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 = get_random_u32_inclusive(900000, 1100000) * (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 = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
 116         do_div(tmp, 1000000);
 117         if ((s32)tmp > mrt) {
 118                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 119                 tmp = get_random_u32_inclusive(900000, 1100000) * (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 static void addrconf_verify(struct net *net);
 151 static void addrconf_verify_rtnl(struct net *net);
 152
 153 static struct workqueue_struct *addrconf_wq;
 154
 155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 157
 158 static void addrconf_type_change(struct net_device *dev,
 159                                  unsigned long event);
 160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 161
 162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 163                                                   int plen,
 164                                                   const struct net_device *dev,
 165                                                   u32 flags, u32 noflags,
 166                                                   bool no_gw);
 167
 168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 169 static void addrconf_dad_work(struct work_struct *w);
 170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 171                                    bool send_na);
 172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 173 static void addrconf_rs_timer(struct timer_list *t);
 174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 176
 177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 178                                 struct prefix_info *pinfo);
 179
 180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 181         .forwarding             = 0,
 182         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 183         .mtu6                   = IPV6_MIN_MTU,
 184         .accept_ra              = 1,
 185         .accept_redirects       = 1,
 186         .autoconf               = 1,
 187         .force_mld_version      = 0,
 188         .mldv1_unsolicited_report_interval = 10 * HZ,
 189         .mldv2_unsolicited_report_interval = HZ,
 190         .dad_transmits          = 1,
 191         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 192         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 193         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 194         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 195         .use_tempaddr           = 0,
 196         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 197         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 198         .regen_max_retry        = REGEN_MAX_RETRY,
 199         .max_desync_factor      = MAX_DESYNC_FACTOR,
 200         .max_addresses          = IPV6_MAX_ADDRESSES,
 201         .accept_ra_defrtr       = 1,
 202         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 203         .accept_ra_from_local   = 0,
 204         .accept_ra_min_hop_limit= 1,
 205         .accept_ra_pinfo        = 1,
 206 #ifdef CONFIG_IPV6_ROUTER_PREF
 207         .accept_ra_rtr_pref     = 1,
 208         .rtr_probe_interval     = 60 * HZ,
 209 #ifdef CONFIG_IPV6_ROUTE_INFO
 210         .accept_ra_rt_info_min_plen = 0,
 211         .accept_ra_rt_info_max_plen = 0,
 212 #endif
 213 #endif
 214         .proxy_ndp              = 0,
 215         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 216         .disable_ipv6           = 0,
 217         .accept_dad             = 0,
 218         .suppress_frag_ndisc    = 1,
 219         .accept_ra_mtu          = 1,
 220         .stable_secret          = {
 221                 .initialized = false,
 222         },
 223         .use_oif_addrs_only     = 0,
 224         .ignore_routes_with_linkdown = 0,
 225         .keep_addr_on_down      = 0,
 226         .seg6_enabled           = 0,
 227 #ifdef CONFIG_IPV6_SEG6_HMAC
 228         .seg6_require_hmac      = 0,
 229 #endif
 230         .enhanced_dad           = 1,
 231         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 232         .disable_policy         = 0,
 233         .rpl_seg_enabled        = 0,
 234         .ioam6_enabled          = 0,
 235         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 236         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 237         .ndisc_evict_nocarrier  = 1,
 238 };
 239
 240 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 241         .forwarding             = 0,
 242         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 243         .mtu6                   = IPV6_MIN_MTU,
 244         .accept_ra              = 1,
 245         .accept_redirects       = 1,
 246         .autoconf               = 1,
 247         .force_mld_version      = 0,
 248         .mldv1_unsolicited_report_interval = 10 * HZ,
 249         .mldv2_unsolicited_report_interval = HZ,
 250         .dad_transmits          = 1,
 251         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 252         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 253         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 254         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 255         .use_tempaddr           = 0,
 256         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 257         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 258         .regen_max_retry        = REGEN_MAX_RETRY,
 259         .max_desync_factor      = MAX_DESYNC_FACTOR,
 260         .max_addresses          = IPV6_MAX_ADDRESSES,
 261         .accept_ra_defrtr       = 1,
 262         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 263         .accept_ra_from_local   = 0,
 264         .accept_ra_min_hop_limit= 1,
 265         .accept_ra_pinfo        = 1,
 266 #ifdef CONFIG_IPV6_ROUTER_PREF
 267         .accept_ra_rtr_pref     = 1,
 268         .rtr_probe_interval     = 60 * HZ,
 269 #ifdef CONFIG_IPV6_ROUTE_INFO
 270         .accept_ra_rt_info_min_plen = 0,
 271         .accept_ra_rt_info_max_plen = 0,
 272 #endif
 273 #endif
 274         .proxy_ndp              = 0,
 275         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 276         .disable_ipv6           = 0,
 277         .accept_dad             = 1,
 278         .suppress_frag_ndisc    = 1,
 279         .accept_ra_mtu          = 1,
 280         .stable_secret          = {
 281                 .initialized = false,
 282         },
 283         .use_oif_addrs_only     = 0,
 284         .ignore_routes_with_linkdown = 0,
 285         .keep_addr_on_down      = 0,
 286         .seg6_enabled           = 0,
 287 #ifdef CONFIG_IPV6_SEG6_HMAC
 288         .seg6_require_hmac      = 0,
 289 #endif
 290         .enhanced_dad           = 1,
 291         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 292         .disable_policy         = 0,
 293         .rpl_seg_enabled        = 0,
 294         .ioam6_enabled          = 0,
 295         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 296         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 297         .ndisc_evict_nocarrier  = 1,
 298 };
 299
 300 /* Check if link is ready: is it up and is a valid qdisc available */
 301 static inline bool addrconf_link_ready(const struct net_device *dev)
 302 {
 303         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 304 }
 305
 306 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 307 {
 308         if (del_timer(&idev->rs_timer))
 309                 __in6_dev_put(idev);
 310 }
 311
 312 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 313 {
 314         if (cancel_delayed_work(&ifp->dad_work))
 315                 __in6_ifa_put(ifp);
 316 }
 317
 318 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 319                                   unsigned long when)
 320 {
 321         if (!mod_timer(&idev->rs_timer, jiffies + when))
 322                 in6_dev_hold(idev);
 323 }
 324
 325 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 326                                    unsigned long delay)
 327 {
 328         in6_ifa_hold(ifp);
 329         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 330                 in6_ifa_put(ifp);
 331 }
 332
 333 static int snmp6_alloc_dev(struct inet6_dev *idev)
 334 {
 335         int i;
 336
 337         idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
 338         if (!idev->stats.ipv6)
 339                 goto err_ip;
 340
 341         for_each_possible_cpu(i) {
 342                 struct ipstats_mib *addrconf_stats;
 343                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 344                 u64_stats_init(&addrconf_stats->syncp);
 345         }
 346
 347
 348         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 349                                         GFP_KERNEL);
 350         if (!idev->stats.icmpv6dev)
 351                 goto err_icmp;
 352         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 353                                            GFP_KERNEL_ACCOUNT);
 354         if (!idev->stats.icmpv6msgdev)
 355                 goto err_icmpmsg;
 356
 357         return 0;
 358
 359 err_icmpmsg:
 360         kfree(idev->stats.icmpv6dev);
 361 err_icmp:
 362         free_percpu(idev->stats.ipv6);
 363 err_ip:
 364         return -ENOMEM;
 365 }
 366
 367 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 368 {
 369         struct inet6_dev *ndev;
 370         int err = -ENOMEM;
 371
 372         ASSERT_RTNL();
 373
 374         if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
 375                 return ERR_PTR(-EINVAL);
 376
 377         ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
 378         if (!ndev)
 379                 return ERR_PTR(err);
 380
 381         rwlock_init(&ndev->lock);
 382         ndev->dev = dev;
 383         INIT_LIST_HEAD(&ndev->addr_list);
 384         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 385         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 386
 387         if (ndev->cnf.stable_secret.initialized)
 388                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 389
 390         ndev->cnf.mtu6 = dev->mtu;
 391         ndev->ra_mtu = 0;
 392         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 393         if (!ndev->nd_parms) {
 394                 kfree(ndev);
 395                 return ERR_PTR(err);
 396         }
 397         if (ndev->cnf.forwarding)
 398                 dev_disable_lro(dev);
 399         /* We refer to the device */
 400         netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
 401
 402         if (snmp6_alloc_dev(ndev) < 0) {
 403                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 404                            __func__);
 405                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 406                 netdev_put(dev, &ndev->dev_tracker);
 407                 kfree(ndev);
 408                 return ERR_PTR(err);
 409         }
 410
 411         if (dev != blackhole_netdev) {
 412                 if (snmp6_register_dev(ndev) < 0) {
 413                         netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 414                                    __func__, dev->name);
 415                         goto err_release;
 416                 }
 417         }
 418         /* One reference from device. */
 419         refcount_set(&ndev->refcnt, 1);
 420
 421         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 422                 ndev->cnf.accept_dad = -1;
 423
 424 #if IS_ENABLED(CONFIG_IPV6_SIT)
 425         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 426                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 427                 ndev->cnf.rtr_solicits = 0;
 428         }
 429 #endif
 430
 431         INIT_LIST_HEAD(&ndev->tempaddr_list);
 432         ndev->desync_factor = U32_MAX;
 433         if ((dev->flags&IFF_LOOPBACK) ||
 434             dev->type == ARPHRD_TUNNEL ||
 435             dev->type == ARPHRD_TUNNEL6 ||
 436             dev->type == ARPHRD_SIT ||
 437             dev->type == ARPHRD_NONE) {
 438                 ndev->cnf.use_tempaddr = -1;
 439         }
 440
 441         ndev->token = in6addr_any;
 442
 443         if (netif_running(dev) && addrconf_link_ready(dev))
 444                 ndev->if_flags |= IF_READY;
 445
 446         ipv6_mc_init_dev(ndev);
 447         ndev->tstamp = jiffies;
 448         if (dev != blackhole_netdev) {
 449                 err = addrconf_sysctl_register(ndev);
 450                 if (err) {
 451                         ipv6_mc_destroy_dev(ndev);
 452                         snmp6_unregister_dev(ndev);
 453                         goto err_release;
 454                 }
 455         }
 456         /* protected by rtnl_lock */
 457         rcu_assign_pointer(dev->ip6_ptr, ndev);
 458
 459         if (dev != blackhole_netdev) {
 460                 /* Join interface-local all-node multicast group */
 461                 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 462
 463                 /* Join all-node multicast group */
 464                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 465
 466                 /* Join all-router multicast group if forwarding is set */
 467                 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 468                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 469         }
 470         return ndev;
 471
 472 err_release:
 473         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 474         ndev->dead = 1;
