209
3210         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3211             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3212                 return true;
3213
3214         return false;
3215 }
3216
3217 static int ipv6_generate_stable_address(struct in6_addr *address,
3218                                         u8 dad_count,
3219                                         const struct inet6_dev *idev)
3220 {
3221         static DEFINE_SPINLOCK(lock);
3222         static __u32 digest[SHA_DIGEST_WORDS];
3223         static __u32 workspace[SHA_WORKSPACE_WORDS];
3224
3225         static union {
3226                 char __data[SHA_MESSAGE_BYTES];
3227                 struct {
3228                         struct in6_addr secret;
3229                         __be32 prefix[2];
3230                         unsigned char hwaddr[MAX_ADDR_LEN];
3231                         u8 dad_count;
3232                 } __packed;
3233         } data;
3234
3235         struct in6_addr secret;
3236         struct in6_addr temp;
3237         struct net *net = dev_net(idev->dev);
3238
3239         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3240
3241         if (idev->cnf.stable_secret.initialized)
3242                 secret = idev->cnf.stable_secret.secret;
3243         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3244                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3245         else
3246                 return -1;
3247
3248 retry:
3249         spin_lock_bh(&lock);
3250
3251         sha_init(digest);
3252         memset(&data, 0, sizeof(data));
3253         memset(workspace, 0, sizeof(workspace));
3254         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3255         data.prefix[0] = address->s6_addr32[0];
3256         data.prefix[1] = address->s6_addr32[1];
3257         data.secret = secret;
3258         data.dad_count = dad_count;
3259
3260         sha_transform(digest, data.__data, workspace);
3261
3262         temp = *address;
3263         temp.s6_addr32[2] = (__force __be32)digest[0];
3264         temp.s6_addr32[3] = (__force __be32)digest[1];
3265
3266         spin_unlock_bh(&lock);
3267
3268         if (ipv6_reserved_interfaceid(temp)) {
3269                 dad_count++;
3270                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3271                         return -1;
3272                 goto retry;
3273         }
3274
3275         *address = temp;
3276         return 0;
3277 }
3278
3279 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3280 {
3281         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3282
3283         if (s->initialized)
3284                 return;
3285         s = &idev->cnf.stable_secret;
3286         get_random_bytes(&s->secret, sizeof(s->secret));
3287         s->initialized = true;
3288 }
3289
3290 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3291 {
3292         struct in6_addr addr;
3293
3294         /* no link local addresses on L3 master devices */
3295         if (netif_is_l3_master(idev->dev))
3296                 return;
3297
3298         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3299
3300         switch (idev->cnf.addr_gen_mode) {
3301         case IN6_ADDR_GEN_MODE_RANDOM:
3302                 ipv6_gen_mode_random_init(idev);
3303                 /* fallthrough */
3304         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3305                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3306                         addrconf_add_linklocal(idev, &addr,
3307                                                IFA_F_STABLE_PRIVACY);
3308                 else if (prefix_route)
3309                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3310                                               0, 0, GFP_KERNEL);
3311                 break;
3312         case IN6_ADDR_GEN_MODE_EUI64:
3313                 /* addrconf_add_linklocal also adds a prefix_route and we
3314                  * only need to care about prefix routes if ipv6_generate_eui64
3315                  * couldn't generate one.
3316                  */
3317                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3318                         addrconf_add_linklocal(idev, &addr, 0);
3319                 else if (prefix_route)
3320                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3321                                               0, 0, GFP_KERNEL);
3322                 break;
3323         case IN6_ADDR_GEN_MODE_NONE:
3324         default:
3325                 /* will not add any link local address */
3326                 break;
3327         }
3328 }
3329
3330 static void addrconf_dev_config(struct net_device *dev)
3331 {
3332         struct inet6_dev *idev;
3333
3334         ASSERT_RTNL();
3335
3336         if ((dev->type != ARPHRD_ETHER) &&
3337             (dev->type != ARPHRD_FDDI) &&
3338             (dev->type != ARPHRD_ARCNET) &&
3339             (dev->type != ARPHRD_INFINIBAND) &&
3340             (dev->type != ARPHRD_IEEE1394) &&
3341             (dev->type != ARPHRD_TUNNEL6) &&
3342             (dev->type != ARPHRD_6LOWPAN) &&
3343             (dev->type != ARPHRD_IP6GRE) &&
3344             (dev->type != ARPHRD_IPGRE) &&
3345             (dev->type != ARPHRD_TUNNEL) &&
3346             (dev->type != ARPHRD_NONE) &&
3347             (dev->type != ARPHRD_RAWIP)) {
3348                 /* Alas, we support only Ethernet autoconfiguration. */
3349                 return;
3350         }
3351
3352         idev = addrconf_add_dev(dev);
3353         if (IS_ERR(idev))
3354                 return;
3355
3356         /* this device type has no EUI support */
3357         if (dev->type == ARPHRD_NONE &&
3358             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3359                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3360
3361         addrconf_addr_gen(idev, false);
3362 }
3363
3364 #if IS_ENABLED(CONFIG_IPV6_SIT)
3365 static void addrconf_sit_config(struct net_device *dev)
3366 {
3367         struct inet6_dev *idev;
3368
3369         ASSERT_RTNL();
3370
3371         /*
3372          * Configure the tunnel with one of our IPv4
3373          * addresses... we should configure all of
3374          * our v4 addrs in the tunnel
3375          */
3376
3377         idev = ipv6_find_idev(dev);
3378         if (!idev) {
3379                 pr_debug("%s: add_dev failed\n", __func__);
3380                 return;
3381         }
3382
3383         if (dev->priv_flags & IFF_ISATAP) {
3384                 addrconf_addr_gen(idev, false);
3385                 return;
3386         }
3387
3388         sit_add_v4_addrs(idev);
3389
3390         if (dev->flags&IFF_POINTOPOINT)
3391                 addrconf_add_mroute(dev);
3392 }
3393 #endif
3394
3395 #if IS_ENABLED(CONFIG_NET_IPGRE)
3396 static void addrconf_gre_config(struct net_device *dev)
3397 {
3398         struct inet6_dev *idev;
3399
3400         ASSERT_RTNL();
3401
3402         idev = ipv6_find_idev(dev);
3403         if (!idev) {
3404                 pr_debug("%s: add_dev failed\n", __func__);
3405                 return;
3406         }
3407
3408         addrconf_addr_gen(idev, true);
3409         if (dev->flags & IFF_POINTOPOINT)
3410                 addrconf_add_mroute(dev);
3411 }
3412 #endif
3413
3414 static int fixup_permanent_addr(struct net *net,
3415                                 struct inet6_dev *idev,
3416                                 struct inet6_ifaddr *ifp)
3417 {
3418         /* !fib6_node means the host route was removed from the
3419          * FIB, for example, if 'lo' device is taken down. In that
3420          * case regenerate the host route.
3421          */
3422         if (!ifp->rt || !ifp->rt->fib6_node) {
3423                 struct fib6_info *f6i, *prev;
3424
3425                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3426                                          GFP_ATOMIC);
3427                 if (IS_ERR(f6i))
3428                         return PTR_ERR(f6i);
3429
3430                 /* ifp->rt can be accessed outside of rtnl */
3431                 spin_lock(&ifp->lock);
3432                 prev = ifp->rt;
3433                 ifp->rt = f6i;
3434                 spin_unlock(&ifp->lock);
3435
3436                 fib6_info_release(prev);
3437         }
3438
3439         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3440                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3441                                       ifp->rt_priority, idev->dev, 0, 0,
3442                                       GFP_ATOMIC);
3443         }
3444
3445         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3446                 addrconf_dad_start(ifp);
3447
3448         return 0;
3449 }
3450
3451 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3452 {
3453         struct inet6_ifaddr *ifp, *tmp;
3454         struct inet6_dev *idev;
3455
3456         idev = __in6_dev_get(dev);
3457         if (!idev)
3458                 return;
3459
3460         write_lock_bh(&idev->lock);
3461
3462         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3463                 if ((ifp->flags & IFA_F_PERMANENT) &&
3464                     fixup_permanent_addr(net, idev, ifp) < 0) {
3465                         write_unlock_bh(&idev->lock);
3466                         in6_ifa_hold(ifp);
3467                         ipv6_del_addr(ifp);
3468                         write_lock_bh(&idev->lock);
3469
3470                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3471                                              idev->dev->name, &ifp->addr);
3472                 }
3473         }
3474
3475         write_unlock_bh(&idev->lock);
3476 }
3477
3478 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3479                            void *ptr)
3480 {
3481         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3482         struct netdev_notifier_change_info *change_info;
3483         struct netdev_notifier_changeupper_info *info;
3484         struct inet6_dev *idev = __in6_dev_get(dev);
3485         struct net *net = dev_net(dev);
3486         int run_pending = 0;
3487         int err;
3488
3489         switch (event) {
3490         case NETDEV_REGISTER:
3491                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3492                         idev = ipv6_add_dev(dev);
3493                         if (IS_ERR(idev))
3494                                 return notifier_from_errno(PTR_ERR(idev));
3495                 }
3496                 break;
3497
3498         case NETDEV_CHANGEMTU:
3499                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3500                 if (dev->mtu < IPV6_MIN_MTU) {
3501                         addrconf_ifdown(dev, dev != net->loopback_dev);
3502                         break;
3503                 }
3504
3505                 if (idev) {
3506                         rt6_mtu_change(dev, dev->mtu);
3507                         idev->cnf.mtu6 = dev->mtu;
3508                         break;
3509                 }
3510
3511                 /* allocate new idev */
3512                 idev = ipv6_add_dev(dev);
3513                 if (IS_ERR(idev))
3514                         break;
3515
3516                 /* device is still not ready */
3517                 if (!(idev->if_flags & IF_READY))
3518                         break;
3519
3520                 run_pending = 1;
3521
3522                 /* fall through */
3523
3524         case NETDEV_UP:
3525         case NETDEV_CHANGE:
3526                 if (dev->flags & IFF_SLAVE)
3527                         break;
3528
3529                 if (idev && idev->cnf.disable_ipv6)
3530                         break;
3531
3532                 if (event == NETDEV_UP) {
3533                         /* restore routes for permanent addresses */
3534                         addrconf_permanent_addr(net, dev);
3535
3536                         if (!addrconf_link_ready(dev)) {
3537                                 /* device is not ready yet. */
3538                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3539                                          dev->name);
3540                                 break;
3541                         }
3542
3543                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3544                                 idev = ipv6_add_dev(dev);
3545
3546                         if (!IS_ERR_OR_NULL(idev)) {
3547                                 idev->if_flags |= IF_READY;
3548                                 run_pending = 1;
3549                         }
3550                 } else if (event == NETDEV_CHANGE) {
3551                         if (!addrconf_link_ready(dev)) {
3552                                 /* device is still not ready. */
3553                                 rt6_sync_down_dev(dev, event);
3554                                 break;
3555                         }
3556
3557                         if (!IS_ERR_OR_NULL(idev)) {
3558                                 if (idev->if_flags & IF_READY) {
3559                                         /* device is already configured -
3560                                          * but resend MLD reports, we might
3561                                          * have roamed and need to update
3562                                          * multicast snooping switches
3563                                          */
3564                                         ipv6_mc_up(idev);
3565                                         change_info = ptr;
3566                                         if (change_info->flags_changed & IFF_NOARP)
3567                                                 addrconf_dad_run(idev, true);
3568                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3569                                         break;
3570                                 }
3571                                 idev->if_flags |= IF_READY;
3572                         }
3573
3574                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3575                                 dev->name);
3576
3577                         run_pending = 1;
3578                 }
3579
3580                 switch (dev->type) {
3581 #if IS_ENABLED(CONFIG_IPV6_SIT)
3582                 case ARPHRD_SIT:
3583                         addrconf_sit_config(dev);
3584                         break;
3585 #endif
3586 #if IS_ENABLED(CONFIG_NET_IPGRE)
3587                 case ARPHRD_IPGRE:
3588                         addrconf_gre_config(dev);
3589                         break;
3590 #endif
3591                 case ARPHRD_LOOPBACK:
3592                         init_loopback(dev);
3593                         break;
3594
3595                 default:
3596                         addrconf_dev_config(dev);
3597                         break;
3598                 }
3599
3600                 if (!IS_ERR_OR_NULL(idev)) {
3601                         if (run_pending)
3602                                 addrconf_dad_run(idev, false);
3603
3604                         /* Device has an address by now */
3605                         rt6_sync_up(dev, RTNH_F_DEAD);
3606
3607                         /*
3608                          * If the MTU changed during the interface down,
3609                          * when the interface up, the changed MTU must be
3610                          * reflected in the idev as well as routers.
3611                          */
3612                         if (idev->cnf.mtu6 != dev->mtu &&
3613                             dev->mtu >= IPV6_MIN_MTU) {
3614                                 rt6_mtu_change(dev, dev->mtu);
3615                                 idev->cnf.mtu6 = dev->mtu;
3616                         }
3617                         idev->tstamp = jiffies;
3618                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3619
3620                         /*
3621                          * If the changed mtu during down is lower than
3622                          * IPV6_MIN_MTU stop IPv6 on this interface.
3623                          */
3624                         if (dev->mtu < IPV6_MIN_MTU)
3625                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3626                 }
3627                 break;
3628
3629         case NETDEV_DOWN:
3630         case NETDEV_UNREGISTER:
3631                 /*
3632                  *      Remove all addresses from this interface.
