net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92 #include <linux/ioam6.h>
  93
  94 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  95
  96 #define IPV6_MAX_STRLEN \
  97         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  98
  99 static inline u32 cstamp_delta(unsigned long cstamp)
 100 {
 101         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 102 }
 103
 104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 105 {
 106         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 107         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 108         do_div(tmp, 1000000);
 109         return (s32)tmp;
 110 }
 111
 112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 113 {
 114         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 115         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 116         do_div(tmp, 1000000);
 117         if ((s32)tmp > mrt) {
 118                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 119                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 120                 do_div(tmp, 1000000);
 121         }
 122         return (s32)tmp;
 123 }
 124
 125 #ifdef CONFIG_SYSCTL
 126 static int addrconf_sysctl_register(struct inet6_dev *idev);
 127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 128 #else
 129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 130 {
 131         return 0;
 132 }
 133
 134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 135 {
 136 }
 137 #endif
 138
 139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 140
 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 142 static int ipv6_count_addresses(const struct inet6_dev *idev);
 143 static int ipv6_generate_stable_address(struct in6_addr *addr,
 144                                         u8 dad_count,
 145                                         const struct inet6_dev *idev);
 146
 147 #define IN6_ADDR_HSIZE_SHIFT    8
 148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 149 /*
 150  *      Configured unicast address hash table
 151  */
 152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 153 static DEFINE_SPINLOCK(addrconf_hash_lock);
 154
 155 static void addrconf_verify(void);
 156 static void addrconf_verify_rtnl(void);
 157 static void addrconf_verify_work(struct work_struct *);
 158
 159 static struct workqueue_struct *addrconf_wq;
 160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 161
 162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 164
 165 static void addrconf_type_change(struct net_device *dev,
 166                                  unsigned long event);
 167 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 168
 169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 170                                                   int plen,
 171                                                   const struct net_device *dev,
 172                                                   u32 flags, u32 noflags,
 173                                                   bool no_gw);
 174
 175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 176 static void addrconf_dad_work(struct work_struct *w);
 177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 178                                    bool send_na);
 179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 180 static void addrconf_rs_timer(struct timer_list *t);
 181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 183
 184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 185                                 struct prefix_info *pinfo);
 186
 187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 188         .forwarding             = 0,
 189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 190         .mtu6                   = IPV6_MIN_MTU,
 191         .accept_ra              = 1,
 192         .accept_redirects       = 1,
 193         .autoconf               = 1,
 194         .force_mld_version      = 0,
 195         .mldv1_unsolicited_report_interval = 10 * HZ,
 196         .mldv2_unsolicited_report_interval = HZ,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 210         .accept_ra_from_local   = 0,
 211         .accept_ra_min_hop_limit= 1,
 212         .accept_ra_pinfo        = 1,
 213 #ifdef CONFIG_IPV6_ROUTER_PREF
 214         .accept_ra_rtr_pref     = 1,
 215         .rtr_probe_interval     = 60 * HZ,
 216 #ifdef CONFIG_IPV6_ROUTE_INFO
 217         .accept_ra_rt_info_min_plen = 0,
 218         .accept_ra_rt_info_max_plen = 0,
 219 #endif
 220 #endif
 221         .proxy_ndp              = 0,
 222         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 223         .disable_ipv6           = 0,
 224         .accept_dad             = 0,
 225         .suppress_frag_ndisc    = 1,
 226         .accept_ra_mtu          = 1,
 227         .stable_secret          = {
 228                 .initialized = false,
 229         },
 230         .use_oif_addrs_only     = 0,
 231         .ignore_routes_with_linkdown = 0,
 232         .keep_addr_on_down      = 0,
 233         .seg6_enabled           = 0,
 234 #ifdef CONFIG_IPV6_SEG6_HMAC
 235         .seg6_require_hmac      = 0,
 236 #endif
 237         .enhanced_dad           = 1,
 238         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 239         .disable_policy         = 0,
 240         .rpl_seg_enabled        = 0,
 241         .ioam6_enabled          = 0,
 242         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 243         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 244 };
 245
 246 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 247         .forwarding             = 0,
 248         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 249         .mtu6                   = IPV6_MIN_MTU,
 250         .accept_ra              = 1,
 251         .accept_redirects       = 1,
 252         .autoconf               = 1,
 253         .force_mld_version      = 0,
 254         .mldv1_unsolicited_report_interval = 10 * HZ,
 255         .mldv2_unsolicited_report_interval = HZ,
 256         .dad_transmits          = 1,
 257         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 258         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 259         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 260         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 261         .use_tempaddr           = 0,
 262         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 263         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 264         .regen_max_retry        = REGEN_MAX_RETRY,
 265         .max_desync_factor      = MAX_DESYNC_FACTOR,
 266         .max_addresses          = IPV6_MAX_ADDRESSES,
 267         .accept_ra_defrtr       = 1,
 268         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 269         .accept_ra_from_local   = 0,
 270         .accept_ra_min_hop_limit= 1,
 271         .accept_ra_pinfo        = 1,
 272 #ifdef CONFIG_IPV6_ROUTER_PREF
 273         .accept_ra_rtr_pref     = 1,
 274         .rtr_probe_interval     = 60 * HZ,
 275 #ifdef CONFIG_IPV6_ROUTE_INFO
 276         .accept_ra_rt_info_min_plen = 0,
 277         .accept_ra_rt_info_max_plen = 0,
 278 #endif
 279 #endif
 280         .proxy_ndp              = 0,
 281         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 282         .disable_ipv6           = 0,
 283         .accept_dad             = 1,
 284         .suppress_frag_ndisc    = 1,
 285         .accept_ra_mtu          = 1,
 286         .stable_secret          = {
 287                 .initialized = false,
 288         },
 289         .use_oif_addrs_only     = 0,
 290         .ignore_routes_with_linkdown = 0,
 291         .keep_addr_on_down      = 0,
 292         .seg6_enabled           = 0,
 293 #ifdef CONFIG_IPV6_SEG6_HMAC
 294         .seg6_require_hmac      = 0,
 295 #endif
 296         .enhanced_dad           = 1,
 297         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 298         .disable_policy         = 0,
 299         .rpl_seg_enabled        = 0,
 300         .ioam6_enabled          = 0,
 301         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 302         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 303 };
 304
 305 /* Check if link is ready: is it up and is a valid qdisc available */
 306 static inline bool addrconf_link_ready(const struct net_device *dev)
 307 {
 308         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 309 }
 310
 311 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 312 {
 313         if (del_timer(&idev->rs_timer))
 314                 __in6_dev_put(idev);
 315 }
 316
 317 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 318 {
 319         if (cancel_delayed_work(&ifp->dad_work))
 320                 __in6_ifa_put(ifp);
 321 }
 322
 323 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 324                                   unsigned long when)
 325 {
 326         if (!timer_pending(&idev->rs_timer))
 327                 in6_dev_hold(idev);
 328         mod_timer(&idev->rs_timer, jiffies + when);
 329 }
 330
 331 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 332                                    unsigned long delay)
 333 {
 334         in6_ifa_hold(ifp);
 335         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 336                 in6_ifa_put(ifp);
 337 }
 338
 339 static int snmp6_alloc_dev(struct inet6_dev *idev)
 340 {
 341         int i;
 342
 343         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 344         if (!idev->stats.ipv6)
 345                 goto err_ip;
 346
 347         for_each_possible_cpu(i) {
 348                 struct ipstats_mib *addrconf_stats;
 349                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 350                 u64_stats_init(&addrconf_stats->syncp);
 351         }
 352
 353
 354         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 355                                         GFP_KERNEL);
 356         if (!idev->stats.icmpv6dev)
 357                 goto err_icmp;
 358         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 359                                            GFP_KERNEL);
 360         if (!idev->stats.icmpv6msgdev)
 361                 goto err_icmpmsg;
 362
 363         return 0;
 364
 365 err_icmpmsg:
 366         kfree(idev->stats.icmpv6dev);
 367 err_icmp:
 368         free_percpu(idev->stats.ipv6);
 369 err_ip:
 370         return -ENOMEM;
 371 }
 372
 373 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 374 {
 375         struct inet6_dev *ndev;
 376         int err = -ENOMEM;
 377
 378         ASSERT_RTNL();
 379
 380         if (dev->mtu < IPV6_MIN_MTU)
 381                 return ERR_PTR(-EINVAL);
 382
 383         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 384         if (!ndev)
 385                 return ERR_PTR(err);
 386
 387         rwlock_init(&ndev->lock);
 388         ndev->dev = dev;
 389         INIT_LIST_HEAD(&ndev->addr_list);
 390         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 391         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 392
 393         if (ndev->cnf.stable_secret.initialized)
 394                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 395
 396         ndev->cnf.mtu6 = dev->mtu;
 397         ndev->ra_mtu = 0;
 398         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 399         if (!ndev->nd_parms) {
 400                 kfree(ndev);
 401                 return ERR_PTR(err);
 402         }
 403         if (ndev->cnf.forwarding)
 404                 dev_disable_lro(dev);
 405         /* We refer to the device */
 406         dev_hold(dev);
 407
 408         if (snmp6_alloc_dev(ndev) < 0) {
 409                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 410                            __func__);
 411                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 412                 dev_put(dev);
 413                 kfree(ndev);
 414                 return ERR_PTR(err);
 415         }
 416
 417         if (snmp6_register_dev(ndev) < 0) {
 418                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 419                            __func__, dev->name);
 420                 goto err_release;
 421         }
 422
 423         /* One reference from device. */
 424         refcount_set(&ndev->refcnt, 1);
 425
 426         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 427                 ndev->cnf.accept_dad = -1;
 428
 429 #if IS_ENABLED(CONFIG_IPV6_SIT)
 430         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 431                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 432                 ndev->cnf.rtr_solicits = 0;
 433         }
 434 #endif
 435
 436         INIT_LIST_HEAD(&ndev->tempaddr_list);
 437         ndev->desync_factor = U32_MAX;
 438         if ((dev->flags&IFF_LOOPBACK) ||
 439             dev->type == ARPHRD_TUNNEL ||
 440             dev->type == ARPHRD_TUNNEL6 ||
 441             dev->type == ARPHRD_SIT ||
 442             dev->type == ARPHRD_NONE) {
 443                 ndev->cnf.use_tempaddr = -1;
 444         }
 445
 446         ndev->token = in6addr_any;
 447
 448         if (netif_running(dev) && addrconf_link_ready(dev))
 449                 ndev->if_flags |= IF_READY;
 450
 451         ipv6_mc_init_dev(ndev);
 452         ndev->tstamp = jiffies;
 453         err = addrconf_sysctl_register(ndev);
 454         if (err) {
 455                 ipv6_mc_destroy_dev(ndev);
 456                 snmp6_unregister_dev(ndev);
 457                 goto err_release;
 458         }
 459         /* protected by rtnl_lock */
 460         rcu_assign_pointer(dev->ip6_ptr, ndev);
 461
 462         /* Join interface-local all-node multicast group */
 463         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 464
 465         /* Join all-node multicast group */
 466         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 467
 468         /* Join all-router multicast group if forwarding is set */
 469         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 470                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 471
 472         return ndev;
 473
 474 err_release:
 475         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 476         ndev->dead = 1;
 477         in6_dev_finish_destroy(ndev);
 478         return ERR_PTR(err);
 479 }
 480
 481 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 482 {
 483         struct inet6_dev *idev;
 484
 485         ASSERT_RTNL();
 486
 487         idev = __in6_dev_get(dev);
 488         if (!idev) {
 489                 idev = ipv6_add_dev(dev);
 490                 if (IS_ERR(idev))
 491                         return idev;
 492         }
 493
 494         if (dev->flags&IFF_UP)
 495                 ipv6_mc_up(idev);
 496         return idev;
 497 }
 498
 499 static int inet6_netconf_msgsize_devconf(int type)
 500 {
 501         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 502                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 503         bool all = false;
 504
 505         if (type == NETCONFA_ALL)
 506                 all = true;
 507
 508         if (all || type == NETCONFA_FORWARDING)
 509                 size += nla_total_size(4);
 510 #ifdef CONFIG_IPV6_MROUTE
 511         if (all || type == NETCONFA_MC_FORWARDING)
 512                 size += nla_total_size(4);
 513 #endif
 514         if (all || type == NETCONFA_PROXY_NEIGH)
 515                 size += nla_total_size(4);
 516
 517         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 518                 size += nla_total_size(4);
 519
 520         return size;
 521 }
 522
 523 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 524                                       struct ipv6_devconf *devconf, u32 portid,
 525                                       u32 seq, int event, unsigned int flags,
 526                                       int type)
 527 {
 528         struct nlmsghdr  *nlh;
 529         struct netconfmsg *ncm;
 530         bool all = false;
 531
 532         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 533                         flags);
 534         if (!nlh)
 535                 return -EMSGSIZE;
 536
 537         if (type == NETCONFA_ALL)
 538                 all = true;
 539
 540         ncm = nlmsg_data(nlh);
 541         ncm->ncm_family = AF_INET6;
 542
 543         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 544                 goto nla_put_failure;
 545
 546         if (!devconf)
 547                 goto out;
 548
 549         if ((all || type == NETCONFA_FORWARDING) &&
 550             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 551                 goto nla_put_failure;
 552 #ifdef CONFIG_IPV6_MROUTE
 553         if ((all || type == NETCONFA_MC_FORWARDING) &&
 554             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 555                         devconf->mc_forwarding) < 0)
 556                 goto nla_put_failure;
 557 #endif
 558         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 559             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 560                 goto nla_put_failure;
 561
 562         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 563             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 564                         devconf->ignore_routes_with_linkdown) < 0)
 565                 goto nla_put_failure;
 566
 567 out:
 568         nlmsg_end(skb, nlh);
 569         return 0;
 570
 571 nla_put_failure:
 572         nlmsg_cancel(skb, nlh);
 573         return -EMSGSIZE;
 574 }
 575
 576 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 577                                   int ifindex, struct ipv6_devconf *devconf)
 578 {
 579         struct sk_buff *skb;
 580         int err = -ENOBUFS;
 581
 582         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 583         if (!skb)
 584                 goto errout;
 585
 586         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 587                                          event, 0, type);
 588         if (err < 0) {
 589                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 590                 WARN_ON(err == -EMSGSIZE);
 591                 kfree_skb(skb);
 592                 goto errout;
 593         }
 594         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 595         return;
 596 errout:
 597         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 598 }
 599
 600 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 601         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 602         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 603         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 604         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 605 };
 606
 607 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 608                                        const struct nlmsghdr *nlh,
 609                                        struct nlattr **tb,
 610                                        struct netlink_ext_ack *extack)
 611 {
 612         int i, err;
 613
 614         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 615                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 616                 return -EINVAL;
 617         }
 618
 619         if (!netlink_strict_get_check(skb))
 620                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 621                                               tb, NETCONFA_MAX,
 622                                               devconf_ipv6_policy, extack);
 623
 624         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 625                                             tb, NETCONFA_MAX,
 626                                             devconf_ipv6_policy, extack);
 627         if (err)
 628                 return err;
 629
 630         for (i = 0; i <= NETCONFA_MAX; i++) {
 631                 if (!tb[i])
 632                         continue;
 633
 634                 switch (i) {
 635                 case NETCONFA_IFINDEX:
 636                         break;
 637                 default:
 638                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 639                         return -EINVAL;
 640                 }
 641         }
 642
 643         return 0;
 644 }
 645
 646 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 647                                      struct nlmsghdr *nlh,
 648                                      struct netlink_ext_ack *extack)
 649 {
 650         struct net *net = sock_net(in_skb->sk);
 651         struct nlattr *tb[NETCONFA_MAX+1];
 652         struct inet6_dev *in6_dev = NULL;
 653         struct net_device *dev = NULL;
 654         struct sk_buff *skb;
 655         struct ipv6_devconf *devconf;
 656         int ifindex;
 657         int err;
 658
 659         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 660         if (err < 0)
 661                 return err;
 662
 663         if (!tb[NETCONFA_IFINDEX])
 664                 return -EINVAL;
 665
 666         err = -EINVAL;
 667         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 668         switch (ifindex) {
 669         case NETCONFA_IFINDEX_ALL:
 670                 devconf = net->ipv6.devconf_all;
 671                 break;
 672         case NETCONFA_IFINDEX_DEFAULT:
 673                 devconf = net->ipv6.devconf_dflt;
 674                 break;
 675         default:
 676                 dev = dev_get_by_index(net, ifindex);
 677                 if (!dev)
 678                         return -EINVAL;
 679                 in6_dev = in6_dev_get(dev);
 680                 if (!in6_dev)
 681                         goto errout;
 682                 devconf = &in6_dev->cnf;
 683                 break;
 684         }
 685
 686         err = -ENOBUFS;
 687         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 688         if (!skb)
 689                 goto errout;
 690
 691         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 692                                          NETLINK_CB(in_skb).portid,
 693                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 694                                          NETCONFA_ALL);
 695         if (err < 0) {
 696                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 697                 WARN_ON(err == -EMSGSIZE);
 698                 kfree_skb(skb);
 699                 goto errout;
 700         }
 701         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 702 errout:
 703         if (in6_dev)
 704                 in6_dev_put(in6_dev);
 705         dev_put(dev);
 706         return err;
 707 }
 708
 709 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 710                                       struct netlink_callback *cb)
 711 {
 712         const struct nlmsghdr *nlh = cb->nlh;
 713         struct net *net = sock_net(skb->sk);
 714         int h, s_h;
 715         int idx, s_idx;
 716         struct net_device *dev;
 717         struct inet6_dev *idev;
 718         struct hlist_head *head;
 719
 720         if (cb->strict_check) {
 721                 struct netlink_ext_ack *extack = cb->extack;
 722                 struct netconfmsg *ncm;
 723
 724                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 725                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 726                         return -EINVAL;
 727                 }
 728
 729                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 730                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 731                         return -EINVAL;
 732                 }
 733         }
 734
 735         s_h = cb->args[0];
 736         s_idx = idx = cb->args[1];
 737
 738         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 739                 idx = 0;
 740                 head = &net->dev_index_head[h];
 741                 rcu_read_lock();
 742                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 743                           net->dev_base_seq;
 744                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 745                         if (idx < s_idx)
 746                                 goto cont;
 747                         idev = __in6_dev_get(dev);
 748                         if (!idev)
 749                                 goto cont;
 750
 751                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 752                                                        &idev->cnf,
 753                                                        NETLINK_CB(cb->skb).portid,
 754                                                        nlh->nlmsg_seq,
 755                                                        RTM_NEWNETCONF,
 756                                                        NLM_F_MULTI,
 757                                                        NETCONFA_ALL) < 0) {
 758                                 rcu_read_unlock();
 759                                 goto done;
 760                         }
 761                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 762 cont:
 763                         idx++;
 764                 }
 765                 rcu_read_unlock();
 766         }
 767         if (h == NETDEV_HASHENTRIES) {
 768                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 769                                                net->ipv6.devconf_all,
 770                                                NETLINK_CB(cb->skb).portid,
 771                                                nlh->nlmsg_seq,
 772                                                RTM_NEWNETCONF, NLM_F_MULTI,
 773                                                NETCONFA_ALL) < 0)
 774                         goto done;
 775                 else
 776                         h++;
 777         }
 778         if (h == NETDEV_HASHENTRIES + 1) {
 779                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 780                                                net->ipv6.devconf_dflt,
 781                                                NETLINK_CB(cb->skb).portid,
 782                                                nlh->nlmsg_seq,
 783                                                RTM_NEWNETCONF, NLM_F_MULTI,
 784                                                NETCONFA_ALL) < 0)
 785                         goto done;
 786                 else
 787                         h++;
 788         }
 789 done:
 790         cb->args[0] = h;
 791         cb->args[1] = idx;
 792
 793         return skb->len;
 794 }
 795
 796 #ifdef CONFIG_SYSCTL
 797 static void dev_forward_change(struct inet6_dev *idev)
 798 {
 799         struct net_device *dev;
 800         struct inet6_ifaddr *ifa;
 801
 802         if (!idev)
 803                 return;
 804         dev = idev->dev;
 805         if (idev->cnf.forwarding)
 806                 dev_disable_lro(dev);
 807         if (dev->flags & IFF_MULTICAST) {
 808                 if (idev->cnf.forwarding) {
 809                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 810                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 811                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 812                 } else {
 813                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 814                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 815                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 816                 }
 817         }
 818
 819         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 820                 if (ifa->flags&IFA_F_TENTATIVE)
 821                         continue;
 822                 if (idev->cnf.forwarding)
 823                         addrconf_join_anycast(ifa);
 824                 else
 825                         addrconf_leave_anycast(ifa);
 826         }
 827         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 828                                      NETCONFA_FORWARDING,
 829                                      dev->ifindex, &idev->cnf);
 830 }
 831
 832
 833 static void addrconf_forward_change(struct net *net, __s32 newf)
 834 {
 835         struct net_device *dev;
 836         struct inet6_dev *idev;
 837
 838         for_each_netdev(net, dev) {
 839                 idev = __in6_dev_get(dev);
 840                 if (idev) {
 841                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 842                         idev->cnf.forwarding = newf;
 843                         if (changed)
 844                                 dev_forward_change(idev);
 845                 }
 846         }
 847 }
 848
 849 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 850 {
 851         struct net *net;
 852         int old;
 853
 854         if (!rtnl_trylock())
 855                 return restart_syscall();
 856
 857         net = (struct net *)table->extra2;
 858         old = *p;
 859         *p = newf;
 860
 861         if (p == &net->ipv6.devconf_dflt->forwarding) {
 862                 if ((!newf) ^ (!old))
 863                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 864                                                      NETCONFA_FORWARDING,
 865                                                      NETCONFA_IFINDEX_DEFAULT,
 866                                                      net->ipv6.devconf_dflt);
 867                 rtnl_unlock();
 868                 return 0;
 869         }
 870
 871         if (p == &net->ipv6.devconf_all->forwarding) {
 872                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 873
 874                 net->ipv6.devconf_dflt->forwarding = newf;
 875                 if ((!newf) ^ (!old_dflt))
 876                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 877                                                      NETCONFA_FORWARDING,
 878                                                      NETCONFA_IFINDEX_DEFAULT,
 879                                                      net->ipv6.devconf_dflt);
 880
 881                 addrconf_forward_change(net, newf);
 882                 if ((!newf) ^ (!old))
 883                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 884                                                      NETCONFA_FORWARDING,
 885                                                      NETCONFA_IFINDEX_ALL,
 886                                                      net->ipv6.devconf_all);
 887         } else if ((!newf) ^ (!old))
 888                 dev_forward_change((struct inet6_dev *)table->extra1);
 889         rtnl_unlock();
 890
 891         if (newf)
 892                 rt6_purge_dflt_routers(net);
 893         return 1;
 894 }
 895
 896 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 897 {
 898         struct net_device *dev;
 899         struct inet6_dev *idev;
 900
 901         for_each_netdev(net, dev) {
 902                 idev = __in6_dev_get(dev);
 903                 if (idev) {
 904                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 905
 906                         idev->cnf.ignore_routes_with_linkdown = newf;
 907                         if (changed)
 908                                 inet6_netconf_notify_devconf(dev_net(dev),
 909                                                              RTM_NEWNETCONF,
 910                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 911                                                              dev->ifindex,
 912                                                              &idev->cnf);
 913                 }
 914         }
 915 }
 916
 917 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 918 {
 919         struct net *net;
 920         int old;
 921
 922         if (!rtnl_trylock())
 923                 return restart_syscall();
 924
 925         net = (struct net *)table->extra2;
 926         old = *p;
 927         *p = newf;
 928
 929         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 930                 if ((!newf) ^ (!old))
 931                         inet6_netconf_notify_devconf(net,
 932                                                      RTM_NEWNETCONF,
 933                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 934                                                      NETCONFA_IFINDEX_DEFAULT,
 935                                                      net->ipv6.devconf_dflt);
 936                 rtnl_unlock();
 937                 return 0;
 938         }
 939
 940         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 941                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 942                 addrconf_linkdown_change(net, newf);
 943                 if ((!newf) ^ (!old))
 944                         inet6_netconf_notify_devconf(net,
 945                                                      RTM_NEWNETCONF,
 946                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 947                                                      NETCONFA_IFINDEX_ALL,
 948                                                      net->ipv6.devconf_all);
 949         }
 950         rtnl_unlock();
 951
 952         return 1;
 953 }
 954
 955 #endif
 956
 957 /* Nobody refers to this ifaddr, destroy it */
 958 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 959 {
 960         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 961
 962 #ifdef NET_REFCNT_DEBUG
 963         pr_debug("%s\n", __func__);
 964 #endif
 965
 966         in6_dev_put(ifp->idev);
 967
 968         if (cancel_delayed_work(&ifp->dad_work))
 969                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 970                           ifp);
 971
 972         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 973                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 974                 return;
 975         }
 976
 977         kfree_rcu(ifp, rcu);
 978 }
 979
 980 static void
 981 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 982 {
 983         struct list_head *p;
 984         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 985
 986         /*
 987          * Each device address list is sorted in order of scope -
 988          * global before linklocal.