 475         in6_dev_finish_destroy(ndev);
 476         return ERR_PTR(err);
 477 }
 478
 479 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 480 {
 481         struct inet6_dev *idev;
 482
 483         ASSERT_RTNL();
 484
 485         idev = __in6_dev_get(dev);
 486         if (!idev) {
 487                 idev = ipv6_add_dev(dev);
 488                 if (IS_ERR(idev))
 489                         return idev;
 490         }
 491
 492         if (dev->flags&IFF_UP)
 493                 ipv6_mc_up(idev);
 494         return idev;
 495 }
 496
 497 static int inet6_netconf_msgsize_devconf(int type)
 498 {
 499         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 500                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 501         bool all = false;
 502
 503         if (type == NETCONFA_ALL)
 504                 all = true;
 505
 506         if (all || type == NETCONFA_FORWARDING)
 507                 size += nla_total_size(4);
 508 #ifdef CONFIG_IPV6_MROUTE
 509         if (all || type == NETCONFA_MC_FORWARDING)
 510                 size += nla_total_size(4);
 511 #endif
 512         if (all || type == NETCONFA_PROXY_NEIGH)
 513                 size += nla_total_size(4);
 514
 515         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 516                 size += nla_total_size(4);
 517
 518         return size;
 519 }
 520
 521 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 522                                       struct ipv6_devconf *devconf, u32 portid,
 523                                       u32 seq, int event, unsigned int flags,
 524                                       int type)
 525 {
 526         struct nlmsghdr  *nlh;
 527         struct netconfmsg *ncm;
 528         bool all = false;
 529
 530         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 531                         flags);
 532         if (!nlh)
 533                 return -EMSGSIZE;
 534
 535         if (type == NETCONFA_ALL)
 536                 all = true;
 537
 538         ncm = nlmsg_data(nlh);
 539         ncm->ncm_family = AF_INET6;
 540
 541         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 542                 goto nla_put_failure;
 543
 544         if (!devconf)
 545                 goto out;
 546
 547         if ((all || type == NETCONFA_FORWARDING) &&
 548             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 549                 goto nla_put_failure;
 550 #ifdef CONFIG_IPV6_MROUTE
 551         if ((all || type == NETCONFA_MC_FORWARDING) &&
 552             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 553                         atomic_read(&devconf->mc_forwarding)) < 0)
 554                 goto nla_put_failure;
 555 #endif
 556         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 557             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 558                 goto nla_put_failure;
 559
 560         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 561             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 562                         devconf->ignore_routes_with_linkdown) < 0)
 563                 goto nla_put_failure;
 564
 565 out:
 566         nlmsg_end(skb, nlh);
 567         return 0;
 568
 569 nla_put_failure:
 570         nlmsg_cancel(skb, nlh);
 571         return -EMSGSIZE;
 572 }
 573
 574 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 575                                   int ifindex, struct ipv6_devconf *devconf)
 576 {
 577         struct sk_buff *skb;
 578         int err = -ENOBUFS;
 579
 580         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 581         if (!skb)
 582                 goto errout;
 583
 584         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 585                                          event, 0, type);
 586         if (err < 0) {
 587                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 588                 WARN_ON(err == -EMSGSIZE);
 589                 kfree_skb(skb);
 590                 goto errout;
 591         }
 592         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 593         return;
 594 errout:
 595         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 596 }
 597
 598 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 599         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 600         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 601         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 602         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 603 };
 604
 605 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 606                                        const struct nlmsghdr *nlh,
 607                                        struct nlattr **tb,
 608                                        struct netlink_ext_ack *extack)
 609 {
 610         int i, err;
 611
 612         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 613                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 614                 return -EINVAL;
 615         }
 616
 617         if (!netlink_strict_get_check(skb))
 618                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 619                                               tb, NETCONFA_MAX,
 620                                               devconf_ipv6_policy, extack);
 621
 622         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 623                                             tb, NETCONFA_MAX,
 624                                             devconf_ipv6_policy, extack);
 625         if (err)
 626                 return err;
 627
 628         for (i = 0; i <= NETCONFA_MAX; i++) {
 629                 if (!tb[i])
 630                         continue;
 631
 632                 switch (i) {
 633                 case NETCONFA_IFINDEX:
 634                         break;
 635                 default:
 636                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 637                         return -EINVAL;
 638                 }
 639         }
 640
 641         return 0;
 642 }
 643
 644 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 645                                      struct nlmsghdr *nlh,
 646                                      struct netlink_ext_ack *extack)
 647 {
 648         struct net *net = sock_net(in_skb->sk);
 649         struct nlattr *tb[NETCONFA_MAX+1];
 650         struct inet6_dev *in6_dev = NULL;
 651         struct net_device *dev = NULL;
 652         struct sk_buff *skb;
 653         struct ipv6_devconf *devconf;
 654         int ifindex;
 655         int err;
 656
 657         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 658         if (err < 0)
 659                 return err;
 660
 661         if (!tb[NETCONFA_IFINDEX])
 662                 return -EINVAL;
 663
 664         err = -EINVAL;
 665         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 666         switch (ifindex) {
 667         case NETCONFA_IFINDEX_ALL:
 668                 devconf = net->ipv6.devconf_all;
 669                 break;
 670         case NETCONFA_IFINDEX_DEFAULT:
 671                 devconf = net->ipv6.devconf_dflt;
 672                 break;
 673         default:
 674                 dev = dev_get_by_index(net, ifindex);
 675                 if (!dev)
 676                         return -EINVAL;
 677                 in6_dev = in6_dev_get(dev);
 678                 if (!in6_dev)
 679                         goto errout;
 680                 devconf = &in6_dev->cnf;
 681                 break;
 682         }
 683
 684         err = -ENOBUFS;
 685         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 686         if (!skb)
 687                 goto errout;
 688
 689         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 690                                          NETLINK_CB(in_skb).portid,
 691                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 692                                          NETCONFA_ALL);
 693         if (err < 0) {
 694                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 695                 WARN_ON(err == -EMSGSIZE);
 696                 kfree_skb(skb);
 697                 goto errout;
 698         }
 699         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 700 errout:
 701         if (in6_dev)
 702                 in6_dev_put(in6_dev);
 703         dev_put(dev);
 704         return err;
 705 }
 706
 707 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 708                                       struct netlink_callback *cb)
 709 {
 710         const struct nlmsghdr *nlh = cb->nlh;
 711         struct net *net = sock_net(skb->sk);
 712         int h, s_h;
 713         int idx, s_idx;
 714         struct net_device *dev;
 715         struct inet6_dev *idev;
 716         struct hlist_head *head;
 717
 718         if (cb->strict_check) {
 719                 struct netlink_ext_ack *extack = cb->extack;
 720                 struct netconfmsg *ncm;
 721
 722                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 723                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 724                         return -EINVAL;
 725                 }
 726
 727                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 728                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 729                         return -EINVAL;
 730                 }
 731         }
 732
 733         s_h = cb->args[0];
 734         s_idx = idx = cb->args[1];
 735
 736         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 737                 idx = 0;
 738                 head = &net->dev_index_head[h];
 739                 rcu_read_lock();
 740                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 741                           net->dev_base_seq;
 742                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 743                         if (idx < s_idx)
 744                                 goto cont;
 745                         idev = __in6_dev_get(dev);
 746                         if (!idev)
 747                                 goto cont;
 748
 749                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 750                                                        &idev->cnf,
 751                                                        NETLINK_CB(cb->skb).portid,
 752                                                        nlh->nlmsg_seq,
 753                                                        RTM_NEWNETCONF,
 754                                                        NLM_F_MULTI,
 755                                                        NETCONFA_ALL) < 0) {
 756                                 rcu_read_unlock();
 757                                 goto done;
 758                         }
 759                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 760 cont:
 761                         idx++;
 762                 }
 763                 rcu_read_unlock();
 764         }
 765         if (h == NETDEV_HASHENTRIES) {
 766                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 767                                                net->ipv6.devconf_all,
 768                                                NETLINK_CB(cb->skb).portid,
 769                                                nlh->nlmsg_seq,
 770                                                RTM_NEWNETCONF, NLM_F_MULTI,
 771                                                NETCONFA_ALL) < 0)
 772                         goto done;
 773                 else
 774                         h++;
 775         }
 776         if (h == NETDEV_HASHENTRIES + 1) {
 777                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 778                                                net->ipv6.devconf_dflt,
 779                                                NETLINK_CB(cb->skb).portid,
 780                                                nlh->nlmsg_seq,
 781                                                RTM_NEWNETCONF, NLM_F_MULTI,
 782                                                NETCONFA_ALL) < 0)
 783                         goto done;
 784                 else
 785                         h++;
 786         }
 787 done:
 788         cb->args[0] = h;
 789         cb->args[1] = idx;
 790
 791         return skb->len;
 792 }
 793
 794 #ifdef CONFIG_SYSCTL
 795 static void dev_forward_change(struct inet6_dev *idev)
 796 {
 797         struct net_device *dev;
 798         struct inet6_ifaddr *ifa;
 799         LIST_HEAD(tmp_addr_list);
 800
 801         if (!idev)
 802                 return;
 803         dev = idev->dev;
 804         if (idev->cnf.forwarding)
 805                 dev_disable_lro(dev);
 806         if (dev->flags & IFF_MULTICAST) {
 807                 if (idev->cnf.forwarding) {
 808                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 809                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 810                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 811                 } else {
 812                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 813                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 814                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 815                 }
 816         }
 817
 818         read_lock_bh(&idev->lock);
 819         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 820                 if (ifa->flags&IFA_F_TENTATIVE)
 821                         continue;