3633                  */
3634                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3635                 break;
3636
3637         case NETDEV_CHANGENAME:
3638                 if (idev) {
3639                         snmp6_unregister_dev(idev);
3640                         addrconf_sysctl_unregister(idev);
3641                         err = addrconf_sysctl_register(idev);
3642                         if (err)
3643                                 return notifier_from_errno(err);
3644                         err = snmp6_register_dev(idev);
3645                         if (err) {
3646                                 addrconf_sysctl_unregister(idev);
3647                                 return notifier_from_errno(err);
3648                         }
3649                 }
3650                 break;
3651
3652         case NETDEV_PRE_TYPE_CHANGE:
3653         case NETDEV_POST_TYPE_CHANGE:
3654                 if (idev)
3655                         addrconf_type_change(dev, event);
3656                 break;
3657
3658         case NETDEV_CHANGEUPPER:
3659                 info = ptr;
3660
3661                 /* flush all routes if dev is linked to or unlinked from
3662                  * an L3 master device (e.g., VRF)
3663                  */
3664                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3665                         addrconf_ifdown(dev, 0);
3666         }
3667
3668         return NOTIFY_OK;
3669 }
3670
3671 /*
3672  *      addrconf module should be notified of a device going up
3673  */
3674 static struct notifier_block ipv6_dev_notf = {
3675         .notifier_call = addrconf_notify,
3676         .priority = ADDRCONF_NOTIFY_PRIORITY,
3677 };
3678
3679 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3680 {
3681         struct inet6_dev *idev;
3682         ASSERT_RTNL();
3683
3684         idev = __in6_dev_get(dev);
3685
3686         if (event == NETDEV_POST_TYPE_CHANGE)
3687                 ipv6_mc_remap(idev);
3688         else if (event == NETDEV_PRE_TYPE_CHANGE)
3689                 ipv6_mc_unmap(idev);
3690 }
3691
3692 static bool addr_is_local(const struct in6_addr *addr)
3693 {
3694         return ipv6_addr_type(addr) &
3695                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3696 }
3697
3698 static int addrconf_ifdown(struct net_device *dev, int how)
3699 {
3700         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3701         struct net *net = dev_net(dev);
3702         struct inet6_dev *idev;
3703         struct inet6_ifaddr *ifa, *tmp;
3704         bool keep_addr = false;
3705         int state, i;
3706
3707         ASSERT_RTNL();
3708
3709         rt6_disable_ip(dev, event);
3710
3711         idev = __in6_dev_get(dev);
3712         if (!idev)
3713                 return -ENODEV;
3714
3715         /*
3716          * Step 1: remove reference to ipv6 device from parent device.
3717          *         Do not dev_put!
3718          */
3719         if (how) {
3720                 idev->dead = 1;
3721
3722                 /* protected by rtnl_lock */
3723                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3724
3725                 /* Step 1.5: remove snmp6 entry */
3726                 snmp6_unregister_dev(idev);
3727
3728         }
3729
3730         /* combine the user config with event to determine if permanent
3731          * addresses are to be removed from address hash table
3732          */
3733         if (!how && !idev->cnf.disable_ipv6) {
3734                 /* aggregate the system setting and interface setting */
3735                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3736
3737                 if (!_keep_addr)
3738                         _keep_addr = idev->cnf.keep_addr_on_down;
3739
3740                 keep_addr = (_keep_addr > 0);
3741         }
3742
3743         /* Step 2: clear hash table */
3744         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3745                 struct hlist_head *h = &inet6_addr_lst[i];
3746
3747                 spin_lock_bh(&addrconf_hash_lock);
3748 restart:
3749                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3750                         if (ifa->idev == idev) {
3751                                 addrconf_del_dad_work(ifa);
3752                                 /* combined flag + permanent flag decide if
3753                                  * address is retained on a down event
3754                                  */
3755                                 if (!keep_addr ||
3756                                     !(ifa->flags & IFA_F_PERMANENT) ||
3757                                     addr_is_local(&ifa->addr)) {
3758                                         hlist_del_init_rcu(&ifa->addr_lst);
3759                                         goto restart;
3760                                 }
3761                         }
3762                 }
3763                 spin_unlock_bh(&addrconf_hash_lock);
3764         }
3765
3766         write_lock_bh(&idev->lock);
3767
3768         addrconf_del_rs_timer(idev);
3769
3770         /* Step 2: clear flags for stateless addrconf */
3771         if (!how)
3772                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3773
3774         /* Step 3: clear tempaddr list */
3775         while (!list_empty(&idev->tempaddr_list)) {
3776                 ifa = list_first_entry(&idev->tempaddr_list,
3777                                        struct inet6_ifaddr, tmp_list);
3778                 list_del(&ifa->tmp_list);
3779                 write_unlock_bh(&idev->lock);
3780                 spin_lock_bh(&ifa->lock);
3781
3782                 if (ifa->ifpub) {
3783                         in6_ifa_put(ifa->ifpub);
3784                         ifa->ifpub = NULL;
3785                 }
3786                 spin_unlock_bh(&ifa->lock);
3787                 in6_ifa_put(ifa);
3788                 write_lock_bh(&idev->lock);
3789         }
3790
3791         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3792                 struct fib6_info *rt = NULL;
3793                 bool keep;
3794
3795                 addrconf_del_dad_work(ifa);
3796
3797                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3798                         !addr_is_local(&ifa->addr);
3799
3800                 write_unlock_bh(&idev->lock);
3801                 spin_lock_bh(&ifa->lock);
3802
3803                 if (keep) {
3804                         /* set state to skip the notifier below */
3805                         state = INET6_IFADDR_STATE_DEAD;
3806                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3807                         if (!(ifa->flags & IFA_F_NODAD))
3808                                 ifa->flags |= IFA_F_TENTATIVE;
3809
3810                         rt = ifa->rt;
3811                         ifa->rt = NULL;
3812                 } else {
3813                         state = ifa->state;
3814                         ifa->state = INET6_IFADDR_STATE_DEAD;
3815                 }
3816
3817                 spin_unlock_bh(&ifa->lock);
3818
3819                 if (rt)
3820                         ip6_del_rt(net, rt);
3821
3822                 if (state != INET6_IFADDR_STATE_DEAD) {
3823                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3824                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3825                 } else {
3826                         if (idev->cnf.forwarding)
3827                                 addrconf_leave_anycast(ifa);
3828                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3829                 }
3830
3831                 write_lock_bh(&idev->lock);
3832                 if (!keep) {
3833                         list_del_rcu(&ifa->if_list);
3834                         in6_ifa_put(ifa);
3835                 }
3836         }
3837
3838         write_unlock_bh(&idev->lock);
3839
3840         /* Step 5: Discard anycast and multicast list */
3841         if (how) {
3842                 ipv6_ac_destroy_dev(idev);
3843                 ipv6_mc_destroy_dev(idev);
3844         } else {
3845                 ipv6_mc_down(idev);
3846         }
3847
3848         idev->tstamp = jiffies;
3849
3850         /* Last: Shot the device (if unregistered) */
3851         if (how) {
3852                 addrconf_sysctl_unregister(idev);
3853                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3854                 neigh_ifdown(&nd_tbl, dev);
3855                 in6_dev_put(idev);
3856         }
3857         return 0;
3858 }
3859
3860 static void addrconf_rs_timer(struct timer_list *t)
3861 {
3862         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3863         struct net_device *dev = idev->dev;
3864         struct in6_addr lladdr;
3865
3866         write_lock(&idev->lock);
3867         if (idev->dead || !(idev->if_flags & IF_READY))
3868                 goto out;
3869
3870         if (!ipv6_accept_ra(idev))
3871                 goto out;
3872
3873         /* Announcement received after solicitation was sent */
3874         if (idev->if_flags & IF_RA_RCVD)
3875                 goto out;
3876
3877         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3878                 write_unlock(&idev->lock);
3879                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3880                         ndisc_send_rs(dev, &lladdr,
3881                                       &in6addr_linklocal_allrouters);
3882                 else
3883                         goto put;
3884
3885                 write_lock(&idev->lock);
3886                 idev->rs_interval = rfc3315_s14_backoff_update(
3887                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3888                 /* The wait after the last probe can be shorter */
3889                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3890                                              idev->cnf.rtr_solicits) ?
3891                                       idev->cnf.rtr_solicit_delay :
3892                                       idev->rs_interval);
3893         } else {
3894                 /*
3895                  * Note: we do not support deprecated "all on-link"
3896                  * assumption any longer.
3897                  */
3898                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3899         }
3900
3901 out:
3902         write_unlock(&idev->lock);
3903 put:
3904         in6_dev_put(idev);
3905 }
3906
3907 /*
3908  *      Duplicate Address Detection
3909  */
3910 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3911 {
3912         unsigned long rand_num;
3913         struct inet6_dev *idev = ifp->idev;
3914         u64 nonce;
3915
3916         if (ifp->flags & IFA_F_OPTIMISTIC)
3917                 rand_num = 0;
3918         else
3919                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3920
3921         nonce = 0;
3922         if (idev->cnf.enhanced_dad ||
3923             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3924                 do
3925                         get_random_bytes(&nonce, 6);
3926                 while (nonce == 0);
3927         }
3928         ifp->dad_nonce = nonce;
3929         ifp->dad_probes = idev->cnf.dad_transmits;
3930         addrconf_mod_dad_work(ifp, rand_num);
3931 }
3932
3933 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3934 {
3935         struct inet6_dev *idev = ifp->idev;
3936         struct net_device *dev = idev->dev;
3937         bool bump_id, notify = false;
3938         struct net *net;
3939
3940         addrconf_join_solict(dev, &ifp->addr);
3941
3942         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3943
3944         read_lock_bh(&idev->lock);
3945         spin_lock(&ifp->lock);
3946         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3947                 goto out;
3948
3949         net = dev_net(dev);
3950         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3951             (net->ipv6.devconf_all->accept_dad < 1 &&
3952              idev->cnf.accept_dad < 1) ||
3953             !(ifp->flags&IFA_F_TENTATIVE) ||
3954             ifp->flags & IFA_F_NODAD) {
3955                 bool send_na = false;
3956
3957                 if (ifp->flags & IFA_F_TENTATIVE &&
3958                     !(ifp->flags & IFA_F_OPTIMISTIC))
3959                         send_na = true;
3960                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3961                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3962                 spin_unlock(&ifp->lock);
3963                 read_unlock_bh(&idev->lock);
3964
3965                 addrconf_dad_completed(ifp, bump_id, send_na);
3966                 return;
3967         }
3968
3969         if (!(idev->if_flags & IF_READY)) {
3970                 spin_unlock(&ifp->lock);
3971                 read_unlock_bh(&idev->lock);
3972                 /*
3973                  * If the device is not ready:
3974                  * - keep it tentative if it is a permanent address.
3975                  * - otherwise, kill it.
3976                  */
3977                 in6_ifa_hold(ifp);
3978                 addrconf_dad_stop(ifp, 0);
3979                 return;
3980         }
3981
3982         /*
3983          * Optimistic nodes can start receiving
3984          * Frames right away
3985          */
3986         if (ifp->flags & IFA_F_OPTIMISTIC) {
3987                 ip6_ins_rt(net, ifp->rt);
3988                 if (ipv6_use_optimistic_addr(net, idev)) {
3989                         /* Because optimistic nodes can use this address,
3990                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3991                          */
3992                         notify = true;
3993                 }
3994         }
3995
3996         addrconf_dad_kick(ifp);
3997 out:
3998         spin_unlock(&ifp->lock);
3999         read_unlock_bh(&idev->lock);
4000         if (notify)
4001                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4002 }
4003
4004 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4005 {
4006         bool begin_dad = false;
4007
4008         spin_lock_bh(&ifp->lock);
4009         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4010                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4011                 begin_dad = true;
4012         }
4013         spin_unlock_bh(&ifp->lock);
4014
4015         if (begin_dad)
4016                 addrconf_mod_dad_work(ifp, 0);
4017 }
4018
4019 static void addrconf_dad_work(struct work_struct *w)
4020 {
4021         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4022                                                 struct inet6_ifaddr,
4023                                                 dad_work);
4024         struct inet6_dev *idev = ifp->idev;
4025         bool bump_id, disable_ipv6 = false;
4026         struct in6_addr mcaddr;
4027
4028         enum {
4029                 DAD_PROCESS,
4030                 DAD_BEGIN,
4031                 DAD_ABORT,
4032         } action = DAD_PROCESS;
4033
4034         rtnl_lock();
4035
4036         spin_lock_bh(&ifp->lock);
4037         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4038                 action = DAD_BEGIN;
4039                 ifp->state = INET6_IFADDR_STATE_DAD;
4040         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4041                 action = DAD_ABORT;
4042                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4043
4044                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4045                      idev->cnf.accept_dad > 1) &&
4046                     !idev->cnf.disable_ipv6 &&
4047                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4048                         struct in6_addr addr;
4049
4050                         addr.s6_addr32[0] = htonl(0xfe800000);
4051                         addr.s6_addr32[1] = 0;
4052
4053                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4054                             ipv6_addr_equal(&ifp->addr, &addr)) {
4055                                 /* DAD failed for link-local based on MAC */
4056                                 idev->cnf.disable_ipv6 = 1;
4057
4058                                 pr_info("%s: IPv6 being disabled!\n",
4059                                         ifp->idev->dev->name);
4060                                 disable_ipv6 = true;
4061                         }
4062                 }
4063         }
4064         spin_unlock_bh(&ifp->lock);
4065
4066         if (action == DAD_BEGIN) {
4067                 addrconf_dad_begin(ifp);
4068                 goto out;
4069         } else if (action == DAD_ABORT) {
4070                 in6_ifa_hold(ifp);
4071                 addrconf_dad_stop(ifp, 1);
4072                 if (disable_ipv6)
4073                         addrconf_ifdown(idev->dev, 0);
4074                 goto out;
4075         }
4076
4077         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4078                 goto out;
4079
4080         write_lock_bh(&idev->lock);
4081         if (idev->dead || !(idev->if_flags & IF_READY)) {
4082                 write_unlock_bh(&idev->lock);
4083                 goto out;
4084         }
4085
4086         spin_lock(&ifp->lock);
4087         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4088                 spin_unlock(&ifp->lock);
4089                 write_unlock_bh(&idev->lock);
4090                 goto out;
4091         }
4092
4093         if (ifp->dad_probes == 0) {
4094                 bool send_na = false;
4095
4096                 /*
4097                  * DAD was successful
4098                  */
4099
4100                 if (ifp->flags & IFA_F_TENTATIVE &&
4101                     !(ifp->flags & IFA_F_OPTIMISTIC))
4102                         send_na = true;
4103                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4104                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4105                 spin_unlock(&ifp->lock);
4106                 write_unlock_bh(&idev->lock);
4107
4108                 addrconf_dad_completed(ifp, bump_id, send_na);
4109
4110                 goto out;
4111         }
4112
4113         ifp->dad_probes--;
4114         addrconf_mod_dad_work(ifp,
4115                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4116         spin_unlock(&ifp->lock);
4117         write_unlock_bh(&idev->lock);
4118
4119         /* send a neighbour solicitation for our addr */
4120         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4121         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4122                       ifp->dad_nonce);
4123 out:
4124         in6_ifa_put(ifp);
4125         rtnl_unlock();
4126 }
4127
4128 /* ifp->idev must be at least read locked */
4129 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4130 {
4131         struct inet6_ifaddr *ifpiter;
4132         struct inet6_dev *idev = ifp->idev;
4133
4134         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4135                 if (ifpiter->scope > IFA_LINK)
4136                         break;
4137                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4138                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4139                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4140                     IFA_F_PERMANENT)
4141                         return false;
4142         }
4143         return true;
4144 }
4145
4146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4147                                    bool send_na)
4148 {
4149         struct net_device *dev = ifp->idev->dev;
4150         struct in6_addr lladdr;
4151         bool send_rs, send_mld;
4152
4153         addrconf_del_dad_work(ifp);
4154
4155         /*
4156          *      Configure the address for reception. Now it is valid.