 989          */
 990         list_for_each(p, &idev->addr_list) {
 991                 struct inet6_ifaddr *ifa
 992                         = list_entry(p, struct inet6_ifaddr, if_list);
 993                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 994                         break;
 995         }
 996
 997         list_add_tail_rcu(&ifp->if_list, p);
 998 }
 999
1000 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1001 {
1002         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1003
1004         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1005 }
1006
1007 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1008                                struct net_device *dev, unsigned int hash)
1009 {
1010         struct inet6_ifaddr *ifp;
1011
1012         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1013                 if (!net_eq(dev_net(ifp->idev->dev), net))
1014                         continue;
1015                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1016                         if (!dev || ifp->idev->dev == dev)
1017                                 return true;
1018                 }
1019         }
1020         return false;
1021 }
1022
1023 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1024 {
1025         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1026         int err = 0;
1027
1028         spin_lock(&addrconf_hash_lock);
1029
1030         /* Ignore adding duplicate addresses on an interface */
1031         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1032                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1033                 err = -EEXIST;
1034         } else {
1035                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1036         }
1037
1038         spin_unlock(&addrconf_hash_lock);
1039
1040         return err;
1041 }
1042
1043 /* On success it returns ifp with increased reference count */
1044
1045 static struct inet6_ifaddr *
1046 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1047               bool can_block, struct netlink_ext_ack *extack)
1048 {
1049         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1050         int addr_type = ipv6_addr_type(cfg->pfx);
1051         struct net *net = dev_net(idev->dev);
1052         struct inet6_ifaddr *ifa = NULL;
1053         struct fib6_info *f6i = NULL;
1054         int err = 0;
1055
1056         if (addr_type == IPV6_ADDR_ANY ||
1057             (addr_type & IPV6_ADDR_MULTICAST &&
1058              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1059             (!(idev->dev->flags & IFF_LOOPBACK) &&
1060              !netif_is_l3_master(idev->dev) &&
1061              addr_type & IPV6_ADDR_LOOPBACK))
1062                 return ERR_PTR(-EADDRNOTAVAIL);
1063
1064         if (idev->dead) {
1065                 err = -ENODEV;                  /*XXX*/
1066                 goto out;
1067         }
1068
1069         if (idev->cnf.disable_ipv6) {
1070                 err = -EACCES;
1071                 goto out;
1072         }
1073
1074         /* validator notifier needs to be blocking;
1075          * do not call in atomic context
1076          */
1077         if (can_block) {
1078                 struct in6_validator_info i6vi = {
1079                         .i6vi_addr = *cfg->pfx,
1080                         .i6vi_dev = idev,
1081                         .extack = extack,
1082                 };
1083
1084                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1085                 err = notifier_to_errno(err);
1086                 if (err < 0)
1087                         goto out;
1088         }
1089
1090         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1091         if (!ifa) {
1092                 err = -ENOBUFS;
1093                 goto out;
1094         }
1095
1096         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1097         if (IS_ERR(f6i)) {
1098                 err = PTR_ERR(f6i);
1099                 f6i = NULL;
1100                 goto out;
1101         }
1102
1103         if (net->ipv6.devconf_all->disable_policy ||
1104             idev->cnf.disable_policy)
1105                 f6i->dst_nopolicy = true;
1106
1107         neigh_parms_data_state_setall(idev->nd_parms);
1108
1109         ifa->addr = *cfg->pfx;
1110         if (cfg->peer_pfx)
1111                 ifa->peer_addr = *cfg->peer_pfx;
1112
1113         spin_lock_init(&ifa->lock);
1114         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1115         INIT_HLIST_NODE(&ifa->addr_lst);
1116         ifa->scope = cfg->scope;
1117         ifa->prefix_len = cfg->plen;
1118         ifa->rt_priority = cfg->rt_priority;
1119         ifa->flags = cfg->ifa_flags;
1120         /* No need to add the TENTATIVE flag for addresses with NODAD */
1121         if (!(cfg->ifa_flags & IFA_F_NODAD))
1122                 ifa->flags |= IFA_F_TENTATIVE;
1123         ifa->valid_lft = cfg->valid_lft;
1124         ifa->prefered_lft = cfg->preferred_lft;
1125         ifa->cstamp = ifa->tstamp = jiffies;
1126         ifa->tokenized = false;
1127
1128         ifa->rt = f6i;
1129
1130         ifa->idev = idev;
1131         in6_dev_hold(idev);
1132
1133         /* For caller */
1134         refcount_set(&ifa->refcnt, 1);
1135
1136         rcu_read_lock_bh();
1137
1138         err = ipv6_add_addr_hash(idev->dev, ifa);
1139         if (err < 0) {
1140                 rcu_read_unlock_bh();
1141                 goto out;
1142         }
1143
1144         write_lock(&idev->lock);
1145
1146         /* Add to inet6_dev unicast addr list. */
1147         ipv6_link_dev_addr(idev, ifa);
1148
1149         if (ifa->flags&IFA_F_TEMPORARY) {
1150                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1151                 in6_ifa_hold(ifa);
1152         }
1153
1154         in6_ifa_hold(ifa);
1155         write_unlock(&idev->lock);
1156
1157         rcu_read_unlock_bh();
1158
1159         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1160 out:
1161         if (unlikely(err < 0)) {
1162                 fib6_info_release(f6i);
1163
1164                 if (ifa) {
1165                         if (ifa->idev)
1166                                 in6_dev_put(ifa->idev);
1167                         kfree(ifa);
1168                 }
1169                 ifa = ERR_PTR(err);
1170         }
1171
1172         return ifa;
1173 }
1174
1175 enum cleanup_prefix_rt_t {
1176         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1177         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1178         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1179 };
1180
1181 /*
1182  * Check, whether the prefix for ifp would still need a prefix route
1183  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1184  * constants.
1185  *
1186  * 1) we don't purge prefix if address was not permanent.
1187  *    prefix is managed by its own lifetime.
1188  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1189  * 3) if there are no addresses, delete prefix.
1190  * 4) if there are still other permanent address(es),
1191  *    corresponding prefix is still permanent.
1192  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1193  *    don't purge the prefix, assume user space is managing it.
1194  * 6) otherwise, update prefix lifetime to the
1195  *    longest valid lifetime among the corresponding
1196  *    addresses on the device.
1197  *    Note: subsequent RA will update lifetime.
1198  **/
1199 static enum cleanup_prefix_rt_t
1200 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1201 {
1202         struct inet6_ifaddr *ifa;
1203         struct inet6_dev *idev = ifp->idev;
1204         unsigned long lifetime;
1205         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1206
1207         *expires = jiffies;
1208
1209         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1210                 if (ifa == ifp)
1211                         continue;
1212                 if (ifa->prefix_len != ifp->prefix_len ||
1213                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1214                                        ifp->prefix_len))
1215                         continue;
1216                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1217                         return CLEANUP_PREFIX_RT_NOP;
1218
1219                 action = CLEANUP_PREFIX_RT_EXPIRE;
1220
1221                 spin_lock(&ifa->lock);
1222
1223                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1224                 /*
1225                  * Note: Because this address is
1226                  * not permanent, lifetime <
1227                  * LONG_MAX / HZ here.
1228                  */
1229                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1230                         *expires = ifa->tstamp + lifetime * HZ;
1231                 spin_unlock(&ifa->lock);
1232         }
1233
1234         return action;
1235 }
1236
1237 static void
1238 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1239                      bool del_rt, bool del_peer)
1240 {
1241         struct fib6_info *f6i;
1242
1243         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1244                                         ifp->prefix_len,
1245                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1246         if (f6i) {
1247                 if (del_rt)
1248                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1249                 else {
1250                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1251                                 fib6_set_expires(f6i, expires);
1252                         fib6_info_release(f6i);
1253                 }
1254         }
1255 }
1256
1257
1258 /* This function wants to get referenced ifp and releases it before return */
1259
1260 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1261 {
1262         int state;
1263         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1264         unsigned long expires;
1265
1266         ASSERT_RTNL();
1267
1268         spin_lock_bh(&ifp->lock);
1269         state = ifp->state;
1270         ifp->state = INET6_IFADDR_STATE_DEAD;
1271         spin_unlock_bh(&ifp->lock);
1272
1273         if (state == INET6_IFADDR_STATE_DEAD)
1274                 goto out;
1275
1276         spin_lock_bh(&addrconf_hash_lock);
1277         hlist_del_init_rcu(&ifp->addr_lst);
1278         spin_unlock_bh(&addrconf_hash_lock);
1279
1280         write_lock_bh(&ifp->idev->lock);
1281
1282         if (ifp->flags&IFA_F_TEMPORARY) {
1283                 list_del(&ifp->tmp_list);
1284                 if (ifp->ifpub) {
1285                         in6_ifa_put(ifp->ifpub);
1286                         ifp->ifpub = NULL;
1287                 }
1288                 __in6_ifa_put(ifp);
1289         }
1290
1291         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1292                 action = check_cleanup_prefix_route(ifp, &expires);
1293
1294         list_del_rcu(&ifp->if_list);
1295         __in6_ifa_put(ifp);
1296
1297         write_unlock_bh(&ifp->idev->lock);
1298
1299         addrconf_del_dad_work(ifp);
1300
1301         ipv6_ifa_notify(RTM_DELADDR, ifp);
1302
1303         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1304
1305         if (action != CLEANUP_PREFIX_RT_NOP) {
1306                 cleanup_prefix_route(ifp, expires,
1307                         action == CLEANUP_PREFIX_RT_DEL, false);
1308         }
1309
1310         /* clean up prefsrc entries */
1311         rt6_remove_prefsrc(ifp);
1312 out:
1313         in6_ifa_put(ifp);
1314 }
1315
1316 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1317 {
1318         struct inet6_dev *idev = ifp->idev;
1319         unsigned long tmp_tstamp, age;
1320         unsigned long regen_advance;
1321         unsigned long now = jiffies;
1322         s32 cnf_temp_preferred_lft;
1323         struct inet6_ifaddr *ift;
1324         struct ifa6_config cfg;
1325         long max_desync_factor;
1326         struct in6_addr addr;
1327         int ret = 0;
1328
1329         write_lock_bh(&idev->lock);
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_gen_rnd_iid(&addr);
1354
1355         age = (now - ifp->tstamp) / HZ;
1356
1357         regen_advance = idev->cnf.regen_max_retry *
1358                         idev->cnf.dad_transmits *
1359                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / 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                 write_lock_bh(&idev->lock);
1420                 goto retry;
1421         }
1422
1423         spin_lock_bh(&ift->lock);
1424         ift->ifpub = ifp;
1425         ift->cstamp = now;
1426         ift->tstamp = tmp_tstamp;
1427         spin_unlock_bh(&ift->lock);
1428
1429         addrconf_dad_start(ift);
1430         in6_ifa_put(ift);
1431         in6_dev_put(idev);
1432 out:
1433         return ret;
1434 }
1435
1436 /*
1437  *      Choose an appropriate source address (RFC3484)
1438  */
1439 enum {
1440         IPV6_SADDR_RULE_INIT = 0,
1441         IPV6_SADDR_RULE_LOCAL,
1442         IPV6_SADDR_RULE_SCOPE,
1443         IPV6_SADDR_RULE_PREFERRED,
1444 #ifdef CONFIG_IPV6_MIP6
1445         IPV6_SADDR_RULE_HOA,
1446 #endif
1447         IPV6_SADDR_RULE_OIF,
1448         IPV6_SADDR_RULE_LABEL,
1449         IPV6_SADDR_RULE_PRIVACY,
1450         IPV6_SADDR_RULE_ORCHID,
1451         IPV6_SADDR_RULE_PREFIX,
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1454 #endif
1455         IPV6_SADDR_RULE_MAX
1456 };
1457
1458 struct ipv6_saddr_score {
1459         int                     rule;
1460         int                     addr_type;
1461         struct inet6_ifaddr     *ifa;
1462         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1463         int                     scopedist;
1464         int                     matchlen;
1465 };
1466
1467 struct ipv6_saddr_dst {
1468         const struct in6_addr *addr;
1469         int ifindex;
1470         int scope;
1471         int label;
1472         unsigned int prefs;
1473 };
1474
1475 static inline int ipv6_saddr_preferred(int type)
1476 {
1477         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1478                 return 1;
1479         return 0;
1480 }
1481
1482 static bool ipv6_use_optimistic_addr(struct net *net,
1483                                      struct inet6_dev *idev)
1484 {
1485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486         if (!idev)
1487                 return false;
1488         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1489                 return false;
1490         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1491                 return false;
1492
1493         return true;
1494 #else
1495         return false;
1496 #endif
1497 }
1498
1499 static bool ipv6_allow_optimistic_dad(struct net *net,
1500                                       struct inet6_dev *idev)
1501 {
1502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1503         if (!idev)
1504                 return false;
1505         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1506                 return false;
1507
1508         return true;
1509 #else
1510         return false;
1511 #endif
1512 }
1513
1514 static int ipv6_get_saddr_eval(struct net *net,
1515                                struct ipv6_saddr_score *score,
1516                                struct ipv6_saddr_dst *dst,
1517                                int i)
1518 {
1519         int ret;
1520
1521         if (i <= score->rule) {
1522                 switch (i) {
1523                 case IPV6_SADDR_RULE_SCOPE:
1524                         ret = score->scopedist;
1525                         break;
1526                 case IPV6_SADDR_RULE_PREFIX:
1527                         ret = score->matchlen;
1528                         break;
1529                 default:
1530                         ret = !!test_bit(i, score->scorebits);
1531                 }
1532                 goto out;
1533         }
1534
1535         switch (i) {
1536         case IPV6_SADDR_RULE_INIT:
1537                 /* Rule 0: remember if hiscore is not ready yet */
1538                 ret = !!score->ifa;
1539                 break;
1540         case IPV6_SADDR_RULE_LOCAL:
1541                 /* Rule 1: Prefer same address */
1542                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1543                 break;
1544         case IPV6_SADDR_RULE_SCOPE:
1545                 /* Rule 2: Prefer appropriate scope
1546                  *
1547                  *      ret
1548                  *       ^
1549                  *    -1 |  d 15
1550                  *    ---+--+-+---> scope
1551                  *       |
1552                  *       |             d is scope of the destination.
1553                  *  B-d  |  \
1554                  *       |   \      <- smaller scope is better if
1555                  *  B-15 |    \        if scope is enough for destination.
1556                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1557                  * d-C-1 | /
1558                  *       |/         <- greater is better
1559                  *   -C  /             if scope is not enough for destination.
1560                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1561                  *
1562                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1563                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1564                  * Assume B = 0 and we get C > 29.
1565                  */
1566                 ret = __ipv6_addr_src_scope(score->addr_type);
1567                 if (ret >= dst->scope)
1568                         ret = -ret;
1569                 else
1570                         ret -= 128;     /* 30 is enough */
1571                 score->scopedist = ret;
1572                 break;
1573         case IPV6_SADDR_RULE_PREFERRED:
1574             {
1575                 /* Rule 3: Avoid deprecated and optimistic addresses */
1576                 u8 avoid = IFA_F_DEPRECATED;
1577
1578                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1579                         avoid |= IFA_F_OPTIMISTIC;
1580                 ret = ipv6_saddr_preferred(score->addr_type) ||
1581                       !(score->ifa->flags & avoid);
1582                 break;
1583             }
1584 #ifdef CONFIG_IPV6_MIP6
1585         case IPV6_SADDR_RULE_HOA:
1586             {
1587                 /* Rule 4: Prefer home address */
1588                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1589                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1590                 break;
1591             }
1592 #endif
1593         case IPV6_SADDR_RULE_OIF:
1594                 /* Rule 5: Prefer outgoing interface */
1595                 ret = (!dst->ifindex ||
1596                        dst->ifindex == score->ifa->idev->dev->ifindex);
1597                 break;
1598         case IPV6_SADDR_RULE_LABEL:
1599                 /* Rule 6: Prefer matching label */
1600                 ret = ipv6_addr_label(net,
1601                                       &score->ifa->addr, score->addr_type,
1602                                       score->ifa->idev->dev->ifindex) == dst->label;
1603                 break;
1604         case IPV6_SADDR_RULE_PRIVACY:
1605             {
1606                 /* Rule 7: Prefer public address
1607                  * Note: prefer temporary address if use_tempaddr >= 2
1608                  */
1609                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1610                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1611                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1612                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1613                 break;
1614             }
1615         case IPV6_SADDR_RULE_ORCHID:
1616                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1617                  *          non-ORCHID vs non-ORCHID
1618                  */
1619                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1620                         ipv6_addr_orchid(dst->addr));
1621                 break;
1622         case IPV6_SADDR_RULE_PREFIX:
1623                 /* Rule 8: Use longest matching prefix */
1624                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1625                 if (ret > score->ifa->prefix_len)
1626                         ret = score->ifa->prefix_len;
1627                 score->matchlen = ret;
1628                 break;
1629 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1630         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1631                 /* Optimistic addresses still have lower precedence than other
1632                  * preferred addresses.
1633                  */
1634                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1635                 break;
1636 #endif
1637         default:
1638                 ret = 0;
1639         }
1640
1641         if (ret)
1642                 __set_bit(i, score->scorebits);
1643         score->rule = i;
1644 out:
1645         return ret;
1646 }
1647
1648 static int __ipv6_dev_get_saddr(struct net *net,
1649                                 struct ipv6_saddr_dst *dst,
1650                                 struct inet6_dev *idev,
1651                                 struct ipv6_saddr_score *scores,
1652                                 int hiscore_idx)
1653 {
1654         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1655
1656         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1657                 int i;
1658
1659                 /*
1660                  * - Tentative Address (RFC2462 section 5.4)
1661                  *  - A tentative address is not considered
1662                  *    "assigned to an interface" in the traditional
1663                  *    sense, unless it is also flagged as optimistic.
1664                  * - Candidate Source Address (section 4)
1665                  *  - In any case, anycast addresses, multicast
1666                  *    addresses, and the unspecified address MUST
1667                  *    NOT be included in a candidate set.
1668                  */
1669                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1670                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1671                         continue;
1672
1673                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1674
1675                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1676                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1677                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1678                                             idev->dev->name);
1679                         continue;
1680                 }
1681
1682                 score->rule = -1;
1683                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1684
1685                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1686                         int minihiscore, miniscore;
1687
1688                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1689                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1690
1691                         if (minihiscore > miniscore) {
1692                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1693                                     score->scopedist > 0) {
1694                                         /*
1695                                          * special case:
1696                                          * each remaining entry
1697                                          * has too small (not enough)
1698                                          * scope, because ifa entries
1699                                          * are sorted by their scope
1700                                          * values.
1701                                          */
1702                                         goto out;
1703                                 }
1704                                 break;
1705                         } else if (minihiscore < miniscore) {
1706                                 swap(hiscore, score);
1707                                 hiscore_idx = 1 - hiscore_idx;
1708
1709                                 /* restore our iterator */
1710                                 score->ifa = hiscore->ifa;
1711
1712                                 break;
1713                         }
1714                 }
1715         }
1716 out:
1717         return hiscore_idx;
1718 }
1719
1720 static int ipv6_get_saddr_master(struct net *net,
1721                                  const struct net_device *dst_dev,
1722                                  const struct net_device *master,
1723                                  struct ipv6_saddr_dst *dst,
1724                                  struct ipv6_saddr_score *scores,
1725                                  int hiscore_idx)
1726 {
1727         struct inet6_dev *idev;
1728
1729         idev = __in6_dev_get(dst_dev);
1730         if (idev)
1731                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1732                                                    scores, hiscore_idx);
1733
1734         idev = __in6_dev_get(master);
1735         if (idev)
1736                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1737                                                    scores, hiscore_idx);
1738
1739         return hiscore_idx;
1740 }
1741
1742 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1743                        const struct in6_addr *daddr, unsigned int prefs,
1744                        struct in6_addr *saddr)
1745 {
1746         struct ipv6_saddr_score scores[2], *hiscore;
1747         struct ipv6_saddr_dst dst;
1748         struct inet6_dev *idev;
1749         struct net_device *dev;
1750         int dst_type;
1751         bool use_oif_addr = false;
1752         int hiscore_idx = 0;
1753         int ret = 0;
1754
1755         dst_type = __ipv6_addr_type(daddr);
1756         dst.addr = daddr;
1757         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1758         dst.scope = __ipv6_addr_src_scope(dst_type);
1759         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1760         dst.prefs = prefs;
1761
1762         scores[hiscore_idx].rule = -1;
1763         scores[hiscore_idx].ifa = NULL;
1764
1765         rcu_read_lock();
1766
1767         /* Candidate Source Address (section 4)
1768          *  - multicast and link-local destination address,
1769          *    the set of candidate source address MUST only
1770          *    include addresses assigned to interfaces
1771          *    belonging to the same link as the outgoing
1772          *    interface.
1773          * (- For site-local destination addresses, the
1774          *    set of candidate source addresses MUST only
1775          *    include addresses assigned to interfaces
1776          *    belonging to the same site as the outgoing
1777          *    interface.)
1778          *  - "It is RECOMMENDED that the candidate source addresses
1779          *    be the set of unicast addresses assigned to the
1780          *    interface that will be used to send to the destination
1781          *    (the 'outgoing' interface)." (RFC 6724)
1782          */
1783         if (dst_dev) {
1784                 idev = __in6_dev_get(dst_dev);
1785                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1786                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1787                     (idev && idev->cnf.use_oif_addrs_only)) {
1788                         use_oif_addr = true;
1789                 }
1790         }
1791
1792         if (use_oif_addr) {
1793                 if (idev)
1794                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1795         } else {
1796                 const struct net_device *master;
1797                 int master_idx = 0;
1798
1799                 /* if dst_dev exists and is enslaved to an L3 device, then
1800                  * prefer addresses from dst_dev and then the master over
1801                  * any other enslaved devices in the L3 domain.
1802                  */
1803                 master = l3mdev_master_dev_rcu(dst_dev);
1804                 if (master) {
1805                         master_idx = master->ifindex;
1806
1807                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1808                                                             master, &dst,
1809                                                             scores, hiscore_idx);
1810
1811                         if (scores[hiscore_idx].ifa)
1812                                 goto out;
1813                 }
1814
1815                 for_each_netdev_rcu(net, dev) {
1816                         /* only consider addresses on devices in the
1817                          * same L3 domain
1818                          */
1819                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1820                                 continue;
1821                         idev = __in6_dev_get(dev);
1822                         if (!idev)
1823                                 continue;
1824                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1825                 }
1826         }
1827
1828 out:
1829         hiscore = &scores[hiscore_idx];
1830         if (!hiscore->ifa)
1831                 ret = -EADDRNOTAVAIL;
1832         else
1833                 *saddr = hiscore->ifa->addr;
1834
1835         rcu_read_unlock();
1836         return ret;
1837 }
1838 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1839
1840 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1841                       u32 banned_flags)
1842 {
1843         struct inet6_ifaddr *ifp;
1844         int err = -EADDRNOTAVAIL;
1845
1846         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1847                 if (ifp->scope > IFA_LINK)
1848                         break;
1849                 if (ifp->scope == IFA_LINK &&
1850                     !(ifp->flags & banned_flags)) {
1851                         *addr = ifp->addr;
1852                         err = 0;
1853                         break;
1854                 }
1855         }
1856         return err;
1857 }
1858
1859 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1860                     u32 banned_flags)
1861 {
1862         struct inet6_dev *idev;
1863         int err = -EADDRNOTAVAIL;
1864
1865         rcu_read_lock();
1866         idev = __in6_dev_get(dev);
1867         if (idev) {
1868                 read_lock_bh(&idev->lock);
1869                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1870                 read_unlock_bh(&idev->lock);
1871         }
1872         rcu_read_unlock();
1873         return err;
1874 }
1875
1876 static int ipv6_count_addresses(const struct inet6_dev *idev)
1877 {
1878         const struct inet6_ifaddr *ifp;
1879         int cnt = 0;
1880
1881         rcu_read_lock();
1882         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1883                 cnt++;
1884         rcu_read_unlock();
1885         return cnt;
1886 }
1887
1888 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1889                   const struct net_device *dev, int strict)
1890 {
1891         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1892                                        strict, IFA_F_TENTATIVE);
1893 }
1894 EXPORT_SYMBOL(ipv6_chk_addr);
1895
1896 /* device argument is used to find the L3 domain of interest. If
1897  * skip_dev_check is set, then the ifp device is not checked against
1898  * the passed in dev argument. So the 2 cases for addresses checks are:
1899  *   1. does the address exist in the L3 domain that dev is part of
1900  *      (skip_dev_check = true), or
1901  *
1902  *   2. does the address exist on the specific device
1903  *      (skip_dev_check = false)
1904  */
1905 static struct net_device *
1906 __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         struct net_device *l3mdev, *ndev;
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                 ndev = ifp->idev->dev;
1923                 if (!net_eq(dev_net(ndev), net))
1924                         continue;
1925
1926                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1927                         continue;
1928
1929                 /* Decouple optimistic from tentative for evaluation here.