 822                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
 823         }
 824         read_unlock_bh(&idev->lock);
 825
 826         while (!list_empty(&tmp_addr_list)) {
 827                 ifa = list_first_entry(&tmp_addr_list,
 828                                        struct inet6_ifaddr, if_list_aux);
 829                 list_del(&ifa->if_list_aux);
 830                 if (idev->cnf.forwarding)
 831                         addrconf_join_anycast(ifa);
 832                 else
 833                         addrconf_leave_anycast(ifa);
 834         }
 835
 836         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 837                                      NETCONFA_FORWARDING,
 838                                      dev->ifindex, &idev->cnf);
 839 }
 840
 841
 842 static void addrconf_forward_change(struct net *net, __s32 newf)
 843 {
 844         struct net_device *dev;
 845         struct inet6_dev *idev;
 846
 847         for_each_netdev(net, dev) {
 848                 idev = __in6_dev_get(dev);
 849                 if (idev) {
 850                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 851                         idev->cnf.forwarding = newf;
 852                         if (changed)
 853                                 dev_forward_change(idev);
 854                 }
 855         }
 856 }
 857
 858 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 859 {
 860         struct net *net;
 861         int old;
 862
 863         if (!rtnl_trylock())
 864                 return restart_syscall();
 865
 866         net = (struct net *)table->extra2;
 867         old = *p;
 868         *p = newf;
 869
 870         if (p == &net->ipv6.devconf_dflt->forwarding) {
 871                 if ((!newf) ^ (!old))
 872                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 873                                                      NETCONFA_FORWARDING,
 874                                                      NETCONFA_IFINDEX_DEFAULT,
 875                                                      net->ipv6.devconf_dflt);
 876                 rtnl_unlock();
 877                 return 0;
 878         }
 879
 880         if (p == &net->ipv6.devconf_all->forwarding) {
 881                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 882
 883                 net->ipv6.devconf_dflt->forwarding = newf;
 884                 if ((!newf) ^ (!old_dflt))
 885                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 886                                                      NETCONFA_FORWARDING,
 887                                                      NETCONFA_IFINDEX_DEFAULT,
 888                                                      net->ipv6.devconf_dflt);
 889
 890                 addrconf_forward_change(net, newf);
 891                 if ((!newf) ^ (!old))
 892                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 893                                                      NETCONFA_FORWARDING,
 894                                                      NETCONFA_IFINDEX_ALL,
 895                                                      net->ipv6.devconf_all);
 896         } else if ((!newf) ^ (!old))
 897                 dev_forward_change((struct inet6_dev *)table->extra1);
 898         rtnl_unlock();
 899
 900         if (newf)
 901                 rt6_purge_dflt_routers(net);
 902         return 1;
 903 }
 904
 905 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 906 {
 907         struct net_device *dev;
 908         struct inet6_dev *idev;
 909
 910         for_each_netdev(net, dev) {
 911                 idev = __in6_dev_get(dev);
 912                 if (idev) {
 913                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 914
 915                         idev->cnf.ignore_routes_with_linkdown = newf;
 916                         if (changed)
 917                                 inet6_netconf_notify_devconf(dev_net(dev),
 918                                                              RTM_NEWNETCONF,
 919                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 920                                                              dev->ifindex,
 921                                                              &idev->cnf);
 922                 }
 923         }
 924 }
 925
 926 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 927 {
 928         struct net *net;
 929         int old;
 930
 931         if (!rtnl_trylock())
 932                 return restart_syscall();
 933
 934         net = (struct net *)table->extra2;
 935         old = *p;
 936         *p = newf;
 937
 938         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 939                 if ((!newf) ^ (!old))
 940                         inet6_netconf_notify_devconf(net,
 941                                                      RTM_NEWNETCONF,
 942                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 943                                                      NETCONFA_IFINDEX_DEFAULT,
 944                                                      net->ipv6.devconf_dflt);
 945                 rtnl_unlock();
 946                 return 0;
 947         }
 948
 949         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 950                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 951                 addrconf_linkdown_change(net, newf);
 952                 if ((!newf) ^ (!old))
 953                         inet6_netconf_notify_devconf(net,
 954                                                      RTM_NEWNETCONF,
 955                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 956                                                      NETCONFA_IFINDEX_ALL,
 957                                                      net->ipv6.devconf_all);
 958         }
 959         rtnl_unlock();
 960
 961         return 1;
 962 }
 963
 964 #endif
 965
 966 /* Nobody refers to this ifaddr, destroy it */
 967 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 968 {
 969         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 970
 971 #ifdef NET_REFCNT_DEBUG
 972         pr_debug("%s\n", __func__);
 973 #endif
 974
 975         in6_dev_put(ifp->idev);
 976
 977         if (cancel_delayed_work(&ifp->dad_work))
 978                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 979                           ifp);
 980
 981         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 982                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 983                 return;
 984         }
 985
 986         kfree_rcu(ifp, rcu);
 987 }
 988
 989 static void
 990 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 991 {
 992         struct list_head *p;
 993         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 994
 995         /*
 996          * Each device address list is sorted in order of scope -
 997          * global before linklocal.
 998          */
 999         list_for_each(p, &idev->addr_list) {
1000                 struct inet6_ifaddr *ifa
1001                         = list_entry(p, struct inet6_ifaddr, if_list);
1002                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1003                         break;
1004         }
1005
1006         list_add_tail_rcu(&ifp->if_list, p);
1007 }
1008
1009 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1010 {
1011         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1012
1013         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1014 }
1015
1016 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1017                                struct net_device *dev, unsigned int hash)
1018 {
1019         struct inet6_ifaddr *ifp;
1020
1021         hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1022                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1023                         if (!dev || ifp->idev->dev == dev)
1024                                 return true;
1025                 }
1026         }
1027         return false;
1028 }
1029
1030 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1031 {
1032         struct net *net = dev_net(dev);
1033         unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1034         int err = 0;
1035
1036         spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1037
1038         /* Ignore adding duplicate addresses on an interface */
1039         if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1040                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1041                 err = -EEXIST;
1042         } else {
1043                 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1044         }
1045
1046         spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1047
1048         return err;
1049 }
1050
1051 /* On success it returns ifp with increased reference count */
1052
1053 static struct inet6_ifaddr *
1054 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1055               bool can_block, struct netlink_ext_ack *extack)
1056 {
1057         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1058         int addr_type = ipv6_addr_type(cfg->pfx);
1059         struct net *net = dev_net(idev->dev);
1060         struct inet6_ifaddr *ifa = NULL;
1061         struct fib6_info *f6i = NULL;
1062         int err = 0;
1063
1064         if (addr_type == IPV6_ADDR_ANY ||
1065             (addr_type & IPV6_ADDR_MULTICAST &&
1066              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1067             (!(idev->dev->flags & IFF_LOOPBACK) &&
1068              !netif_is_l3_master(idev->dev) &&
1069              addr_type & IPV6_ADDR_LOOPBACK))
1070                 return ERR_PTR(-EADDRNOTAVAIL);
1071
1072         if (idev->dead) {
1073                 err = -ENODEV;                  /*XXX*/
1074                 goto out;
1075         }
1076
1077         if (idev->cnf.disable_ipv6) {
1078                 err = -EACCES;
1079                 goto out;
1080         }
1081
1082         /* validator notifier needs to be blocking;
1083          * do not call in atomic context
1084          */
1085         if (can_block) {
1086                 struct in6_validator_info i6vi = {
1087                         .i6vi_addr = *cfg->pfx,
1088                         .i6vi_dev = idev,
1089                         .extack = extack,
1090                 };
1091
1092                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1093                 err = notifier_to_errno(err);
1094                 if (err < 0)
1095                         goto out;
1096         }
1097
1098         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1099         if (!ifa) {
1100                 err = -ENOBUFS;
1101                 goto out;
1102         }
1103
1104         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1105         if (IS_ERR(f6i)) {
1106                 err = PTR_ERR(f6i);
1107                 f6i = NULL;
1108                 goto out;
1109         }
1110
1111         neigh_parms_data_state_setall(idev->nd_parms);
1112
1113         ifa->addr = *cfg->pfx;
1114         if (cfg->peer_pfx)
1115                 ifa->peer_addr = *cfg->peer_pfx;
1116
1117         spin_lock_init(&ifa->lock);
1118         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1119         INIT_HLIST_NODE(&ifa->addr_lst);
1120         ifa->scope = cfg->scope;
1121         ifa->prefix_len = cfg->plen;
1122         ifa->rt_priority = cfg->rt_priority;
1123         ifa->flags = cfg->ifa_flags;
1124         ifa->ifa_proto = cfg->ifa_proto;
1125         /* No need to add the TENTATIVE flag for addresses with NODAD */
1126         if (!(cfg->ifa_flags & IFA_F_NODAD))
1127                 ifa->flags |= IFA_F_TENTATIVE;
1128         ifa->valid_lft = cfg->valid_lft;
1129         ifa->prefered_lft = cfg->preferred_lft;
1130         ifa->cstamp = ifa->tstamp = jiffies;
1131         ifa->tokenized = false;
1132
1133         ifa->rt = f6i;
1134
1135         ifa->idev = idev;
1136         in6_dev_hold(idev);
1137
1138         /* For caller */
1139         refcount_set(&ifa->refcnt, 1);
1140
1141         rcu_read_lock();
1142
1143         err = ipv6_add_addr_hash(idev->dev, ifa);
1144         if (err < 0) {
1145                 rcu_read_unlock();
1146                 goto out;
1147         }
1148
1149         write_lock_bh(&idev->lock);
1150
1151         /* Add to inet6_dev unicast addr list. */
1152         ipv6_link_dev_addr(idev, ifa);
1153
1154         if (ifa->flags&IFA_F_TEMPORARY) {
1155                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1156                 in6_ifa_hold(ifa);
1157         }
1158
1159         in6_ifa_hold(ifa);
1160         write_unlock_bh(&idev->lock);
1161
1162         rcu_read_unlock();
1163
1164         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1165 out:
1166         if (unlikely(err < 0)) {
1167                 fib6_info_release(f6i);
1168
1169                 if (ifa) {
1170                         if (ifa->idev)
1171                                 in6_dev_put(ifa->idev);
1172                         kfree(ifa);
1173                 }
1174                 ifa = ERR_PTR(err);
1175         }
1176
1177         return ifa;
1178 }
1179
1180 enum cleanup_prefix_rt_t {
1181         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1182         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1183         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1184 };
1185
1186 /*
1187  * Check, whether the prefix for ifp would still need a prefix route
1188  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1189  * constants.