4157          */
4158
4159         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4160
4161         /* If added prefix is link local and we are prepared to process
4162            router advertisements, start sending router solicitations.
4163          */
4164
4165         read_lock_bh(&ifp->idev->lock);
4166         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4167         send_rs = send_mld &&
4168                   ipv6_accept_ra(ifp->idev) &&
4169                   ifp->idev->cnf.rtr_solicits != 0 &&
4170                   (dev->flags&IFF_LOOPBACK) == 0;
4171         read_unlock_bh(&ifp->idev->lock);
4172
4173         /* While dad is in progress mld report's source address is in6_addrany.
4174          * Resend with proper ll now.
4175          */
4176         if (send_mld)
4177                 ipv6_mc_dad_complete(ifp->idev);
4178
4179         /* send unsolicited NA if enabled */
4180         if (send_na &&
4181             (ifp->idev->cnf.ndisc_notify ||
4182              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4183                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4184                               /*router=*/ !!ifp->idev->cnf.forwarding,
4185                               /*solicited=*/ false, /*override=*/ true,
4186                               /*inc_opt=*/ true);
4187         }
4188
4189         if (send_rs) {
4190                 /*
4191                  *      If a host as already performed a random delay
4192                  *      [...] as part of DAD [...] there is no need
4193                  *      to delay again before sending the first RS
4194                  */
4195                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4196                         return;
4197                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4198
4199                 write_lock_bh(&ifp->idev->lock);
4200                 spin_lock(&ifp->lock);
4201                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4202                         ifp->idev->cnf.rtr_solicit_interval);
4203                 ifp->idev->rs_probes = 1;
4204                 ifp->idev->if_flags |= IF_RS_SENT;
4205                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4206                 spin_unlock(&ifp->lock);
4207                 write_unlock_bh(&ifp->idev->lock);
4208         }
4209
4210         if (bump_id)
4211                 rt_genid_bump_ipv6(dev_net(dev));
4212
4213         /* Make sure that a new temporary address will be created
4214          * before this temporary address becomes deprecated.
4215          */
4216         if (ifp->flags & IFA_F_TEMPORARY)
4217                 addrconf_verify_rtnl();
4218 }
4219
4220 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4221 {
4222         struct inet6_ifaddr *ifp;
4223
4224         read_lock_bh(&idev->lock);
4225         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4226                 spin_lock(&ifp->lock);
4227                 if ((ifp->flags & IFA_F_TENTATIVE &&
4228                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4229                         if (restart)
4230                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4231                         addrconf_dad_kick(ifp);
4232                 }
4233                 spin_unlock(&ifp->lock);
4234         }
4235         read_unlock_bh(&idev->lock);
4236 }
4237
4238 #ifdef CONFIG_PROC_FS
4239 struct if6_iter_state {
4240         struct seq_net_private p;
4241         int bucket;
4242         int offset;
4243 };
4244
4245 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4246 {
4247         struct if6_iter_state *state = seq->private;
4248         struct net *net = seq_file_net(seq);
4249         struct inet6_ifaddr *ifa = NULL;
4250         int p = 0;
4251
4252         /* initial bucket if pos is 0 */
4253         if (pos == 0) {
4254                 state->bucket = 0;
4255                 state->offset = 0;
4256         }
4257
4258         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4259                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4260                                          addr_lst) {
4261                         if (!net_eq(dev_net(ifa->idev->dev), net))
4262                                 continue;
4263                         /* sync with offset */
4264                         if (p < state->offset) {
4265                                 p++;
4266                                 continue;
4267                         }
4268                         return ifa;
4269                 }
4270
4271                 /* prepare for next bucket */
4272                 state->offset = 0;
4273                 p = 0;
4274         }
4275         return NULL;
4276 }
4277
4278 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4279                                          struct inet6_ifaddr *ifa)
4280 {
4281         struct if6_iter_state *state = seq->private;
4282         struct net *net = seq_file_net(seq);
4283
4284         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4285                 if (!net_eq(dev_net(ifa->idev->dev), net))
4286                         continue;
4287                 state->offset++;
4288                 return ifa;
4289         }
4290
4291         state->offset = 0;
4292         while (++state->bucket < IN6_ADDR_HSIZE) {
4293                 hlist_for_each_entry_rcu(ifa,
4294                                      &inet6_addr_lst[state->bucket], addr_lst) {
4295                         if (!net_eq(dev_net(ifa->idev->dev), net))
4296                                 continue;
4297                         return ifa;
4298                 }
4299         }
4300
4301         return NULL;
4302 }
4303
4304 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4305         __acquires(rcu)
4306 {
4307         rcu_read_lock();
4308         return if6_get_first(seq, *pos);
4309 }
4310
4311 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4312 {
4313         struct inet6_ifaddr *ifa;
4314
4315         ifa = if6_get_next(seq, v);
4316         ++*pos;
4317         return ifa;
4318 }
4319
4320 static void if6_seq_stop(struct seq_file *seq, void *v)
4321         __releases(rcu)
4322 {
4323         rcu_read_unlock();
4324 }
4325
4326 static int if6_seq_show(struct seq_file *seq, void *v)
4327 {
4328         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4329         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4330                    &ifp->addr,
4331                    ifp->idev->dev->ifindex,
4332                    ifp->prefix_len,
4333                    ifp->scope,
4334                    (u8) ifp->flags,
4335                    ifp->idev->dev->name);
4336         return 0;
4337 }
4338
4339 static const struct seq_operations if6_seq_ops = {
4340         .start  = if6_seq_start,
4341         .next   = if6_seq_next,
4342         .show   = if6_seq_show,
4343         .stop   = if6_seq_stop,
4344 };
4345
4346 static int __net_init if6_proc_net_init(struct net *net)
4347 {
4348         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4349                         sizeof(struct if6_iter_state)))
4350                 return -ENOMEM;
4351         return 0;
4352 }
4353
4354 static void __net_exit if6_proc_net_exit(struct net *net)
4355 {
4356         remove_proc_entry("if_inet6", net->proc_net);
4357 }
4358
4359 static struct pernet_operations if6_proc_net_ops = {
4360         .init = if6_proc_net_init,
4361         .exit = if6_proc_net_exit,
4362 };
4363
4364 int __init if6_proc_init(void)
4365 {
4366         return register_pernet_subsys(&if6_proc_net_ops);
4367 }
4368
4369 void if6_proc_exit(void)
4370 {
4371         unregister_pernet_subsys(&if6_proc_net_ops);
4372 }
4373 #endif  /* CONFIG_PROC_FS */
4374
4375 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4376 /* Check if address is a home address configured on any interface. */
4377 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4378 {
4379         unsigned int hash = inet6_addr_hash(net, addr);
4380         struct inet6_ifaddr *ifp = NULL;
4381         int ret = 0;
4382
4383         rcu_read_lock();
4384         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4385                 if (!net_eq(dev_net(ifp->idev->dev), net))
4386                         continue;
4387                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4388                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4389                         ret = 1;
4390                         break;
4391                 }
4392         }
4393         rcu_read_unlock();
4394         return ret;
4395 }
4396 #endif
4397
4398 /*
4399  *      Periodic address status verification
4400  */
4401
4402 static void addrconf_verify_rtnl(void)
4403 {
4404         unsigned long now, next, next_sec, next_sched;
4405         struct inet6_ifaddr *ifp;
4406         int i;
4407
4408         ASSERT_RTNL();
4409
4410         rcu_read_lock_bh();
4411         now = jiffies;
4412         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4413
4414         cancel_delayed_work(&addr_chk_work);
4415
4416         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4417 restart:
4418                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4419                         unsigned long age;
4420
4421                         /* When setting preferred_lft to a value not zero or
4422                          * infinity, while valid_lft is infinity
4423                          * IFA_F_PERMANENT has a non-infinity life time.
4424                          */
4425                         if ((ifp->flags & IFA_F_PERMANENT) &&
4426                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4427                                 continue;
4428
4429                         spin_lock(&ifp->lock);
4430                         /* We try to batch several events at once. */
4431                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4432
4433                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4434                             age >= ifp->valid_lft) {
4435                                 spin_unlock(&ifp->lock);
4436                                 in6_ifa_hold(ifp);
4437                                 ipv6_del_addr(ifp);
4438                                 goto restart;
4439                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4440                                 spin_unlock(&ifp->lock);
4441                                 continue;
4442                         } else if (age >= ifp->prefered_lft) {
4443                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4444                                 int deprecate = 0;
4445
4446                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4447                                         deprecate = 1;
4448                                         ifp->flags |= IFA_F_DEPRECATED;
4449                                 }
4450
4451                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4452                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4453                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4454
4455                                 spin_unlock(&ifp->lock);
4456
4457                                 if (deprecate) {
4458                                         in6_ifa_hold(ifp);
4459
4460                                         ipv6_ifa_notify(0, ifp);
4461                                         in6_ifa_put(ifp);
4462                                         goto restart;
4463                                 }
4464                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4465                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4466                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4467                                         ifp->idev->cnf.dad_transmits *
4468                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4469
4470                                 if (age >= ifp->prefered_lft - regen_advance) {
4471                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4472                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4473                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4474                                         if (!ifp->regen_count && ifpub) {
4475                                                 ifp->regen_count++;
4476                                                 in6_ifa_hold(ifp);
4477                                                 in6_ifa_hold(ifpub);
4478                                                 spin_unlock(&ifp->lock);
4479
4480                                                 spin_lock(&ifpub->lock);
4481                                                 ifpub->regen_count = 0;
4482                                                 spin_unlock(&ifpub->lock);
4483                                                 rcu_read_unlock_bh();
4484                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4485                                                 in6_ifa_put(ifpub);
4486                                                 in6_ifa_put(ifp);
4487                                                 rcu_read_lock_bh();
4488                                                 goto restart;
4489                                         }
4490                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4491                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4492                                 spin_unlock(&ifp->lock);
4493                         } else {
4494                                 /* ifp->prefered_lft <= ifp->valid_lft */
4495                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4496                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4497                                 spin_unlock(&ifp->lock);
4498                         }
4499                 }
4500         }
4501
4502         next_sec = round_jiffies_up(next);
4503         next_sched = next;
4504
4505         /* If rounded timeout is accurate enough, accept it. */
4506         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4507                 next_sched = next_sec;
4508
4509         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4510         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4511                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4512
4513         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4514                  now, next, next_sec, next_sched);
4515         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4516         rcu_read_unlock_bh();
4517 }
4518
4519 static void addrconf_verify_work(struct work_struct *w)
4520 {
4521         rtnl_lock();
4522         addrconf_verify_rtnl();
4523         rtnl_unlock();
4524 }
4525
4526 static void addrconf_verify(void)
4527 {
4528         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4529 }
4530
4531 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4532                                      struct in6_addr **peer_pfx)
4533 {
4534         struct in6_addr *pfx = NULL;
4535
4536         *peer_pfx = NULL;
4537
4538         if (addr)
4539                 pfx = nla_data(addr);
4540
4541         if (local) {
4542                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4543                         *peer_pfx = pfx;
4544                 pfx = nla_data(local);
4545         }
4546
4547         return pfx;
4548 }
4549
4550 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4551         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4552         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4553         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4554         [IFA_FLAGS]             = { .len = sizeof(u32) },
4555         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4556         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4557 };
4558
4559 static int
4560 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4561                   struct netlink_ext_ack *extack)
4562 {
4563         struct net *net = sock_net(skb->sk);
4564         struct ifaddrmsg *ifm;
4565         struct nlattr *tb[IFA_MAX+1];
4566         struct in6_addr *pfx, *peer_pfx;
4567         u32 ifa_flags;
4568         int err;
4569
4570         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4571                                      ifa_ipv6_policy, extack);
4572         if (err < 0)
4573                 return err;
4574
4575         ifm = nlmsg_data(nlh);
4576         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4577         if (!pfx)
4578                 return -EINVAL;
4579
4580         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4581
4582         /* We ignore other flags so far. */
4583         ifa_flags &= IFA_F_MANAGETEMPADDR;
4584
4585         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4586                               ifm->ifa_prefixlen);
4587 }
4588
4589 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4590                                unsigned long expires, u32 flags)
4591 {
4592         struct fib6_info *f6i;
4593         u32 prio;
4594
4595         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
4596                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4597         if (!f6i)
4598                 return -ENOENT;
4599
4600         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4601         if (f6i->fib6_metric != prio) {
4602                 /* delete old one */
4603                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4604
4605                 /* add new one */
4606                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4607                                       ifp->rt_priority, ifp->idev->dev,
4608                                       expires, flags, GFP_KERNEL);
4609         } else {
4610                 if (!expires)
4611                         fib6_clean_expires(f6i);
4612                 else
4613                         fib6_set_expires(f6i, expires);
4614
4615                 fib6_info_release(f6i);
4616         }
4617
4618         return 0;
4619 }
4620
4621 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4622 {
4623         u32 flags;
4624         clock_t expires;
4625         unsigned long timeout;
4626         bool was_managetempaddr;
4627         bool had_prefixroute;
4628
4629         ASSERT_RTNL();
4630
4631         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4632                 return -EINVAL;
4633
4634         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4635             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4636                 return -EINVAL;
4637
4638         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4639                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4640
4641         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4642         if (addrconf_finite_timeout(timeout)) {
4643                 expires = jiffies_to_clock_t(timeout * HZ);