1930                  * Ban optimistic addresses explicitly, when required.
1931                  */
1932                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1933                             ? (ifp->flags&~IFA_F_TENTATIVE)
1934                             : ifp->flags;
1935                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1936                     !(ifp_flags&banned_flags) &&
1937                     (!dev || ndev == dev ||
1938                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1939                         rcu_read_unlock();
1940                         return ndev;
1941                 }
1942         }
1943
1944         rcu_read_unlock();
1945         return NULL;
1946 }
1947
1948 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1949                             const struct net_device *dev, bool skip_dev_check,
1950                             int strict, u32 banned_flags)
1951 {
1952         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1953                                          strict, banned_flags) ? 1 : 0;
1954 }
1955 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1956
1957
1958 /* Compares an address/prefix_len with addresses on device @dev.
1959  * If one is found it returns true.
1960  */
1961 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1962         const unsigned int prefix_len, struct net_device *dev)
1963 {
1964         const struct inet6_ifaddr *ifa;
1965         const struct inet6_dev *idev;
1966         bool ret = false;
1967
1968         rcu_read_lock();
1969         idev = __in6_dev_get(dev);
1970         if (idev) {
1971                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1972                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1973                         if (ret)
1974                                 break;
1975                 }
1976         }
1977         rcu_read_unlock();
1978
1979         return ret;
1980 }
1981 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1982
1983 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1984 {
1985         const struct inet6_ifaddr *ifa;
1986         const struct inet6_dev *idev;
1987         int     onlink;
1988
1989         onlink = 0;
1990         rcu_read_lock();
1991         idev = __in6_dev_get(dev);
1992         if (idev) {
1993                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1994                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1995                                                    ifa->prefix_len);
1996                         if (onlink)
1997                                 break;
1998                 }
1999         }
2000         rcu_read_unlock();
2001         return onlink;
2002 }
2003 EXPORT_SYMBOL(ipv6_chk_prefix);
2004
2005 /**
2006  * ipv6_dev_find - find the first device with a given source address.
2007  * @net: the net namespace
2008  * @addr: the source address
2009  * @dev: used to find the L3 domain of interest
2010  *
2011  * The caller should be protected by RCU, or RTNL.
2012  */
2013 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2014                                  struct net_device *dev)
2015 {
2016         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2017                                          IFA_F_TENTATIVE);
2018 }
2019 EXPORT_SYMBOL(ipv6_dev_find);
2020
2021 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2022                                      struct net_device *dev, int strict)
2023 {
2024         unsigned int hash = inet6_addr_hash(net, addr);
2025         struct inet6_ifaddr *ifp, *result = NULL;
2026
2027         rcu_read_lock();
2028         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2029                 if (!net_eq(dev_net(ifp->idev->dev), net))
2030                         continue;
2031                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2032                         if (!dev || ifp->idev->dev == dev ||
2033                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2034                                 result = ifp;
2035                                 in6_ifa_hold(ifp);
2036                                 break;
2037                         }
2038                 }
2039         }
2040         rcu_read_unlock();
2041
2042         return result;
2043 }
2044
2045 /* Gets referenced address, destroys ifaddr */
2046
2047 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2048 {
2049         if (dad_failed)
2050                 ifp->flags |= IFA_F_DADFAILED;
2051
2052         if (ifp->flags&IFA_F_TEMPORARY) {
2053                 struct inet6_ifaddr *ifpub;
2054                 spin_lock_bh(&ifp->lock);
2055                 ifpub = ifp->ifpub;
2056                 if (ifpub) {
2057                         in6_ifa_hold(ifpub);
2058                         spin_unlock_bh(&ifp->lock);
2059                         ipv6_create_tempaddr(ifpub, true);
2060                         in6_ifa_put(ifpub);
2061                 } else {
2062                         spin_unlock_bh(&ifp->lock);
2063                 }
2064                 ipv6_del_addr(ifp);
2065         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2066                 spin_lock_bh(&ifp->lock);
2067                 addrconf_del_dad_work(ifp);
2068                 ifp->flags |= IFA_F_TENTATIVE;
2069                 if (dad_failed)
2070                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2071                 spin_unlock_bh(&ifp->lock);
2072                 if (dad_failed)
2073                         ipv6_ifa_notify(0, ifp);
2074                 in6_ifa_put(ifp);
2075         } else {
2076                 ipv6_del_addr(ifp);
2077         }
2078 }
2079
2080 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2081 {
2082         int err = -ENOENT;
2083
2084         spin_lock_bh(&ifp->lock);
2085         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2086                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2087                 err = 0;
2088         }
2089         spin_unlock_bh(&ifp->lock);
2090
2091         return err;
2092 }
2093
2094 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2095 {
2096         struct inet6_dev *idev = ifp->idev;
2097         struct net *net = dev_net(ifp->idev->dev);
2098
2099         if (addrconf_dad_end(ifp)) {
2100                 in6_ifa_put(ifp);
2101                 return;
2102         }
2103
2104         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2105                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2106
2107         spin_lock_bh(&ifp->lock);
2108
2109         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2110                 struct in6_addr new_addr;
2111                 struct inet6_ifaddr *ifp2;
2112                 int retries = ifp->stable_privacy_retry + 1;
2113                 struct ifa6_config cfg = {
2114                         .pfx = &new_addr,
2115                         .plen = ifp->prefix_len,
2116                         .ifa_flags = ifp->flags,
2117                         .valid_lft = ifp->valid_lft,
2118                         .preferred_lft = ifp->prefered_lft,
2119                         .scope = ifp->scope,
2120                 };
2121
2122                 if (retries > net->ipv6.sysctl.idgen_retries) {
2123                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2124                                              ifp->idev->dev->name);
2125                         goto errdad;
2126                 }
2127
2128                 new_addr = ifp->addr;
2129                 if (ipv6_generate_stable_address(&new_addr, retries,
2130                                                  idev))
2131                         goto errdad;
2132
2133                 spin_unlock_bh(&ifp->lock);
2134
2135                 if (idev->cnf.max_addresses &&
2136                     ipv6_count_addresses(idev) >=
2137                     idev->cnf.max_addresses)
2138                         goto lock_errdad;
2139
2140                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2141                                      ifp->idev->dev->name);
2142
2143                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2144                 if (IS_ERR(ifp2))
2145                         goto lock_errdad;
2146
2147                 spin_lock_bh(&ifp2->lock);
2148                 ifp2->stable_privacy_retry = retries;
2149                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2150                 spin_unlock_bh(&ifp2->lock);
2151
2152                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2153                 in6_ifa_put(ifp2);
2154 lock_errdad:
2155                 spin_lock_bh(&ifp->lock);
2156         }
2157
2158 errdad:
2159         /* transition from _POSTDAD to _ERRDAD */
2160         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2161         spin_unlock_bh(&ifp->lock);
2162
2163         addrconf_mod_dad_work(ifp, 0);
2164         in6_ifa_put(ifp);
2165 }
2166
2167 /* Join to solicited addr multicast group.
2168  * caller must hold RTNL */
2169 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2170 {
2171         struct in6_addr maddr;
2172
2173         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2174                 return;
2175
2176         addrconf_addr_solict_mult(addr, &maddr);
2177         ipv6_dev_mc_inc(dev, &maddr);
2178 }
2179
2180 /* caller must hold RTNL */
2181 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2182 {
2183         struct in6_addr maddr;
2184
2185         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2186                 return;
2187
2188         addrconf_addr_solict_mult(addr, &maddr);
2189         __ipv6_dev_mc_dec(idev, &maddr);
2190 }
2191
2192 /* caller must hold RTNL */
2193 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2194 {
2195         struct in6_addr addr;
2196
2197         if (ifp->prefix_len >= 127) /* RFC 6164 */
2198                 return;
2199         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2200         if (ipv6_addr_any(&addr))
2201                 return;
2202         __ipv6_dev_ac_inc(ifp->idev, &addr);
2203 }
2204
2205 /* caller must hold RTNL */
2206 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2207 {
2208         struct in6_addr addr;
2209
2210         if (ifp->prefix_len >= 127) /* RFC 6164 */
2211                 return;
2212         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2213         if (ipv6_addr_any(&addr))
2214                 return;
2215         __ipv6_dev_ac_dec(ifp->idev, &addr);
2216 }
2217
2218 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2219 {
2220         switch (dev->addr_len) {
2221         case ETH_ALEN:
2222                 memcpy(eui, dev->dev_addr, 3);
2223                 eui[3] = 0xFF;
2224                 eui[4] = 0xFE;
2225                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2226                 break;
2227         case EUI64_ADDR_LEN:
2228                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2229                 eui[0] ^= 2;
2230                 break;
2231         default:
2232                 return -1;
2233         }
2234
2235         return 0;
2236 }
2237
2238 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2239 {
2240         union fwnet_hwaddr *ha;
2241
2242         if (dev->addr_len != FWNET_ALEN)
2243                 return -1;
2244
2245         ha = (union fwnet_hwaddr *)dev->dev_addr;
2246
2247         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2248         eui[0] ^= 2;
2249         return 0;
2250 }
2251
2252 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2253 {
2254         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2255         if (dev->addr_len != ARCNET_ALEN)
2256                 return -1;
2257         memset(eui, 0, 7);
2258         eui[7] = *(u8 *)dev->dev_addr;
2259         return 0;
2260 }
2261
2262 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2263 {
2264         if (dev->addr_len != INFINIBAND_ALEN)
2265                 return -1;
2266         memcpy(eui, dev->dev_addr + 12, 8);
2267         eui[0] |= 2;
2268         return 0;
2269 }
2270
2271 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2272 {
2273         if (addr == 0)
2274                 return -1;
2275         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2276                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2277                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2278                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2279                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2280                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2281         eui[1] = 0;
2282         eui[2] = 0x5E;
2283         eui[3] = 0xFE;
2284         memcpy(eui + 4, &addr, 4);
2285         return 0;
2286 }
2287
2288 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2289 {
2290         if (dev->priv_flags & IFF_ISATAP)
2291                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2292         return -1;
2293 }
2294
2295 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2296 {
2297         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2298 }
2299
2300 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2301 {
2302         memcpy(eui, dev->perm_addr, 3);
2303         memcpy(eui + 5, dev->perm_addr + 3, 3);
2304         eui[3] = 0xFF;
2305         eui[4] = 0xFE;
2306         eui[0] ^= 2;
2307         return 0;
2308 }
2309
2310 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2311 {
2312         switch (dev->type) {
2313         case ARPHRD_ETHER:
2314         case ARPHRD_FDDI:
2315                 return addrconf_ifid_eui48(eui, dev);
2316         case ARPHRD_ARCNET:
2317                 return addrconf_ifid_arcnet(eui, dev);
2318         case ARPHRD_INFINIBAND:
2319                 return addrconf_ifid_infiniband(eui, dev);
2320         case ARPHRD_SIT:
2321                 return addrconf_ifid_sit(eui, dev);
2322         case ARPHRD_IPGRE:
2323         case ARPHRD_TUNNEL:
2324                 return addrconf_ifid_gre(eui, dev);
2325         case ARPHRD_6LOWPAN:
2326                 return addrconf_ifid_6lowpan(eui, dev);
2327         case ARPHRD_IEEE1394:
2328                 return addrconf_ifid_ieee1394(eui, dev);
2329         case ARPHRD_TUNNEL6:
2330         case ARPHRD_IP6GRE:
2331         case ARPHRD_RAWIP:
2332                 return addrconf_ifid_ip6tnl(eui, dev);
2333         }
2334         return -1;
2335 }
2336
2337 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2338 {
2339         int err = -1;
2340         struct inet6_ifaddr *ifp;
2341
2342         read_lock_bh(&idev->lock);
2343         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2344                 if (ifp->scope > IFA_LINK)
2345                         break;
2346                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2347                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2348                         err = 0;
2349                         break;
2350                 }
2351         }
2352         read_unlock_bh(&idev->lock);
2353         return err;
2354 }
2355
2356 /* Generation of a randomized Interface Identifier
2357  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2358  */
2359
2360 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2361 {
2362 regen:
2363         get_random_bytes(&addr->s6_addr[8], 8);
2364
2365         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2366          * check if generated address is not inappropriate:
2367          *
2368          * - Reserved IPv6 Interface Identifiers
2369          * - XXX: already assigned to an address on the device
2370          */
2371
2372         /* Subnet-router anycast: 0000:0000:0000:0000 */
2373         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2374                 goto regen;
2375
2376         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2377          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2378          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2379          */
2380         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2381             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2382                 goto regen;
2383
2384         /* Reserved subnet anycast addresses */
2385         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2386             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2387                 goto regen;
2388 }
2389
2390 /*
2391  *      Add prefix route.
2392  */
2393
2394 static void
2395 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2396                       struct net_device *dev, unsigned long expires,
2397                       u32 flags, gfp_t gfp_flags)
2398 {
2399         struct fib6_config cfg = {
2400                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2401                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2402                 .fc_ifindex = dev->ifindex,
2403                 .fc_expires = expires,
2404                 .fc_dst_len = plen,
2405                 .fc_flags = RTF_UP | flags,
2406                 .fc_nlinfo.nl_net = dev_net(dev),
2407                 .fc_protocol = RTPROT_KERNEL,
2408                 .fc_type = RTN_UNICAST,
2409         };
2410
2411         cfg.fc_dst = *pfx;
2412
2413         /* Prevent useless cloning on PtP SIT.
2414            This thing is done here expecting that the whole
2415            class of non-broadcast devices need not cloning.
2416          */
2417 #if IS_ENABLED(CONFIG_IPV6_SIT)
2418         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2419                 cfg.fc_flags |= RTF_NONEXTHOP;
2420 #endif
2421
2422         ip6_route_add(&cfg, gfp_flags, NULL);
2423 }
2424
2425
2426 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2427                                                   int plen,
2428                                                   const struct net_device *dev,
2429                                                   u32 flags, u32 noflags,
2430                                                   bool no_gw)
2431 {
2432         struct fib6_node *fn;
2433         struct fib6_info *rt = NULL;
2434         struct fib6_table *table;
2435         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2436
2437         table = fib6_get_table(dev_net(dev), tb_id);
2438         if (!table)
2439                 return NULL;
2440
2441         rcu_read_lock();
2442         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2443         if (!fn)
2444                 goto out;
2445
2446         for_each_fib6_node_rt_rcu(fn) {
2447                 /* prefix routes only use builtin fib6_nh */
2448                 if (rt->nh)
2449                         continue;
2450
2451                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2452                         continue;
2453                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2454                         continue;
2455                 if ((rt->fib6_flags & flags) != flags)
2456                         continue;
2457                 if ((rt->fib6_flags & noflags) != 0)
2458                         continue;
2459                 if (!fib6_info_hold_safe(rt))
2460                         continue;
2461                 break;
2462         }
2463 out:
2464         rcu_read_unlock();
2465         return rt;
2466 }
2467
2468
2469 /* Create "default" multicast route to the interface */
2470
2471 static void addrconf_add_mroute(struct net_device *dev)
2472 {
2473         struct fib6_config cfg = {
2474                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2475                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2476                 .fc_ifindex = dev->ifindex,
2477                 .fc_dst_len = 8,
2478                 .fc_flags = RTF_UP,
2479                 .fc_type = RTN_MULTICAST,
2480                 .fc_nlinfo.nl_net = dev_net(dev),
2481                 .fc_protocol = RTPROT_KERNEL,
2482         };
2483
2484         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2485
2486         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2487 }
2488
2489 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2490 {
2491         struct inet6_dev *idev;
2492
2493         ASSERT_RTNL();
2494
2495         idev = ipv6_find_idev(dev);
2496         if (IS_ERR(idev))
2497                 return idev;
2498
2499         if (idev->cnf.disable_ipv6)
2500                 return ERR_PTR(-EACCES);
2501
2502         /* Add default multicast route */
2503         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2504                 addrconf_add_mroute(dev);
2505
2506         return idev;
2507 }
2508
2509 static void manage_tempaddrs(struct inet6_dev *idev,
2510                              struct inet6_ifaddr *ifp,
2511                              __u32 valid_lft, __u32 prefered_lft,
2512                              bool create, unsigned long now)
2513 {
2514         u32 flags;
2515         struct inet6_ifaddr *ift;
2516
2517         read_lock_bh(&idev->lock);
2518         /* update all temporary addresses in the list */
2519         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2520                 int age, max_valid, max_prefered;
2521
2522                 if (ifp != ift->ifpub)
2523                         continue;
2524
2525                 /* RFC 4941 section 3.3:
2526                  * If a received option will extend the lifetime of a public
2527                  * address, the lifetimes of temporary addresses should
2528                  * be extended, subject to the overall constraint that no
2529                  * temporary addresses should ever remain "valid" or "preferred"
2530                  * for a time longer than (TEMP_VALID_LIFETIME) or
2531                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2532                  */
2533                 age = (now - ift->cstamp) / HZ;
2534                 max_valid = idev->cnf.temp_valid_lft - age;
2535                 if (max_valid < 0)
2536                         max_valid = 0;
2537
2538                 max_prefered = idev->cnf.temp_prefered_lft -
2539                                idev->desync_factor - age;
2540                 if (max_prefered < 0)
2541                         max_prefered = 0;
2542
2543                 if (valid_lft > max_valid)
2544                         valid_lft = max_valid;
2545
2546                 if (prefered_lft > max_prefered)
2547                         prefered_lft = max_prefered;
2548
2549                 spin_lock(&ift->lock);
2550                 flags = ift->flags;
2551                 ift->valid_lft = valid_lft;
2552                 ift->prefered_lft = prefered_lft;
2553                 ift->tstamp = now;
2554                 if (prefered_lft > 0)
2555                         ift->flags &= ~IFA_F_DEPRECATED;
2556
2557                 spin_unlock(&ift->lock);
2558                 if (!(flags&IFA_F_TENTATIVE))
2559                         ipv6_ifa_notify(0, ift);
2560         }
2561
2562         if ((create || list_empty(&idev->tempaddr_list)) &&
2563             idev->cnf.use_tempaddr > 0) {
2564                 /* When a new public address is created as described
2565                  * in [ADDRCONF], also create a new temporary address.
2566                  * Also create a temporary address if it's enabled but
2567                  * no temporary address currently exists.
2568                  */
2569                 read_unlock_bh(&idev->lock);
2570                 ipv6_create_tempaddr(ifp, false);
2571         } else {
2572                 read_unlock_bh(&idev->lock);
2573         }
2574 }
2575
2576 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2577 {
2578         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2579                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2580 }
2581
2582 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2583                                  const struct prefix_info *pinfo,
2584                                  struct inet6_dev *in6_dev,
2585                                  const struct in6_addr *addr, int addr_type,
2586                                  u32 addr_flags, bool sllao, bool tokenized,
2587                                  __u32 valid_lft, u32 prefered_lft)
2588 {
2589         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2590         int create = 0;
2591
2592         if (!ifp && valid_lft) {
2593                 int max_addresses = in6_dev->cnf.max_addresses;
2594                 struct ifa6_config cfg = {
2595                         .pfx = addr,
2596                         .plen = pinfo->prefix_len,
2597                         .ifa_flags = addr_flags,
2598                         .valid_lft = valid_lft,
2599                         .preferred_lft = prefered_lft,
2600                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2601                 };
2602
2603 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2604                 if ((net->ipv6.devconf_all->optimistic_dad ||
2605                      in6_dev->cnf.optimistic_dad) &&
2606                     !net->ipv6.devconf_all->forwarding && sllao)
2607                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2608 #endif
2609
2610                 /* Do not allow to create too much of autoconfigured
2611                  * addresses; this would be too easy way to crash kernel.
2612                  */
2613                 if (!max_addresses ||
2614                     ipv6_count_addresses(in6_dev) < max_addresses)
2615                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2616
2617                 if (IS_ERR_OR_NULL(ifp))
2618                         return -1;
2619
2620                 create = 1;
2621                 spin_lock_bh(&ifp->lock);
2622                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2623                 ifp->cstamp = jiffies;
2624                 ifp->tokenized = tokenized;
2625                 spin_unlock_bh(&ifp->lock);
2626                 addrconf_dad_start(ifp);
2627         }
2628
2629         if (ifp) {
2630                 u32 flags;
2631                 unsigned long now;
2632                 u32 stored_lft;
2633
2634                 /* Update lifetime (RFC4862 5.5.3 e)
2635                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2636                  * improve the reaction of SLAAC to renumbering events
2637                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2638                  */
2639                 spin_lock_bh(&ifp->lock);
2640                 now = jiffies;
2641                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2642                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2643                 else
2644                         stored_lft = 0;
2645
2646                 if (!create && stored_lft) {
2647                         ifp->valid_lft = valid_lft;
2648                         ifp->prefered_lft = prefered_lft;
2649                         ifp->tstamp = now;
2650                         flags = ifp->flags;
2651                         ifp->flags &= ~IFA_F_DEPRECATED;
2652                         spin_unlock_bh(&ifp->lock);
2653
2654                         if (!(flags&IFA_F_TENTATIVE))
2655                                 ipv6_ifa_notify(0, ifp);
2656                 } else
2657                         spin_unlock_bh(&ifp->lock);
2658
2659                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2660                                  create, now);
2661
2662                 in6_ifa_put(ifp);
2663                 addrconf_verify();
2664         }
2665
2666         return 0;
2667 }
2668 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2669
2670 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2671 {
2672         struct prefix_info *pinfo;
2673         __u32 valid_lft;
2674         __u32 prefered_lft;
2675         int addr_type, err;
2676         u32 addr_flags = 0;
2677         struct inet6_dev *in6_dev;
2678         struct net *net = dev_net(dev);
2679
2680         pinfo = (struct prefix_info *) opt;
2681
2682         if (len < sizeof(struct prefix_info)) {
2683                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2684                 return;
2685         }
2686
2687         /*
2688          *      Validation checks ([ADDRCONF], page 19)
2689          */
2690
2691         addr_type = ipv6_addr_type(&pinfo->prefix);
2692
2693         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2694                 return;
2695
2696         valid_lft = ntohl(pinfo->valid);
2697         prefered_lft = ntohl(pinfo->prefered);
2698
2699         if (prefered_lft > valid_lft) {
2700                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2701                 return;
2702         }
2703
2704         in6_dev = in6_dev_get(dev);
2705
2706         if (!in6_dev) {
2707                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2708                                     dev->name);
2709                 return;
2710         }
2711
2712         /*
2713          *      Two things going on here:
2714          *      1) Add routes for on-link prefixes
2715          *      2) Configure prefixes with the auto flag set
2716          */
2717
2718         if (pinfo->onlink) {
2719                 struct fib6_info *rt;
2720                 unsigned long rt_expires;
2721
2722                 /* Avoid arithmetic overflow. Really, we could
2723                  * save rt_expires in seconds, likely valid_lft,
2724                  * but it would require division in fib gc, that it
2725                  * not good.
2726                  */
2727                 if (HZ > USER_HZ)
2728                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2729                 else
2730                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2731
2732                 if (addrconf_finite_timeout(rt_expires))
2733                         rt_expires *= HZ;
2734
2735                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2736                                                pinfo->prefix_len,
2737                                                dev,
2738                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2739                                                RTF_DEFAULT, true);
2740
2741                 if (rt) {
2742                         /* Autoconf prefix route */
2743                         if (valid_lft == 0) {
2744                                 ip6_del_rt(net, rt, false);
2745                                 rt = NULL;
2746                         } else if (addrconf_finite_timeout(rt_expires)) {
2747                                 /* not infinity */
2748                                 fib6_set_expires(rt, jiffies + rt_expires);
2749                         } else {
2750                                 fib6_clean_expires(rt);
2751                         }
2752                 } else if (valid_lft) {
2753                         clock_t expires = 0;
2754                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2755                         if (addrconf_finite_timeout(rt_expires)) {
2756                                 /* not infinity */
2757                                 flags |= RTF_EXPIRES;
2758                                 expires = jiffies_to_clock_t(rt_expires);
2759                         }
2760                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2761                                               0, dev, expires, flags,
2762                                               GFP_ATOMIC);
2763                 }
2764                 fib6_info_release(rt);
2765         }
2766
2767         /* Try to figure out our local address for this prefix */
2768
2769         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2770                 struct in6_addr addr;
2771                 bool tokenized = false, dev_addr_generated = false;
2772
2773                 if (pinfo->prefix_len == 64) {
2774                         memcpy(&addr, &pinfo->prefix, 8);
2775
2776                         if (!ipv6_addr_any(&in6_dev->token)) {
2777                                 read_lock_bh(&in6_dev->lock);
2778                                 memcpy(addr.s6_addr + 8,
2779                                        in6_dev->token.s6_addr + 8, 8);
2780                                 read_unlock_bh(&in6_dev->lock);
2781                                 tokenized = true;
2782                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2783                                    !ipv6_generate_stable_address(&addr, 0,
2784                                                                  in6_dev)) {
2785                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2786                                 goto ok;
2787                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2788                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2789                                 goto put;
2790                         } else {
2791                                 dev_addr_generated = true;
2792                         }
2793                         goto ok;
2794                 }
2795                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2796                                     pinfo->prefix_len);
2797                 goto put;
2798
2799 ok:
2800                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2801                                                    &addr, addr_type,
2802                                                    addr_flags, sllao,
2803                                                    tokenized, valid_lft,
2804                                                    prefered_lft);
2805                 if (err)
2806                         goto put;
2807
2808                 /* Ignore error case here because previous prefix add addr was
2809                  * successful which will be notified.