1190  *
1191  * 1) we don't purge prefix if address was not permanent.
1192  *    prefix is managed by its own lifetime.
1193  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1194  * 3) if there are no addresses, delete prefix.
1195  * 4) if there are still other permanent address(es),
1196  *    corresponding prefix is still permanent.
1197  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1198  *    don't purge the prefix, assume user space is managing it.
1199  * 6) otherwise, update prefix lifetime to the
1200  *    longest valid lifetime among the corresponding
1201  *    addresses on the device.
1202  *    Note: subsequent RA will update lifetime.
1203  **/
1204 static enum cleanup_prefix_rt_t
1205 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1206 {
1207         struct inet6_ifaddr *ifa;
1208         struct inet6_dev *idev = ifp->idev;
1209         unsigned long lifetime;
1210         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1211
1212         *expires = jiffies;
1213
1214         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1215                 if (ifa == ifp)
1216                         continue;
1217                 if (ifa->prefix_len != ifp->prefix_len ||
1218                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1219                                        ifp->prefix_len))
1220                         continue;
1221                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1222                         return CLEANUP_PREFIX_RT_NOP;
1223
1224                 action = CLEANUP_PREFIX_RT_EXPIRE;
1225
1226                 spin_lock(&ifa->lock);
1227
1228                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1229                 /*
1230                  * Note: Because this address is
1231                  * not permanent, lifetime <
1232                  * LONG_MAX / HZ here.
1233                  */
1234                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1235                         *expires = ifa->tstamp + lifetime * HZ;
1236                 spin_unlock(&ifa->lock);
1237         }
1238
1239         return action;
1240 }
1241
1242 static void
1243 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1244                      bool del_rt, bool del_peer)
1245 {
1246         struct fib6_info *f6i;
1247
1248         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1249                                         ifp->prefix_len,
1250                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1251         if (f6i) {
1252                 if (del_rt)
1253                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1254                 else {
1255                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1256                                 fib6_set_expires(f6i, expires);
1257                         fib6_info_release(f6i);
1258                 }
1259         }
1260 }
1261
1262
1263 /* This function wants to get referenced ifp and releases it before return */
1264
1265 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1266 {
1267         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1268         struct net *net = dev_net(ifp->idev->dev);
1269         unsigned long expires;
1270         int state;
1271
1272         ASSERT_RTNL();
1273
1274         spin_lock_bh(&ifp->lock);
1275         state = ifp->state;
1276         ifp->state = INET6_IFADDR_STATE_DEAD;
1277         spin_unlock_bh(&ifp->lock);
1278
1279         if (state == INET6_IFADDR_STATE_DEAD)
1280                 goto out;
1281
1282         spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1283         hlist_del_init_rcu(&ifp->addr_lst);
1284         spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1285
1286         write_lock_bh(&ifp->idev->lock);
1287
1288         if (ifp->flags&IFA_F_TEMPORARY) {
1289                 list_del(&ifp->tmp_list);
1290                 if (ifp->ifpub) {
1291                         in6_ifa_put(ifp->ifpub);
1292                         ifp->ifpub = NULL;
1293                 }
1294                 __in6_ifa_put(ifp);
1295         }
1296
1297         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1298                 action = check_cleanup_prefix_route(ifp, &expires);
1299
1300         list_del_rcu(&ifp->if_list);
1301         __in6_ifa_put(ifp);
1302
1303         write_unlock_bh(&ifp->idev->lock);
1304
1305         addrconf_del_dad_work(ifp);
1306
1307         ipv6_ifa_notify(RTM_DELADDR, ifp);
1308
1309         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1310
1311         if (action != CLEANUP_PREFIX_RT_NOP) {
1312                 cleanup_prefix_route(ifp, expires,
1313                         action == CLEANUP_PREFIX_RT_DEL, false);
1314         }
1315
1316         /* clean up prefsrc entries */
1317         rt6_remove_prefsrc(ifp);
1318 out:
1319         in6_ifa_put(ifp);
1320 }
1321
1322 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1323 {
1324         struct inet6_dev *idev = ifp->idev;
1325         unsigned long tmp_tstamp, age;
1326         unsigned long regen_advance;
1327         unsigned long now = jiffies;
1328         s32 cnf_temp_preferred_lft;
1329         struct inet6_ifaddr *ift;
1330         struct ifa6_config cfg;
1331         long max_desync_factor;
1332         struct in6_addr addr;
1333         int ret = 0;
1334
1335         write_lock_bh(&idev->lock);
1336
1337 retry:
1338         in6_dev_hold(idev);
1339         if (idev->cnf.use_tempaddr <= 0) {
1340                 write_unlock_bh(&idev->lock);
1341                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1342                 in6_dev_put(idev);
1343                 ret = -1;
1344                 goto out;
1345         }
1346         spin_lock_bh(&ifp->lock);
1347         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1348                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1349                 spin_unlock_bh(&ifp->lock);
1350                 write_unlock_bh(&idev->lock);
1351                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1352                         __func__);
1353                 in6_dev_put(idev);
1354                 ret = -1;
1355                 goto out;
1356         }
1357         in6_ifa_hold(ifp);
1358         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1359         ipv6_gen_rnd_iid(&addr);
1360
1361         age = (now - ifp->tstamp) / HZ;
1362
1363         regen_advance = idev->cnf.regen_max_retry *
1364                         idev->cnf.dad_transmits *
1365                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1366
1367         /* recalculate max_desync_factor each time and update
1368          * idev->desync_factor if it's larger
1369          */
1370         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1371         max_desync_factor = min_t(__u32,
1372                                   idev->cnf.max_desync_factor,
1373                                   cnf_temp_preferred_lft - regen_advance);
1374
1375         if (unlikely(idev->desync_factor > max_desync_factor)) {
1376                 if (max_desync_factor > 0) {
1377                         get_random_bytes(&idev->desync_factor,
1378                                          sizeof(idev->desync_factor));
1379                         idev->desync_factor %= max_desync_factor;
1380                 } else {
1381                         idev->desync_factor = 0;
1382                 }
1383         }
1384
1385         memset(&cfg, 0, sizeof(cfg));
1386         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1387                               idev->cnf.temp_valid_lft + age);
1388         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1389         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1390
1391         cfg.plen = ifp->prefix_len;
1392         tmp_tstamp = ifp->tstamp;
1393         spin_unlock_bh(&ifp->lock);
1394
1395         write_unlock_bh(&idev->lock);
1396
1397         /* A temporary address is created only if this calculated Preferred
1398          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1399          * an implementation must not create a temporary address with a zero
1400          * Preferred Lifetime.
1401          * Use age calculation as in addrconf_verify to avoid unnecessary
1402          * temporary addresses being generated.