4644                 cfg->valid_lft = timeout;
4645                 flags = RTF_EXPIRES;
4646         } else {
4647                 expires = 0;
4648                 flags = 0;
4649                 cfg->ifa_flags |= IFA_F_PERMANENT;
4650         }
4651
4652         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4653         if (addrconf_finite_timeout(timeout)) {
4654                 if (timeout == 0)
4655                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4656                 cfg->preferred_lft = timeout;
4657         }
4658
4659         spin_lock_bh(&ifp->lock);
4660         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4661         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4662                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4663         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4664                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4665                         IFA_F_NOPREFIXROUTE);
4666         ifp->flags |= cfg->ifa_flags;
4667         ifp->tstamp = jiffies;
4668         ifp->valid_lft = cfg->valid_lft;
4669         ifp->prefered_lft = cfg->preferred_lft;
4670
4671         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4672                 ifp->rt_priority = cfg->rt_priority;
4673
4674         spin_unlock_bh(&ifp->lock);
4675         if (!(ifp->flags&IFA_F_TENTATIVE))
4676                 ipv6_ifa_notify(0, ifp);
4677
4678         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4679                 int rc = -ENOENT;
4680
4681                 if (had_prefixroute)
4682                         rc = modify_prefix_route(ifp, expires, flags);
4683
4684                 /* prefix route could have been deleted; if so restore it */
4685                 if (rc == -ENOENT) {
4686                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4687                                               ifp->rt_priority, ifp->idev->dev,
4688                                               expires, flags, GFP_KERNEL);
4689                 }
4690         } else if (had_prefixroute) {
4691                 enum cleanup_prefix_rt_t action;
4692                 unsigned long rt_expires;
4693
4694                 write_lock_bh(&ifp->idev->lock);
4695                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4696                 write_unlock_bh(&ifp->idev->lock);
4697
4698                 if (action != CLEANUP_PREFIX_RT_NOP) {
4699                         cleanup_prefix_route(ifp, rt_expires,
4700                                 action == CLEANUP_PREFIX_RT_DEL);
4701                 }
4702         }
4703
4704         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4705                 if (was_managetempaddr &&
4706                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4707                         cfg->valid_lft = 0;
4708                         cfg->preferred_lft = 0;
4709                 }
4710                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4711                                  cfg->preferred_lft, !was_managetempaddr,
4712                                  jiffies);
4713         }
4714
4715         addrconf_verify_rtnl();
4716
4717         return 0;
4718 }
4719
4720 static int
4721 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4722                   struct netlink_ext_ack *extack)
4723 {
4724         struct net *net = sock_net(skb->sk);
4725         struct ifaddrmsg *ifm;
4726         struct nlattr *tb[IFA_MAX+1];
4727         struct in6_addr *peer_pfx;
4728         struct inet6_ifaddr *ifa;
4729         struct net_device *dev;
4730         struct inet6_dev *idev;
4731         struct ifa6_config cfg;
4732         int err;
4733
4734         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4735                                      ifa_ipv6_policy, extack);
4736         if (err < 0)
4737                 return err;
4738
4739         memset(&cfg, 0, sizeof(cfg));
4740
4741         ifm = nlmsg_data(nlh);
4742         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4743         if (!cfg.pfx)
4744                 return -EINVAL;
4745
4746         cfg.peer_pfx = peer_pfx;
4747         cfg.plen = ifm->ifa_prefixlen;
4748         if (tb[IFA_RT_PRIORITY])
4749                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4750
4751         cfg.valid_lft = INFINITY_LIFE_TIME;
4752         cfg.preferred_lft = INFINITY_LIFE_TIME;
4753
4754         if (tb[IFA_CACHEINFO]) {
4755                 struct ifa_cacheinfo *ci;
4756
4757                 ci = nla_data(tb[IFA_CACHEINFO]);
4758                 cfg.valid_lft = ci->ifa_valid;
4759                 cfg.preferred_lft = ci->ifa_prefered;
4760         }
4761
4762         dev =  __dev_get_by_index(net, ifm->ifa_index);
4763         if (!dev)
4764                 return -ENODEV;
4765
4766         if (tb[IFA_FLAGS])
4767                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4768         else
4769                 cfg.ifa_flags = ifm->ifa_flags;
4770
4771         /* We ignore other flags so far. */
4772         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4773                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4774                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4775
4776         idev = ipv6_find_idev(dev);
4777         if (!idev)
4778                 return -ENOBUFS;
4779
4780         if (!ipv6_allow_optimistic_dad(net, idev))
4781                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4782
4783         if (cfg.ifa_flags & IFA_F_NODAD &&
4784             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4785                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4786                 return -EINVAL;
4787         }
4788
4789         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4790         if (!ifa) {
4791                 /*
4792                  * It would be best to check for !NLM_F_CREATE here but
4793                  * userspace already relies on not having to provide this.
4794                  */
4795                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4796         }
4797
4798         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4799             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4800                 err = -EEXIST;
4801         else
4802                 err = inet6_addr_modify(ifa, &cfg);
4803
4804         in6_ifa_put(ifa);
4805
4806         return err;
4807 }
4808
4809 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4810                           u8 scope, int ifindex)
4811 {
4812         struct ifaddrmsg *ifm;
4813
4814         ifm = nlmsg_data(nlh);
4815         ifm->ifa_family = AF_INET6;
4816         ifm->ifa_prefixlen = prefixlen;
4817         ifm->ifa_flags = flags;
4818         ifm->ifa_scope = scope;
4819         ifm->ifa_index = ifindex;
4820 }
4821
4822 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4823                          unsigned long tstamp, u32 preferred, u32 valid)
4824 {
4825         struct ifa_cacheinfo ci;
4826
4827         ci.cstamp = cstamp_delta(cstamp);
4828         ci.tstamp = cstamp_delta(tstamp);
4829         ci.ifa_prefered = preferred;
4830         ci.ifa_valid = valid;
4831
4832         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4833 }
4834
4835 static inline int rt_scope(int ifa_scope)
4836 {
4837         if (ifa_scope & IFA_HOST)
4838                 return RT_SCOPE_HOST;
4839         else if (ifa_scope & IFA_LINK)
4840                 return RT_SCOPE_LINK;
4841         else if (ifa_scope & IFA_SITE)
4842                 return RT_SCOPE_SITE;
4843         else
4844                 return RT_SCOPE_UNIVERSE;
4845 }
4846
4847 static inline int inet6_ifaddr_msgsize(void)
4848 {
4849         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4850                + nla_total_size(16) /* IFA_LOCAL */
4851                + nla_total_size(16) /* IFA_ADDRESS */
4852                + nla_total_size(sizeof(struct ifa_cacheinfo))
4853                + nla_total_size(4)  /* IFA_FLAGS */
4854                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4855 }
4856
4857 enum addr_type_t {
4858         UNICAST_ADDR,
4859         MULTICAST_ADDR,
4860         ANYCAST_ADDR,
4861 };
4862
4863 struct inet6_fill_args {
4864         u32 portid;
4865         u32 seq;
4866         int event;
4867         unsigned int flags;
4868         int netnsid;
4869         int ifindex;
4870         enum addr_type_t type;
4871 };
4872
4873 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4874                              struct inet6_fill_args *args)
4875 {
4876         struct nlmsghdr  *nlh;
4877         u32 preferred, valid;
4878
4879         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4880                         sizeof(struct ifaddrmsg), args->flags);
4881         if (!nlh)
4882                 return -EMSGSIZE;
4883
4884         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4885                       ifa->idev->dev->ifindex);
4886
4887         if (args->netnsid >= 0 &&
4888             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4889                 goto error;
4890
4891         if (!((ifa->flags&IFA_F_PERMANENT) &&
4892               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4893                 preferred = ifa->prefered_lft;
4894                 valid = ifa->valid_lft;
4895                 if (preferred != INFINITY_LIFE_TIME) {
4896                         long tval = (jiffies - ifa->tstamp)/HZ;
4897                         if (preferred > tval)
4898                                 preferred -= tval;
4899                         else
4900                                 preferred = 0;
4901                         if (valid != INFINITY_LIFE_TIME) {
4902                                 if (valid > tval)
4903                                         valid -= tval;
4904                                 else
4905                                         valid = 0;
4906                         }
4907                 }
4908         } else {
4909                 preferred = INFINITY_LIFE_TIME;
4910                 valid = INFINITY_LIFE_TIME;
4911         }
4912
4913         if (!ipv6_addr_any(&ifa->peer_addr)) {
4914                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4915                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4916                         goto error;
4917         } else
4918                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4919                         goto error;
4920
4921         if (ifa->rt_priority &&
4922             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4923                 goto error;
4924
4925         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4926                 goto error;
4927
4928         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4929                 goto error;
4930
4931         nlmsg_end(skb, nlh);
4932         return 0;
4933
4934 error:
4935         nlmsg_cancel(skb, nlh);
4936         return -EMSGSIZE;
4937 }
4938
4939 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4940                                struct inet6_fill_args *args)
4941 {
4942         struct nlmsghdr  *nlh;
4943         u8 scope = RT_SCOPE_UNIVERSE;
4944         int ifindex = ifmca->idev->dev->ifindex;
4945
4946         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4947                 scope = RT_SCOPE_SITE;
4948
4949         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4950                         sizeof(struct ifaddrmsg), args->flags);
4951         if (!nlh)
4952                 return -EMSGSIZE;
4953
4954         if (args->netnsid >= 0 &&
4955             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4956                 return -EMSGSIZE;
4957
4958         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4959         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4960             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4961                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4962                 nlmsg_cancel(skb, nlh);
4963                 return -EMSGSIZE;
4964         }
4965
4966         nlmsg_end(skb, nlh);
4967         return 0;
4968 }
4969
4970 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4971                                struct inet6_fill_args *args)
4972 {
4973         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4974         int ifindex = dev ? dev->ifindex : 1;
4975         struct nlmsghdr  *nlh;
4976         u8 scope = RT_SCOPE_UNIVERSE;
4977
4978         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4979                 scope = RT_SCOPE_SITE;
4980
4981         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4982                         sizeof(struct ifaddrmsg), args->flags);
4983         if (!nlh)
4984                 return -EMSGSIZE;
4985
4986         if (args->netnsid >= 0 &&
4987             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4988                 return -EMSGSIZE;
4989
4990         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4991         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4992             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4993                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4994                 nlmsg_cancel(skb, nlh);
4995                 return -EMSGSIZE;
4996         }
4997
4998         nlmsg_end(skb, nlh);
4999         return 0;
5000 }
5001
5002 /* called with rcu_read_lock() */
5003 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5004                           struct netlink_callback *cb, int s_ip_idx,
5005                           struct inet6_fill_args *fillargs)
5006 {
5007         struct ifmcaddr6 *ifmca;
5008         struct ifacaddr6 *ifaca;
5009         int ip_idx = 0;
5010         int err = 1;
5011
5012         read_lock_bh(&idev->lock);
5013         switch (fillargs->type) {
5014         case UNICAST_ADDR: {
5015                 struct inet6_ifaddr *ifa;
5016                 fillargs->event = RTM_NEWADDR;
5017
5018                 /* unicast address incl. temp addr */
5019                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5020                         if (ip_idx < s_ip_idx)
5021                                 goto next;
5022                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5023                         if (err < 0)
5024                                 break;
5025                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5026 next:
5027                         ip_idx++;
5028                 }
5029                 break;
5030         }
5031         case MULTICAST_ADDR:
5032                 fillargs->event = RTM_GETMULTICAST;
5033
5034                 /* multicast address */
5035                 for (ifmca = idev->mc_list; ifmca;
5036                      ifmca = ifmca->next, ip_idx++) {
5037                         if (ip_idx < s_ip_idx)
5038                                 continue;
5039                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5040                         if (err < 0)
5041                                 break;
5042                 }
5043                 break;
5044         case ANYCAST_ADDR:
5045                 fillargs->event = RTM_GETANYCAST;
5046                 /* anycast address */
5047                 for (ifaca = idev->ac_list; ifaca;
5048                      ifaca = ifaca->aca_next, ip_idx++) {
5049                         if (ip_idx < s_ip_idx)
5050                                 continue;
5051                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5052                         if (err < 0)
5053                                 break;
5054                 }
5055                 break;
5056         default:
5057                 break;
5058         }
5059         read_unlock_bh(&idev->lock);
5060         cb->args[2] = ip_idx;
5061         return err;
5062 }
5063
5064 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5065                                        struct inet6_fill_args *fillargs,
5066                                        struct net **tgt_net, struct sock *sk,
5067                                        struct netlink_callback *cb)
5068 {
5069         struct netlink_ext_ack *extack = cb->extack;
5070         struct nlattr *tb[IFA_MAX+1];
5071         struct ifaddrmsg *ifm;
5072         int err, i;
5073
5074         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5075                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5076                 return -EINVAL;
5077         }
5078
5079         ifm = nlmsg_data(nlh);
5080         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5081                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5082                 return -EINVAL;
5083         }
5084
5085         fillargs->ifindex = ifm->ifa_index;
5086         if (fillargs->ifindex) {
5087                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5088                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5089         }
5090
5091         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5092                                             ifa_ipv6_policy, extack);
5093         if (err < 0)
5094                 return err;
5095
5096         for (i = 0; i <= IFA_MAX; ++i) {
5097                 if (!tb[i])
5098                         continue;
5099
5100                 if (i == IFA_TARGET_NETNSID) {
5101                         struct net *net;
5102
5103                         fillargs->netnsid = nla_get_s32(tb[i]);
5104                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5105                         if (IS_ERR(net)) {
5106                                 fillargs->netnsid = -1;
5107                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5108                                 return PTR_ERR(net);
5109                         }
5110                         *tgt_net = net;
5111                 } else {
5112                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5113                         return -EINVAL;
5114                 }
5115         }
5116
5117         return 0;
5118 }
5119
5120 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5121                            enum addr_type_t type)
5122 {
5123         const struct nlmsghdr *nlh = cb->nlh;
5124         struct inet6_fill_args fillargs = {
5125                 .portid = NETLINK_CB(cb->skb).portid,
5126                 .seq = cb->nlh->nlmsg_seq,
5127                 .flags = NLM_F_MULTI,
5128                 .netnsid = -1,
5129                 .type = type,
5130         };
5131         struct net *net = sock_net(skb->sk);
5132         struct net *tgt_net = net;
5133         int idx, s_idx, s_ip_idx;
5134         int h, s_h;
5135         struct net_device *dev;
5136         struct inet6_dev *idev;
5137         struct hlist_head *head;
5138         int err = 0;
5139
5140         s_h = cb->args[0];
5141         s_idx = idx = cb->args[1];
5142         s_ip_idx = cb->args[2];
5143
5144         if (cb->strict_check) {