2810                  */
2811                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2812                                               addr_type, addr_flags, sllao,
2813                                               tokenized, valid_lft,
2814                                               prefered_lft,
2815                                               dev_addr_generated);
2816         }
2817         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2818 put:
2819         in6_dev_put(in6_dev);
2820 }
2821
2822 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2823                 struct in6_ifreq *ireq)
2824 {
2825         struct ip_tunnel_parm p = { };
2826         int err;
2827
2828         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2829                 return -EADDRNOTAVAIL;
2830
2831         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2832         p.iph.version = 4;
2833         p.iph.ihl = 5;
2834         p.iph.protocol = IPPROTO_IPV6;
2835         p.iph.ttl = 64;
2836
2837         if (!dev->netdev_ops->ndo_tunnel_ctl)
2838                 return -EOPNOTSUPP;
2839         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2840         if (err)
2841                 return err;
2842
2843         dev = __dev_get_by_name(net, p.name);
2844         if (!dev)
2845                 return -ENOBUFS;
2846         return dev_open(dev, NULL);
2847 }
2848
2849 /*
2850  *      Set destination address.
2851  *      Special case for SIT interfaces where we create a new "virtual"
2852  *      device.
2853  */
2854 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2855 {
2856         struct net_device *dev;
2857         struct in6_ifreq ireq;
2858         int err = -ENODEV;
2859
2860         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2861                 return -ENODEV;
2862         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2863                 return -EFAULT;
2864
2865         rtnl_lock();
2866         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2867         if (dev && dev->type == ARPHRD_SIT)
2868                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
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                         int flag = scope;
3131
3132                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3133                                 addr.s6_addr32[3] = ifa->ifa_local;
3134
3135                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3136                                         continue;
3137                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3138                                         if (idev->dev->flags&IFF_POINTOPOINT)
3139                                                 continue;
3140                                         flag |= IFA_HOST;
3141                                 }
3142
3143                                 add_addr(idev, &addr, plen, flag);
3144                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3145                                                       0, pflags, GFP_KERNEL);
3146                         }
3147                 }
3148         }
3149 }
3150 #endif
3151
3152 static void init_loopback(struct net_device *dev)
3153 {
3154         struct inet6_dev  *idev;
3155
3156         /* ::1 */
3157
3158         ASSERT_RTNL();
3159
3160         idev = ipv6_find_idev(dev);
3161         if (IS_ERR(idev)) {
3162                 pr_debug("%s: add_dev failed\n", __func__);
3163                 return;
3164         }
3165
3166         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3167 }
3168
3169 void addrconf_add_linklocal(struct inet6_dev *idev,
3170                             const struct in6_addr *addr, u32 flags)
3171 {
3172         struct ifa6_config cfg = {
3173                 .pfx = addr,
3174                 .plen = 64,
3175                 .ifa_flags = flags | IFA_F_PERMANENT,
3176                 .valid_lft = INFINITY_LIFE_TIME,
3177                 .preferred_lft = INFINITY_LIFE_TIME,
3178                 .scope = IFA_LINK
3179         };
3180         struct inet6_ifaddr *ifp;
3181
3182 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3183         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3184              idev->cnf.optimistic_dad) &&
3185             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3186                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3187 #endif
3188
3189         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3190         if (!IS_ERR(ifp)) {
3191                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3192                                       0, 0, GFP_ATOMIC);
3193                 addrconf_dad_start(ifp);
3194                 in6_ifa_put(ifp);
3195         }
3196 }
3197 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3198
3199 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3200 {
3201         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3202                 return true;
3203
3204         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3205             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3206                 return true;
3207
3208         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3209             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3210                 return true;
3211
3212         return false;
3213 }
3214
3215 static int ipv6_generate_stable_address(struct in6_addr *address,
3216                                         u8 dad_count,
3217                                         const struct inet6_dev *idev)
3218 {
3219         static DEFINE_SPINLOCK(lock);
3220         static __u32 digest[SHA1_DIGEST_WORDS];
3221         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3222
3223         static union {
3224                 char __data[SHA1_BLOCK_SIZE];
3225                 struct {
3226                         struct in6_addr secret;
3227                         __be32 prefix[2];
3228                         unsigned char hwaddr[MAX_ADDR_LEN];
3229                         u8 dad_count;
3230                 } __packed;
3231         } data;
3232
3233         struct in6_addr secret;
3234         struct in6_addr temp;
3235         struct net *net = dev_net(idev->dev);
3236
3237         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3238
3239         if (idev->cnf.stable_secret.initialized)
3240                 secret = idev->cnf.stable_secret.secret;
3241         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3242                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3243         else
3244                 return -1;
3245
3246 retry:
3247         spin_lock_bh(&lock);
3248
3249         sha1_init(digest);
3250         memset(&data, 0, sizeof(data));
3251         memset(workspace, 0, sizeof(workspace));
3252         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3253         data.prefix[0] = address->s6_addr32[0];
3254         data.prefix[1] = address->s6_addr32[1];
3255         data.secret = secret;
3256         data.dad_count = dad_count;
3257
3258         sha1_transform(digest, data.__data, workspace);
3259
3260         temp = *address;
3261         temp.s6_addr32[2] = (__force __be32)digest[0];
3262         temp.s6_addr32[3] = (__force __be32)digest[1];
3263
3264         spin_unlock_bh(&lock);
3265
3266         if (ipv6_reserved_interfaceid(temp)) {
3267                 dad_count++;
3268                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3269                         return -1;
3270                 goto retry;
3271         }
3272
3273         *address = temp;
3274         return 0;
3275 }
3276
3277 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3278 {
3279         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3280
3281         if (s->initialized)
3282                 return;
3283         s = &idev->cnf.stable_secret;
3284         get_random_bytes(&s->secret, sizeof(s->secret));
3285         s->initialized = true;
3286 }
3287
3288 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3289 {
3290         struct in6_addr addr;
3291
3292         /* no link local addresses on L3 master devices */
3293         if (netif_is_l3_master(idev->dev))
3294                 return;
3295
3296         /* no link local addresses on devices flagged as slaves */
3297         if (idev->dev->flags & IFF_SLAVE)
3298                 return;
3299
3300         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3301
3302         switch (idev->cnf.addr_gen_mode) {
3303         case IN6_ADDR_GEN_MODE_RANDOM:
3304                 ipv6_gen_mode_random_init(idev);
3305                 fallthrough;
3306         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3307                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3308                         addrconf_add_linklocal(idev, &addr,
3309                                                IFA_F_STABLE_PRIVACY);
3310                 else if (prefix_route)
3311                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3312                                               0, 0, GFP_KERNEL);
3313                 break;
3314         case IN6_ADDR_GEN_MODE_EUI64:
3315                 /* addrconf_add_linklocal also adds a prefix_route and we
3316                  * only need to care about prefix routes if ipv6_generate_eui64
3317                  * couldn't generate one.
3318                  */
3319                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3320                         addrconf_add_linklocal(idev, &addr, 0);
3321                 else if (prefix_route)
3322                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3323                                               0, 0, GFP_KERNEL);
3324                 break;
3325         case IN6_ADDR_GEN_MODE_NONE:
3326         default:
3327                 /* will not add any link local address */
3328                 break;
3329         }
3330 }
3331
3332 static void addrconf_dev_config(struct net_device *dev)
3333 {
3334         struct inet6_dev *idev;
3335
3336         ASSERT_RTNL();
3337
3338         if ((dev->type != ARPHRD_ETHER) &&
3339             (dev->type != ARPHRD_FDDI) &&
3340             (dev->type != ARPHRD_ARCNET) &&
3341             (dev->type != ARPHRD_INFINIBAND) &&
3342             (dev->type != ARPHRD_IEEE1394) &&
3343             (dev->type != ARPHRD_TUNNEL6) &&
3344             (dev->type != ARPHRD_6LOWPAN) &&
3345             (dev->type != ARPHRD_IP6GRE) &&
3346             (dev->type != ARPHRD_IPGRE) &&
3347             (dev->type != ARPHRD_TUNNEL) &&
3348             (dev->type != ARPHRD_NONE) &&
3349             (dev->type != ARPHRD_RAWIP)) {
3350                 /* Alas, we support only Ethernet autoconfiguration. */
3351                 idev = __in6_dev_get(dev);
3352                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3353                     dev->flags & IFF_MULTICAST)
3354                         ipv6_mc_up(idev);
3355                 return;
3356         }
3357
3358         idev = addrconf_add_dev(dev);
3359         if (IS_ERR(idev))
3360                 return;
3361
3362         /* this device type has no EUI support */
3363         if (dev->type == ARPHRD_NONE &&
3364             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3365                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3366
3367         addrconf_addr_gen(idev, false);
3368 }
3369
3370 #if IS_ENABLED(CONFIG_IPV6_SIT)
3371 static void addrconf_sit_config(struct net_device *dev)
3372 {
3373         struct inet6_dev *idev;
3374
3375         ASSERT_RTNL();
3376
3377         /*
3378          * Configure the tunnel with one of our IPv4
3379          * addresses... we should configure all of
3380          * our v4 addrs in the tunnel
3381          */
3382
3383         idev = ipv6_find_idev(dev);
3384         if (IS_ERR(idev)) {
3385                 pr_debug("%s: add_dev failed\n", __func__);
3386                 return;
3387         }
3388
3389         if (dev->priv_flags & IFF_ISATAP) {
3390                 addrconf_addr_gen(idev, false);
3391                 return;
3392         }
3393
3394         sit_add_v4_addrs(idev);
3395
3396         if (dev->flags&IFF_POINTOPOINT)
3397                 addrconf_add_mroute(dev);
3398 }
3399 #endif
3400
3401 #if IS_ENABLED(CONFIG_NET_IPGRE)
3402 static void addrconf_gre_config(struct net_device *dev)
3403 {
3404         struct inet6_dev *idev;
3405
3406         ASSERT_RTNL();
3407
3408         idev = ipv6_find_idev(dev);
3409         if (IS_ERR(idev)) {
3410                 pr_debug("%s: add_dev failed\n", __func__);
3411                 return;
3412         }
3413
3414         addrconf_addr_gen(idev, true);
3415         if (dev->flags & IFF_POINTOPOINT)
3416                 addrconf_add_mroute(dev);
3417 }
3418 #endif
3419
3420 static int fixup_permanent_addr(struct net *net,
3421                                 struct inet6_dev *idev,
3422                                 struct inet6_ifaddr *ifp)
3423 {
3424         /* !fib6_node means the host route was removed from the
3425          * FIB, for example, if 'lo' device is taken down. In that
3426          * case regenerate the host route.
3427          */
3428         if (!ifp->rt || !ifp->rt->fib6_node) {
3429                 struct fib6_info *f6i, *prev;
3430
3431                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3432                                          GFP_ATOMIC);
3433                 if (IS_ERR(f6i))
3434                         return PTR_ERR(f6i);
3435
3436                 /* ifp->rt can be accessed outside of rtnl */
3437                 spin_lock(&ifp->lock);
3438                 prev = ifp->rt;
3439                 ifp->rt = f6i;
3440                 spin_unlock(&ifp->lock);
3441
3442                 fib6_info_release(prev);
3443         }
3444
3445         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3446                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3447                                       ifp->rt_priority, idev->dev, 0, 0,
3448                                       GFP_ATOMIC);
3449         }
3450
3451         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3452                 addrconf_dad_start(ifp);
3453
3454         return 0;
3455 }
3456
3457 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3458 {
3459         struct inet6_ifaddr *ifp, *tmp;
3460         struct inet6_dev *idev;
3461
3462         idev = __in6_dev_get(dev);
3463         if (!idev)
3464                 return;
3465
3466         write_lock_bh(&idev->lock);
3467
3468         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3469                 if ((ifp->flags & IFA_F_PERMANENT) &&
3470                     fixup_permanent_addr(net, idev, ifp) < 0) {
3471                         write_unlock_bh(&idev->lock);
3472                         in6_ifa_hold(ifp);
3473                         ipv6_del_addr(ifp);
3474                         write_lock_bh(&idev->lock);
3475
3476                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3477                                              idev->dev->name, &ifp->addr);
3478                 }
3479         }
3480
3481         write_unlock_bh(&idev->lock);
3482 }
3483
3484 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3485                            void *ptr)
3486 {
3487         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3488         struct netdev_notifier_change_info *change_info;
3489         struct netdev_notifier_changeupper_info *info;
3490         struct inet6_dev *idev = __in6_dev_get(dev);
3491         struct net *net = dev_net(dev);
3492         int run_pending = 0;
3493         int err;
3494
3495         switch (event) {
3496         case NETDEV_REGISTER:
3497                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3498                         idev = ipv6_add_dev(dev);
3499                         if (IS_ERR(idev))
3500                                 return notifier_from_errno(PTR_ERR(idev));
3501                 }
3502                 break;
3503
3504         case NETDEV_CHANGEMTU:
3505                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3506                 if (dev->mtu < IPV6_MIN_MTU) {
3507                         addrconf_ifdown(dev, dev != net->loopback_dev);
3508                         break;
3509                 }
3510
3511                 if (idev) {
3512                         rt6_mtu_change(dev, dev->mtu);
3513                         idev->cnf.mtu6 = dev->mtu;
3514                         break;
3515                 }
3516
3517                 /* allocate new idev */
3518                 idev = ipv6_add_dev(dev);
3519                 if (IS_ERR(idev))
3520                         break;
3521
3522                 /* device is still not ready */
3523                 if (!(idev->if_flags & IF_READY))
3524                         break;
3525
3526                 run_pending = 1;
3527                 fallthrough;
3528         case NETDEV_UP:
3529         case NETDEV_CHANGE:
3530                 if (dev->flags & IFF_SLAVE)
3531                         break;
3532
3533                 if (idev && idev->cnf.disable_ipv6)
3534                         break;
3535
3536                 if (event == NETDEV_UP) {
3537                         /* restore routes for permanent addresses */
3538                         addrconf_permanent_addr(net, dev);
3539
3540                         if (!addrconf_link_ready(dev)) {
3541                                 /* device is not ready yet. */
3542                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3543                                          dev->name);
3544                                 break;
3545                         }
3546
3547                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3548                                 idev = ipv6_add_dev(dev);
3549
3550                         if (!IS_ERR_OR_NULL(idev)) {
3551                                 idev->if_flags |= IF_READY;
3552                                 run_pending = 1;
3553                         }
3554                 } else if (event == NETDEV_CHANGE) {
3555                         if (!addrconf_link_ready(dev)) {
3556                                 /* device is still not ready. */
3557                                 rt6_sync_down_dev(dev, event);
3558                                 break;
3559                         }
3560
3561                         if (!IS_ERR_OR_NULL(idev)) {
3562                                 if (idev->if_flags & IF_READY) {
3563                                         /* device is already configured -
3564                                          * but resend MLD reports, we might
3565                                          * have roamed and need to update
3566                                          * multicast snooping switches
3567                                          */
3568                                         ipv6_mc_up(idev);
3569                                         change_info = ptr;
3570                                         if (change_info->flags_changed & IFF_NOARP)
3571                                                 addrconf_dad_run(idev, true);
3572                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3573                                         break;
3574                                 }
3575                                 idev->if_flags |= IF_READY;
3576                         }
3577
3578                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3579                                 dev->name);
3580
3581                         run_pending = 1;
3582                 }
3583
3584                 switch (dev->type) {
3585 #if IS_ENABLED(CONFIG_IPV6_SIT)
3586                 case ARPHRD_SIT:
3587                         addrconf_sit_config(dev);
3588                         break;
3589 #endif
3590 #if IS_ENABLED(CONFIG_NET_IPGRE)
3591                 case ARPHRD_IPGRE:
3592                         addrconf_gre_config(dev);
3593                         break;
3594 #endif
3595                 case ARPHRD_LOOPBACK:
3596                         init_loopback(dev);
3597                         break;
3598
3599                 default:
3600                         addrconf_dev_config(dev);
3601                         break;
3602                 }
3603
3604                 if (!IS_ERR_OR_NULL(idev)) {
3605                         if (run_pending)
3606                                 addrconf_dad_run(idev, false);
3607
3608                         /* Device has an address by now */
3609                         rt6_sync_up(dev, RTNH_F_DEAD);
3610
3611                         /*
3612                          * If the MTU changed during the interface down,
3613                          * when the interface up, the changed MTU must be
3614                          * reflected in the idev as well as routers.
3615                          */
3616                         if (idev->cnf.mtu6 != dev->mtu &&
3617                             dev->mtu >= IPV6_MIN_MTU) {
3618                                 rt6_mtu_change(dev, dev->mtu);
3619                                 idev->cnf.mtu6 = dev->mtu;
3620                         }
3621                         idev->tstamp = jiffies;
3622                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3623
3624                         /*
3625                          * If the changed mtu during down is lower than
3626                          * IPV6_MIN_MTU stop IPv6 on this interface.
3627                          */
3628                         if (dev->mtu < IPV6_MIN_MTU)
3629                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3630                 }
3631                 break;
3632
3633         case NETDEV_DOWN:
3634         case NETDEV_UNREGISTER:
3635                 /*
3636                  *      Remove all addresses from this interface.
3637                  */
3638                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3639                 break;
3640
3641         case NETDEV_CHANGENAME:
3642                 if (idev) {
3643                         snmp6_unregister_dev(idev);
3644                         addrconf_sysctl_unregister(idev);
3645                         err = addrconf_sysctl_register(idev);
3646                         if (err)
3647                                 return notifier_from_errno(err);
3648                         err = snmp6_register_dev(idev);
3649                         if (err) {
3650                                 addrconf_sysctl_unregister(idev);
3651                                 return notifier_from_errno(err);
3652                         }
3653                 }
3654                 break;
3655
3656         case NETDEV_PRE_TYPE_CHANGE:
3657         case NETDEV_POST_TYPE_CHANGE:
3658                 if (idev)
3659                         addrconf_type_change(dev, event);
3660                 break;
3661
3662         case NETDEV_CHANGEUPPER:
3663                 info = ptr;
3664
3665                 /* flush all routes if dev is linked to or unlinked from
3666                  * an L3 master device (e.g., VRF)
3667                  */
3668                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3669                         addrconf_ifdown(dev, false);
3670         }
3671
3672         return NOTIFY_OK;
3673 }
3674
3675 /*
3676  *      addrconf module should be notified of a device going up
3677  */
3678 static struct notifier_block ipv6_dev_notf = {
3679         .notifier_call = addrconf_notify,
3680         .priority = ADDRCONF_NOTIFY_PRIORITY,
3681 };
3682
3683 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3684 {
3685         struct inet6_dev *idev;
3686         ASSERT_RTNL();
3687
3688         idev = __in6_dev_get(dev);
3689
3690         if (event == NETDEV_POST_TYPE_CHANGE)
3691                 ipv6_mc_remap(idev);
3692         else if (event == NETDEV_PRE_TYPE_CHANGE)
3693                 ipv6_mc_unmap(idev);
3694 }
3695
3696 static bool addr_is_local(const struct in6_addr *addr)
3697 {
3698         return ipv6_addr_type(addr) &
3699                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3700 }
3701
3702 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3703 {
3704         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3705         struct net *net = dev_net(dev);
3706         struct inet6_dev *idev;
3707         struct inet6_ifaddr *ifa, *tmp;
3708         bool keep_addr = false;
3709         int state, i;
3710
3711         ASSERT_RTNL();
3712
3713         rt6_disable_ip(dev, event);
3714
3715         idev = __in6_dev_get(dev);
3716         if (!idev)
3717                 return -ENODEV;
3718
3719         /*
3720          * Step 1: remove reference to ipv6 device from parent device.
3721          *         Do not dev_put!
3722          */
3723         if (unregister) {
3724                 idev->dead = 1;
3725
3726                 /* protected by rtnl_lock */
3727                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3728
3729                 /* Step 1.5: remove snmp6 entry */
3730                 snmp6_unregister_dev(idev);
3731
3732         }
3733
3734         /* combine the user config with event to determine if permanent
3735          * addresses are to be removed from address hash table
3736          */
3737         if (!unregister && !idev->cnf.disable_ipv6) {
3738                 /* aggregate the system setting and interface setting */
3739                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3740
3741                 if (!_keep_addr)
3742                         _keep_addr = idev->cnf.keep_addr_on_down;
3743
3744                 keep_addr = (_keep_addr > 0);
3745         }
3746
3747         /* Step 2: clear hash table */
3748         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3749                 struct hlist_head *h = &inet6_addr_lst[i];
3750
3751                 spin_lock_bh(&addrconf_hash_lock);
3752 restart:
3753                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3754                         if (ifa->idev == idev) {
3755                                 addrconf_del_dad_work(ifa);
3756                                 /* combined flag + permanent flag decide if
3757                                  * address is retained on a down event
3758                                  */
3759                                 if (!keep_addr ||
3760                                     !(ifa->flags & IFA_F_PERMANENT) ||
3761                                     addr_is_local(&ifa->addr)) {
3762                                         hlist_del_init_rcu(&ifa->addr_lst);
3763                                         goto restart;
3764                                 }
3765                         }
3766                 }
3767                 spin_unlock_bh(&addrconf_hash_lock);
3768         }
3769
3770         write_lock_bh(&idev->lock);
3771
3772         addrconf_del_rs_timer(idev);
3773
3774         /* Step 2: clear flags for stateless addrconf */
3775         if (!unregister)
3776                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3777
3778         /* Step 3: clear tempaddr list */
3779         while (!list_empty(&idev->tempaddr_list)) {
3780                 ifa = list_first_entry(&idev->tempaddr_list,
3781                                        struct inet6_ifaddr, tmp_list);
3782                 list_del(&ifa->tmp_list);
3783                 write_unlock_bh(&idev->lock);
3784                 spin_lock_bh(&ifa->lock);
3785
3786                 if (ifa->ifpub) {
3787                         in6_ifa_put(ifa->ifpub);
3788                         ifa->ifpub = NULL;
3789                 }
3790                 spin_unlock_bh(&ifa->lock);
3791                 in6_ifa_put(ifa);
3792                 write_lock_bh(&idev->lock);
3793         }
3794
3795         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3796                 struct fib6_info *rt = NULL;
3797                 bool keep;
3798
3799                 addrconf_del_dad_work(ifa);
3800
3801                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3802                         !addr_is_local(&ifa->addr);
3803
3804                 write_unlock_bh(&idev->lock);
3805                 spin_lock_bh(&ifa->lock);
3806
3807                 if (keep) {
3808                         /* set state to skip the notifier below */
3809                         state = INET6_IFADDR_STATE_DEAD;
3810                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3811                         if (!(ifa->flags & IFA_F_NODAD))
3812                                 ifa->flags |= IFA_F_TENTATIVE;
3813
3814                         rt = ifa->rt;
3815                         ifa->rt = NULL;
3816                 } else {
3817                         state = ifa->state;
3818                         ifa->state = INET6_IFADDR_STATE_DEAD;
3819                 }
3820
3821                 spin_unlock_bh(&ifa->lock);
3822
3823                 if (rt)
3824                         ip6_del_rt(net, rt, false);
3825
3826                 if (state != INET6_IFADDR_STATE_DEAD) {
3827                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3828                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3829                 } else {
3830                         if (idev->cnf.forwarding)
3831                                 addrconf_leave_anycast(ifa);
3832                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3833                 }
3834
3835                 write_lock_bh(&idev->lock);
3836                 if (!keep) {
3837                         list_del_rcu(&ifa->if_list);
3838                         in6_ifa_put(ifa);
3839                 }
3840         }
3841
3842         write_unlock_bh(&idev->lock);
3843
3844         /* Step 5: Discard anycast and multicast list */
3845         if (unregister) {
3846                 ipv6_ac_destroy_dev(idev);
3847                 ipv6_mc_destroy_dev(idev);
3848         } else {
3849                 ipv6_mc_down(idev);
3850         }
3851
3852         idev->tstamp = jiffies;
3853         idev->ra_mtu = 0;
3854
3855         /* Last: Shot the device (if unregistered) */
3856         if (unregister) {
3857                 addrconf_sysctl_unregister(idev);
3858                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3859                 neigh_ifdown(&nd_tbl, dev);
3860                 in6_dev_put(idev);
3861         }
3862         return 0;
3863 }
3864
3865 static void addrconf_rs_timer(struct timer_list *t)
3866 {
3867         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3868         struct net_device *dev = idev->dev;
3869         struct in6_addr lladdr;
3870
3871         write_lock(&idev->lock);
3872         if (idev->dead || !(idev->if_flags & IF_READY))
3873                 goto out;
3874
3875         if (!ipv6_accept_ra(idev))
3876                 goto out;
3877
3878         /* Announcement received after solicitation was sent */
3879         if (idev->if_flags & IF_RA_RCVD)
3880                 goto out;
3881
3882         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3883                 write_unlock(&idev->lock);
3884                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3885                         ndisc_send_rs(dev, &lladdr,
3886                                       &in6addr_linklocal_allrouters);
3887                 else
3888                         goto put;
3889
3890                 write_lock(&idev->lock);
3891                 idev->rs_interval = rfc3315_s14_backoff_update(
3892                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3893                 /* The wait after the last probe can be shorter */
3894                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3895                                              idev->cnf.rtr_solicits) ?