1403          */
1404         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1405         if (cfg.preferred_lft <= regen_advance + age) {
1406                 in6_ifa_put(ifp);
1407                 in6_dev_put(idev);
1408                 ret = -1;
1409                 goto out;
1410         }
1411
1412         cfg.ifa_flags = IFA_F_TEMPORARY;
1413         /* set in addrconf_prefix_rcv() */
1414         if (ifp->flags & IFA_F_OPTIMISTIC)
1415                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1416
1417         cfg.pfx = &addr;
1418         cfg.scope = ipv6_addr_scope(cfg.pfx);
1419
1420         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1421         if (IS_ERR(ift)) {
1422                 in6_ifa_put(ifp);
1423                 in6_dev_put(idev);
1424                 pr_info("%s: retry temporary address regeneration\n", __func__);
1425                 write_lock_bh(&idev->lock);
1426                 goto retry;
1427         }
1428
1429         spin_lock_bh(&ift->lock);
1430         ift->ifpub = ifp;
1431         ift->cstamp = now;
1432         ift->tstamp = tmp_tstamp;
1433         spin_unlock_bh(&ift->lock);
1434
1435         addrconf_dad_start(ift);
1436         in6_ifa_put(ift);
1437         in6_dev_put(idev);
1438 out:
1439         return ret;
1440 }
1441
1442 /*
1443  *      Choose an appropriate source address (RFC3484)
1444  */
1445 enum {
1446         IPV6_SADDR_RULE_INIT = 0,
1447         IPV6_SADDR_RULE_LOCAL,
1448         IPV6_SADDR_RULE_SCOPE,
1449         IPV6_SADDR_RULE_PREFERRED,
1450 #ifdef CONFIG_IPV6_MIP6
1451         IPV6_SADDR_RULE_HOA,
1452 #endif
1453         IPV6_SADDR_RULE_OIF,
1454         IPV6_SADDR_RULE_LABEL,
1455         IPV6_SADDR_RULE_PRIVACY,
1456         IPV6_SADDR_RULE_ORCHID,
1457         IPV6_SADDR_RULE_PREFIX,
1458 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1459         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1460 #endif
1461         IPV6_SADDR_RULE_MAX
1462 };
1463
1464 struct ipv6_saddr_score {
1465         int                     rule;
1466         int                     addr_type;
1467         struct inet6_ifaddr     *ifa;
1468         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1469         int                     scopedist;
1470         int                     matchlen;
1471 };
1472
1473 struct ipv6_saddr_dst {
1474         const struct in6_addr *addr;
1475         int ifindex;
1476         int scope;
1477         int label;
1478         unsigned int prefs;
1479 };
1480
1481 static inline int ipv6_saddr_preferred(int type)
1482 {
1483         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1484                 return 1;
1485         return 0;
1486 }
1487
1488 static bool ipv6_use_optimistic_addr(struct net *net,
1489                                      struct inet6_dev *idev)
1490 {
1491 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1492         if (!idev)
1493                 return false;
1494         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1495                 return false;
1496         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1497                 return false;
1498
1499         return true;
1500 #else
1501         return false;
1502 #endif
1503 }
1504
1505 static bool ipv6_allow_optimistic_dad(struct net *net,
1506                                       struct inet6_dev *idev)
1507 {
1508 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1509         if (!idev)
1510                 return false;
1511         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1512                 return false;
1513
1514         return true;
1515 #else
1516         return false;
1517 #endif
1518 }
1519
1520 static int ipv6_get_saddr_eval(struct net *net,
1521                                struct ipv6_saddr_score *score,
1522                                struct ipv6_saddr_dst *dst,
1523                                int i)
1524 {
1525         int ret;
1526
1527         if (i <= score->rule) {
1528                 switch (i) {
1529                 case IPV6_SADDR_RULE_SCOPE:
1530                         ret = score->scopedist;
1531                         break;
1532                 case IPV6_SADDR_RULE_PREFIX:
1533                         ret = score->matchlen;
1534                         break;
1535                 default:
1536                         ret = !!test_bit(i, score->scorebits);
1537                 }
1538                 goto out;
1539         }
1540
1541         switch (i) {
1542         case IPV6_SADDR_RULE_INIT:
1543                 /* Rule 0: remember if hiscore is not ready yet */
1544                 ret = !!score->ifa;
1545                 break;
1546         case IPV6_SADDR_RULE_LOCAL:
1547                 /* Rule 1: Prefer same address */
1548                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1549                 break;
1550         case IPV6_SADDR_RULE_SCOPE:
1551                 /* Rule 2: Prefer appropriate scope
1552                  *
1553                  *      ret
1554                  *       ^
1555                  *    -1 |  d 15
1556                  *    ---+--+-+---> scope
1557                  *       |
1558                  *       |             d is scope of the destination.
1559                  *  B-d  |  \
1560                  *       |   \      <- smaller scope is better if
1561                  *  B-15 |    \        if scope is enough for destination.
1562                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1563                  * d-C-1 | /
1564                  *       |/         <- greater is better
1565                  *   -C  /             if scope is not enough for destination.
1566                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1567                  *
1568                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1569                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1570                  * Assume B = 0 and we get C > 29.
1571                  */
1572                 ret = __ipv6_addr_src_scope(score->addr_type);
1573                 if (ret >= dst->scope)
1574                         ret = -ret;
1575                 else
1576                         ret -= 128;     /* 30 is enough */
1577                 score->scopedist = ret;
1578                 break;
1579         case IPV6_SADDR_RULE_PREFERRED:
1580             {
1581                 /* Rule 3: Avoid deprecated and optimistic addresses */
1582                 u8 avoid = IFA_F_DEPRECATED;
1583
1584                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1585                         avoid |= IFA_F_OPTIMISTIC;
1586                 ret = ipv6_saddr_preferred(score->addr_type) ||
1587                       !(score->ifa->flags & avoid);
1588                 break;
1589             }
1590 #ifdef CONFIG_IPV6_MIP6
1591         case IPV6_SADDR_RULE_HOA:
1592             {
1593                 /* Rule 4: Prefer home address */
1594                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1595                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1596                 break;
1597             }
1598 #endif
1599         case IPV6_SADDR_RULE_OIF:
1600                 /* Rule 5: Prefer outgoing interface */
1601                 ret = (!dst->ifindex ||
1602                        dst->ifindex == score->ifa->idev->dev->ifindex);
1603                 break;
1604         case IPV6_SADDR_RULE_LABEL:
1605                 /* Rule 6: Prefer matching label */
1606                 ret = ipv6_addr_label(net,
1607                                       &score->ifa->addr, score->addr_type,
1608                                       score->ifa->idev->dev->ifindex) == dst->label;
1609                 break;
1610         case IPV6_SADDR_RULE_PRIVACY:
1611             {
1612                 /* Rule 7: Prefer public address
1613                  * Note: prefer temporary address if use_tempaddr >= 2
1614                  */
1615                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1616                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1617                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1618                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1619                 break;
1620             }
1621         case IPV6_SADDR_RULE_ORCHID:
1622                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1623                  *          non-ORCHID vs non-ORCHID
1624                  */
1625                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1626                         ipv6_addr_orchid(dst->addr));
1627                 break;
1628         case IPV6_SADDR_RULE_PREFIX:
1629                 /* Rule 8: Use longest matching prefix */
1630                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1631                 if (ret > score->ifa->prefix_len)
1632                         ret = score->ifa->prefix_len;
1633                 score->matchlen = ret;
1634                 break;
1635 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1636         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1637                 /* Optimistic addresses still have lower precedence than other
1638                  * preferred addresses.
1639                  */
1640                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1641                 break;
1642 #endif
1643         default:
1644                 ret = 0;
1645         }
1646
1647         if (ret)
1648                 __set_bit(i, score->scorebits);
1649         score->rule = i;
1650 out:
1651         return ret;
1652 }
1653
1654 static int __ipv6_dev_get_saddr(struct net *net,
1655                                 struct ipv6_saddr_dst *dst,
1656                                 struct inet6_dev *idev,
1657                                 struct ipv6_saddr_score *scores,
1658                                 int hiscore_idx)
1659 {
1660         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1661
1662         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1663                 int i;
1664
1665                 /*
1666                  * - Tentative Address (RFC2462 section 5.4)
1667                  *  - A tentative address is not considered
1668                  *    "assigned to an interface" in the traditional
1669                  *    sense, unless it is also flagged as optimistic.
1670                  * - Candidate Source Address (section 4)
1671                  *  - In any case, anycast addresses, multicast
1672                  *    addresses, and the unspecified address MUST
1673                  *    NOT be included in a candidate set.
1674                  */
1675                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1676                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1677                         continue;
1678
1679                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1680
1681                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1682                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1683                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1684                                             idev->dev->name);
1685                         continue;
1686                 }
1687
1688                 score->rule = -1;
1689                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1690
1691                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1692                         int minihiscore, miniscore;
1693
1694                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1695                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1696
1697                         if (minihiscore > miniscore) {
1698                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1699                                     score->scopedist > 0) {
1700                                         /*
1701                                          * special case:
1702                                          * each remaining entry
1703                                          * has too small (not enough)
1704                                          * scope, because ifa entries
1705                                          * are sorted by their scope
1706                                          * values.
1707                                          */
1708                                         goto out;
1709                                 }
1710                                 break;
1711                         } else if (minihiscore < miniscore) {
1712                                 swap(hiscore, score);
1713                                 hiscore_idx = 1 - hiscore_idx;
1714
1715                                 /* restore our iterator */
1716                                 score->ifa = hiscore->ifa;
1717
1718                                 break;
1719                         }
1720                 }
1721         }
1722 out:
1723         return hiscore_idx;
1724 }
1725
1726 static int ipv6_get_saddr_master(struct net *net,
1727                                  const struct net_device *dst_dev,
1728                                  const struct net_device *master,
1729                                  struct ipv6_saddr_dst *dst,
1730                                  struct ipv6_saddr_score *scores,
1731                                  int hiscore_idx)
1732 {
1733         struct inet6_dev *idev;
1734
1735         idev = __in6_dev_get(dst_dev);
1736         if (idev)
1737                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1738                                                    scores, hiscore_idx);
1739
1740         idev = __in6_dev_get(master);
1741         if (idev)
1742                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1743                                                    scores, hiscore_idx);
1744
1745         return hiscore_idx;
1746 }
1747
1748 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1749                        const struct in6_addr *daddr, unsigned int prefs,
1750                        struct in6_addr *saddr)
1751 {
1752         struct ipv6_saddr_score scores[2], *hiscore;
1753         struct ipv6_saddr_dst dst;
1754         struct inet6_dev *idev;
1755         struct net_device *dev;
1756         int dst_type;
1757         bool use_oif_addr = false;
1758         int hiscore_idx = 0;
1759         int ret = 0;
1760
1761         dst_type = __ipv6_addr_type(daddr);
1762         dst.addr = daddr;
1763         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1764         dst.scope = __ipv6_addr_src_scope(dst_type);
1765         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1766         dst.prefs = prefs;
1767
1768         scores[hiscore_idx].rule = -1;
1769         scores[hiscore_idx].ifa = NULL;
1770
1771         rcu_read_lock();
1772
1773         /* Candidate Source Address (section 4)
1774          *  - multicast and link-local destination address,
1775          *    the set of candidate source address MUST only
1776          *    include addresses assigned to interfaces
1777          *    belonging to the same link as the outgoing
1778          *    interface.
1779          * (- For site-local destination addresses, the
1780          *    set of candidate source addresses MUST only
1781          *    include addresses assigned to interfaces
1782          *    belonging to the same site as the outgoing
1783          *    interface.)