5145                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5146                                                   skb->sk, cb);
5147                 if (err < 0)
5148                         goto put_tgt_net;
5149
5150                 err = 0;
5151                 if (fillargs.ifindex) {
5152                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5153                         if (!dev) {
5154                                 err = -ENODEV;
5155                                 goto put_tgt_net;
5156                         }
5157                         idev = __in6_dev_get(dev);
5158                         if (idev) {
5159                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5160                                                      &fillargs);
5161                                 if (err > 0)
5162                                         err = 0;
5163                         }
5164                         goto put_tgt_net;
5165                 }
5166         }
5167
5168         rcu_read_lock();
5169         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5170         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5171                 idx = 0;
5172                 head = &tgt_net->dev_index_head[h];
5173                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5174                         if (idx < s_idx)
5175                                 goto cont;
5176                         if (h > s_h || idx > s_idx)
5177                                 s_ip_idx = 0;
5178                         idev = __in6_dev_get(dev);
5179                         if (!idev)
5180                                 goto cont;
5181
5182                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5183                                            &fillargs) < 0)
5184                                 goto done;
5185 cont:
5186                         idx++;
5187                 }
5188         }
5189 done:
5190         rcu_read_unlock();
5191         cb->args[0] = h;
5192         cb->args[1] = idx;
5193 put_tgt_net:
5194         if (fillargs.netnsid >= 0)
5195                 put_net(tgt_net);
5196
5197         return skb->len ? : err;
5198 }
5199
5200 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5201 {
5202         enum addr_type_t type = UNICAST_ADDR;
5203
5204         return inet6_dump_addr(skb, cb, type);
5205 }
5206
5207 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5208 {
5209         enum addr_type_t type = MULTICAST_ADDR;
5210
5211         return inet6_dump_addr(skb, cb, type);
5212 }
5213
5214
5215 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5216 {
5217         enum addr_type_t type = ANYCAST_ADDR;
5218
5219         return inet6_dump_addr(skb, cb, type);
5220 }
5221
5222 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5223                                        const struct nlmsghdr *nlh,
5224                                        struct nlattr **tb,
5225                                        struct netlink_ext_ack *extack)
5226 {
5227         struct ifaddrmsg *ifm;
5228         int i, err;
5229
5230         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5231                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5232                 return -EINVAL;
5233         }
5234
5235         ifm = nlmsg_data(nlh);
5236         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5237                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5238                 return -EINVAL;
5239         }
5240
5241         if (!netlink_strict_get_check(skb))
5242                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5243                                               ifa_ipv6_policy, extack);
5244
5245         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5246                                             ifa_ipv6_policy, extack);
5247         if (err)
5248                 return err;
5249
5250         for (i = 0; i <= IFA_MAX; i++) {
5251                 if (!tb[i])
5252                         continue;
5253
5254                 switch (i) {
5255                 case IFA_TARGET_NETNSID:
5256                 case IFA_ADDRESS:
5257                 case IFA_LOCAL:
5258                         break;
5259                 default:
5260                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5261                         return -EINVAL;
5262                 }
5263         }
5264
5265         return 0;
5266 }
5267
5268 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5269                              struct netlink_ext_ack *extack)
5270 {
5271         struct net *net = sock_net(in_skb->sk);
5272         struct inet6_fill_args fillargs = {
5273                 .portid = NETLINK_CB(in_skb).portid,
5274                 .seq = nlh->nlmsg_seq,
5275                 .event = RTM_NEWADDR,
5276                 .flags = 0,
5277                 .netnsid = -1,
5278         };
5279         struct net *tgt_net = net;
5280         struct ifaddrmsg *ifm;
5281         struct nlattr *tb[IFA_MAX+1];
5282         struct in6_addr *addr = NULL, *peer;
5283         struct net_device *dev = NULL;
5284         struct inet6_ifaddr *ifa;
5285         struct sk_buff *skb;
5286         int err;
5287
5288         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5289         if (err < 0)
5290                 return err;
5291
5292         if (tb[IFA_TARGET_NETNSID]) {
5293                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5294
5295                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5296                                                   fillargs.netnsid);
5297                 if (IS_ERR(tgt_net))
5298                         return PTR_ERR(tgt_net);
5299         }
5300
5301         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5302         if (!addr)
5303                 return -EINVAL;
5304
5305         ifm = nlmsg_data(nlh);
5306         if (ifm->ifa_index)
5307                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5308
5309         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5310         if (!ifa) {
5311                 err = -EADDRNOTAVAIL;
5312                 goto errout;
5313         }
5314
5315         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5316         if (!skb) {
5317                 err = -ENOBUFS;
5318                 goto errout_ifa;
5319         }
5320
5321         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5322         if (err < 0) {
5323                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5324                 WARN_ON(err == -EMSGSIZE);
5325                 kfree_skb(skb);
5326                 goto errout_ifa;
5327         }
5328         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5329 errout_ifa:
5330         in6_ifa_put(ifa);
5331 errout:
5332         if (dev)
5333                 dev_put(dev);
5334         if (fillargs.netnsid >= 0)
5335                 put_net(tgt_net);
5336
5337         return err;
5338 }
5339
5340 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5341 {
5342         struct sk_buff *skb;
5343         struct net *net = dev_net(ifa->idev->dev);
5344         struct inet6_fill_args fillargs = {
5345                 .portid = 0,
5346                 .seq = 0,
5347                 .event = event,
5348                 .flags = 0,
5349                 .netnsid = -1,
5350         };
5351         int err = -ENOBUFS;
5352
5353         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5354         if (!skb)
5355                 goto errout;
5356
5357         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5358         if (err < 0) {
5359                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5360                 WARN_ON(err == -EMSGSIZE);
5361                 kfree_skb(skb);
5362                 goto errout;
5363         }
5364         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5365         return;
5366 errout:
5367         if (err < 0)
5368                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5369 }
5370
5371 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5372                                 __s32 *array, int bytes)
5373 {
5374         BUG_ON(bytes < (DEVCONF_MAX * 4));
5375
5376         memset(array, 0, bytes);
5377         array[DEVCONF_FORWARDING] = cnf->forwarding;
5378         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5379         array[DEVCONF_MTU6] = cnf->mtu6;
5380         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5381         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5382         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5383         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5384         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5385         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5386                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5387         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5388                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5389         array[DEVCONF_RTR_SOLICIT_DELAY] =
5390                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5391         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5392         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5393                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5394         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5395                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5396         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5397         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5398         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5399         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5400         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5401         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5402         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5403         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5404         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5405 #ifdef CONFIG_IPV6_ROUTER_PREF
5406         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5407         array[DEVCONF_RTR_PROBE_INTERVAL] =
5408                 jiffies_to_msecs(cnf->rtr_probe_interval);
5409 #ifdef CONFIG_IPV6_ROUTE_INFO
5410         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5411         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5412 #endif
5413 #endif
5414         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5415         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5416 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5417         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5418         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5419 #endif
5420 #ifdef CONFIG_IPV6_MROUTE
5421         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5422 #endif
5423         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5424         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5425         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5426         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5427         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5428         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5429         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5430         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5431         /* we omit DEVCONF_STABLE_SECRET for now */
5432         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5433         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5434         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5435         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5436         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5437 #ifdef CONFIG_IPV6_SEG6_HMAC
5438         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5439 #endif
5440         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5441         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5442         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5443         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5444 }
5445
5446 static inline size_t inet6_ifla6_size(void)
5447 {
5448         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5449              + nla_total_size(sizeof(struct ifla_cacheinfo))
5450              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5451              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5452              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5453              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5454              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5455              + 0;
5456 }
5457
5458 static inline size_t inet6_if_nlmsg_size(void)
5459 {
5460         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5461                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5462                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5463                + nla_total_size(4) /* IFLA_MTU */
5464                + nla_total_size(4) /* IFLA_LINK */
5465                + nla_total_size(1) /* IFLA_OPERSTATE */
5466                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5467 }
5468
5469 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5470                                         int bytes)
5471 {
5472         int i;
5473         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5474         BUG_ON(pad < 0);
5475
5476         /* Use put_unaligned() because stats may not be aligned for u64. */
5477         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5478         for (i = 1; i < ICMP6_MIB_MAX; i++)
5479                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5480
5481         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5482 }
5483
5484 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5485                                         int bytes, size_t syncpoff)
5486 {
5487         int i, c;
5488         u64 buff[IPSTATS_MIB_MAX];
5489         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5490
5491         BUG_ON(pad < 0);
5492
5493         memset(buff, 0, sizeof(buff));
5494         buff[0] = IPSTATS_MIB_MAX;
5495
5496         for_each_possible_cpu(c) {
5497                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5498                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5499         }
5500
5501         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5502         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5503 }
5504
5505 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5506                              int bytes)
5507 {
5508         switch (attrtype) {
5509         case IFLA_INET6_STATS:
5510                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5511                                      offsetof(struct ipstats_mib, syncp));
5512                 break;
5513         case IFLA_INET6_ICMP6STATS:
5514                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5515                 break;
5516         }
5517 }
5518
5519 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5520                                   u32 ext_filter_mask)
5521 {
5522         struct nlattr *nla;
5523         struct ifla_cacheinfo ci;
5524
5525         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5526                 goto nla_put_failure;
5527         ci.max_reasm_len = IPV6_MAXPLEN;
5528         ci.tstamp = cstamp_delta(idev->tstamp);
5529         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5530         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5531         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5532                 goto nla_put_failure;
5533         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5534         if (!nla)
5535                 goto nla_put_failure;
5536         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5537
5538         /* XXX - MC not implemented */
5539
5540         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5541                 return 0;
5542
5543         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5544         if (!nla)
5545                 goto nla_put_failure;
5546         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5547
5548         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5549         if (!nla)
5550                 goto nla_put_failure;
5551         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5552
5553         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5554         if (!nla)
5555                 goto nla_put_failure;
5556
5557         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5558                 goto nla_put_failure;
5559
5560         read_lock_bh(&idev->lock);
5561         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5562         read_unlock_bh(&idev->lock);
5563
5564         return 0;
5565
5566 nla_put_failure:
5567         return -EMSGSIZE;
5568 }
5569
5570 static size_t inet6_get_link_af_size(const struct net_device *dev,
5571                                      u32 ext_filter_mask)
5572 {
5573         if (!__in6_dev_get(dev))
5574                 return 0;
5575
5576         return inet6_ifla6_size();
5577 }
5578
5579 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5580                               u32 ext_filter_mask)
5581 {
5582         struct inet6_dev *idev = __in6_dev_get(dev);
5583
5584         if (!idev)
5585                 return -ENODATA;
5586
5587         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5588                 return -EMSGSIZE;
5589
5590         return 0;
5591 }
5592
5593 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5594 {
5595         struct inet6_ifaddr *ifp;
5596         struct net_device *dev = idev->dev;
5597         bool clear_token, update_rs = false;
5598         struct in6_addr ll_addr;
5599
5600         ASSERT_RTNL();
5601
5602         if (!token)
5603                 return -EINVAL;
5604         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5605                 return -EINVAL;
5606         if (!ipv6_accept_ra(idev))
5607                 return -EINVAL;
5608         if (idev->cnf.rtr_solicits == 0)
5609                 return -EINVAL;
5610
5611         write_lock_bh(&idev->lock);
5612
5613         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5614         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5615
5616         write_unlock_bh(&idev->lock);
5617
5618         clear_token = ipv6_addr_any(token);
5619         if (clear_token)
5620                 goto update_lft;
5621
5622         if (!idev->dead && (idev->if_flags & IF_READY) &&
5623             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5624                              IFA_F_OPTIMISTIC)) {
5625                 /* If we're not ready, then normal ifup will take care
5626                  * of this. Otherwise, we need to request our rs here.