3896                                       idev->cnf.rtr_solicit_delay :
3897                                       idev->rs_interval);
3898         } else {
3899                 /*
3900                  * Note: we do not support deprecated "all on-link"
3901                  * assumption any longer.
3902                  */
3903                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3904         }
3905
3906 out:
3907         write_unlock(&idev->lock);
3908 put:
3909         in6_dev_put(idev);
3910 }
3911
3912 /*
3913  *      Duplicate Address Detection
3914  */
3915 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3916 {
3917         unsigned long rand_num;
3918         struct inet6_dev *idev = ifp->idev;
3919         u64 nonce;
3920
3921         if (ifp->flags & IFA_F_OPTIMISTIC)
3922                 rand_num = 0;
3923         else
3924                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3925
3926         nonce = 0;
3927         if (idev->cnf.enhanced_dad ||
3928             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3929                 do
3930                         get_random_bytes(&nonce, 6);
3931                 while (nonce == 0);
3932         }
3933         ifp->dad_nonce = nonce;
3934         ifp->dad_probes = idev->cnf.dad_transmits;
3935         addrconf_mod_dad_work(ifp, rand_num);
3936 }
3937
3938 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3939 {
3940         struct inet6_dev *idev = ifp->idev;
3941         struct net_device *dev = idev->dev;
3942         bool bump_id, notify = false;
3943         struct net *net;
3944
3945         addrconf_join_solict(dev, &ifp->addr);
3946
3947         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3948
3949         read_lock_bh(&idev->lock);
3950         spin_lock(&ifp->lock);
3951         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3952                 goto out;
3953
3954         net = dev_net(dev);
3955         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3956             (net->ipv6.devconf_all->accept_dad < 1 &&
3957              idev->cnf.accept_dad < 1) ||
3958             !(ifp->flags&IFA_F_TENTATIVE) ||
3959             ifp->flags & IFA_F_NODAD) {
3960                 bool send_na = false;
3961
3962                 if (ifp->flags & IFA_F_TENTATIVE &&
3963                     !(ifp->flags & IFA_F_OPTIMISTIC))
3964                         send_na = true;
3965                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3966                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3967                 spin_unlock(&ifp->lock);
3968                 read_unlock_bh(&idev->lock);
3969
3970                 addrconf_dad_completed(ifp, bump_id, send_na);
3971                 return;
3972         }
3973
3974         if (!(idev->if_flags & IF_READY)) {
3975                 spin_unlock(&ifp->lock);
3976                 read_unlock_bh(&idev->lock);
3977                 /*
3978                  * If the device is not ready:
3979                  * - keep it tentative if it is a permanent address.
3980                  * - otherwise, kill it.
3981                  */
3982                 in6_ifa_hold(ifp);
3983                 addrconf_dad_stop(ifp, 0);
3984                 return;
3985         }
3986
3987         /*
3988          * Optimistic nodes can start receiving
3989          * Frames right away
3990          */
3991         if (ifp->flags & IFA_F_OPTIMISTIC) {
3992                 ip6_ins_rt(net, ifp->rt);
3993                 if (ipv6_use_optimistic_addr(net, idev)) {
3994                         /* Because optimistic nodes can use this address,
3995                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3996                          */
3997                         notify = true;
3998                 }
3999         }
4000
4001         addrconf_dad_kick(ifp);
4002 out:
4003         spin_unlock(&ifp->lock);
4004         read_unlock_bh(&idev->lock);
4005         if (notify)
4006                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4007 }
4008
4009 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4010 {
4011         bool begin_dad = false;
4012
4013         spin_lock_bh(&ifp->lock);
4014         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4015                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4016                 begin_dad = true;
4017         }
4018         spin_unlock_bh(&ifp->lock);
4019
4020         if (begin_dad)
4021                 addrconf_mod_dad_work(ifp, 0);
4022 }
4023
4024 static void addrconf_dad_work(struct work_struct *w)
4025 {
4026         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4027                                                 struct inet6_ifaddr,
4028                                                 dad_work);
4029         struct inet6_dev *idev = ifp->idev;
4030         bool bump_id, disable_ipv6 = false;
4031         struct in6_addr mcaddr;
4032
4033         enum {
4034                 DAD_PROCESS,
4035                 DAD_BEGIN,
4036                 DAD_ABORT,
4037         } action = DAD_PROCESS;
4038
4039         rtnl_lock();
4040
4041         spin_lock_bh(&ifp->lock);
4042         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4043                 action = DAD_BEGIN;
4044                 ifp->state = INET6_IFADDR_STATE_DAD;
4045         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4046                 action = DAD_ABORT;
4047                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4048
4049                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4050                      idev->cnf.accept_dad > 1) &&
4051                     !idev->cnf.disable_ipv6 &&
4052                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4053                         struct in6_addr addr;
4054
4055                         addr.s6_addr32[0] = htonl(0xfe800000);
4056                         addr.s6_addr32[1] = 0;
4057
4058                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4059                             ipv6_addr_equal(&ifp->addr, &addr)) {
4060                                 /* DAD failed for link-local based on MAC */
4061                                 idev->cnf.disable_ipv6 = 1;
4062
4063                                 pr_info("%s: IPv6 being disabled!\n",
4064                                         ifp->idev->dev->name);
4065                                 disable_ipv6 = true;
4066                         }
4067                 }
4068         }
4069         spin_unlock_bh(&ifp->lock);
4070
4071         if (action == DAD_BEGIN) {
4072                 addrconf_dad_begin(ifp);
4073                 goto out;
4074         } else if (action == DAD_ABORT) {
4075                 in6_ifa_hold(ifp);
4076                 addrconf_dad_stop(ifp, 1);
4077                 if (disable_ipv6)
4078                         addrconf_ifdown(idev->dev, false);
4079                 goto out;
4080         }
4081
4082         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4083                 goto out;
4084
4085         write_lock_bh(&idev->lock);
4086         if (idev->dead || !(idev->if_flags & IF_READY)) {
4087                 write_unlock_bh(&idev->lock);
4088                 goto out;
4089         }
4090
4091         spin_lock(&ifp->lock);
4092         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4093                 spin_unlock(&ifp->lock);
4094                 write_unlock_bh(&idev->lock);
4095                 goto out;
4096         }
4097
4098         if (ifp->dad_probes == 0) {
4099                 bool send_na = false;
4100
4101                 /*
4102                  * DAD was successful
4103                  */
4104
4105                 if (ifp->flags & IFA_F_TENTATIVE &&
4106                     !(ifp->flags & IFA_F_OPTIMISTIC))
4107                         send_na = true;
4108                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4109                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4110                 spin_unlock(&ifp->lock);
4111                 write_unlock_bh(&idev->lock);
4112
4113                 addrconf_dad_completed(ifp, bump_id, send_na);
4114
4115                 goto out;
4116         }
4117
4118         ifp->dad_probes--;
4119         addrconf_mod_dad_work(ifp,
4120                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4121                                   HZ/100));
4122         spin_unlock(&ifp->lock);
4123         write_unlock_bh(&idev->lock);
4124
4125         /* send a neighbour solicitation for our addr */
4126         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4127         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4128                       ifp->dad_nonce);
4129 out:
4130         in6_ifa_put(ifp);
4131         rtnl_unlock();
4132 }
4133
4134 /* ifp->idev must be at least read locked */
4135 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4136 {
4137         struct inet6_ifaddr *ifpiter;
4138         struct inet6_dev *idev = ifp->idev;
4139
4140         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4141                 if (ifpiter->scope > IFA_LINK)
4142                         break;
4143                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4144                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4145                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4146                     IFA_F_PERMANENT)
4147                         return false;
4148         }
4149         return true;
4150 }
4151
4152 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4153                                    bool send_na)
4154 {
4155         struct net_device *dev = ifp->idev->dev;
4156         struct in6_addr lladdr;
4157         bool send_rs, send_mld;
4158
4159         addrconf_del_dad_work(ifp);
4160
4161         /*
4162          *      Configure the address for reception. Now it is valid.
4163          */
4164
4165         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4166
4167         /* If added prefix is link local and we are prepared to process
4168            router advertisements, start sending router solicitations.
4169          */
4170
4171         read_lock_bh(&ifp->idev->lock);
4172         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4173         send_rs = send_mld &&
4174                   ipv6_accept_ra(ifp->idev) &&
4175                   ifp->idev->cnf.rtr_solicits != 0 &&
4176                   (dev->flags&IFF_LOOPBACK) == 0;
4177         read_unlock_bh(&ifp->idev->lock);
4178
4179         /* While dad is in progress mld report's source address is in6_addrany.
4180          * Resend with proper ll now.
4181          */
4182         if (send_mld)
4183                 ipv6_mc_dad_complete(ifp->idev);
4184
4185         /* send unsolicited NA if enabled */
4186         if (send_na &&
4187             (ifp->idev->cnf.ndisc_notify ||
4188              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4189                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4190                               /*router=*/ !!ifp->idev->cnf.forwarding,
4191                               /*solicited=*/ false, /*override=*/ true,
4192                               /*inc_opt=*/ true);
4193         }
4194
4195         if (send_rs) {
4196                 /*
4197                  *      If a host as already performed a random delay
4198                  *      [...] as part of DAD [...] there is no need
4199                  *      to delay again before sending the first RS
4200                  */
4201                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4202                         return;
4203                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4204
4205                 write_lock_bh(&ifp->idev->lock);
4206                 spin_lock(&ifp->lock);
4207                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4208                         ifp->idev->cnf.rtr_solicit_interval);
4209                 ifp->idev->rs_probes = 1;
4210                 ifp->idev->if_flags |= IF_RS_SENT;
4211                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4212                 spin_unlock(&ifp->lock);
4213                 write_unlock_bh(&ifp->idev->lock);
4214         }
4215
4216         if (bump_id)
4217                 rt_genid_bump_ipv6(dev_net(dev));
4218
4219         /* Make sure that a new temporary address will be created
4220          * before this temporary address becomes deprecated.
4221          */
4222         if (ifp->flags & IFA_F_TEMPORARY)
4223                 addrconf_verify_rtnl();
4224 }
4225
4226 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4227 {
4228         struct inet6_ifaddr *ifp;
4229
4230         read_lock_bh(&idev->lock);
4231         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4232                 spin_lock(&ifp->lock);
4233                 if ((ifp->flags & IFA_F_TENTATIVE &&
4234                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4235                         if (restart)
4236                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4237                         addrconf_dad_kick(ifp);
4238                 }
4239                 spin_unlock(&ifp->lock);
4240         }
4241         read_unlock_bh(&idev->lock);
4242 }
4243
4244 #ifdef CONFIG_PROC_FS
4245 struct if6_iter_state {
4246         struct seq_net_private p;
4247         int bucket;
4248         int offset;
4249 };
4250
4251 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4252 {
4253         struct if6_iter_state *state = seq->private;
4254         struct net *net = seq_file_net(seq);
4255         struct inet6_ifaddr *ifa = NULL;
4256         int p = 0;
4257
4258         /* initial bucket if pos is 0 */
4259         if (pos == 0) {
4260                 state->bucket = 0;
4261                 state->offset = 0;
4262         }
4263
4264         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4265                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4266                                          addr_lst) {
4267                         if (!net_eq(dev_net(ifa->idev->dev), net))
4268                                 continue;
4269                         /* sync with offset */
4270                         if (p < state->offset) {
4271                                 p++;
4272                                 continue;
4273                         }
4274                         return ifa;
4275                 }
4276
4277                 /* prepare for next bucket */
4278                 state->offset = 0;
4279                 p = 0;
4280         }
4281         return NULL;
4282 }
4283
4284 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4285                                          struct inet6_ifaddr *ifa)
4286 {
4287         struct if6_iter_state *state = seq->private;
4288         struct net *net = seq_file_net(seq);
4289
4290         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4291                 if (!net_eq(dev_net(ifa->idev->dev), net))
4292                         continue;
4293                 state->offset++;
4294                 return ifa;
4295         }
4296
4297         state->offset = 0;
4298         while (++state->bucket < IN6_ADDR_HSIZE) {
4299                 hlist_for_each_entry_rcu(ifa,
4300                                      &inet6_addr_lst[state->bucket], addr_lst) {
4301                         if (!net_eq(dev_net(ifa->idev->dev), net))
4302                                 continue;
4303                         return ifa;
4304                 }
4305         }
4306
4307         return NULL;
4308 }
4309
4310 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4311         __acquires(rcu)
4312 {
4313         rcu_read_lock();
4314         return if6_get_first(seq, *pos);
4315 }
4316
4317 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4318 {
4319         struct inet6_ifaddr *ifa;
4320
4321         ifa = if6_get_next(seq, v);
4322         ++*pos;
4323         return ifa;
4324 }
4325
4326 static void if6_seq_stop(struct seq_file *seq, void *v)
4327         __releases(rcu)
4328 {
4329         rcu_read_unlock();
4330 }
4331
4332 static int if6_seq_show(struct seq_file *seq, void *v)
4333 {
4334         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4335         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4336                    &ifp->addr,
4337                    ifp->idev->dev->ifindex,
4338                    ifp->prefix_len,
4339                    ifp->scope,
4340                    (u8) ifp->flags,
4341                    ifp->idev->dev->name);
4342         return 0;
4343 }
4344
4345 static const struct seq_operations if6_seq_ops = {
4346         .start  = if6_seq_start,
4347         .next   = if6_seq_next,
4348         .show   = if6_seq_show,
4349         .stop   = if6_seq_stop,
4350 };
4351
4352 static int __net_init if6_proc_net_init(struct net *net)
4353 {
4354         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4355                         sizeof(struct if6_iter_state)))
4356                 return -ENOMEM;
4357         return 0;
4358 }
4359
4360 static void __net_exit if6_proc_net_exit(struct net *net)
4361 {
4362         remove_proc_entry("if_inet6", net->proc_net);
4363 }
4364
4365 static struct pernet_operations if6_proc_net_ops = {
4366         .init = if6_proc_net_init,
4367         .exit = if6_proc_net_exit,
4368 };
4369
4370 int __init if6_proc_init(void)
4371 {
4372         return register_pernet_subsys(&if6_proc_net_ops);
4373 }
4374
4375 void if6_proc_exit(void)
4376 {
4377         unregister_pernet_subsys(&if6_proc_net_ops);
4378 }
4379 #endif  /* CONFIG_PROC_FS */
4380
4381 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4382 /* Check if address is a home address configured on any interface. */
4383 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4384 {
4385         unsigned int hash = inet6_addr_hash(net, addr);
4386         struct inet6_ifaddr *ifp = NULL;
4387         int ret = 0;
4388
4389         rcu_read_lock();
4390         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4391                 if (!net_eq(dev_net(ifp->idev->dev), net))
4392                         continue;
4393                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4394                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4395                         ret = 1;
4396                         break;
4397                 }
4398         }
4399         rcu_read_unlock();
4400         return ret;
4401 }
4402 #endif
4403
4404 /* RFC6554 has some algorithm to avoid loops in segment routing by
4405  * checking if the segments contains any of a local interface address.
4406  *
4407  * Quote:
4408  *
4409  * To detect loops in the SRH, a router MUST determine if the SRH
4410  * includes multiple addresses assigned to any interface on that router.
4411  * If such addresses appear more than once and are separated by at least
4412  * one address not assigned to that router.
4413  */
4414 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4415                           unsigned char nsegs)
4416 {
4417         const struct in6_addr *addr;
4418         int i, ret = 0, found = 0;
4419         struct inet6_ifaddr *ifp;
4420         bool separated = false;
4421         unsigned int hash;
4422         bool hash_found;
4423
4424         rcu_read_lock();
4425         for (i = 0; i < nsegs; i++) {
4426                 addr = &segs[i];
4427                 hash = inet6_addr_hash(net, addr);
4428
4429                 hash_found = false;
4430                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4431                         if (!net_eq(dev_net(ifp->idev->dev), net))
4432                                 continue;
4433
4434                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4435                                 hash_found = true;
4436                                 break;
4437                         }
4438                 }
4439
4440                 if (hash_found) {
4441                         if (found > 1 && separated) {
4442                                 ret = 1;
4443                                 break;
4444                         }
4445
4446                         separated = false;
4447                         found++;
4448                 } else {
4449                         separated = true;
4450                 }
4451         }
4452         rcu_read_unlock();
4453
4454         return ret;
4455 }
4456
4457 /*
4458  *      Periodic address status verification
4459  */
4460
4461 static void addrconf_verify_rtnl(void)
4462 {
4463         unsigned long now, next, next_sec, next_sched;
4464         struct inet6_ifaddr *ifp;
4465         int i;
4466
4467         ASSERT_RTNL();
4468
4469         rcu_read_lock_bh();
4470         now = jiffies;
4471         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4472
4473         cancel_delayed_work(&addr_chk_work);
4474
4475         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4476 restart:
4477                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4478                         unsigned long age;
4479
4480                         /* When setting preferred_lft to a value not zero or
4481                          * infinity, while valid_lft is infinity
4482                          * IFA_F_PERMANENT has a non-infinity life time.
4483                          */
4484                         if ((ifp->flags & IFA_F_PERMANENT) &&
4485                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4486                                 continue;
4487
4488                         spin_lock(&ifp->lock);
4489                         /* We try to batch several events at once. */
4490                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4491
4492                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4493                             age >= ifp->valid_lft) {
4494                                 spin_unlock(&ifp->lock);
4495                                 in6_ifa_hold(ifp);
4496                                 rcu_read_unlock_bh();
4497                                 ipv6_del_addr(ifp);
4498                                 rcu_read_lock_bh();
4499                                 goto restart;
4500                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4501                                 spin_unlock(&ifp->lock);
4502                                 continue;
4503                         } else if (age >= ifp->prefered_lft) {
4504                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4505                                 int deprecate = 0;
4506
4507                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4508                                         deprecate = 1;
4509                                         ifp->flags |= IFA_F_DEPRECATED;
4510                                 }
4511
4512                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4513                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4514                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4515
4516                                 spin_unlock(&ifp->lock);
4517
4518                                 if (deprecate) {
4519                                         in6_ifa_hold(ifp);
4520
4521                                         ipv6_ifa_notify(0, ifp);
4522                                         in6_ifa_put(ifp);
4523                                         goto restart;
4524                                 }
4525                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4526                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4527                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4528                                         ifp->idev->cnf.dad_transmits *
4529                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4530
4531                                 if (age >= ifp->prefered_lft - regen_advance) {
4532                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4533                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4534                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4535                                         if (!ifp->regen_count && ifpub) {
4536                                                 ifp->regen_count++;
4537                                                 in6_ifa_hold(ifp);
4538                                                 in6_ifa_hold(ifpub);
4539                                                 spin_unlock(&ifp->lock);
4540
4541                                                 spin_lock(&ifpub->lock);
4542                                                 ifpub->regen_count = 0;
4543                                                 spin_unlock(&ifpub->lock);
4544                                                 rcu_read_unlock_bh();
4545                                                 ipv6_create_tempaddr(ifpub, true);
4546                                                 in6_ifa_put(ifpub);
4547                                                 in6_ifa_put(ifp);
4548                                                 rcu_read_lock_bh();
4549                                                 goto restart;
4550                                         }
4551                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4552                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4553                                 spin_unlock(&ifp->lock);
4554                         } else {
4555                                 /* ifp->prefered_lft <= ifp->valid_lft */
4556                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4557                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4558                                 spin_unlock(&ifp->lock);
4559                         }
4560                 }
4561         }
4562
4563         next_sec = round_jiffies_up(next);
4564         next_sched = next;
4565
4566         /* If rounded timeout is accurate enough, accept it. */
4567         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4568                 next_sched = next_sec;
4569
4570         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4571         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4572                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4573
4574         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4575                  now, next, next_sec, next_sched);
4576         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4577         rcu_read_unlock_bh();
4578 }
4579
4580 static void addrconf_verify_work(struct work_struct *w)
4581 {
4582         rtnl_lock();
4583         addrconf_verify_rtnl();
4584         rtnl_unlock();
4585 }
4586
4587 static void addrconf_verify(void)
4588 {
4589         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4590 }
4591
4592 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4593                                      struct in6_addr **peer_pfx)
4594 {
4595         struct in6_addr *pfx = NULL;
4596
4597         *peer_pfx = NULL;
4598
4599         if (addr)
4600                 pfx = nla_data(addr);
4601
4602         if (local) {
4603                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4604                         *peer_pfx = pfx;
4605                 pfx = nla_data(local);
4606         }
4607
4608         return pfx;
4609 }
4610
4611 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4612         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4613         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4614         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4615         [IFA_FLAGS]             = { .len = sizeof(u32) },
4616         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4617         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4618 };
4619
4620 static int
4621 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4622                   struct netlink_ext_ack *extack)
4623 {
4624         struct net *net = sock_net(skb->sk);
4625         struct ifaddrmsg *ifm;
4626         struct nlattr *tb[IFA_MAX+1];
4627         struct in6_addr *pfx, *peer_pfx;
4628         u32 ifa_flags;
4629         int err;
4630
4631         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4632                                      ifa_ipv6_policy, extack);
4633         if (err < 0)
4634                 return err;
4635
4636         ifm = nlmsg_data(nlh);
4637         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4638         if (!pfx)
4639                 return -EINVAL;
4640
4641         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4642
4643         /* We ignore other flags so far. */
4644         ifa_flags &= IFA_F_MANAGETEMPADDR;
4645
4646         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4647                               ifm->ifa_prefixlen);
4648 }
4649
4650 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4651                                unsigned long expires, u32 flags,
4652                                bool modify_peer)
4653 {
4654         struct fib6_info *f6i;
4655         u32 prio;
4656
4657         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4658                                         ifp->prefix_len,
4659                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4660         if (!f6i)
4661                 return -ENOENT;
4662
4663         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4664         if (f6i->fib6_metric != prio) {
4665                 /* delete old one */
4666                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4667
4668                 /* add new one */
4669                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4670                                       ifp->prefix_len,
4671                                       ifp->rt_priority, ifp->idev->dev,
4672                                       expires, flags, GFP_KERNEL);
4673         } else {
4674                 if (!expires)
4675                         fib6_clean_expires(f6i);
4676                 else
4677                         fib6_set_expires(f6i, expires);
4678
4679                 fib6_info_release(f6i);
4680         }
4681
4682         return 0;
4683 }
4684
4685 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4686 {
4687         u32 flags;
4688         clock_t expires;
4689         unsigned long timeout;
4690         bool was_managetempaddr;
4691         bool had_prefixroute;
4692         bool new_peer = false;
4693
4694         ASSERT_RTNL();
4695
4696         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4697                 return -EINVAL;
4698
4699         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4700             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4701                 return -EINVAL;
4702
4703         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4704                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4705
4706         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4707         if (addrconf_finite_timeout(timeout)) {
4708                 expires = jiffies_to_clock_t(timeout * HZ);
4709                 cfg->valid_lft = timeout;
4710                 flags = RTF_EXPIRES;
4711         } else {
4712                 expires = 0;
4713                 flags = 0;
4714                 cfg->ifa_flags |= IFA_F_PERMANENT;
4715         }
4716
4717         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4718         if (addrconf_finite_timeout(timeout)) {
4719                 if (timeout == 0)
4720                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4721                 cfg->preferred_lft = timeout;
4722         }
4723
4724         if (cfg->peer_pfx &&
4725             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4726                 if (!ipv6_addr_any(&ifp->peer_addr))
4727                         cleanup_prefix_route(ifp, expires, true, true);
4728                 new_peer = true;
4729         }
4730
4731         spin_lock_bh(&ifp->lock);
4732         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4733         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4734                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4735         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4736                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4737                         IFA_F_NOPREFIXROUTE);
4738         ifp->flags |= cfg->ifa_flags;
4739         ifp->tstamp = jiffies;
4740         ifp->valid_lft = cfg->valid_lft;
4741         ifp->prefered_lft = cfg->preferred_lft;
4742
4743         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4744                 ifp->rt_priority = cfg->rt_priority;
4745
4746         if (new_peer)
4747                 ifp->peer_addr = *cfg->peer_pfx;
4748
4749         spin_unlock_bh(&ifp->lock);
4750         if (!(ifp->flags&IFA_F_TENTATIVE))
4751                 ipv6_ifa_notify(0, ifp);
4752
4753         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4754                 int rc = -ENOENT;
4755
4756                 if (had_prefixroute)
4757                         rc = modify_prefix_route(ifp, expires, flags, false);
4758
4759                 /* prefix route could have been deleted; if so restore it */
4760                 if (rc == -ENOENT) {
4761                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4762                                               ifp->rt_priority, ifp->idev->dev,
4763                                               expires, flags, GFP_KERNEL);
4764                 }
4765
4766                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4767                         rc = modify_prefix_route(ifp, expires, flags, true);
4768
4769                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4770                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4771                                               ifp->rt_priority, ifp->idev->dev,
4772                                               expires, flags, GFP_KERNEL);
4773                 }
4774         } else if (had_prefixroute) {
4775                 enum cleanup_prefix_rt_t action;
4776                 unsigned long rt_expires;
4777
4778                 write_lock_bh(&ifp->idev->lock);
4779                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4780                 write_unlock_bh(&ifp->idev->lock);
4781
4782                 if (action != CLEANUP_PREFIX_RT_NOP) {
4783                         cleanup_prefix_route(ifp, rt_expires,
4784                                 action == CLEANUP_PREFIX_RT_DEL, false);
4785                 }
4786         }
4787
4788         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4789                 if (was_managetempaddr &&
4790                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4791                         cfg->valid_lft = 0;
4792                         cfg->preferred_lft = 0;
4793                 }
4794                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4795                                  cfg->preferred_lft, !was_managetempaddr,
4796                                  jiffies);
4797         }
4798
4799         addrconf_verify_rtnl();
4800
4801         return 0;
4802 }
4803
4804 static int
4805 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4806                   struct netlink_ext_ack *extack)
4807 {
4808         struct net *net = sock_net(skb->sk);
4809         struct ifaddrmsg *ifm;
4810         struct nlattr *tb[IFA_MAX+1];
4811         struct in6_addr *peer_pfx;
4812         struct inet6_ifaddr *ifa;
4813         struct net_device *dev;
4814         struct inet6_dev *idev;
4815         struct ifa6_config cfg;
4816         int err;
4817
4818         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4819                                      ifa_ipv6_policy, extack);
4820         if (err < 0)
4821                 return err;
4822
4823         memset(&cfg, 0, sizeof(cfg));
4824
4825         ifm = nlmsg_data(nlh);
4826         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4827         if (!cfg.pfx)
4828                 return -EINVAL;
4829
4830         cfg.peer_pfx = peer_pfx;
4831         cfg.plen = ifm->ifa_prefixlen;
4832         if (tb[IFA_RT_PRIORITY])
4833                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4834
4835         cfg.valid_lft = INFINITY_LIFE_TIME;
4836         cfg.preferred_lft = INFINITY_LIFE_TIME;
4837
4838         if (tb[IFA_CACHEINFO]) {
4839                 struct ifa_cacheinfo *ci;
4840
4841                 ci = nla_data(tb[IFA_CACHEINFO]);
4842                 cfg.valid_lft = ci->ifa_valid;
4843                 cfg.preferred_lft = ci->ifa_prefered;
4844         }
4845
4846         dev =  __dev_get_by_index(net, ifm->ifa_index);
4847         if (!dev)
4848                 return -ENODEV;
4849
4850         if (tb[IFA_FLAGS])
4851                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4852         else
4853                 cfg.ifa_flags = ifm->ifa_flags;
4854
4855         /* We ignore other flags so far. */
4856         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4857                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4858                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4859
4860         idev = ipv6_find_idev(dev);
4861         if (IS_ERR(idev))
4862                 return PTR_ERR(idev);
4863
4864         if (!ipv6_allow_optimistic_dad(net, idev))
4865                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4866
4867         if (cfg.ifa_flags & IFA_F_NODAD &&
4868             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4869                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4870                 return -EINVAL;
4871         }
4872
4873         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4874         if (!ifa) {
4875                 /*
4876                  * It would be best to check for !NLM_F_CREATE here but
4877                  * userspace already relies on not having to provide this.