1784          *  - "It is RECOMMENDED that the candidate source addresses
1785          *    be the set of unicast addresses assigned to the
1786          *    interface that will be used to send to the destination
1787          *    (the 'outgoing' interface)." (RFC 6724)
1788          */
1789         if (dst_dev) {
1790                 idev = __in6_dev_get(dst_dev);
1791                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1792                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1793                     (idev && idev->cnf.use_oif_addrs_only)) {
1794                         use_oif_addr = true;
1795                 }
1796         }
1797
1798         if (use_oif_addr) {
1799                 if (idev)
1800                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1801         } else {
1802                 const struct net_device *master;
1803                 int master_idx = 0;
1804
1805                 /* if dst_dev exists and is enslaved to an L3 device, then
1806                  * prefer addresses from dst_dev and then the master over
1807                  * any other enslaved devices in the L3 domain.
1808                  */
1809                 master = l3mdev_master_dev_rcu(dst_dev);
1810                 if (master) {
1811                         master_idx = master->ifindex;
1812
1813                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1814                                                             master, &dst,
1815                                                             scores, hiscore_idx);
1816
1817                         if (scores[hiscore_idx].ifa)
1818                                 goto out;
1819                 }
1820
1821                 for_each_netdev_rcu(net, dev) {
1822                         /* only consider addresses on devices in the
1823                          * same L3 domain
1824                          */
1825                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1826                                 continue;
1827                         idev = __in6_dev_get(dev);
1828                         if (!idev)
1829                                 continue;
1830                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1831                 }
1832         }
1833
1834 out:
1835         hiscore = &scores[hiscore_idx];
1836         if (!hiscore->ifa)
1837                 ret = -EADDRNOTAVAIL;
1838         else
1839                 *saddr = hiscore->ifa->addr;
1840
1841         rcu_read_unlock();
1842         return ret;
1843 }
1844 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1845
1846 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1847                               u32 banned_flags)
1848 {
1849         struct inet6_ifaddr *ifp;
1850         int err = -EADDRNOTAVAIL;
1851
1852         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1853                 if (ifp->scope > IFA_LINK)
1854                         break;
1855                 if (ifp->scope == IFA_LINK &&
1856                     !(ifp->flags & banned_flags)) {
1857                         *addr = ifp->addr;
1858                         err = 0;
1859                         break;
1860                 }
1861         }
1862         return err;
1863 }
1864
1865 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1866                     u32 banned_flags)
1867 {
1868         struct inet6_dev *idev;
1869         int err = -EADDRNOTAVAIL;
1870
1871         rcu_read_lock();
1872         idev = __in6_dev_get(dev);
1873         if (idev) {
1874                 read_lock_bh(&idev->lock);
1875                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1876                 read_unlock_bh(&idev->lock);
1877         }
1878         rcu_read_unlock();
1879         return err;
1880 }
1881
1882 static int ipv6_count_addresses(const struct inet6_dev *idev)
1883 {
1884         const struct inet6_ifaddr *ifp;
1885         int cnt = 0;
1886
1887         rcu_read_lock();
1888         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1889                 cnt++;
1890         rcu_read_unlock();
1891         return cnt;
1892 }
1893
1894 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1895                   const struct net_device *dev, int strict)
1896 {
1897         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1898                                        strict, IFA_F_TENTATIVE);
1899 }
1900 EXPORT_SYMBOL(ipv6_chk_addr);
1901
1902 /* device argument is used to find the L3 domain of interest. If
1903  * skip_dev_check is set, then the ifp device is not checked against
1904  * the passed in dev argument. So the 2 cases for addresses checks are:
1905  *   1. does the address exist in the L3 domain that dev is part of
1906  *      (skip_dev_check = true), or
1907  *
1908  *   2. does the address exist on the specific device
1909  *      (skip_dev_check = false)
1910  */
1911 static struct net_device *
1912 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1913                           const struct net_device *dev, bool skip_dev_check,
1914                           int strict, u32 banned_flags)
1915 {
1916         unsigned int hash = inet6_addr_hash(net, addr);
1917         struct net_device *l3mdev, *ndev;
1918         struct inet6_ifaddr *ifp;
1919         u32 ifp_flags;
1920
1921         rcu_read_lock();
1922
1923         l3mdev = l3mdev_master_dev_rcu(dev);
1924         if (skip_dev_check)
1925                 dev = NULL;
1926
1927         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1928                 ndev = ifp->idev->dev;
1929
1930                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1931                         continue;
1932
1933                 /* Decouple optimistic from tentative for evaluation here.
1934                  * Ban optimistic addresses explicitly, when required.
1935                  */
1936                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1937                             ? (ifp->flags&~IFA_F_TENTATIVE)
1938                             : ifp->flags;
1939                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1940                     !(ifp_flags&banned_flags) &&
1941                     (!dev || ndev == dev ||
1942                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1943                         rcu_read_unlock();
1944                         return ndev;
1945                 }
1946         }
1947
1948         rcu_read_unlock();
1949         return NULL;
1950 }
1951
1952 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1953                             const struct net_device *dev, bool skip_dev_check,
1954                             int strict, u32 banned_flags)
1955 {
1956         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1957                                          strict, banned_flags) ? 1 : 0;
1958 }
1959 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1960
1961
1962 /* Compares an address/prefix_len with addresses on device @dev.
1963  * If one is found it returns true.
1964  */
1965 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1966         const unsigned int prefix_len, struct net_device *dev)
1967 {
1968         const struct inet6_ifaddr *ifa;
1969         const struct inet6_dev *idev;
1970         bool ret = false;
1971
1972         rcu_read_lock();
1973         idev = __in6_dev_get(dev);
1974         if (idev) {
1975                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1976                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1977                         if (ret)
1978                                 break;
1979                 }
1980         }
1981         rcu_read_unlock();
1982
1983         return ret;
1984 }
1985 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1986
1987 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1988 {
1989         const struct inet6_ifaddr *ifa;
1990         const struct inet6_dev *idev;
1991         int     onlink;
1992
1993         onlink = 0;
1994         rcu_read_lock();
1995         idev = __in6_dev_get(dev);
1996         if (idev) {
1997                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1998                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1999                                                    ifa->prefix_len);
2000                         if (onlink)
2001                                 break;
2002                 }
2003         }
2004         rcu_read_unlock();
2005         return onlink;
2006 }
2007 EXPORT_SYMBOL(ipv6_chk_prefix);
2008
2009 /**
2010  * ipv6_dev_find - find the first device with a given source address.
2011  * @net: the net namespace
2012  * @addr: the source address
2013  * @dev: used to find the L3 domain of interest
2014  *
2015  * The caller should be protected by RCU, or RTNL.
2016  */
2017 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2018                                  struct net_device *dev)
2019 {
2020         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2021                                          IFA_F_TENTATIVE);
2022 }
2023 EXPORT_SYMBOL(ipv6_dev_find);
2024
2025 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2026                                      struct net_device *dev, int strict)
2027 {
2028         unsigned int hash = inet6_addr_hash(net, addr);
2029         struct inet6_ifaddr *ifp, *result = NULL;
2030
2031         rcu_read_lock();
2032         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2033                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2034                         if (!dev || ifp->idev->dev == dev ||
2035                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2036                                 result = ifp;
2037                                 in6_ifa_hold(ifp);
2038                                 break;
2039                         }
2040                 }
2041         }
2042         rcu_read_unlock();
2043
2044         return result;
2045 }
2046
2047 /* Gets referenced address, destroys ifaddr */
2048
2049 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2050 {
2051         if (dad_failed)
2052                 ifp->flags |= IFA_F_DADFAILED;
2053
2054         if (ifp->flags&IFA_F_TEMPORARY) {
2055                 struct inet6_ifaddr *ifpub;
2056                 spin_lock_bh(&ifp->lock);
2057                 ifpub = ifp->ifpub;
2058                 if (ifpub) {
2059                         in6_ifa_hold(ifpub);
2060                         spin_unlock_bh(&ifp->lock);
2061                         ipv6_create_tempaddr(ifpub, true);
2062                         in6_ifa_put(ifpub);
2063                 } else {
2064                         spin_unlock_bh(&ifp->lock);
2065                 }
2066                 ipv6_del_addr(ifp);
2067         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2068                 spin_lock_bh(&ifp->lock);
2069                 addrconf_del_dad_work(ifp);
2070                 ifp->flags |= IFA_F_TENTATIVE;
2071                 if (dad_failed)
2072                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2073                 spin_unlock_bh(&ifp->lock);
2074                 if (dad_failed)
2075                         ipv6_ifa_notify(0, ifp);
2076                 in6_ifa_put(ifp);
2077         } else {
2078                 ipv6_del_addr(ifp);
2079         }
2080 }
2081
2082 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2083 {
2084         int err = -ENOENT;
2085
2086         spin_lock_bh(&ifp->lock);
2087         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2088                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2089                 err = 0;
2090         }
2091         spin_unlock_bh(&ifp->lock);
2092
2093         return err;
2094 }
2095
2096 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2097 {
2098         struct inet6_dev *idev = ifp->idev;
2099         struct net *net = dev_net(idev->dev);
2100
2101         if (addrconf_dad_end(ifp)) {
2102                 in6_ifa_put(ifp);
2103                 return;
2104         }
2105
2106         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2107                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2108
2109         spin_lock_bh(&ifp->lock);
2110
2111         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2112                 struct in6_addr new_addr;
2113                 struct inet6_ifaddr *ifp2;
2114                 int retries = ifp->stable_privacy_retry + 1;
2115                 struct ifa6_config cfg = {
2116                         .pfx = &new_addr,
2117                         .plen = ifp->prefix_len,
2118                         .ifa_flags = ifp->flags,
2119                         .valid_lft = ifp->valid_lft,
2120                         .preferred_lft = ifp->prefered_lft,
2121                         .scope = ifp->scope,
2122                 };
2123
2124                 if (retries > net->ipv6.sysctl.idgen_retries) {
2125                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2126                                              ifp->idev->dev->name);
2127                         goto errdad;
2128                 }
2129
2130                 new_addr = ifp->addr;
2131                 if (ipv6_generate_stable_address(&new_addr, retries,
2132                                                  idev))
2133                         goto errdad;
2134
2135                 spin_unlock_bh(&ifp->lock);
2136
2137                 if (idev->cnf.max_addresses &&
2138                     ipv6_count_addresses(idev) >=
2139                     idev->cnf.max_addresses)
2140                         goto lock_errdad;
2141
2142                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2143                                      ifp->idev->dev->name);
2144
2145                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2146                 if (IS_ERR(ifp2))
2147                         goto lock_errdad;
2148
2149                 spin_lock_bh(&ifp2->lock);
2150                 ifp2->stable_privacy_retry = retries;
2151                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2152                 spin_unlock_bh(&ifp2->lock);
2153
2154                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2155                 in6_ifa_put(ifp2);
2156 lock_errdad:
2157                 spin_lock_bh(&ifp->lock);
2158         }
2159
2160 errdad:
2161         /* transition from _POSTDAD to _ERRDAD */
2162         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2163         spin_unlock_bh(&ifp->lock);
2164
2165         addrconf_mod_dad_work(ifp, 0);
2166         in6_ifa_put(ifp);
2167 }
2168
2169 /* Join to solicited addr multicast group.