5627                  */
5628                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5629                 update_rs = true;
5630         }
5631
5632 update_lft:
5633         write_lock_bh(&idev->lock);
5634
5635         if (update_rs) {
5636                 idev->if_flags |= IF_RS_SENT;
5637                 idev->rs_interval = rfc3315_s14_backoff_init(
5638                         idev->cnf.rtr_solicit_interval);
5639                 idev->rs_probes = 1;
5640                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5641         }
5642
5643         /* Well, that's kinda nasty ... */
5644         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5645                 spin_lock(&ifp->lock);
5646                 if (ifp->tokenized) {
5647                         ifp->valid_lft = 0;
5648                         ifp->prefered_lft = 0;
5649                 }
5650                 spin_unlock(&ifp->lock);
5651         }
5652
5653         write_unlock_bh(&idev->lock);
5654         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5655         addrconf_verify_rtnl();
5656         return 0;
5657 }
5658
5659 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5660         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5661         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5662 };
5663
5664 static int inet6_validate_link_af(const struct net_device *dev,
5665                                   const struct nlattr *nla)
5666 {
5667         struct nlattr *tb[IFLA_INET6_MAX + 1];
5668
5669         if (dev && !__in6_dev_get(dev))
5670                 return -EAFNOSUPPORT;
5671
5672         return nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5673                                            inet6_af_policy, NULL);
5674 }
5675
5676 static int check_addr_gen_mode(int mode)
5677 {
5678         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5679             mode != IN6_ADDR_GEN_MODE_NONE &&
5680             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5681             mode != IN6_ADDR_GEN_MODE_RANDOM)
5682                 return -EINVAL;
5683         return 1;
5684 }
5685
5686 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5687                                 int mode)
5688 {
5689         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5690             !idev->cnf.stable_secret.initialized &&
5691             !net->ipv6.devconf_dflt->stable_secret.initialized)
5692                 return -EINVAL;
5693         return 1;
5694 }
5695
5696 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5697 {
5698         int err = -EINVAL;
5699         struct inet6_dev *idev = __in6_dev_get(dev);
5700         struct nlattr *tb[IFLA_INET6_MAX + 1];
5701
5702         if (!idev)
5703                 return -EAFNOSUPPORT;
5704
5705         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5706                 BUG();
5707
5708         if (tb[IFLA_INET6_TOKEN]) {
5709                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5710                 if (err)
5711                         return err;
5712         }
5713
5714         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5715                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5716
5717                 if (check_addr_gen_mode(mode) < 0 ||
5718                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5719                         return -EINVAL;
5720
5721                 idev->cnf.addr_gen_mode = mode;
5722                 err = 0;
5723         }
5724
5725         return err;
5726 }
5727
5728 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5729                              u32 portid, u32 seq, int event, unsigned int flags)
5730 {
5731         struct net_device *dev = idev->dev;
5732         struct ifinfomsg *hdr;
5733         struct nlmsghdr *nlh;
5734         void *protoinfo;
5735
5736         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5737         if (!nlh)
5738                 return -EMSGSIZE;
5739
5740         hdr = nlmsg_data(nlh);
5741         hdr->ifi_family = AF_INET6;
5742         hdr->__ifi_pad = 0;
5743         hdr->ifi_type = dev->type;
5744         hdr->ifi_index = dev->ifindex;
5745         hdr->ifi_flags = dev_get_flags(dev);
5746         hdr->ifi_change = 0;
5747
5748         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5749             (dev->addr_len &&
5750              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5751             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5752             (dev->ifindex != dev_get_iflink(dev) &&
5753              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5754             nla_put_u8(skb, IFLA_OPERSTATE,
5755                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5756                 goto nla_put_failure;
5757         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5758         if (!protoinfo)
5759                 goto nla_put_failure;
5760
5761         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5762                 goto nla_put_failure;
5763
5764         nla_nest_end(skb, protoinfo);
5765         nlmsg_end(skb, nlh);
5766         return 0;
5767
5768 nla_put_failure:
5769         nlmsg_cancel(skb, nlh);
5770         return -EMSGSIZE;
5771 }
5772
5773 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5774                                    struct netlink_ext_ack *extack)
5775 {
5776         struct ifinfomsg *ifm;
5777
5778         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5779                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5780                 return -EINVAL;
5781         }
5782
5783         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5784                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5785                 return -EINVAL;
5786         }
5787
5788         ifm = nlmsg_data(nlh);
5789         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5790             ifm->ifi_change || ifm->ifi_index) {
5791                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5792                 return -EINVAL;
5793         }
5794
5795         return 0;
5796 }
5797
5798 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5799 {
5800         struct net *net = sock_net(skb->sk);
5801         int h, s_h;
5802         int idx = 0, s_idx;
5803         struct net_device *dev;
5804         struct inet6_dev *idev;
5805         struct hlist_head *head;
5806
5807         /* only requests using strict checking can pass data to
5808          * influence the dump
5809          */
5810         if (cb->strict_check) {
5811                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5812
5813                 if (err < 0)
5814                         return err;
5815         }
5816
5817         s_h = cb->args[0];
5818         s_idx = cb->args[1];
5819
5820         rcu_read_lock();
5821         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5822                 idx = 0;
5823                 head = &net->dev_index_head[h];
5824                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5825                         if (idx < s_idx)
5826                                 goto cont;
5827                         idev = __in6_dev_get(dev);
5828                         if (!idev)
5829                                 goto cont;
5830                         if (inet6_fill_ifinfo(skb, idev,
5831                                               NETLINK_CB(cb->skb).portid,
5832                                               cb->nlh->nlmsg_seq,
5833                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5834                                 goto out;
5835 cont:
5836                         idx++;
5837                 }
5838         }
5839 out:
5840         rcu_read_unlock();
5841         cb->args[1] = idx;
5842         cb->args[0] = h;
5843
5844         return skb->len;
5845 }
5846
5847 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5848 {
5849         struct sk_buff *skb;
5850         struct net *net = dev_net(idev->dev);
5851         int err = -ENOBUFS;
5852
5853         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5854         if (!skb)
5855                 goto errout;
5856
5857         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5858         if (err < 0) {
5859                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5860                 WARN_ON(err == -EMSGSIZE);
5861                 kfree_skb(skb);
5862                 goto errout;
5863         }
5864         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5865         return;
5866 errout:
5867         if (err < 0)
5868                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5869 }
5870
5871 static inline size_t inet6_prefix_nlmsg_size(void)
5872 {
5873         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5874                + nla_total_size(sizeof(struct in6_addr))
5875                + nla_total_size(sizeof(struct prefix_cacheinfo));
5876 }
5877
5878 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5879                              struct prefix_info *pinfo, u32 portid, u32 seq,
5880                              int event, unsigned int flags)
5881 {
5882         struct prefixmsg *pmsg;
5883         struct nlmsghdr *nlh;
5884         struct prefix_cacheinfo ci;
5885
5886         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5887         if (!nlh)
5888                 return -EMSGSIZE;
5889
5890         pmsg = nlmsg_data(nlh);
5891         pmsg->prefix_family = AF_INET6;
5892         pmsg->prefix_pad1 = 0;
5893         pmsg->prefix_pad2 = 0;
5894         pmsg->prefix_ifindex = idev->dev->ifindex;
5895         pmsg->prefix_len = pinfo->prefix_len;
5896         pmsg->prefix_type = pinfo->type;
5897         pmsg->prefix_pad3 = 0;
5898         pmsg->prefix_flags = 0;
5899         if (pinfo->onlink)
5900                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5901         if (pinfo->autoconf)
5902                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5903
5904         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5905                 goto nla_put_failure;
5906         ci.preferred_time = ntohl(pinfo->prefered);
5907         ci.valid_time = ntohl(pinfo->valid);
5908         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5909                 goto nla_put_failure;
5910         nlmsg_end(skb, nlh);
5911         return 0;
5912
5913 nla_put_failure:
5914         nlmsg_cancel(skb, nlh);
5915         return -EMSGSIZE;
5916 }
5917
5918 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5919                          struct prefix_info *pinfo)
5920 {
5921         struct sk_buff *skb;
5922         struct net *net = dev_net(idev->dev);
5923         int err = -ENOBUFS;
5924
5925         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5926         if (!skb)
5927                 goto errout;
5928
5929         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5930         if (err < 0) {
5931                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5932                 WARN_ON(err == -EMSGSIZE);
5933                 kfree_skb(skb);
5934                 goto errout;
5935         }
5936         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5937         return;
5938 errout:
5939         if (err < 0)
5940                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5941 }
5942
5943 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5944 {
5945         struct net *net = dev_net(ifp->idev->dev);
5946
5947         if (event)
5948                 ASSERT_RTNL();
5949
5950         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5951
5952         switch (event) {
5953         case RTM_NEWADDR:
5954                 /*
5955                  * If the address was optimistic
5956                  * we inserted the route at the start of
5957                  * our DAD process, so we don't need
5958                  * to do it again
5959                  */
5960                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5961                         ip6_ins_rt(net, ifp->rt);
5962                 if (ifp->idev->cnf.forwarding)
5963                         addrconf_join_anycast(ifp);
5964                 if (!ipv6_addr_any(&ifp->peer_addr))
5965                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5966                                               ifp->idev->dev, 0, 0,
5967                                               GFP_ATOMIC);
5968                 break;
5969         case RTM_DELADDR:
5970                 if (ifp->idev->cnf.forwarding)
5971                         addrconf_leave_anycast(ifp);
5972                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5973                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5974                         struct fib6_info *rt;
5975
5976                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5977                                                        ifp->idev->dev, 0, 0,
5978                                                        false);
5979                         if (rt)
5980                                 ip6_del_rt(net, rt);
5981                 }
5982                 if (ifp->rt) {
5983                         ip6_del_rt(net, ifp->rt);
5984                         ifp->rt = NULL;
5985                 }
5986                 rt_genid_bump_ipv6(net);
5987                 break;
5988         }
5989         atomic_inc(&net->ipv6.dev_addr_genid);
5990 }
5991
5992 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5993 {
5994         rcu_read_lock_bh();
5995         if (likely(ifp->idev->dead == 0))
5996                 __ipv6_ifa_notify(event, ifp);
5997         rcu_read_unlock_bh();
5998 }
5999
6000 #ifdef CONFIG_SYSCTL
6001
6002 static
6003 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6004                            void __user *buffer, size_t *lenp, loff_t *ppos)
6005 {
6006         int *valp = ctl->data;
6007         int val = *valp;
6008         loff_t pos = *ppos;
6009         struct ctl_table lctl;
6010         int ret;
6011
6012         /*
6013          * ctl->data points to idev->cnf.forwarding, we should
6014          * not modify it until we get the rtnl lock.
6015          */
6016         lctl = *ctl;
6017         lctl.data = &val;
6018
6019         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6020
6021         if (write)
6022                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6023         if (ret)
6024                 *ppos = pos;
6025         return ret;
6026 }
6027
6028 static
6029 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6030                         void __user *buffer, size_t *lenp, loff_t *ppos)
6031 {
6032         struct inet6_dev *idev = ctl->extra1;
6033         int min_mtu = IPV6_MIN_MTU;
6034         struct ctl_table lctl;
6035
6036         lctl = *ctl;
6037         lctl.extra1 = &min_mtu;
6038         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6039
6040         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6041 }
6042
6043 static void dev_disable_change(struct inet6_dev *idev)
6044 {
6045         struct netdev_notifier_info info;
6046
6047         if (!idev || !idev->dev)
6048                 return;
6049
6050         netdev_notifier_info_init(&info, idev->dev);
6051         if (idev->cnf.disable_ipv6)
6052                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6053         else
6054                 addrconf_notify(NULL, NETDEV_UP, &info);
6055 }
6056
6057 static void addrconf_disable_change(struct net *net, __s32 newf)
6058 {
6059         struct net_device *dev;
6060         struct inet6_dev *idev;
6061
6062         for_each_netdev(net, dev) {
6063                 idev = __in6_dev_get(dev);
6064                 if (idev) {
6065                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6066                         idev->cnf.disable_ipv6 = newf;
6067                         if (changed)
6068                                 dev_disable_change(idev);
6069                 }
6070         }
6071 }
6072
6073 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6074 {
6075         struct net *net;
6076         int old;
6077
6078         if (!rtnl_trylock())
6079                 return restart_syscall();
6080
6081         net = (struct net *)table->extra2;
6082         old = *p;
6083         *p = newf;
6084
6085         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6086                 rtnl_unlock();
6087                 return 0;
6088         }
6089
6090         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6091                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6092                 addrconf_disable_change(net, newf);
6093         } else if ((!newf) ^ (!old))
6094                 dev_disable_change((struct inet6_dev *)table->extra1);
6095
6096         rtnl_unlock();
6097         return 0;
6098 }
6099
6100 static
6101 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6102                             void __user *buffer, size_t *lenp, loff_t *ppos)
6103 {
6104         int *valp = ctl->data;
6105         int val = *valp;
6106         loff_t pos = *ppos;
6107         struct ctl_table lctl;
6108         int ret;
6109
6110         /*
6111          * ctl->data points to idev->cnf.disable_ipv6, we should
6112          * not modify it until we get the rtnl lock.
6113          */
6114         lctl = *ctl;
6115         lctl.data = &val;
6116
6117         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6118
6119         if (write)
6120                 ret = addrconf_disable_ipv6(ctl, valp, val);
6121         if (ret)
6122                 *ppos = pos;
6123         return ret;
6124 }
6125
6126 static
6127 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6128                               void __user *buffer, size_t *lenp, loff_t *ppos)
6129 {
6130         int *valp = ctl->data;
6131         int ret;
6132         int old, new;
6133
6134         old = *valp;
6135         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6136         new = *valp;
6137
6138         if (write && old != new) {
6139                 struct net *net = ctl->extra2;
6140
6141                 if (!rtnl_trylock())
6142                         return restart_syscall();
6143
6144                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6145                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6146                                                      NETCONFA_PROXY_NEIGH,
6147                                                      NETCONFA_IFINDEX_DEFAULT,
6148                                                      net->ipv6.devconf_dflt);
6149                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6150                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6151                                                      NETCONFA_PROXY_NEIGH,
6152                                                      NETCONFA_IFINDEX_ALL,
6153                                                      net->ipv6.devconf_all);
6154                 else {
6155                         struct inet6_dev *idev = ctl->extra1;
6156
6157                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6158                                                      NETCONFA_PROXY_NEIGH,
6159                                                      idev->dev->ifindex,
6160                                                      &idev->cnf);
6161                 }
6162                 rtnl_unlock();
6163         }
6164
6165         return ret;
6166 }
6167
6168 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6169                                          void __user *buffer, size_t *lenp,
6170                                          loff_t *ppos)
6171 {
6172         int ret = 0;
6173         u32 new_val;
6174         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6175         struct net *net = (struct net *)ctl->extra2;
6176         struct ctl_table tmp = {
6177                 .data = &new_val,
6178                 .maxlen = sizeof(new_val),
6179                 .mode = ctl->mode,
6180         };
6181
6182         if (!rtnl_trylock())
6183                 return restart_syscall();
6184
6185         new_val = *((u32 *)ctl->data);
6186
6187         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6188         if (ret != 0)
6189                 goto out;
6190
6191         if (write) {
6192                 if (check_addr_gen_mode(new_val) < 0) {
6193                         ret = -EINVAL;
6194                         goto out;
6195                 }
6196
6197                 if (idev) {
6198                         if (check_stable_privacy(idev, net, new_val) < 0) {
6199                                 ret = -EINVAL;
6200                                 goto out;
6201                         }
6202
6203                         if (idev->cnf.addr_gen_mode != new_val) {
6204                                 idev->cnf.addr_gen_mode = new_val;
6205                                 addrconf_dev_config(idev->dev);
6206                         }
6207                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6208                         struct net_device *dev;
6209
6210                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6211                         for_each_netdev(net, dev) {
6212                                 idev = __in6_dev_get(dev);
6213                                 if (idev &&
6214                                     idev->cnf.addr_gen_mode != new_val) {
6215                                         idev->cnf.addr_gen_mode = new_val;
6216                                         addrconf_dev_config(idev->dev);
6217                                 }
6218                         }
6219                 }
6220
6221                 *((u32 *)ctl->data) = new_val;
6222         }
6223
6224 out:
6225         rtnl_unlock();
6226
6227         return ret;
6228 }
6229
6230 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6231                                          void __user *buffer, size_t *lenp,
6232                                          loff_t *ppos)
6233 {
6234         int err;
6235         struct in6_addr addr;
6236         char str[IPV6_MAX_STRLEN];
6237         struct ctl_table lctl = *ctl;
6238         struct net *net = ctl->extra2;
6239         struct ipv6_stable_secret *secret = ctl->data;
6240
6241         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6242                 return -EIO;
6243
6244         lctl.maxlen = IPV6_MAX_STRLEN;
6245         lctl.data = str;
6246
6247         if (!rtnl_trylock())
6248                 return restart_syscall();
6249
6250         if (!write && !secret->initialized) {
6251                 err = -EIO;
6252                 goto out;
6253         }
6254
6255         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6256         if (err >= sizeof(str)) {
6257                 err = -EIO;
6258                 goto out;
6259         }
6260
6261         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6262         if (err || !write)
6263                 goto out;
6264
6265         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6266                 err = -EIO;
6267                 goto out;
6268         }
6269
6270         secret->initialized = true;
6271         secret->secret = addr;
6272
6273         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6274                 struct net_device *dev;
6275
6276                 for_each_netdev(net, dev) {
6277                         struct inet6_dev *idev = __in6_dev_get(dev);
6278
6279                         if (idev) {
6280                                 idev->cnf.addr_gen_mode =
6281                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6282                         }
6283                 }
6284         } else {
6285                 struct inet6_dev *idev = ctl->extra1;
6286
6287                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6288         }
6289
6290 out:
6291         rtnl_unlock();
6292
6293         return err;
6294 }
6295
6296 static
6297 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6298                                                 int write,
6299                                                 void __user *buffer,
6300                                                 size_t *lenp,
6301                                                 loff_t *ppos)
6302 {
6303         int *valp = ctl->data;
6304         int val = *valp;
6305         loff_t pos = *ppos;
6306         struct ctl_table lctl;
6307         int ret;
6308
6309         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6310          * we should not modify it until we get the rtnl lock.