4878                  */
4879                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4880         }
4881
4882         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4883             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4884                 err = -EEXIST;
4885         else
4886                 err = inet6_addr_modify(ifa, &cfg);
4887
4888         in6_ifa_put(ifa);
4889
4890         return err;
4891 }
4892
4893 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4894                           u8 scope, int ifindex)
4895 {
4896         struct ifaddrmsg *ifm;
4897
4898         ifm = nlmsg_data(nlh);
4899         ifm->ifa_family = AF_INET6;
4900         ifm->ifa_prefixlen = prefixlen;
4901         ifm->ifa_flags = flags;
4902         ifm->ifa_scope = scope;
4903         ifm->ifa_index = ifindex;
4904 }
4905
4906 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4907                          unsigned long tstamp, u32 preferred, u32 valid)
4908 {
4909         struct ifa_cacheinfo ci;
4910
4911         ci.cstamp = cstamp_delta(cstamp);
4912         ci.tstamp = cstamp_delta(tstamp);
4913         ci.ifa_prefered = preferred;
4914         ci.ifa_valid = valid;
4915
4916         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4917 }
4918
4919 static inline int rt_scope(int ifa_scope)
4920 {
4921         if (ifa_scope & IFA_HOST)
4922                 return RT_SCOPE_HOST;
4923         else if (ifa_scope & IFA_LINK)
4924                 return RT_SCOPE_LINK;
4925         else if (ifa_scope & IFA_SITE)
4926                 return RT_SCOPE_SITE;
4927         else
4928                 return RT_SCOPE_UNIVERSE;
4929 }
4930
4931 static inline int inet6_ifaddr_msgsize(void)
4932 {
4933         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4934                + nla_total_size(16) /* IFA_LOCAL */
4935                + nla_total_size(16) /* IFA_ADDRESS */
4936                + nla_total_size(sizeof(struct ifa_cacheinfo))
4937                + nla_total_size(4)  /* IFA_FLAGS */
4938                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4939 }
4940
4941 enum addr_type_t {
4942         UNICAST_ADDR,
4943         MULTICAST_ADDR,
4944         ANYCAST_ADDR,
4945 };
4946
4947 struct inet6_fill_args {
4948         u32 portid;
4949         u32 seq;
4950         int event;
4951         unsigned int flags;
4952         int netnsid;
4953         int ifindex;
4954         enum addr_type_t type;
4955 };
4956
4957 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4958                              struct inet6_fill_args *args)
4959 {
4960         struct nlmsghdr  *nlh;
4961         u32 preferred, valid;
4962
4963         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4964                         sizeof(struct ifaddrmsg), args->flags);
4965         if (!nlh)
4966                 return -EMSGSIZE;
4967
4968         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4969                       ifa->idev->dev->ifindex);
4970
4971         if (args->netnsid >= 0 &&
4972             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4973                 goto error;
4974
4975         if (!((ifa->flags&IFA_F_PERMANENT) &&
4976               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4977                 preferred = ifa->prefered_lft;
4978                 valid = ifa->valid_lft;
4979                 if (preferred != INFINITY_LIFE_TIME) {
4980                         long tval = (jiffies - ifa->tstamp)/HZ;
4981                         if (preferred > tval)
4982                                 preferred -= tval;
4983                         else
4984                                 preferred = 0;
4985                         if (valid != INFINITY_LIFE_TIME) {
4986                                 if (valid > tval)
4987                                         valid -= tval;
4988                                 else
4989                                         valid = 0;
4990                         }
4991                 }
4992         } else {
4993                 preferred = INFINITY_LIFE_TIME;
4994                 valid = INFINITY_LIFE_TIME;
4995         }
4996
4997         if (!ipv6_addr_any(&ifa->peer_addr)) {
4998                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4999                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5000                         goto error;
5001         } else
5002                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5003                         goto error;
5004
5005         if (ifa->rt_priority &&
5006             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5007                 goto error;
5008
5009         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5010                 goto error;
5011
5012         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5013                 goto error;
5014
5015         nlmsg_end(skb, nlh);
5016         return 0;
5017
5018 error:
5019         nlmsg_cancel(skb, nlh);
5020         return -EMSGSIZE;
5021 }
5022
5023 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5024                                struct inet6_fill_args *args)
5025 {
5026         struct nlmsghdr  *nlh;
5027         u8 scope = RT_SCOPE_UNIVERSE;
5028         int ifindex = ifmca->idev->dev->ifindex;
5029
5030         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5031                 scope = RT_SCOPE_SITE;
5032
5033         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5034                         sizeof(struct ifaddrmsg), args->flags);
5035         if (!nlh)
5036                 return -EMSGSIZE;
5037
5038         if (args->netnsid >= 0 &&
5039             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5040                 nlmsg_cancel(skb, nlh);
5041                 return -EMSGSIZE;
5042         }
5043
5044         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5045         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5046             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5047                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5048                 nlmsg_cancel(skb, nlh);
5049                 return -EMSGSIZE;
5050         }
5051
5052         nlmsg_end(skb, nlh);
5053         return 0;
5054 }
5055
5056 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5057                                struct inet6_fill_args *args)
5058 {
5059         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5060         int ifindex = dev ? dev->ifindex : 1;
5061         struct nlmsghdr  *nlh;
5062         u8 scope = RT_SCOPE_UNIVERSE;
5063
5064         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5065                 scope = RT_SCOPE_SITE;
5066
5067         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5068                         sizeof(struct ifaddrmsg), args->flags);
5069         if (!nlh)
5070                 return -EMSGSIZE;
5071
5072         if (args->netnsid >= 0 &&
5073             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5074                 nlmsg_cancel(skb, nlh);
5075                 return -EMSGSIZE;
5076         }
5077
5078         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5079         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5080             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5081                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5082                 nlmsg_cancel(skb, nlh);
5083                 return -EMSGSIZE;
5084         }
5085
5086         nlmsg_end(skb, nlh);
5087         return 0;
5088 }
5089
5090 /* called with rcu_read_lock() */
5091 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5092                           struct netlink_callback *cb, int s_ip_idx,
5093                           struct inet6_fill_args *fillargs)
5094 {
5095         struct ifmcaddr6 *ifmca;
5096         struct ifacaddr6 *ifaca;
5097         int ip_idx = 0;
5098         int err = 1;
5099
5100         read_lock_bh(&idev->lock);
5101         switch (fillargs->type) {
5102         case UNICAST_ADDR: {
5103                 struct inet6_ifaddr *ifa;
5104                 fillargs->event = RTM_NEWADDR;
5105
5106                 /* unicast address incl. temp addr */
5107                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5108                         if (ip_idx < s_ip_idx)
5109                                 goto next;
5110                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5111                         if (err < 0)
5112                                 break;
5113                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5114 next:
5115                         ip_idx++;
5116                 }
5117                 break;
5118         }
5119         case MULTICAST_ADDR:
5120                 read_unlock_bh(&idev->lock);
5121                 fillargs->event = RTM_GETMULTICAST;
5122
5123                 /* multicast address */
5124                 for (ifmca = rcu_dereference(idev->mc_list);
5125                      ifmca;
5126                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5127                         if (ip_idx < s_ip_idx)
5128                                 continue;
5129                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5130                         if (err < 0)
5131                                 break;
5132                 }
5133                 read_lock_bh(&idev->lock);
5134                 break;
5135         case ANYCAST_ADDR:
5136                 fillargs->event = RTM_GETANYCAST;
5137                 /* anycast address */
5138                 for (ifaca = idev->ac_list; ifaca;
5139                      ifaca = ifaca->aca_next, ip_idx++) {
5140                         if (ip_idx < s_ip_idx)
5141                                 continue;
5142                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5143                         if (err < 0)
5144                                 break;
5145                 }
5146                 break;
5147         default:
5148                 break;
5149         }
5150         read_unlock_bh(&idev->lock);
5151         cb->args[2] = ip_idx;
5152         return err;
5153 }
5154
5155 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5156                                        struct inet6_fill_args *fillargs,
5157                                        struct net **tgt_net, struct sock *sk,
5158                                        struct netlink_callback *cb)
5159 {
5160         struct netlink_ext_ack *extack = cb->extack;
5161         struct nlattr *tb[IFA_MAX+1];
5162         struct ifaddrmsg *ifm;
5163         int err, i;
5164
5165         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5166                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5167                 return -EINVAL;
5168         }
5169
5170         ifm = nlmsg_data(nlh);
5171         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5172                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5173                 return -EINVAL;
5174         }
5175
5176         fillargs->ifindex = ifm->ifa_index;
5177         if (fillargs->ifindex) {
5178                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5179                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5180         }
5181
5182         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5183                                             ifa_ipv6_policy, extack);
5184         if (err < 0)
5185                 return err;
5186
5187         for (i = 0; i <= IFA_MAX; ++i) {
5188                 if (!tb[i])
5189                         continue;
5190
5191                 if (i == IFA_TARGET_NETNSID) {
5192                         struct net *net;
5193
5194                         fillargs->netnsid = nla_get_s32(tb[i]);
5195                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5196                         if (IS_ERR(net)) {
5197                                 fillargs->netnsid = -1;
5198                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5199                                 return PTR_ERR(net);
5200                         }
5201                         *tgt_net = net;
5202                 } else {
5203                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5204                         return -EINVAL;
5205                 }
5206         }
5207
5208         return 0;
5209 }
5210
5211 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5212                            enum addr_type_t type)
5213 {
5214         const struct nlmsghdr *nlh = cb->nlh;
5215         struct inet6_fill_args fillargs = {
5216                 .portid = NETLINK_CB(cb->skb).portid,
5217                 .seq = cb->nlh->nlmsg_seq,
5218                 .flags = NLM_F_MULTI,
5219                 .netnsid = -1,
5220                 .type = type,
5221         };
5222         struct net *tgt_net = sock_net(skb->sk);
5223         int idx, s_idx, s_ip_idx;
5224         int h, s_h;
5225         struct net_device *dev;
5226         struct inet6_dev *idev;
5227         struct hlist_head *head;
5228         int err = 0;
5229
5230         s_h = cb->args[0];
5231         s_idx = idx = cb->args[1];
5232         s_ip_idx = cb->args[2];
5233
5234         if (cb->strict_check) {
5235                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5236                                                   skb->sk, cb);
5237                 if (err < 0)
5238                         goto put_tgt_net;
5239
5240                 err = 0;
5241                 if (fillargs.ifindex) {
5242                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5243                         if (!dev) {
5244                                 err = -ENODEV;
5245                                 goto put_tgt_net;
5246                         }
5247                         idev = __in6_dev_get(dev);
5248                         if (idev) {
5249                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5250                                                      &fillargs);
5251                                 if (err > 0)
5252                                         err = 0;
5253                         }
5254                         goto put_tgt_net;
5255                 }
5256         }
5257
5258         rcu_read_lock();
5259         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5260         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5261                 idx = 0;
5262                 head = &tgt_net->dev_index_head[h];
5263                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5264                         if (idx < s_idx)
5265                                 goto cont;
5266                         if (h > s_h || idx > s_idx)
5267                                 s_ip_idx = 0;
5268                         idev = __in6_dev_get(dev);
5269                         if (!idev)
5270                                 goto cont;
5271
5272                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5273                                            &fillargs) < 0)
5274                                 goto done;
5275 cont:
5276                         idx++;
5277                 }
5278         }
5279 done:
5280         rcu_read_unlock();
5281         cb->args[0] = h;
5282         cb->args[1] = idx;
5283 put_tgt_net:
5284         if (fillargs.netnsid >= 0)
5285                 put_net(tgt_net);
5286
5287         return skb->len ? : err;
5288 }
5289
5290 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5291 {
5292         enum addr_type_t type = UNICAST_ADDR;
5293
5294         return inet6_dump_addr(skb, cb, type);
5295 }
5296
5297 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5298 {
5299         enum addr_type_t type = MULTICAST_ADDR;
5300
5301         return inet6_dump_addr(skb, cb, type);
5302 }
5303
5304
5305 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5306 {
5307         enum addr_type_t type = ANYCAST_ADDR;
5308
5309         return inet6_dump_addr(skb, cb, type);
5310 }
5311
5312 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5313                                        const struct nlmsghdr *nlh,
5314                                        struct nlattr **tb,
5315                                        struct netlink_ext_ack *extack)
5316 {
5317         struct ifaddrmsg *ifm;
5318         int i, err;
5319
5320         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5321                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5322                 return -EINVAL;
5323         }
5324
5325         if (!netlink_strict_get_check(skb))
5326                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5327                                               ifa_ipv6_policy, extack);
5328
5329         ifm = nlmsg_data(nlh);
5330         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5331                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5332                 return -EINVAL;
5333         }
5334
5335         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5336                                             ifa_ipv6_policy, extack);
5337         if (err)
5338                 return err;
5339
5340         for (i = 0; i <= IFA_MAX; i++) {
5341                 if (!tb[i])
5342                         continue;
5343
5344                 switch (i) {
5345                 case IFA_TARGET_NETNSID:
5346                 case IFA_ADDRESS:
5347                 case IFA_LOCAL:
5348                         break;
5349                 default:
5350                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5351                         return -EINVAL;
5352                 }
5353         }
5354
5355         return 0;
5356 }
5357
5358 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5359                              struct netlink_ext_ack *extack)
5360 {
5361         struct net *tgt_net = sock_net(in_skb->sk);
5362         struct inet6_fill_args fillargs = {
5363                 .portid = NETLINK_CB(in_skb).portid,
5364                 .seq = nlh->nlmsg_seq,
5365                 .event = RTM_NEWADDR,
5366                 .flags = 0,
5367                 .netnsid = -1,
5368         };
5369         struct ifaddrmsg *ifm;
5370         struct nlattr *tb[IFA_MAX+1];
5371         struct in6_addr *addr = NULL, *peer;
5372         struct net_device *dev = NULL;
5373         struct inet6_ifaddr *ifa;
5374         struct sk_buff *skb;
5375         int err;
5376
5377         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5378         if (err < 0)
5379                 return err;
5380
5381         if (tb[IFA_TARGET_NETNSID]) {
5382                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5383
5384                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5385                                                   fillargs.netnsid);
5386                 if (IS_ERR(tgt_net))
5387                         return PTR_ERR(tgt_net);
5388         }
5389
5390         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5391         if (!addr)
5392                 return -EINVAL;
5393
5394         ifm = nlmsg_data(nlh);
5395         if (ifm->ifa_index)
5396                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5397
5398         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5399         if (!ifa) {
5400                 err = -EADDRNOTAVAIL;
5401                 goto errout;
5402         }
5403
5404         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5405         if (!skb) {
5406                 err = -ENOBUFS;
5407                 goto errout_ifa;
5408         }
5409
5410         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5411         if (err < 0) {
5412                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5413                 WARN_ON(err == -EMSGSIZE);
5414                 kfree_skb(skb);
5415                 goto errout_ifa;
5416         }
5417         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5418 errout_ifa:
5419         in6_ifa_put(ifa);
5420 errout:
5421         dev_put(dev);
5422         if (fillargs.netnsid >= 0)
5423                 put_net(tgt_net);
5424
5425         return err;
5426 }
5427
5428 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5429 {
5430         struct sk_buff *skb;
5431         struct net *net = dev_net(ifa->idev->dev);
5432         struct inet6_fill_args fillargs = {
5433                 .portid = 0,
5434                 .seq = 0,
5435                 .event = event,
5436                 .flags = 0,
5437                 .netnsid = -1,
5438         };
5439         int err = -ENOBUFS;
5440
5441         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5442         if (!skb)
5443                 goto errout;
5444
5445         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5446         if (err < 0) {
5447                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5448                 WARN_ON(err == -EMSGSIZE);
5449                 kfree_skb(skb);
5450                 goto errout;
5451         }
5452         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5453         return;
5454 errout:
5455         if (err < 0)
5456                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5457 }
5458
5459 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5460                                 __s32 *array, int bytes)
5461 {
5462         BUG_ON(bytes < (DEVCONF_MAX * 4));
5463
5464         memset(array, 0, bytes);
5465         array[DEVCONF_FORWARDING] = cnf->forwarding;
5466         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5467         array[DEVCONF_MTU6] = cnf->mtu6;
5468         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5469         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5470         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5471         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5472         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5473         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5474                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5475         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5476                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5477         array[DEVCONF_RTR_SOLICIT_DELAY] =
5478                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5479         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5480         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5481                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5482         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5483                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5484         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5485         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5486         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5487         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5488         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5489         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5490         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5491         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5492         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5493         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5494 #ifdef CONFIG_IPV6_ROUTER_PREF
5495         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5496         array[DEVCONF_RTR_PROBE_INTERVAL] =
5497                 jiffies_to_msecs(cnf->rtr_probe_interval);
5498 #ifdef CONFIG_IPV6_ROUTE_INFO
5499         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5500         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5501 #endif
5502 #endif
5503         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5504         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5505 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5506         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5507         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5508 #endif
5509 #ifdef CONFIG_IPV6_MROUTE
5510         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5511 #endif
5512         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5513         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5514         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5515         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5516         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5517         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5518         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5519         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5520         /* we omit DEVCONF_STABLE_SECRET for now */
5521         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5522         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5523         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5524         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5525         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5526 #ifdef CONFIG_IPV6_SEG6_HMAC
5527         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5528 #endif
5529         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5530         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5531         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5532         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5533         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5534         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5535         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5536         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5537 }
5538
5539 static inline size_t inet6_ifla6_size(void)
5540 {
5541         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5542              + nla_total_size(sizeof(struct ifla_cacheinfo))
5543              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5544              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5545              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5546              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5547              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5548              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5549              + 0;
5550 }
5551
5552 static inline size_t inet6_if_nlmsg_size(void)
5553 {
5554         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5555                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5556                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5557                + nla_total_size(4) /* IFLA_MTU */
5558                + nla_total_size(4) /* IFLA_LINK */
5559                + nla_total_size(1) /* IFLA_OPERSTATE */
5560                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5561 }
5562
5563 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5564                                         int bytes)
5565 {
5566         int i;
5567         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5568         BUG_ON(pad < 0);
5569
5570         /* Use put_unaligned() because stats may not be aligned for u64. */
5571         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5572         for (i = 1; i < ICMP6_MIB_MAX; i++)
5573                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5574
5575         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5576 }
5577
5578 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5579                                         int bytes, size_t syncpoff)
5580 {
5581         int i, c;
5582         u64 buff[IPSTATS_MIB_MAX];
5583         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5584
5585         BUG_ON(pad < 0);
5586
5587         memset(buff, 0, sizeof(buff));
5588         buff[0] = IPSTATS_MIB_MAX;
5589
5590         for_each_possible_cpu(c) {
5591                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5592                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5593         }
5594
5595         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5596         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5597 }
5598
5599 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5600                              int bytes)
5601 {
5602         switch (attrtype) {
5603         case IFLA_INET6_STATS:
5604                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5605                                      offsetof(struct ipstats_mib, syncp));
5606                 break;
5607         case IFLA_INET6_ICMP6STATS:
5608                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5609                 break;
5610         }
5611 }
5612
5613 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5614                                   u32 ext_filter_mask)
5615 {
5616         struct nlattr *nla;
5617         struct ifla_cacheinfo ci;
5618
5619         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5620                 goto nla_put_failure;
5621         ci.max_reasm_len = IPV6_MAXPLEN;
5622         ci.tstamp = cstamp_delta(idev->tstamp);
5623         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5624         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5625         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5626                 goto nla_put_failure;
5627         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5628         if (!nla)
5629                 goto nla_put_failure;
5630         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5631
5632         /* XXX - MC not implemented */
5633
5634         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5635                 return 0;
5636
5637         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5638         if (!nla)
5639                 goto nla_put_failure;
5640         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5641
5642         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5643         if (!nla)
5644                 goto nla_put_failure;
5645         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5646
5647         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5648         if (!nla)
5649                 goto nla_put_failure;
5650         read_lock_bh(&idev->lock);
5651         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5652         read_unlock_bh(&idev->lock);
5653
5654         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5655                 goto nla_put_failure;
5656
5657         if (idev->ra_mtu &&
5658             nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5659                 goto nla_put_failure;
5660
5661         return 0;
5662
5663 nla_put_failure:
5664         return -EMSGSIZE;
5665 }
5666
5667 static size_t inet6_get_link_af_size(const struct net_device *dev,
5668                                      u32 ext_filter_mask)
5669 {
5670         if (!__in6_dev_get(dev))
5671                 return 0;
5672
5673         return inet6_ifla6_size();
5674 }
5675
5676 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5677                               u32 ext_filter_mask)
5678 {
5679         struct inet6_dev *idev = __in6_dev_get(dev);
5680
5681         if (!idev)
5682                 return -ENODATA;
5683
5684         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5685                 return -EMSGSIZE;
5686
5687         return 0;
5688 }
5689
5690 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5691                              struct netlink_ext_ack *extack)
5692 {
5693         struct inet6_ifaddr *ifp;
5694         struct net_device *dev = idev->dev;
5695         bool clear_token, update_rs = false;
5696         struct in6_addr ll_addr;
5697
5698         ASSERT_RTNL();
5699
5700         if (!token)
5701                 return -EINVAL;
5702
5703         if (dev->flags & IFF_LOOPBACK) {
5704                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5705                 return -EINVAL;
5706         }
5707
5708         if (dev->flags & IFF_NOARP) {
5709                 NL_SET_ERR_MSG_MOD(extack,
5710                                    "Device does not do neighbour discovery");
5711                 return -EINVAL;
5712         }
5713
5714         if (!ipv6_accept_ra(idev)) {
5715                 NL_SET_ERR_MSG_MOD(extack,
5716                                    "Router advertisement is disabled on device");
5717                 return -EINVAL;
5718         }
5719
5720         if (idev->cnf.rtr_solicits == 0) {
5721                 NL_SET_ERR_MSG(extack,
5722                                "Router solicitation is disabled on device");
5723                 return -EINVAL;
5724         }
5725
5726         write_lock_bh(&idev->lock);
5727
5728         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5729         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5730
5731         write_unlock_bh(&idev->lock);
5732
5733         clear_token = ipv6_addr_any(token);
5734         if (clear_token)
5735                 goto update_lft;
5736
5737         if (!idev->dead && (idev->if_flags & IF_READY) &&
5738             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5739                              IFA_F_OPTIMISTIC)) {
5740                 /* If we're not ready, then normal ifup will take care
5741                  * of this. Otherwise, we need to request our rs here.