2170  * caller must hold RTNL */
2171 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2172 {
2173         struct in6_addr maddr;
2174
2175         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2176                 return;
2177
2178         addrconf_addr_solict_mult(addr, &maddr);
2179         ipv6_dev_mc_inc(dev, &maddr);
2180 }
2181
2182 /* caller must hold RTNL */
2183 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2184 {
2185         struct in6_addr maddr;
2186
2187         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2188                 return;
2189
2190         addrconf_addr_solict_mult(addr, &maddr);
2191         __ipv6_dev_mc_dec(idev, &maddr);
2192 }
2193
2194 /* caller must hold RTNL */
2195 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2196 {
2197         struct in6_addr addr;
2198
2199         if (ifp->prefix_len >= 127) /* RFC 6164 */
2200                 return;
2201         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2202         if (ipv6_addr_any(&addr))
2203                 return;
2204         __ipv6_dev_ac_inc(ifp->idev, &addr);
2205 }
2206
2207 /* caller must hold RTNL */
2208 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2209 {
2210         struct in6_addr addr;
2211
2212         if (ifp->prefix_len >= 127) /* RFC 6164 */
2213                 return;
2214         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2215         if (ipv6_addr_any(&addr))
2216                 return;
2217         __ipv6_dev_ac_dec(ifp->idev, &addr);
2218 }
2219
2220 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2221 {
2222         switch (dev->addr_len) {
2223         case ETH_ALEN:
2224                 memcpy(eui, dev->dev_addr, 3);
2225                 eui[3] = 0xFF;
2226                 eui[4] = 0xFE;
2227                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2228                 break;
2229         case EUI64_ADDR_LEN:
2230                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2231                 eui[0] ^= 2;
2232                 break;
2233         default:
2234                 return -1;
2235         }
2236
2237         return 0;
2238 }
2239
2240 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2241 {
2242         const union fwnet_hwaddr *ha;
2243
2244         if (dev->addr_len != FWNET_ALEN)
2245                 return -1;
2246
2247         ha = (const union fwnet_hwaddr *)dev->dev_addr;
2248
2249         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2250         eui[0] ^= 2;
2251         return 0;
2252 }
2253
2254 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2255 {
2256         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2257         if (dev->addr_len != ARCNET_ALEN)
2258                 return -1;
2259         memset(eui, 0, 7);
2260         eui[7] = *(u8 *)dev->dev_addr;
2261         return 0;
2262 }
2263
2264 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2265 {
2266         if (dev->addr_len != INFINIBAND_ALEN)
2267                 return -1;
2268         memcpy(eui, dev->dev_addr + 12, 8);
2269         eui[0] |= 2;
2270         return 0;
2271 }
2272
2273 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2274 {
2275         if (addr == 0)
2276                 return -1;
2277         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2278                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2279                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2280                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2281                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2282                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2283         eui[1] = 0;
2284         eui[2] = 0x5E;
2285         eui[3] = 0xFE;
2286         memcpy(eui + 4, &addr, 4);
2287         return 0;
2288 }
2289
2290 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2291 {
2292         if (dev->priv_flags & IFF_ISATAP)
2293                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2294         return -1;
2295 }
2296
2297 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2298 {
2299         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2300 }
2301
2302 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2303 {
2304         memcpy(eui, dev->perm_addr, 3);
2305         memcpy(eui + 5, dev->perm_addr + 3, 3);
2306         eui[3] = 0xFF;
2307         eui[4] = 0xFE;
2308         eui[0] ^= 2;
2309         return 0;
2310 }
2311
2312 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2313 {
2314         switch (dev->type) {
2315         case ARPHRD_ETHER:
2316         case ARPHRD_FDDI:
2317                 return addrconf_ifid_eui48(eui, dev);
2318         case ARPHRD_ARCNET:
2319                 return addrconf_ifid_arcnet(eui, dev);
2320         case ARPHRD_INFINIBAND:
2321                 return addrconf_ifid_infiniband(eui, dev);
2322         case ARPHRD_SIT:
2323                 return addrconf_ifid_sit(eui, dev);
2324         case ARPHRD_IPGRE:
2325         case ARPHRD_TUNNEL:
2326                 return addrconf_ifid_gre(eui, dev);
2327         case ARPHRD_6LOWPAN:
2328                 return addrconf_ifid_6lowpan(eui, dev);
2329         case ARPHRD_IEEE1394:
2330                 return addrconf_ifid_ieee1394(eui, dev);
2331         case ARPHRD_TUNNEL6:
2332         case ARPHRD_IP6GRE:
2333         case ARPHRD_RAWIP:
2334                 return addrconf_ifid_ip6tnl(eui, dev);
2335         }
2336         return -1;
2337 }
2338
2339 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2340 {
2341         int err = -1;
2342         struct inet6_ifaddr *ifp;
2343
2344         read_lock_bh(&idev->lock);
2345         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2346                 if (ifp->scope > IFA_LINK)
2347                         break;
2348                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2349                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2350                         err = 0;
2351                         break;
2352                 }
2353         }
2354         read_unlock_bh(&idev->lock);
2355         return err;
2356 }
2357
2358 /* Generation of a randomized Interface Identifier
2359  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2360  */
2361
2362 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2363 {
2364 regen:
2365         get_random_bytes(&addr->s6_addr[8], 8);
2366
2367         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2368          * check if generated address is not inappropriate:
2369          *
2370          * - Reserved IPv6 Interface Identifiers
2371          * - XXX: already assigned to an address on the device
2372          */
2373
2374         /* Subnet-router anycast: 0000:0000:0000:0000 */
2375         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2376                 goto regen;
2377
2378         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2379          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2380          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2381          */
2382         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2383             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2384                 goto regen;
2385
2386         /* Reserved subnet anycast addresses */
2387         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2388             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2389                 goto regen;
2390 }
2391
2392 /*
2393  *      Add prefix route.
2394  */
2395
2396 static void
2397 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2398                       struct net_device *dev, unsigned long expires,
2399                       u32 flags, gfp_t gfp_flags)
2400 {
2401         struct fib6_config cfg = {
2402                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2403                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2404                 .fc_ifindex = dev->ifindex,
2405                 .fc_expires = expires,
2406                 .fc_dst_len = plen,
2407                 .fc_flags = RTF_UP | flags,
2408                 .fc_nlinfo.nl_net = dev_net(dev),
2409                 .fc_protocol = RTPROT_KERNEL,
2410                 .fc_type = RTN_UNICAST,
2411         };
2412
2413         cfg.fc_dst = *pfx;
2414
2415         /* Prevent useless cloning on PtP SIT.
2416            This thing is done here expecting that the whole
2417            class of non-broadcast devices need not cloning.