6311          */
6312         lctl = *ctl;
6313         lctl.data = &val;
6314
6315         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6316
6317         if (write)
6318                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6319         if (ret)
6320                 *ppos = pos;
6321         return ret;
6322 }
6323
6324 static
6325 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6326 {
6327         if (rt) {
6328                 if (action)
6329                         rt->dst.flags |= DST_NOPOLICY;
6330                 else
6331                         rt->dst.flags &= ~DST_NOPOLICY;
6332         }
6333 }
6334
6335 static
6336 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6337 {
6338         struct inet6_ifaddr *ifa;
6339
6340         read_lock_bh(&idev->lock);
6341         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6342                 spin_lock(&ifa->lock);
6343                 if (ifa->rt) {
6344                         struct fib6_nh *nh = &ifa->rt->fib6_nh;
6345                         int cpu;
6346
6347                         rcu_read_lock();
6348                         ifa->rt->dst_nopolicy = val ? true : false;
6349                         if (nh->rt6i_pcpu) {
6350                                 for_each_possible_cpu(cpu) {
6351                                         struct rt6_info **rtp;
6352
6353                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6354                                         addrconf_set_nopolicy(*rtp, val);
6355                                 }
6356                         }
6357                         rcu_read_unlock();
6358                 }
6359                 spin_unlock(&ifa->lock);
6360         }
6361         read_unlock_bh(&idev->lock);
6362 }
6363
6364 static
6365 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6366 {
6367         struct inet6_dev *idev;
6368         struct net *net;
6369
6370         if (!rtnl_trylock())
6371                 return restart_syscall();
6372
6373         *valp = val;
6374
6375         net = (struct net *)ctl->extra2;
6376         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6377                 rtnl_unlock();
6378                 return 0;
6379         }
6380
6381         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6382                 struct net_device *dev;
6383
6384                 for_each_netdev(net, dev) {
6385                         idev = __in6_dev_get(dev);
6386                         if (idev)
6387                                 addrconf_disable_policy_idev(idev, val);
6388                 }
6389         } else {
6390                 idev = (struct inet6_dev *)ctl->extra1;
6391                 addrconf_disable_policy_idev(idev, val);
6392         }
6393
6394         rtnl_unlock();
6395         return 0;
6396 }
6397
6398 static
6399 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6400                                    void __user *buffer, size_t *lenp,
6401                                    loff_t *ppos)
6402 {
6403         int *valp = ctl->data;
6404         int val = *valp;
6405         loff_t pos = *ppos;
6406         struct ctl_table lctl;
6407         int ret;
6408
6409         lctl = *ctl;
6410         lctl.data = &val;
6411         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6412
6413         if (write && (*valp != val))
6414                 ret = addrconf_disable_policy(ctl, valp, val);
6415
6416         if (ret)
6417                 *ppos = pos;
6418
6419         return ret;
6420 }
6421
6422 static int minus_one = -1;
6423 static const int zero = 0;
6424 static const int one = 1;
6425 static const int two_five_five = 255;
6426
6427 static const struct ctl_table addrconf_sysctl[] = {
6428         {
6429                 .procname       = "forwarding",
6430                 .data           = &ipv6_devconf.forwarding,
6431                 .maxlen         = sizeof(int),
6432                 .mode           = 0644,
6433                 .proc_handler   = addrconf_sysctl_forward,
6434         },
6435         {
6436                 .procname       = "hop_limit",
6437                 .data           = &ipv6_devconf.hop_limit,
6438                 .maxlen         = sizeof(int),
6439                 .mode           = 0644,
6440                 .proc_handler   = proc_dointvec_minmax,
6441                 .extra1         = (void *)&one,
6442                 .extra2         = (void *)&two_five_five,
6443         },
6444         {
6445                 .procname       = "mtu",
6446                 .data           = &ipv6_devconf.mtu6,
6447                 .maxlen         = sizeof(int),
6448                 .mode           = 0644,
6449                 .proc_handler   = addrconf_sysctl_mtu,
6450         },
6451         {
6452                 .procname       = "accept_ra",
6453                 .data           = &ipv6_devconf.accept_ra,
6454                 .maxlen         = sizeof(int),
6455                 .mode           = 0644,
6456                 .proc_handler   = proc_dointvec,
6457         },
6458         {
6459                 .procname       = "accept_redirects",
6460                 .data           = &ipv6_devconf.accept_redirects,
6461                 .maxlen         = sizeof(int),
6462                 .mode           = 0644,
6463                 .proc_handler   = proc_dointvec,
6464         },
6465         {
6466                 .procname       = "autoconf",
6467                 .data           = &ipv6_devconf.autoconf,
6468                 .maxlen         = sizeof(int),
6469                 .mode           = 0644,
6470                 .proc_handler   = proc_dointvec,
6471         },
6472         {
6473                 .procname       = "dad_transmits",
6474                 .data           = &ipv6_devconf.dad_transmits,
6475                 .maxlen         = sizeof(int),
6476                 .mode           = 0644,
6477                 .proc_handler   = proc_dointvec,
6478         },
6479         {
6480                 .procname       = "router_solicitations",
6481                 .data           = &ipv6_devconf.rtr_solicits,
6482                 .maxlen         = sizeof(int),
6483                 .mode           = 0644,
6484                 .proc_handler   = proc_dointvec_minmax,
6485                 .extra1         = &minus_one,
6486         },
6487         {
6488                 .procname       = "router_solicitation_interval",
6489                 .data           = &ipv6_devconf.rtr_solicit_interval,
6490                 .maxlen         = sizeof(int),
6491                 .mode           = 0644,
6492                 .proc_handler   = proc_dointvec_jiffies,
6493         },
6494         {
6495                 .procname       = "router_solicitation_max_interval",
6496                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6497                 .maxlen         = sizeof(int),
6498                 .mode           = 0644,
6499                 .proc_handler   = proc_dointvec_jiffies,
6500         },
6501         {
6502                 .procname       = "router_solicitation_delay",
6503                 .data           = &ipv6_devconf.rtr_solicit_delay,
6504                 .maxlen         = sizeof(int),
6505                 .mode           = 0644,
6506                 .proc_handler   = proc_dointvec_jiffies,
6507         },
6508         {
6509                 .procname       = "force_mld_version",
6510                 .data           = &ipv6_devconf.force_mld_version,
6511                 .maxlen         = sizeof(int),
6512                 .mode           = 0644,
6513                 .proc_handler   = proc_dointvec,
6514         },
6515         {
6516                 .procname       = "mldv1_unsolicited_report_interval",
6517                 .data           =
6518                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6519                 .maxlen         = sizeof(int),
6520                 .mode           = 0644,
6521                 .proc_handler   = proc_dointvec_ms_jiffies,
6522         },
6523         {
6524                 .procname       = "mldv2_unsolicited_report_interval",
6525                 .data           =
6526                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6527                 .maxlen         = sizeof(int),
6528                 .mode           = 0644,
6529                 .proc_handler   = proc_dointvec_ms_jiffies,
6530         },
6531         {
6532                 .procname       = "use_tempaddr",
6533                 .data           = &ipv6_devconf.use_tempaddr,
6534                 .maxlen         = sizeof(int),
6535                 .mode           = 0644,
6536                 .proc_handler   = proc_dointvec,
6537         },
6538         {
6539                 .procname       = "temp_valid_lft",
6540                 .data           = &ipv6_devconf.temp_valid_lft,
6541                 .maxlen         = sizeof(int),
6542                 .mode           = 0644,
6543                 .proc_handler   = proc_dointvec,
6544         },
6545         {
6546                 .procname       = "temp_prefered_lft",
6547                 .data           = &ipv6_devconf.temp_prefered_lft,
6548                 .maxlen         = sizeof(int),
6549                 .mode           = 0644,
6550                 .proc_handler   = proc_dointvec,
6551         },
6552         {
6553                 .procname       = "regen_max_retry",
6554                 .data           = &ipv6_devconf.regen_max_retry,
6555                 .maxlen         = sizeof(int),
6556                 .mode           = 0644,
6557                 .proc_handler   = proc_dointvec,
6558         },
6559         {
6560                 .procname       = "max_desync_factor",
6561                 .data           = &ipv6_devconf.max_desync_factor,
6562                 .maxlen         = sizeof(int),
6563                 .mode           = 0644,
6564                 .proc_handler   = proc_dointvec,
6565         },
6566         {
6567                 .procname       = "max_addresses",
6568                 .data           = &ipv6_devconf.max_addresses,
6569                 .maxlen         = sizeof(int),
6570                 .mode           = 0644,
6571                 .proc_handler   = proc_dointvec,
6572         },
6573         {
6574                 .procname       = "accept_ra_defrtr",
6575                 .data           = &ipv6_devconf.accept_ra_defrtr,
6576                 .maxlen         = sizeof(int),
6577                 .mode           = 0644,
6578                 .proc_handler   = proc_dointvec,
6579         },
6580         {
6581                 .procname       = "accept_ra_min_hop_limit",
6582                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6583                 .maxlen         = sizeof(int),
6584                 .mode           = 0644,
6585                 .proc_handler   = proc_dointvec,
6586         },
6587         {
6588                 .procname       = "accept_ra_pinfo",
6589                 .data           = &ipv6_devconf.accept_ra_pinfo,
6590                 .maxlen         = sizeof(int),
6591                 .mode           = 0644,
6592                 .proc_handler   = proc_dointvec,
6593         },
6594 #ifdef CONFIG_IPV6_ROUTER_PREF
6595         {
6596                 .procname       = "accept_ra_rtr_pref",
6597                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6598                 .maxlen         = sizeof(int),
6599                 .mode           = 0644,
6600                 .proc_handler   = proc_dointvec,
6601         },
6602         {
6603                 .procname       = "router_probe_interval",
6604                 .data           = &ipv6_devconf.rtr_probe_interval,
6605                 .maxlen         = sizeof(int),
6606                 .mode           = 0644,
6607                 .proc_handler   = proc_dointvec_jiffies,
6608         },
6609 #ifdef CONFIG_IPV6_ROUTE_INFO
6610         {
6611                 .procname       = "accept_ra_rt_info_min_plen",
6612                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6613                 .maxlen         = sizeof(int),
6614                 .mode           = 0644,
6615                 .proc_handler   = proc_dointvec,
6616         },
6617         {
6618                 .procname       = "accept_ra_rt_info_max_plen",
6619                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6620                 .maxlen         = sizeof(int),
6621                 .mode           = 0644,
6622                 .proc_handler   = proc_dointvec,
6623         },
6624 #endif
6625 #endif
6626         {
6627                 .procname       = "proxy_ndp",
6628                 .data           = &ipv6_devconf.proxy_ndp,
6629                 .maxlen         = sizeof(int),
6630                 .mode           = 0644,
6631                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6632         },
6633         {
6634                 .procname       = "accept_source_route",
6635                 .data           = &ipv6_devconf.accept_source_route,
6636                 .maxlen         = sizeof(int),
6637                 .mode           = 0644,
6638                 .proc_handler   = proc_dointvec,
6639         },
6640 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6641         {
6642                 .procname       = "optimistic_dad",
6643                 .data           = &ipv6_devconf.optimistic_dad,
6644                 .maxlen         = sizeof(int),
6645                 .mode           = 0644,
6646                 .proc_handler   = proc_dointvec,
6647         },
6648         {
6649                 .procname       = "use_optimistic",
6650                 .data           = &ipv6_devconf.use_optimistic,
6651                 .maxlen         = sizeof(int),
6652                 .mode           = 0644,
6653                 .proc_handler   = proc_dointvec,
6654         },
6655 #endif
6656 #ifdef CONFIG_IPV6_MROUTE
6657         {
6658                 .procname       = "mc_forwarding",
6659                 .data           = &ipv6_devconf.mc_forwarding,
6660                 .maxlen         = sizeof(int),
6661                 .mode           = 0444,
6662                 .proc_handler   = proc_dointvec,
6663         },
6664 #endif
6665         {
6666                 .procname       = "disable_ipv6",
6667                 .data           = &ipv6_devconf.disable_ipv6,
6668                 .maxlen         = sizeof(int),
6669                 .mode           = 0644,
6670                 .proc_handler   = addrconf_sysctl_disable,
6671         },
6672         {
6673                 .procname       = "accept_dad",
6674                 .data           = &ipv6_devconf.accept_dad,
6675                 .maxlen         = sizeof(int),
6676                 .mode           = 0644,
6677                 .proc_handler   = proc_dointvec,
6678         },
6679         {
6680                 .procname       = "force_tllao",
6681                 .data           = &ipv6_devconf.force_tllao,
6682                 .maxlen         = sizeof(int),
6683                 .mode           = 0644,
6684                 .proc_handler   = proc_dointvec
6685         },
6686         {
6687                 .procname       = "ndisc_notify",
6688                 .data           = &ipv6_devconf.ndisc_notify,
6689                 .maxlen         = sizeof(int),
6690                 .mode           = 0644,
6691                 .proc_handler   = proc_dointvec
6692         },
6693         {
6694                 .procname       = "suppress_frag_ndisc",
6695                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6696                 .maxlen         = sizeof(int),
6697                 .mode           = 0644,
6698                 .proc_handler   = proc_dointvec
6699         },
6700         {
6701                 .procname       = "accept_ra_from_local",
6702                 .data           = &ipv6_devconf.accept_ra_from_local,