5742                  */
5743                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5744                 update_rs = true;
5745         }
5746
5747 update_lft:
5748         write_lock_bh(&idev->lock);
5749
5750         if (update_rs) {
5751                 idev->if_flags |= IF_RS_SENT;
5752                 idev->rs_interval = rfc3315_s14_backoff_init(
5753                         idev->cnf.rtr_solicit_interval);
5754                 idev->rs_probes = 1;
5755                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5756         }
5757
5758         /* Well, that's kinda nasty ... */
5759         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5760                 spin_lock(&ifp->lock);
5761                 if (ifp->tokenized) {
5762                         ifp->valid_lft = 0;
5763                         ifp->prefered_lft = 0;
5764                 }
5765                 spin_unlock(&ifp->lock);
5766         }
5767
5768         write_unlock_bh(&idev->lock);
5769         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5770         addrconf_verify_rtnl();
5771         return 0;
5772 }
5773
5774 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5775         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5776         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5777         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5778                                             .reject_message =
5779                                                 "IFLA_INET6_RA_MTU can not be set" },
5780 };
5781
5782 static int check_addr_gen_mode(int mode)
5783 {
5784         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5785             mode != IN6_ADDR_GEN_MODE_NONE &&
5786             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5787             mode != IN6_ADDR_GEN_MODE_RANDOM)
5788                 return -EINVAL;
5789         return 1;
5790 }
5791
5792 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5793                                 int mode)
5794 {
5795         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5796             !idev->cnf.stable_secret.initialized &&
5797             !net->ipv6.devconf_dflt->stable_secret.initialized)
5798                 return -EINVAL;
5799         return 1;
5800 }
5801
5802 static int inet6_validate_link_af(const struct net_device *dev,
5803                                   const struct nlattr *nla,
5804                                   struct netlink_ext_ack *extack)
5805 {
5806         struct nlattr *tb[IFLA_INET6_MAX + 1];
5807         struct inet6_dev *idev = NULL;
5808         int err;
5809
5810         if (dev) {
5811                 idev = __in6_dev_get(dev);
5812                 if (!idev)
5813                         return -EAFNOSUPPORT;
5814         }
5815
5816         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5817                                           inet6_af_policy, extack);
5818         if (err)
5819                 return err;
5820
5821         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5822                 return -EINVAL;
5823
5824         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5825                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5826
5827                 if (check_addr_gen_mode(mode) < 0)
5828                         return -EINVAL;
5829                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5830                         return -EINVAL;
5831         }
5832
5833         return 0;
5834 }
5835
5836 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5837                              struct netlink_ext_ack *extack)
5838 {
5839         struct inet6_dev *idev = __in6_dev_get(dev);
5840         struct nlattr *tb[IFLA_INET6_MAX + 1];
5841         int err;
5842
5843         if (!idev)
5844                 return -EAFNOSUPPORT;
5845
5846         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5847                 return -EINVAL;
5848
5849         if (tb[IFLA_INET6_TOKEN]) {
5850                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5851                                         extack);
5852                 if (err)
5853                         return err;
5854         }
5855
5856         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5857                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5858
5859                 idev->cnf.addr_gen_mode = mode;
5860         }
5861
5862         return 0;
5863 }
5864
5865 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5866                              u32 portid, u32 seq, int event, unsigned int flags)
5867 {
5868         struct net_device *dev = idev->dev;
5869         struct ifinfomsg *hdr;
5870         struct nlmsghdr *nlh;
5871         void *protoinfo;
5872
5873         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5874         if (!nlh)
5875                 return -EMSGSIZE;
5876
5877         hdr = nlmsg_data(nlh);
5878         hdr->ifi_family = AF_INET6;
5879         hdr->__ifi_pad = 0;
5880         hdr->ifi_type = dev->type;
5881         hdr->ifi_index = dev->ifindex;
5882         hdr->ifi_flags = dev_get_flags(dev);
5883         hdr->ifi_change = 0;
5884
5885         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5886             (dev->addr_len &&
5887              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5888             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5889             (dev->ifindex != dev_get_iflink(dev) &&
5890              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5891             nla_put_u8(skb, IFLA_OPERSTATE,
5892                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5893                 goto nla_put_failure;
5894         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5895         if (!protoinfo)
5896                 goto nla_put_failure;
5897
5898         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5899                 goto nla_put_failure;
5900
5901         nla_nest_end(skb, protoinfo);
5902         nlmsg_end(skb, nlh);
5903         return 0;
5904
5905 nla_put_failure:
5906         nlmsg_cancel(skb, nlh);
5907         return -EMSGSIZE;
5908 }
5909
5910 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5911                                    struct netlink_ext_ack *extack)
5912 {
5913         struct ifinfomsg *ifm;
5914
5915         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5916                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5917                 return -EINVAL;
5918         }
5919
5920         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5921                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5922                 return -EINVAL;
5923         }
5924
5925         ifm = nlmsg_data(nlh);
5926         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5927             ifm->ifi_change || ifm->ifi_index) {
5928                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5929                 return -EINVAL;
5930         }
5931
5932         return 0;
5933 }
5934
5935 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5936 {
5937         struct net *net = sock_net(skb->sk);
5938         int h, s_h;
5939         int idx = 0, s_idx;
5940         struct net_device *dev;
5941         struct inet6_dev *idev;
5942         struct hlist_head *head;
5943
5944         /* only requests using strict checking can pass data to
5945          * influence the dump
5946          */
5947         if (cb->strict_check) {
5948                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5949
5950                 if (err < 0)
5951                         return err;
5952         }
5953
5954         s_h = cb->args[0];
5955         s_idx = cb->args[1];
5956
5957         rcu_read_lock();
5958         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5959                 idx = 0;
5960                 head = &net->dev_index_head[h];
5961                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5962                         if (idx < s_idx)
5963                                 goto cont;
5964                         idev = __in6_dev_get(dev);
5965                         if (!idev)
5966                                 goto cont;
5967                         if (inet6_fill_ifinfo(skb, idev,
5968                                               NETLINK_CB(cb->skb).portid,
5969                                               cb->nlh->nlmsg_seq,
5970                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5971                                 goto out;
5972 cont:
5973                         idx++;
5974                 }
5975         }
5976 out:
5977         rcu_read_unlock();
5978         cb->args[1] = idx;
5979         cb->args[0] = h;
5980
5981         return skb->len;
5982 }
5983
5984 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5985 {
5986         struct sk_buff *skb;
5987         struct net *net = dev_net(idev->dev);
5988         int err = -ENOBUFS;
5989
5990         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5991         if (!skb)
5992                 goto errout;
5993
5994         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5995         if (err < 0) {
5996                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5997                 WARN_ON(err == -EMSGSIZE);
5998                 kfree_skb(skb);
5999                 goto errout;
6000         }
6001         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6002         return;
6003 errout:
6004         if (err < 0)
6005                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6006 }
6007
6008 static inline size_t inet6_prefix_nlmsg_size(void)
6009 {
6010         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6011                + nla_total_size(sizeof(struct in6_addr))
6012                + nla_total_size(sizeof(struct prefix_cacheinfo));
6013 }
6014
6015 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6016                              struct prefix_info *pinfo, u32 portid, u32 seq,
6017                              int event, unsigned int flags)
6018 {
6019         struct prefixmsg *pmsg;
6020         struct nlmsghdr *nlh;
6021         struct prefix_cacheinfo ci;
6022
6023         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6024         if (!nlh)
6025                 return -EMSGSIZE;
6026
6027         pmsg = nlmsg_data(nlh);
6028         pmsg->prefix_family = AF_INET6;
6029         pmsg->prefix_pad1 = 0;
6030         pmsg->prefix_pad2 = 0;
6031         pmsg->prefix_ifindex = idev->dev->ifindex;
6032         pmsg->prefix_len = pinfo->prefix_len;
6033         pmsg->prefix_type = pinfo->type;
6034         pmsg->prefix_pad3 = 0;
6035         pmsg->prefix_flags = 0;
6036         if (pinfo->onlink)
6037                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6038         if (pinfo->autoconf)
6039                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6040
6041         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6042                 goto nla_put_failure;
6043         ci.preferred_time = ntohl(pinfo->prefered);
6044         ci.valid_time = ntohl(pinfo->valid);
6045         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6046                 goto nla_put_failure;
6047         nlmsg_end(skb, nlh);
6048         return 0;
6049
6050 nla_put_failure:
6051         nlmsg_cancel(skb, nlh);
6052         return -EMSGSIZE;
6053 }
6054
6055 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6056                          struct prefix_info *pinfo)
6057 {
6058         struct sk_buff *skb;
6059         struct net *net = dev_net(idev->dev);
6060         int err = -ENOBUFS;
6061
6062         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6063         if (!skb)
6064                 goto errout;
6065
6066         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6067         if (err < 0) {
6068                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6069                 WARN_ON(err == -EMSGSIZE);
6070                 kfree_skb(skb);
6071                 goto errout;
6072         }
6073         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6074         return;
6075 errout:
6076         if (err < 0)
6077                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6078 }
6079
6080 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6081 {
6082         struct net *net = dev_net(ifp->idev->dev);
6083
6084         if (event)
6085                 ASSERT_RTNL();
6086
6087         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6088
6089         switch (event) {
6090         case RTM_NEWADDR:
6091                 /*
6092                  * If the address was optimistic we inserted the route at the
6093                  * start of our DAD process, so we don't need to do it again.
6094                  * If the device was taken down in the middle of the DAD
6095                  * cycle there is a race where we could get here without a
6096                  * host route, so nothing to insert. That will be fixed when
6097                  * the device is brought up.
6098                  */
6099                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6100                         ip6_ins_rt(net, ifp->rt);
6101                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6102                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6103                                 &ifp->addr, ifp->idev->dev->name);
6104                 }
6105
6106                 if (ifp->idev->cnf.forwarding)
6107                         addrconf_join_anycast(ifp);
6108                 if (!ipv6_addr_any(&ifp->peer_addr))
6109                         addrconf_prefix_route(&ifp->peer_addr, 128,
6110                                               ifp->rt_priority, ifp->idev->dev,
6111                                               0, 0, GFP_ATOMIC);
6112                 break;
6113         case RTM_DELADDR:
6114                 if (ifp->idev->cnf.forwarding)
6115                         addrconf_leave_anycast(ifp);
6116                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6117                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6118                         struct fib6_info *rt;
6119
6120                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6121                                                        ifp->idev->dev, 0, 0,
6122                                                        false);
6123                         if (rt)
6124                                 ip6_del_rt(net, rt, false);
6125                 }
6126                 if (ifp->rt) {
6127                         ip6_del_rt(net, ifp->rt, false);
6128                         ifp->rt = NULL;
6129                 }
6130                 rt_genid_bump_ipv6(net);
6131                 break;
6132         }
6133         atomic_inc(&net->ipv6.dev_addr_genid);
6134 }
6135
6136 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6137 {
6138         if (likely(ifp->idev->dead == 0))
6139                 __ipv6_ifa_notify(event, ifp);
6140 }
6141
6142 #ifdef CONFIG_SYSCTL
6143
6144 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6145                 void *buffer, size_t *lenp, loff_t *ppos)
6146 {
6147         int *valp = ctl->data;
6148         int val = *valp;
6149         loff_t pos = *ppos;
6150         struct ctl_table lctl;
6151         int ret;
6152
6153         /*
6154          * ctl->data points to idev->cnf.forwarding, we should
6155          * not modify it until we get the rtnl lock.
6156          */
6157         lctl = *ctl;
6158         lctl.data = &val;
6159
6160         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6161
6162         if (write)
6163                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6164         if (ret)
6165                 *ppos = pos;
6166         return ret;
6167 }
6168
6169 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6170                 void *buffer, size_t *lenp, loff_t *ppos)
6171 {
6172         struct inet6_dev *idev = ctl->extra1;
6173         int min_mtu = IPV6_MIN_MTU;
6174         struct ctl_table lctl;
6175
6176         lctl = *ctl;
6177         lctl.extra1 = &min_mtu;
6178         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6179
6180         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6181 }
6182
6183 static void dev_disable_change(struct inet6_dev *idev)
6184 {
6185         struct netdev_notifier_info info;
6186
6187         if (!idev || !idev->dev)
6188                 return;
6189
6190         netdev_notifier_info_init(&info, idev->dev);
6191         if (idev->cnf.disable_ipv6)
6192                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6193         else
6194                 addrconf_notify(NULL, NETDEV_UP, &info);
6195 }
6196
6197 static void addrconf_disable_change(struct net *net, __s32 newf)
6198 {
6199         struct net_device *dev;
6200         struct inet6_dev *idev;
6201
6202         for_each_netdev(net, dev) {
6203                 idev = __in6_dev_get(dev);
6204                 if (idev) {
6205                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6206                         idev->cnf.disable_ipv6 = newf;
6207                         if (changed)
6208                                 dev_disable_change(idev);
6209                 }
6210         }
6211 }
6212
6213 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6214 {
6215         struct net *net;
6216         int old;
6217
6218         if (!rtnl_trylock())
6219                 return restart_syscall();
6220
6221         net = (struct net *)table->extra2;
6222         old = *p;
6223         *p = newf;
6224
6225         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6226                 rtnl_unlock();
6227                 return 0;
6228         }
6229
6230         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6231                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6232                 addrconf_disable_change(net, newf);
6233         } else if ((!newf) ^ (!old))
6234                 dev_disable_change((struct inet6_dev *)table->extra1);
6235
6236         rtnl_unlock();
6237         return 0;
6238 }
6239
6240 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6241                 void *buffer, size_t *lenp, loff_t *ppos)
6242 {
6243         int *valp = ctl->data;
6244         int val = *valp;
6245         loff_t pos = *ppos;
6246         struct ctl_table lctl;
6247         int ret;
6248
6249         /*
6250          * ctl->data points to idev->cnf.disable_ipv6, we should
6251          * not modify it until we get the rtnl lock.
6252          */
6253         lctl = *ctl;
6254         lctl.data = &val;
6255
6256         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6257
6258         if (write)
6259                 ret = addrconf_disable_ipv6(ctl, valp, val);
6260         if (ret)
6261                 *ppos = pos;
6262         return ret;
6263 }
6264
6265 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6266                 void *buffer, size_t *lenp, loff_t *ppos)
6267 {
6268         int *valp = ctl->data;
6269         int ret;
6270         int old, new;
6271
6272         old = *valp;
6273         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6274         new = *valp;
6275
6276         if (write && old != new) {
6277                 struct net *net = ctl->extra2;
6278
6279                 if (!rtnl_trylock())
6280                         return restart_syscall();
6281
6282                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6283                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6284                                                      NETCONFA_PROXY_NEIGH,
6285                                                      NETCONFA_IFINDEX_DEFAULT,
6286                                                      net->ipv6.devconf_dflt);
6287                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6288                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6289                                                      NETCONFA_PROXY_NEIGH,
6290                                                      NETCONFA_IFINDEX_ALL,
6291                                                      net->ipv6.devconf_all);
6292                 else {
6293                         struct inet6_dev *idev = ctl->extra1;
6294
6295                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6296                                                      NETCONFA_PROXY_NEIGH,
6297                                                      idev->dev->ifindex,
6298                                                      &idev->cnf);
6299                 }
6300                 rtnl_unlock();
6301         }
6302
6303         return ret;
6304 }
6305
6306 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6307                                          void *buffer, size_t *lenp,
6308                                          loff_t *ppos)
6309 {
6310         int ret = 0;
6311         u32 new_val;
6312         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6313         struct net *net = (struct net *)ctl->extra2;
6314         struct ctl_table tmp = {
6315                 .data = &new_val,
6316                 .maxlen = sizeof(new_val),
6317                 .mode = ctl->mode,
6318         };
6319
6320         if (!rtnl_trylock())
6321                 return restart_syscall();
6322
6323         new_val = *((u32 *)ctl->data);
6324
6325         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6326         if (ret != 0)
6327                 goto out;
6328
6329         if (write) {
6330                 if (check_addr_gen_mode(new_val) < 0) {
6331                         ret = -EINVAL;
6332                         goto out;
6333                 }
6334
6335                 if (idev) {
6336                         if (check_stable_privacy(idev, net, new_val) < 0) {
6337                                 ret = -EINVAL;
6338                                 goto out;
6339                         }
6340
6341                         if (idev->cnf.addr_gen_mode != new_val) {
6342                                 idev->cnf.addr_gen_mode = new_val;
6343                                 addrconf_dev_config(idev->dev);
6344                         }
6345                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6346                         struct net_device *dev;
6347
6348                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6349                         for_each_netdev(net, dev) {
6350                                 idev = __in6_dev_get(dev);
6351                                 if (idev &&
6352                                     idev->cnf.addr_gen_mode != new_val) {
6353                                         idev->cnf.addr_gen_mode = new_val;
6354                                         addrconf_dev_config(idev->dev);
6355                                 }
6356                         }
6357                 }
6358
6359                 *((u32 *)ctl->data) = new_val;
6360         }
6361
6362 out:
6363         rtnl_unlock();
6364
6365         return ret;
6366 }
6367
6368 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6369                                          void *buffer, size_t *lenp,
6370                                          loff_t *ppos)
6371 {
6372         int err;
6373         struct in6_addr addr;
6374         char str[IPV6_MAX_STRLEN];
6375         struct ctl_table lctl = *ctl;
6376         struct net *net = ctl->extra2;
6377         struct ipv6_stable_secret *secret = ctl->data;
6378
6379         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6380                 return -EIO;
6381
6382         lctl.maxlen = IPV6_MAX_STRLEN;
6383         lctl.data = str;
6384
6385         if (!rtnl_trylock())
6386                 return restart_syscall();
6387
6388         if (!write && !secret->initialized) {
6389                 err = -EIO;
6390                 goto out;
6391         }
6392
6393         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6394         if (err >= sizeof(str)) {
6395                 err = -EIO;
6396                 goto out;
6397         }
6398
6399         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6400         if (err || !write)
6401                 goto out;
6402
6403         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6404                 err = -EIO;
6405                 goto out;
6406         }
6407
6408         secret->initialized = true;
6409         secret->secret = addr;
6410
6411         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6412                 struct net_device *dev;
6413
6414                 for_each_netdev(net, dev) {
6415                         struct inet6_dev *idev = __in6_dev_get(dev);
6416
6417                         if (idev) {
6418                                 idev->cnf.addr_gen_mode =
6419                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6420                         }
6421                 }
6422         } else {
6423                 struct inet6_dev *idev = ctl->extra1;
6424
6425                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6426         }
6427
6428 out:
6429         rtnl_unlock();
6430
6431         return err;
6432 }
6433
6434 static
6435 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6436                                                 int write, void *buffer,
6437                                                 size_t *lenp,
6438                                                 loff_t *ppos)
6439 {
6440         int *valp = ctl->data;
6441         int val = *valp;
6442         loff_t pos = *ppos;
6443         struct ctl_table lctl;
6444         int ret;
6445
6446         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6447          * we should not modify it until we get the rtnl lock.