6703                 .maxlen         = sizeof(int),
6704                 .mode           = 0644,
6705                 .proc_handler   = proc_dointvec,
6706         },
6707         {
6708                 .procname       = "accept_ra_mtu",
6709                 .data           = &ipv6_devconf.accept_ra_mtu,
6710                 .maxlen         = sizeof(int),
6711                 .mode           = 0644,
6712                 .proc_handler   = proc_dointvec,
6713         },
6714         {
6715                 .procname       = "stable_secret",
6716                 .data           = &ipv6_devconf.stable_secret,
6717                 .maxlen         = IPV6_MAX_STRLEN,
6718                 .mode           = 0600,
6719                 .proc_handler   = addrconf_sysctl_stable_secret,
6720         },
6721         {
6722                 .procname       = "use_oif_addrs_only",
6723                 .data           = &ipv6_devconf.use_oif_addrs_only,
6724                 .maxlen         = sizeof(int),
6725                 .mode           = 0644,
6726                 .proc_handler   = proc_dointvec,
6727         },
6728         {
6729                 .procname       = "ignore_routes_with_linkdown",
6730                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6731                 .maxlen         = sizeof(int),
6732                 .mode           = 0644,
6733                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6734         },
6735         {
6736                 .procname       = "drop_unicast_in_l2_multicast",
6737                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6738                 .maxlen         = sizeof(int),
6739                 .mode           = 0644,
6740                 .proc_handler   = proc_dointvec,
6741         },
6742         {
6743                 .procname       = "drop_unsolicited_na",
6744                 .data           = &ipv6_devconf.drop_unsolicited_na,
6745                 .maxlen         = sizeof(int),
6746                 .mode           = 0644,
6747                 .proc_handler   = proc_dointvec,
6748         },
6749         {
6750                 .procname       = "keep_addr_on_down",
6751                 .data           = &ipv6_devconf.keep_addr_on_down,
6752                 .maxlen         = sizeof(int),
6753                 .mode           = 0644,
6754                 .proc_handler   = proc_dointvec,
6755
6756         },
6757         {
6758                 .procname       = "seg6_enabled",
6759                 .data           = &ipv6_devconf.seg6_enabled,
6760                 .maxlen         = sizeof(int),
6761                 .mode           = 0644,
6762                 .proc_handler   = proc_dointvec,
6763         },
6764 #ifdef CONFIG_IPV6_SEG6_HMAC
6765         {
6766                 .procname       = "seg6_require_hmac",
6767                 .data           = &ipv6_devconf.seg6_require_hmac,
6768                 .maxlen         = sizeof(int),
6769                 .mode           = 0644,
6770                 .proc_handler   = proc_dointvec,
6771         },
6772 #endif
6773         {
6774                 .procname       = "enhanced_dad",
6775                 .data           = &ipv6_devconf.enhanced_dad,
6776                 .maxlen         = sizeof(int),
6777                 .mode           = 0644,
6778                 .proc_handler   = proc_dointvec,
6779         },
6780         {
6781                 .procname               = "addr_gen_mode",
6782                 .data                   = &ipv6_devconf.addr_gen_mode,
6783                 .maxlen                 = sizeof(int),
6784                 .mode                   = 0644,
6785                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6786         },
6787         {
6788                 .procname       = "disable_policy",
6789                 .data           = &ipv6_devconf.disable_policy,
6790                 .maxlen         = sizeof(int),
6791                 .mode           = 0644,
6792                 .proc_handler   = addrconf_sysctl_disable_policy,
6793         },
6794         {
6795                 .procname       = "ndisc_tclass",
6796                 .data           = &ipv6_devconf.ndisc_tclass,
6797                 .maxlen         = sizeof(int),
6798                 .mode           = 0644,
6799                 .proc_handler   = proc_dointvec_minmax,
6800                 .extra1         = (void *)&zero,
6801                 .extra2         = (void *)&two_five_five,
6802         },
6803         {
6804                 /* sentinel */
6805         }
6806 };
6807
6808 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6809                 struct inet6_dev *idev, struct ipv6_devconf *p)
6810 {
6811         int i, ifindex;
6812         struct ctl_table *table;
6813         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6814
6815         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6816         if (!table)
6817                 goto out;
6818
6819         for (i = 0; table[i].data; i++) {
6820                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6821                 /* If one of these is already set, then it is not safe to
6822                  * overwrite either of them: this makes proc_dointvec_minmax
6823                  * usable.
6824                  */
6825                 if (!table[i].extra1 && !table[i].extra2) {
6826                         table[i].extra1 = idev; /* embedded; no ref */
6827                         table[i].extra2 = net;
6828                 }
6829         }
6830
6831         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6832
6833         p->sysctl_header = register_net_sysctl(net, path, table);
6834         if (!p->sysctl_header)
6835                 goto free;
6836
6837         if (!strcmp(dev_name, "all"))
6838                 ifindex = NETCONFA_IFINDEX_ALL;
6839         else if (!strcmp(dev_name, "default"))
6840                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6841         else
6842                 ifindex = idev->dev->ifindex;
6843         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6844                                      ifindex, p);
6845         return 0;
6846
6847 free:
6848         kfree(table);
6849 out:
6850         return -ENOBUFS;
6851 }
6852
6853 static void __addrconf_sysctl_unregister(struct net *net,
6854                                          struct ipv6_devconf *p, int ifindex)
6855 {
6856         struct ctl_table *table;
6857
6858         if (!p->sysctl_header)
6859                 return;
6860
6861         table = p->sysctl_header->ctl_table_arg;
6862         unregister_net_sysctl_table(p->sysctl_header);
6863         p->sysctl_header = NULL;
6864         kfree(table);
6865
6866         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6867 }
6868
6869 static int addrconf_sysctl_register(struct inet6_dev *idev)
6870 {
6871         int err;
6872
6873         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6874                 return -EINVAL;
6875
6876         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6877                                     &ndisc_ifinfo_sysctl_change);
6878         if (err)
6879                 return err;
6880         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6881                                          idev, &idev->cnf);
6882         if (err)
6883                 neigh_sysctl_unregister(idev->nd_parms);
6884
6885         return err;
6886 }
6887
6888 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6889 {
6890         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6891                                      idev->dev->ifindex);
6892         neigh_sysctl_unregister(idev->nd_parms);
6893 }
6894
6895
6896 #endif
6897
6898 static int __net_init addrconf_init_net(struct net *net)
6899 {
6900         int err = -ENOMEM;
6901         struct ipv6_devconf *all, *dflt;
6902
6903         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6904         if (!all)
6905                 goto err_alloc_all;
6906
6907         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6908         if (!dflt)
6909                 goto err_alloc_dflt;
6910
6911         if (IS_ENABLED(CONFIG_SYSCTL) &&
6912             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6913                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6914                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6915         }
6916
6917         /* these will be inherited by all namespaces */
6918         dflt->autoconf = ipv6_defaults.autoconf;
6919         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6920
6921         dflt->stable_secret.initialized = false;
6922         all->stable_secret.initialized = false;
6923
6924         net->ipv6.devconf_all = all;
6925         net->ipv6.devconf_dflt = dflt;
6926
6927 #ifdef CONFIG_SYSCTL
6928         err = __addrconf_sysctl_register(net, "all", NULL, all);
6929         if (err < 0)
6930                 goto err_reg_all;
6931
6932         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6933         if (err < 0)
6934                 goto err_reg_dflt;
6935 #endif
6936         return 0;
6937
6938 #ifdef CONFIG_SYSCTL
6939 err_reg_dflt:
6940         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6941 err_reg_all:
6942         kfree(dflt);
6943 #endif
6944 err_alloc_dflt:
6945         kfree(all);
6946 err_alloc_all:
6947         return err;
6948 }
6949
6950 static void __net_exit addrconf_exit_net(struct net *net)
6951 {
6952 #ifdef CONFIG_SYSCTL
6953         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6954                                      NETCONFA_IFINDEX_DEFAULT);
6955         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6956                                      NETCONFA_IFINDEX_ALL);
6957 #endif
6958         kfree(net->ipv6.devconf_dflt);
6959         kfree(net->ipv6.devconf_all);
6960 }
6961
6962 static struct pernet_operations addrconf_ops = {
6963         .init = addrconf_init_net,
6964         .exit = addrconf_exit_net,
6965 };
6966
6967 static struct rtnl_af_ops inet6_ops __read_mostly = {
6968         .family           = AF_INET6,
6969         .fill_link_af     = inet6_fill_link_af,
6970         .get_link_af_size = inet6_get_link_af_size,
6971         .validate_link_af = inet6_validate_link_af,
6972         .set_link_af      = inet6_set_link_af,
6973 };
6974
6975 /*
6976  *      Init / cleanup code
6977  */
6978
6979 int __init addrconf_init(void)
6980 {
6981         struct inet6_dev *idev;
6982         int i, err;
6983
6984         err = ipv6_addr_label_init();
6985         if (err < 0) {
6986                 pr_crit("%s: cannot initialize default policy table: %d\n",
6987                         __func__, err);
6988                 goto out;
6989         }
6990
6991         err = register_pernet_subsys(&addrconf_ops);
6992         if (err < 0)
6993                 goto out_addrlabel;
6994
6995         addrconf_wq = create_workqueue("ipv6_addrconf");
6996         if (!addrconf_wq) {
6997                 err = -ENOMEM;
6998                 goto out_nowq;
6999         }
7000
7001         /* The addrconf netdev notifier requires that loopback_dev
7002          * has it's ipv6 private information allocated and setup
7003          * before it can bring up and give link-local addresses
7004          * to other devices which are up.
7005          *
7006          * Unfortunately, loopback_dev is not necessarily the first
7007          * entry in the global dev_base list of net devices.  In fact,
7008          * it is likely to be the very last entry on that list.
7009          * So this causes the notifier registry below to try and
7010          * give link-local addresses to all devices besides loopback_dev
7011          * first, then loopback_dev, which cases all the non-loopback_dev
7012          * devices to fail to get a link-local address.
7013          *
7014          * So, as a temporary fix, allocate the ipv6 structure for
7015          * loopback_dev first by hand.
7016          * Longer term, all of the dependencies ipv6 has upon the loopback
7017          * device and it being up should be removed.
7018          */
7019         rtnl_lock();
7020         idev = ipv6_add_dev(init_net.loopback_dev);
7021         rtnl_unlock();
7022         if (IS_ERR(idev)) {
7023                 err = PTR_ERR(idev);
7024                 goto errlo;
7025         }
7026
7027         ip6_route_init_special_entries();
7028
7029         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7030                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7031
7032         register_netdevice_notifier(&ipv6_dev_notf);
7033
7034         addrconf_verify();
7035
7036         rtnl_af_register(&inet6_ops);
7037
7038         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7039                                    NULL, inet6_dump_ifinfo, 0);
7040         if (err < 0)
7041                 goto errout;
7042
7043         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7044                                    inet6_rtm_newaddr, NULL, 0);
7045         if (err < 0)
7046                 goto errout;
7047         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7048                                    inet6_rtm_deladdr, NULL, 0);
7049         if (err < 0)
7050                 goto errout;
7051         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7052                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7053                                    RTNL_FLAG_DOIT_UNLOCKED);
7054         if (err < 0)
7055                 goto errout;
7056         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7057                                    NULL, inet6_dump_ifmcaddr, 0);
7058         if (err < 0)
7059                 goto errout;
7060         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7061                                    NULL, inet6_dump_ifacaddr, 0);
7062         if (err < 0)
7063                 goto errout;
7064         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7065                                    inet6_netconf_get_devconf,
7066                                    inet6_netconf_dump_devconf,
7067                                    RTNL_FLAG_DOIT_UNLOCKED);
7068         if (err < 0)
7069                 goto errout;
7070         err = ipv6_addr_label_rtnl_register();
7071         if (err < 0)
7072                 goto errout;
7073
7074         return 0;
7075 errout:
7076         rtnl_unregister_all(PF_INET6);
7077         rtnl_af_unregister(&inet6_ops);
7078         unregister_netdevice_notifier(&ipv6_dev_notf);
7079 errlo:
7080         destroy_workqueue(addrconf_wq);
7081 out_nowq:
7082         unregister_pernet_subsys(&addrconf_ops);
7083 out_addrlabel:
7084         ipv6_addr_label_cleanup();
7085 out:
7086         return err;
7087 }
7088
7089 void addrconf_cleanup(void)
7090 {
7091         struct net_device *dev;
7092         int i;
7093
7094         unregister_netdevice_notifier(&ipv6_dev_notf);
7095         unregister_pernet_subsys(&addrconf_ops);
7096         ipv6_addr_label_cleanup();
7097
7098         rtnl_af_unregister(&inet6_ops);
7099
7100         rtnl_lock();
7101
7102         /* clean dev list */
7103         for_each_netdev(&init_net, dev) {
7104                 if (__in6_dev_get(dev) == NULL)
7105                         continue;
7106                 addrconf_ifdown(dev, 1);
7107         }
7108         addrconf_ifdown(init_net.loopback_dev, 2);
7109
7110         /*
7111          *      Check hash table.
7112          */
7113         spin_lock_bh(&addrconf_hash_lock);
7114         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7115                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7116         spin_unlock_bh(&addrconf_hash_lock);
7117         cancel_delayed_work(&addr_chk_work);
7118         rtnl_unlock();
7119
7120         destroy_workqueue(addrconf_wq);
7121 }