6448          */
6449         lctl = *ctl;
6450         lctl.data = &val;
6451
6452         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6453
6454         if (write)
6455                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6456         if (ret)
6457                 *ppos = pos;
6458         return ret;
6459 }
6460
6461 static
6462 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6463 {
6464         if (rt) {
6465                 if (action)
6466                         rt->dst.flags |= DST_NOPOLICY;
6467                 else
6468                         rt->dst.flags &= ~DST_NOPOLICY;
6469         }
6470 }
6471
6472 static
6473 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6474 {
6475         struct inet6_ifaddr *ifa;
6476
6477         read_lock_bh(&idev->lock);
6478         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6479                 spin_lock(&ifa->lock);
6480                 if (ifa->rt) {
6481                         /* host routes only use builtin fib6_nh */
6482                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6483                         int cpu;
6484
6485                         rcu_read_lock();
6486                         ifa->rt->dst_nopolicy = val ? true : false;
6487                         if (nh->rt6i_pcpu) {
6488                                 for_each_possible_cpu(cpu) {
6489                                         struct rt6_info **rtp;
6490
6491                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6492                                         addrconf_set_nopolicy(*rtp, val);
6493                                 }
6494                         }
6495                         rcu_read_unlock();
6496                 }
6497                 spin_unlock(&ifa->lock);
6498         }
6499         read_unlock_bh(&idev->lock);
6500 }
6501
6502 static
6503 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6504 {
6505         struct inet6_dev *idev;
6506         struct net *net;
6507
6508         if (!rtnl_trylock())
6509                 return restart_syscall();
6510
6511         *valp = val;
6512
6513         net = (struct net *)ctl->extra2;
6514         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6515                 rtnl_unlock();
6516                 return 0;
6517         }
6518
6519         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6520                 struct net_device *dev;
6521
6522                 for_each_netdev(net, dev) {
6523                         idev = __in6_dev_get(dev);
6524                         if (idev)
6525                                 addrconf_disable_policy_idev(idev, val);
6526                 }
6527         } else {
6528                 idev = (struct inet6_dev *)ctl->extra1;
6529                 addrconf_disable_policy_idev(idev, val);
6530         }
6531
6532         rtnl_unlock();
6533         return 0;
6534 }
6535
6536 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6537                                    void *buffer, size_t *lenp, loff_t *ppos)
6538 {
6539         int *valp = ctl->data;
6540         int val = *valp;
6541         loff_t pos = *ppos;
6542         struct ctl_table lctl;
6543         int ret;
6544
6545         lctl = *ctl;
6546         lctl.data = &val;
6547         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6548
6549         if (write && (*valp != val))
6550                 ret = addrconf_disable_policy(ctl, valp, val);
6551
6552         if (ret)
6553                 *ppos = pos;
6554
6555         return ret;
6556 }
6557
6558 static int minus_one = -1;
6559 static const int two_five_five = 255;
6560 static u32 ioam6_if_id_max = U16_MAX;
6561
6562 static const struct ctl_table addrconf_sysctl[] = {
6563         {
6564                 .procname       = "forwarding",
6565                 .data           = &ipv6_devconf.forwarding,
6566                 .maxlen         = sizeof(int),
6567                 .mode           = 0644,
6568                 .proc_handler   = addrconf_sysctl_forward,
6569         },
6570         {
6571                 .procname       = "hop_limit",
6572                 .data           = &ipv6_devconf.hop_limit,
6573                 .maxlen         = sizeof(int),
6574                 .mode           = 0644,
6575                 .proc_handler   = proc_dointvec_minmax,
6576                 .extra1         = (void *)SYSCTL_ONE,
6577                 .extra2         = (void *)&two_five_five,
6578         },
6579         {
6580                 .procname       = "mtu",
6581                 .data           = &ipv6_devconf.mtu6,
6582                 .maxlen         = sizeof(int),
6583                 .mode           = 0644,
6584                 .proc_handler   = addrconf_sysctl_mtu,
6585         },
6586         {
6587                 .procname       = "accept_ra",
6588                 .data           = &ipv6_devconf.accept_ra,
6589                 .maxlen         = sizeof(int),
6590                 .mode           = 0644,
6591                 .proc_handler   = proc_dointvec,
6592         },
6593         {
6594                 .procname       = "accept_redirects",
6595                 .data           = &ipv6_devconf.accept_redirects,
6596                 .maxlen         = sizeof(int),
6597                 .mode           = 0644,
6598                 .proc_handler   = proc_dointvec,
6599         },
6600         {
6601                 .procname       = "autoconf",
6602                 .data           = &ipv6_devconf.autoconf,
6603                 .maxlen         = sizeof(int),
6604                 .mode           = 0644,
6605                 .proc_handler   = proc_dointvec,
6606         },
6607         {
6608                 .procname       = "dad_transmits",
6609                 .data           = &ipv6_devconf.dad_transmits,
6610                 .maxlen         = sizeof(int),
6611                 .mode           = 0644,
6612                 .proc_handler   = proc_dointvec,
6613         },
6614         {
6615                 .procname       = "router_solicitations",
6616                 .data           = &ipv6_devconf.rtr_solicits,
6617                 .maxlen         = sizeof(int),
6618                 .mode           = 0644,
6619                 .proc_handler   = proc_dointvec_minmax,
6620                 .extra1         = &minus_one,
6621         },
6622         {
6623                 .procname       = "router_solicitation_interval",
6624                 .data           = &ipv6_devconf.rtr_solicit_interval,
6625                 .maxlen         = sizeof(int),
6626                 .mode           = 0644,
6627                 .proc_handler   = proc_dointvec_jiffies,
6628         },
6629         {
6630                 .procname       = "router_solicitation_max_interval",
6631                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6632                 .maxlen         = sizeof(int),
6633                 .mode           = 0644,
6634                 .proc_handler   = proc_dointvec_jiffies,
6635         },
6636         {
6637                 .procname       = "router_solicitation_delay",
6638                 .data           = &ipv6_devconf.rtr_solicit_delay,
6639                 .maxlen         = sizeof(int),
6640                 .mode           = 0644,
6641                 .proc_handler   = proc_dointvec_jiffies,
6642         },
6643         {
6644                 .procname       = "force_mld_version",
6645                 .data           = &ipv6_devconf.force_mld_version,
6646                 .maxlen         = sizeof(int),
6647                 .mode           = 0644,
6648                 .proc_handler   = proc_dointvec,
6649         },
6650         {
6651                 .procname       = "mldv1_unsolicited_report_interval",
6652                 .data           =
6653                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6654                 .maxlen         = sizeof(int),
6655                 .mode           = 0644,
6656                 .proc_handler   = proc_dointvec_ms_jiffies,
6657         },
6658         {
6659                 .procname       = "mldv2_unsolicited_report_interval",
6660                 .data           =
6661                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6662                 .maxlen         = sizeof(int),
6663                 .mode           = 0644,
6664                 .proc_handler   = proc_dointvec_ms_jiffies,
6665         },
6666         {
6667                 .procname       = "use_tempaddr",
6668                 .data           = &ipv6_devconf.use_tempaddr,
6669                 .maxlen         = sizeof(int),
6670                 .mode           = 0644,
6671                 .proc_handler   = proc_dointvec,
6672         },
6673         {
6674                 .procname       = "temp_valid_lft",
6675                 .data           = &ipv6_devconf.temp_valid_lft,
6676                 .maxlen         = sizeof(int),
6677                 .mode           = 0644,
6678                 .proc_handler   = proc_dointvec,
6679         },
6680         {
6681                 .procname       = "temp_prefered_lft",
6682                 .data           = &ipv6_devconf.temp_prefered_lft,
6683                 .maxlen         = sizeof(int),
6684                 .mode           = 0644,
6685                 .proc_handler   = proc_dointvec,
6686         },
6687         {
6688                 .procname       = "regen_max_retry",
6689                 .data           = &ipv6_devconf.regen_max_retry,
6690                 .maxlen         = sizeof(int),
6691                 .mode           = 0644,
6692                 .proc_handler   = proc_dointvec,
6693         },
6694         {
6695                 .procname       = "max_desync_factor",
6696                 .data           = &ipv6_devconf.max_desync_factor,
6697                 .maxlen         = sizeof(int),
6698                 .mode           = 0644,
6699                 .proc_handler   = proc_dointvec,
6700         },
6701         {
6702                 .procname       = "max_addresses",
6703                 .data           = &ipv6_devconf.max_addresses,
6704                 .maxlen         = sizeof(int),
6705                 .mode           = 0644,
6706                 .proc_handler   = proc_dointvec,
6707         },
6708         {
6709                 .procname       = "accept_ra_defrtr",
6710                 .data           = &ipv6_devconf.accept_ra_defrtr,
6711                 .maxlen         = sizeof(int),
6712                 .mode           = 0644,
6713                 .proc_handler   = proc_dointvec,
6714         },
6715         {
6716                 .procname       = "ra_defrtr_metric",
6717                 .data           = &ipv6_devconf.ra_defrtr_metric,
6718                 .maxlen         = sizeof(u32),
6719                 .mode           = 0644,
6720                 .proc_handler   = proc_douintvec_minmax,
6721                 .extra1         = (void *)SYSCTL_ONE,
6722         },
6723         {
6724                 .procname       = "accept_ra_min_hop_limit",
6725                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6726                 .maxlen         = sizeof(int),
6727                 .mode           = 0644,
6728                 .proc_handler   = proc_dointvec,
6729         },
6730         {
6731                 .procname       = "accept_ra_pinfo",
6732                 .data           = &ipv6_devconf.accept_ra_pinfo,
6733                 .maxlen         = sizeof(int),
6734                 .mode           = 0644,
6735                 .proc_handler   = proc_dointvec,
6736         },
6737 #ifdef CONFIG_IPV6_ROUTER_PREF
6738         {
6739                 .procname       = "accept_ra_rtr_pref",
6740                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6741                 .maxlen         = sizeof(int),
6742                 .mode           = 0644,
6743                 .proc_handler   = proc_dointvec,
6744         },
6745         {
6746                 .procname       = "router_probe_interval",
6747                 .data           = &ipv6_devconf.rtr_probe_interval,
6748                 .maxlen         = sizeof(int),
6749                 .mode           = 0644,
6750                 .proc_handler   = proc_dointvec_jiffies,
6751         },
6752 #ifdef CONFIG_IPV6_ROUTE_INFO
6753         {
6754                 .procname       = "accept_ra_rt_info_min_plen",
6755                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6756                 .maxlen         = sizeof(int),
6757                 .mode           = 0644,
6758                 .proc_handler   = proc_dointvec,
6759         },
6760         {
6761                 .procname       = "accept_ra_rt_info_max_plen",
6762                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6763                 .maxlen         = sizeof(int),
6764                 .mode           = 0644,
6765                 .proc_handler   = proc_dointvec,
6766         },
6767 #endif
6768 #endif
6769         {
6770                 .procname       = "proxy_ndp",
6771                 .data           = &ipv6_devconf.proxy_ndp,
6772                 .maxlen         = sizeof(int),
6773                 .mode           = 0644,
6774                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6775         },
6776         {
6777                 .procname       = "accept_source_route",
6778                 .data           = &ipv6_devconf.accept_source_route,
6779                 .maxlen         = sizeof(int),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_dointvec,
6782         },
6783 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6784         {
6785                 .procname       = "optimistic_dad",
6786                 .data           = &ipv6_devconf.optimistic_dad,
6787                 .maxlen         = sizeof(int),
6788                 .mode           = 0644,
6789                 .proc_handler   = proc_dointvec,
6790         },
6791         {
6792                 .procname       = "use_optimistic",
6793                 .data           = &ipv6_devconf.use_optimistic,
6794                 .maxlen         = sizeof(int),
6795                 .mode           = 0644,
6796                 .proc_handler   = proc_dointvec,
6797         },
6798 #endif
6799 #ifdef CONFIG_IPV6_MROUTE
6800         {
6801                 .procname       = "mc_forwarding",
6802                 .data           = &ipv6_devconf.mc_forwarding,
6803                 .maxlen         = sizeof(int),
6804                 .mode           = 0444,
6805                 .proc_handler   = proc_dointvec,
6806         },
6807 #endif
6808         {
6809                 .procname       = "disable_ipv6",
6810                 .data           = &ipv6_devconf.disable_ipv6,
6811                 .maxlen         = sizeof(int),
6812                 .mode           = 0644,
6813                 .proc_handler   = addrconf_sysctl_disable,
6814         },
6815         {
6816                 .procname       = "accept_dad",
6817                 .data           = &ipv6_devconf.accept_dad,
6818                 .maxlen         = sizeof(int),
6819                 .mode           = 0644,
6820                 .proc_handler   = proc_dointvec,
6821         },
6822         {
6823                 .procname       = "force_tllao",
6824                 .data           = &ipv6_devconf.force_tllao,
6825                 .maxlen         = sizeof(int),
6826                 .mode           = 0644,
6827                 .proc_handler   = proc_dointvec
6828         },
6829         {
6830                 .procname       = "ndisc_notify",
6831                 .data           = &ipv6_devconf.ndisc_notify,
6832                 .maxlen         = sizeof(int),
6833                 .mode           = 0644,
6834                 .proc_handler   = proc_dointvec
6835         },
6836         {
6837                 .procname       = "suppress_frag_ndisc",
6838                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6839                 .maxlen         = sizeof(int),
6840                 .mode           = 0644,
6841                 .proc_handler   = proc_dointvec
6842         },
6843         {
6844                 .procname       = "accept_ra_from_local",
6845                 .data           = &ipv6_devconf.accept_ra_from_local,
6846                 .maxlen         = sizeof(int),
6847                 .mode           = 0644,
6848                 .proc_handler   = proc_dointvec,
6849         },
6850         {
6851                 .procname       = "accept_ra_mtu",
6852                 .data           = &ipv6_devconf.accept_ra_mtu,
6853                 .maxlen         = sizeof(int),
6854                 .mode           = 0644,
6855                 .proc_handler   = proc_dointvec,
6856         },
6857         {
6858                 .procname       = "stable_secret",
6859                 .data           = &ipv6_devconf.stable_secret,
6860                 .maxlen         = IPV6_MAX_STRLEN,
6861                 .mode           = 0600,
6862                 .proc_handler   = addrconf_sysctl_stable_secret,
6863         },
6864         {
6865                 .procname       = "use_oif_addrs_only",
6866                 .data           = &ipv6_devconf.use_oif_addrs_only,
6867                 .maxlen         = sizeof(int),
6868                 .mode           = 0644,
6869                 .proc_handler   = proc_dointvec,
6870         },
6871         {
6872                 .procname       = "ignore_routes_with_linkdown",
6873                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6874                 .maxlen         = sizeof(int),
6875                 .mode           = 0644,
6876                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6877         },
6878         {
6879                 .procname       = "drop_unicast_in_l2_multicast",
6880                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6881                 .maxlen         = sizeof(int),
6882                 .mode           = 0644,
6883                 .proc_handler   = proc_dointvec,
6884         },
6885         {
6886                 .procname       = "drop_unsolicited_na",
6887                 .data           = &ipv6_devconf.drop_unsolicited_na,
6888                 .maxlen         = sizeof(int),
6889                 .mode           = 0644,
6890                 .proc_handler   = proc_dointvec,
6891         },
6892         {
6893                 .procname       = "keep_addr_on_down",
6894                 .data           = &ipv6_devconf.keep_addr_on_down,
6895                 .maxlen         = sizeof(int),
6896                 .mode           = 0644,
6897                 .proc_handler   = proc_dointvec,
6898
6899         },
6900         {
6901                 .procname       = "seg6_enabled",
6902                 .data           = &ipv6_devconf.seg6_enabled,
6903                 .maxlen         = sizeof(int),
6904                 .mode           = 0644,
6905                 .proc_handler   = proc_dointvec,
6906         },
6907 #ifdef CONFIG_IPV6_SEG6_HMAC
6908         {
6909                 .procname       = "seg6_require_hmac",
6910                 .data           = &ipv6_devconf.seg6_require_hmac,
6911                 .maxlen         = sizeof(int),
6912                 .mode           = 0644,
6913                 .proc_handler   = proc_dointvec,
6914         },
6915 #endif
6916         {
6917                 .procname       = "enhanced_dad",
6918                 .data           = &ipv6_devconf.enhanced_dad,
6919                 .maxlen         = sizeof(int),
6920                 .mode           = 0644,
6921                 .proc_handler   = proc_dointvec,
6922         },
6923         {
6924                 .procname       = "addr_gen_mode",
6925                 .data           = &ipv6_devconf.addr_gen_mode,
6926                 .maxlen         = sizeof(int),
6927                 .mode           = 0644,
6928                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6929         },
6930         {
6931                 .procname       = "disable_policy",
6932                 .data           = &ipv6_devconf.disable_policy,
6933                 .maxlen         = sizeof(int),
6934                 .mode           = 0644,
6935                 .proc_handler   = addrconf_sysctl_disable_policy,
6936         },
6937         {
6938                 .procname       = "ndisc_tclass",
6939                 .data           = &ipv6_devconf.ndisc_tclass,
6940                 .maxlen         = sizeof(int),
6941                 .mode           = 0644,
6942                 .proc_handler   = proc_dointvec_minmax,
6943                 .extra1         = (void *)SYSCTL_ZERO,
6944                 .extra2         = (void *)&two_five_five,
6945         },
6946         {
6947                 .procname       = "rpl_seg_enabled",
6948                 .data           = &ipv6_devconf.rpl_seg_enabled,
6949                 .maxlen         = sizeof(int),
6950                 .mode           = 0644,
6951                 .proc_handler   = proc_dointvec,
6952         },
6953         {
6954                 .procname       = "ioam6_enabled",
6955                 .data           = &ipv6_devconf.ioam6_enabled,
6956                 .maxlen         = sizeof(u8),
6957                 .mode           = 0644,
6958                 .proc_handler   = proc_dou8vec_minmax,
6959                 .extra1         = (void *)SYSCTL_ZERO,
6960                 .extra2         = (void *)SYSCTL_ONE,
6961         },
6962         {
6963                 .procname       = "ioam6_id",
6964                 .data           = &ipv6_devconf.ioam6_id,
6965                 .maxlen         = sizeof(u32),
6966                 .mode           = 0644,
6967                 .proc_handler   = proc_douintvec_minmax,
6968                 .extra1         = (void *)SYSCTL_ZERO,
6969                 .extra2         = (void *)&ioam6_if_id_max,
6970         },
6971         {
6972                 .procname       = "ioam6_id_wide",
6973                 .data           = &ipv6_devconf.ioam6_id_wide,
6974                 .maxlen         = sizeof(u32),
6975                 .mode           = 0644,
6976                 .proc_handler   = proc_douintvec,
6977         },
6978         {
6979                 /* sentinel */
6980         }
6981 };
6982
6983 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6984                 struct inet6_dev *idev, struct ipv6_devconf *p)
6985 {
6986         int i, ifindex;
6987         struct ctl_table *table;
6988         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6989
6990         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6991         if (!table)
6992                 goto out;
6993
6994         for (i = 0; table[i].data; i++) {
6995                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6996                 /* If one of these is already set, then it is not safe to
6997                  * overwrite either of them: this makes proc_dointvec_minmax
6998                  * usable.
6999                  */
7000                 if (!table[i].extra1 && !table[i].extra2) {
7001                         table[i].extra1 = idev; /* embedded; no ref */
7002                         table[i].extra2 = net;
7003                 }
7004         }
7005
7006         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7007
7008         p->sysctl_header = register_net_sysctl(net, path, table);
7009         if (!p->sysctl_header)
7010                 goto free;
7011
7012         if (!strcmp(dev_name, "all"))
7013                 ifindex = NETCONFA_IFINDEX_ALL;
7014         else if (!strcmp(dev_name, "default"))
7015                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7016         else
7017                 ifindex = idev->dev->ifindex;
7018         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7019                                      ifindex, p);
7020         return 0;
7021
7022 free:
7023         kfree(table);
7024 out:
7025         return -ENOBUFS;
7026 }
7027
7028 static void __addrconf_sysctl_unregister(struct net *net,
7029                                          struct ipv6_devconf *p, int ifindex)
7030 {
7031         struct ctl_table *table;
7032
7033         if (!p->sysctl_header)
7034                 return;
7035
7036         table = p->sysctl_header->ctl_table_arg;
7037         unregister_net_sysctl_table(p->sysctl_header);
7038         p->sysctl_header = NULL;
7039         kfree(table);
7040
7041         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7042 }
7043
7044 static int addrconf_sysctl_register(struct inet6_dev *idev)
7045 {
7046         int err;
7047
7048         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7049                 return -EINVAL;
7050
7051         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7052                                     &ndisc_ifinfo_sysctl_change);
7053         if (err)
7054                 return err;
7055         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7056                                          idev, &idev->cnf);
7057         if (err)
7058                 neigh_sysctl_unregister(idev->nd_parms);
7059
7060         return err;
7061 }
7062
7063 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7064 {
7065         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7066                                      idev->dev->ifindex);
7067         neigh_sysctl_unregister(idev->nd_parms);
7068 }
7069
7070
7071 #endif
7072
7073 static int __net_init addrconf_init_net(struct net *net)
7074 {
7075         int err = -ENOMEM;
7076         struct ipv6_devconf *all, *dflt;
7077
7078         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7079         if (!all)
7080                 goto err_alloc_all;
7081
7082         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7083         if (!dflt)
7084                 goto err_alloc_dflt;
7085
7086         if (IS_ENABLED(CONFIG_SYSCTL) &&
7087             !net_eq(net, &init_net)) {
7088                 switch (sysctl_devconf_inherit_init_net) {
7089                 case 1:  /* copy from init_net */
7090                         memcpy(all, init_net.ipv6.devconf_all,
7091                                sizeof(ipv6_devconf));
7092                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7093                                sizeof(ipv6_devconf_dflt));
7094                         break;
7095                 case 3: /* copy from the current netns */
7096                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7097                                sizeof(ipv6_devconf));
7098                         memcpy(dflt,
7099                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7100                                sizeof(ipv6_devconf_dflt));
7101                         break;
7102                 case 0:
7103                 case 2:
7104                         /* use compiled values */
7105                         break;
7106                 }
7107         }
7108
7109         /* these will be inherited by all namespaces */
7110         dflt->autoconf = ipv6_defaults.autoconf;
7111         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7112
7113         dflt->stable_secret.initialized = false;
7114         all->stable_secret.initialized = false;
7115
7116         net->ipv6.devconf_all = all;
7117         net->ipv6.devconf_dflt = dflt;
7118
7119 #ifdef CONFIG_SYSCTL
7120         err = __addrconf_sysctl_register(net, "all", NULL, all);
7121         if (err < 0)
7122                 goto err_reg_all;
7123
7124         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7125         if (err < 0)
7126                 goto err_reg_dflt;
7127 #endif
7128         return 0;
7129
7130 #ifdef CONFIG_SYSCTL
7131 err_reg_dflt:
7132         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7133 err_reg_all:
7134         kfree(dflt);
7135 #endif
7136 err_alloc_dflt:
7137         kfree(all);
7138 err_alloc_all:
7139         return err;
7140 }
7141
7142 static void __net_exit addrconf_exit_net(struct net *net)
7143 {
7144 #ifdef CONFIG_SYSCTL
7145         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7146                                      NETCONFA_IFINDEX_DEFAULT);
7147         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7148                                      NETCONFA_IFINDEX_ALL);
7149 #endif
7150         kfree(net->ipv6.devconf_dflt);
7151         kfree(net->ipv6.devconf_all);
7152 }
7153
7154 static struct pernet_operations addrconf_ops = {
7155         .init = addrconf_init_net,
7156         .exit = addrconf_exit_net,
7157 };
7158
7159 static struct rtnl_af_ops inet6_ops __read_mostly = {
7160         .family           = AF_INET6,
7161         .fill_link_af     = inet6_fill_link_af,
7162         .get_link_af_size = inet6_get_link_af_size,
7163         .validate_link_af = inet6_validate_link_af,
7164         .set_link_af      = inet6_set_link_af,
7165 };
7166
7167 /*
7168  *      Init / cleanup code
7169  */
7170
7171 int __init addrconf_init(void)
7172 {
7173         struct inet6_dev *idev;
7174         int i, err;
7175
7176         err = ipv6_addr_label_init();
7177         if (err < 0) {
7178                 pr_crit("%s: cannot initialize default policy table: %d\n",
7179                         __func__, err);
7180                 goto out;
7181         }
7182
7183         err = register_pernet_subsys(&addrconf_ops);
7184         if (err < 0)
7185                 goto out_addrlabel;
7186
7187         addrconf_wq = create_workqueue("ipv6_addrconf");
7188         if (!addrconf_wq) {
7189                 err = -ENOMEM;
7190                 goto out_nowq;
7191         }
7192
7193         /* The addrconf netdev notifier requires that loopback_dev
7194          * has it's ipv6 private information allocated and setup
7195          * before it can bring up and give link-local addresses
7196          * to other devices which are up.
7197          *
7198          * Unfortunately, loopback_dev is not necessarily the first
7199          * entry in the global dev_base list of net devices.  In fact,
7200          * it is likely to be the very last entry on that list.
7201          * So this causes the notifier registry below to try and
7202          * give link-local addresses to all devices besides loopback_dev
7203          * first, then loopback_dev, which cases all the non-loopback_dev
7204          * devices to fail to get a link-local address.
7205          *
7206          * So, as a temporary fix, allocate the ipv6 structure for
7207          * loopback_dev first by hand.
7208          * Longer term, all of the dependencies ipv6 has upon the loopback
7209          * device and it being up should be removed.
7210          */
7211         rtnl_lock();
7212         idev = ipv6_add_dev(init_net.loopback_dev);
7213         rtnl_unlock();
7214         if (IS_ERR(idev)) {
7215                 err = PTR_ERR(idev);
7216                 goto errlo;
7217         }
7218
7219         ip6_route_init_special_entries();
7220
7221         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7222                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7223
7224         register_netdevice_notifier(&ipv6_dev_notf);
7225
7226         addrconf_verify();
7227
7228         rtnl_af_register(&inet6_ops);
7229
7230         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7231                                    NULL, inet6_dump_ifinfo, 0);
7232         if (err < 0)
7233                 goto errout;
7234
7235         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7236                                    inet6_rtm_newaddr, NULL, 0);
7237         if (err < 0)
7238                 goto errout;
7239         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7240                                    inet6_rtm_deladdr, NULL, 0);
7241         if (err < 0)
7242                 goto errout;
7243         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7244                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7245                                    RTNL_FLAG_DOIT_UNLOCKED);
7246         if (err < 0)
7247                 goto errout;
7248         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7249                                    NULL, inet6_dump_ifmcaddr, 0);
7250         if (err < 0)
7251                 goto errout;
7252         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7253                                    NULL, inet6_dump_ifacaddr, 0);
7254         if (err < 0)
7255                 goto errout;
7256         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7257                                    inet6_netconf_get_devconf,
7258                                    inet6_netconf_dump_devconf,
7259                                    RTNL_FLAG_DOIT_UNLOCKED);
7260         if (err < 0)
7261                 goto errout;
7262         err = ipv6_addr_label_rtnl_register();
7263         if (err < 0)
7264                 goto errout;
7265
7266         return 0;
7267 errout:
7268         rtnl_unregister_all(PF_INET6);
7269         rtnl_af_unregister(&inet6_ops);
7270         unregister_netdevice_notifier(&ipv6_dev_notf);
7271 errlo:
7272         destroy_workqueue(addrconf_wq);
7273 out_nowq:
7274         unregister_pernet_subsys(&addrconf_ops);
7275 out_addrlabel:
7276         ipv6_addr_label_cleanup();
7277 out:
7278         return err;
7279 }
7280
7281 void addrconf_cleanup(void)
7282 {
7283         struct net_device *dev;
7284         int i;
7285
7286         unregister_netdevice_notifier(&ipv6_dev_notf);
7287         unregister_pernet_subsys(&addrconf_ops);
7288         ipv6_addr_label_cleanup();
7289
7290         rtnl_af_unregister(&inet6_ops);
7291
7292         rtnl_lock();
7293
7294         /* clean dev list */
7295         for_each_netdev(&init_net, dev) {
7296                 if (__in6_dev_get(dev) == NULL)
7297                         continue;
7298                 addrconf_ifdown(dev, true);
7299         }
7300         addrconf_ifdown(init_net.loopback_dev, true);
7301
7302         /*
7303          *      Check hash table.
7304          */
7305         spin_lock_bh(&addrconf_hash_lock);
7306         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7307                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7308         spin_unlock_bh(&addrconf_hash_lock);
7309         cancel_delayed_work(&addr_chk_work);
7310         rtnl_unlock();
7311
7312         destroy_workqueue(addrconf_wq);
7313 }