2418          */
2419 #if IS_ENABLED(CONFIG_IPV6_SIT)
2420         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2421                 cfg.fc_flags |= RTF_NONEXTHOP;
2422 #endif
2423
2424         ip6_route_add(&cfg, gfp_flags, NULL);
2425 }
2426
2427
2428 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2429                                                   int plen,
2430                                                   const struct net_device *dev,
2431                                                   u32 flags, u32 noflags,
2432                                                   bool no_gw)
2433 {
2434         struct fib6_node *fn;
2435         struct fib6_info *rt = NULL;
2436         struct fib6_table *table;
2437         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2438
2439         table = fib6_get_table(dev_net(dev), tb_id);
2440         if (!table)
2441                 return NULL;
2442
2443         rcu_read_lock();
2444         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2445         if (!fn)
2446                 goto out;
2447
2448         for_each_fib6_node_rt_rcu(fn) {
2449                 /* prefix routes only use builtin fib6_nh */
2450                 if (rt->nh)
2451                         continue;
2452
2453                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2454                         continue;
2455                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2456                         continue;
2457                 if ((rt->fib6_flags & flags) != flags)
2458                         continue;
2459                 if ((rt->fib6_flags & noflags) != 0)
2460                         continue;
2461                 if (!fib6_info_hold_safe(rt))
2462                         continue;
2463                 break;
2464         }
2465 out:
2466         rcu_read_unlock();
2467         return rt;
2468 }
2469
2470
2471 /* Create "default" multicast route to the interface */
2472
2473 static void addrconf_add_mroute(struct net_device *dev)
2474 {
2475         struct fib6_config cfg = {
2476                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2477                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2478                 .fc_ifindex = dev->ifindex,
2479                 .fc_dst_len = 8,
2480                 .fc_flags = RTF_UP,
2481                 .fc_type = RTN_MULTICAST,
2482                 .fc_nlinfo.nl_net = dev_net(dev),
2483                 .fc_protocol = RTPROT_KERNEL,
2484         };
2485
2486         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2487
2488         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2489 }
2490
2491 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2492 {
2493         struct inet6_dev *idev;
2494
2495         ASSERT_RTNL();
2496
2497         idev = ipv6_find_idev(dev);
2498         if (IS_ERR(idev))
2499                 return idev;
2500
2501         if (idev->cnf.disable_ipv6)
2502                 return ERR_PTR(-EACCES);
2503
2504         /* Add default multicast route */
2505         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2506                 addrconf_add_mroute(dev);
2507
2508         return idev;
2509 }
2510
2511 static void manage_tempaddrs(struct inet6_dev *idev,
2512                              struct inet6_ifaddr *ifp,
2513                              __u32 valid_lft, __u32 prefered_lft,
2514                              bool create, unsigned long now)
2515 {
2516         u32 flags;
2517         struct inet6_ifaddr *ift;
2518
2519         read_lock_bh(&idev->lock);
2520         /* update all temporary addresses in the list */
2521         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2522                 int age, max_valid, max_prefered;
2523
2524                 if (ifp != ift->ifpub)
2525                         continue;
2526
2527                 /* RFC 4941 section 3.3:
2528                  * If a received option will extend the lifetime of a public
2529                  * address, the lifetimes of temporary addresses should
2530                  * be extended, subject to the overall constraint that no
2531                  * temporary addresses should ever remain "valid" or "preferred"
2532                  * for a time longer than (TEMP_VALID_LIFETIME) or
2533                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2534                  */
2535                 age = (now - ift->cstamp) / HZ;
2536                 max_valid = idev->cnf.temp_valid_lft - age;
2537                 if (max_valid < 0)
2538                         max_valid = 0;
2539
2540                 max_prefered = idev->cnf.temp_prefered_lft -
2541                                idev->desync_factor - age;
2542                 if (max_prefered < 0)
2543                         max_prefered = 0;
2544
2545                 if (valid_lft > max_valid)
2546                         valid_lft = max_valid;
2547
2548                 if (prefered_lft > max_prefered)
2549                         prefered_lft = max_prefered;
2550
2551                 spin_lock(&ift->lock);
2552                 flags = ift->flags;
2553                 ift->valid_lft = valid_lft;
2554                 ift->prefered_lft = prefered_lft;
2555                 ift->tstamp = now;
2556                 if (prefered_lft > 0)
2557                         ift->flags &= ~IFA_F_DEPRECATED;
2558
2559                 spin_unlock(&ift->lock);
2560                 if (!(flags&IFA_F_TENTATIVE))
2561                         ipv6_ifa_notify(0, ift);
2562         }
2563
2564         /* Also create a temporary address if it's enabled but no temporary
2565          * address currently exists.
2566          * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2567          * as part of cleanup (ie. deleting the mngtmpaddr).
2568          * We don't want that to result in creating a new temporary ip address.
2569          */
2570         if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2571                 create = true;
2572
2573         if (create && idev->cnf.use_tempaddr > 0) {
2574                 /* When a new public address is created as described
2575                  * in [ADDRCONF], also create a new temporary address.
2576                  */
2577                 read_unlock_bh(&idev->lock);
2578                 ipv6_create_tempaddr(ifp, false);
2579         } else {
2580                 read_unlock_bh(&idev->lock);
2581         }
2582 }
2583
2584 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2585 {
2586         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2587                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2588 }
2589
2590 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2591                                  const struct prefix_info *pinfo,
2592                                  struct inet6_dev *in6_dev,
2593                                  const struct in6_addr *addr, int addr_type,
2594                                  u32 addr_flags, bool sllao, bool tokenized,
2595                                  __u32 valid_lft, u32 prefered_lft)
2596 {
2597         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2598         int create = 0, update_lft = 0;
2599
2600         if (!ifp && valid_lft) {
2601                 int max_addresses = in6_dev->cnf.max_addresses;
2602                 struct ifa6_config cfg = {
2603                         .pfx = addr,
2604                         .plen = pinfo->prefix_len,
2605                         .ifa_flags = addr_flags,
2606                         .valid_lft = valid_lft,
2607                         .preferred_lft = prefered_lft,
2608                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2609                         .ifa_proto = IFAPROT_KERNEL_RA
2610                 };
2611
2612 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2613                 if ((net->ipv6.devconf_all->optimistic_dad ||
2614                      in6_dev->cnf.optimistic_dad) &&
2615                     !net->ipv6.devconf_all->forwarding && sllao)
2616                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2617 #endif
2618
2619                 /* Do not allow to create too much of autoconfigured
2620                  * addresses; this would be too easy way to crash kernel.
2621                  */
2622                 if (!max_addresses ||
2623                     ipv6_count_addresses(in6_dev) < max_addresses)
2624                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2625
2626                 if (IS_ERR_OR_NULL(ifp))
2627                         return -1;
2628
2629                 create = 1;
2630                 spin_lock_bh(&ifp->lock);
2631                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2632                 ifp->cstamp = jiffies;
2633                 ifp->tokenized = tokenized;
2634                 spin_unlock_bh(&ifp->lock);
2635                 addrconf_dad_start(ifp);
2636         }
2637
2638         if (ifp) {
2639                 u32 flags;
2640                 unsigned long now;
2641                 u32 stored_lft;
2642
2643                 /* update lifetime (RFC2462 5.5.3 e) */
2644                 spin_lock_bh(&ifp->lock);
2645                 now = jiffies;
2646                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2647                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2648                 else
2649                         stored_lft = 0;
2650                 if (!create && stored_lft) {
2651                         const u32 minimum_lft = min_t(u32,
2652                                 stored_lft, MIN_VALID_LIFETIME);
2653                         valid_lft = max(valid_lft, minimum_lft);
2654
2655                         /* RFC4862 Section 5.5.3e:
2656                          * "Note that the preferred lifetime of the
2657                          *  corresponding address is always reset to
2658                          *  the Preferred Lifetime in the received
2659                          *  Prefix Information option, regardless of
2660                          *  whether the valid lifetime is also reset or
2661                          *  ignored."
2662                          *
2663                          * So we should always update prefered_lft here.
2664                          */
2665                         update_lft = 1;
2666                 }
2667
2668                 if (update_lft) {
2669                         ifp->valid_lft = valid_lft;
2670                         ifp->prefered_lft = prefered_lft;
2671                         ifp->tstamp = now;
2672                         flags = ifp->flags;
2673                         ifp->flags &= ~IFA_F_DEPRECATED;
2674                         spin_unlock_bh(&ifp->lock);
2675
2676                         if (!(flags&IFA_F_TENTATIVE))
2677                                 ipv6_ifa_notify(0, ifp);
2678                 } else
2679                         spin_unlock_bh(&ifp->lock);
2680
2681                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2682                                  create, now);
2683
2684                 in6_ifa_put(ifp);
2685                 addrconf_verify(net);
2686         }
2687
2688         return 0;
2689 }
2690 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2691
2692 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2693 {
2694         struct prefix_info *pinfo;
2695         __u32 valid_lft;
2696         __u32 prefered_lft;
2697         int addr_type, err;
2698         u32 addr_flags = 0;
2699         struct inet6_dev *in6_dev;
2700         struct net *net = dev_net(dev);
2701
2702         pinfo = (struct prefix_info *) opt;
2703
2704         if (len < sizeof(struct prefix_info)) {
2705                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2706                 return;
2707         }
2708
2709         /*
2710          *      Validation checks ([ADDRCONF], page 19)
2711          */
2712
2713         addr_type = ipv6_addr_type(&pinfo->prefix);
2714
2715         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2716                 return;
2717
2718         valid_lft = ntohl(pinfo->valid);
2719         prefered_lft = ntohl(pinfo->prefered);
2720
2721         if (prefered_lft > valid_lft) {
2722                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2723                 return;
2724         }
2725
2726         in6_dev = in6_dev_get(dev);
2727
2728         if (!in6_dev) {
2729                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2730                                     dev->name);
2731                 return;
2732         }
2733
2734         /*
2735          *      Two things going on here:
2736          *      1) Add routes for on-link prefixes
2737          *      2) Configure prefixes with the auto flag set
2738          */
2739
2740         if (pinfo->onlink) {
2741                 struct fib6_info *rt;
2742                 unsigned long rt_expires;
2743
2744                 /* Avoid arithmetic overflow. Really, we could
2745                  * save rt_expires in seconds, likely valid_lft,
2746                  * but it would require division in fib gc, that it
2747                  * not good.
2748                  */
2749                 if (HZ > USER_HZ)
2750                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2751                 else
2752                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2753
2754                 if (addrconf_finite_timeout(rt_expires))
2755                         rt_expires *= HZ;
2756
2757                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2758                                                pinfo->prefix_len,
2759                                                dev,
2760                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2761                                                RTF_DEFAULT, true);
2762
2763                 if (rt) {
2764                         /* Autoconf prefix route */
2765                         if (valid_lft == 0) {
2766                                 ip6_del_rt(net, rt, false);
2767                                 rt = NULL;
2768                         } else if (addrconf_finite_timeout(rt_expires)) {
2769                                 /* not infinity */
2770                                 fib6_set_expires(rt, jiffies + rt_expires);
2771                         } else {
2772                                 fib6_clean_expires(rt);
2773                         }
2774                 } else if (valid_lft) {
2775                         clock_t expires = 0;
2776                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2777                         if (addrconf_finite_timeout(rt_expires)) {
2778                                 /* not infinity */
2779                                 flags |= RTF_EXPIRES;
2780                                 expires = jiffies_to_clock_t(rt_expires);
2781                         }
2782                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2783                                               0, dev, expires, flags,
2784                                               GFP_ATOMIC);
2785                 }
2786                 fib6_info_release(rt);
2787         }
2788
2789         /* Try to figure out our local address for this prefix */
2790
2791         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2792                 struct in6_addr addr;
2793                 bool tokenized = false, dev_addr_generated = false;
2794
2795                 if (pinfo->prefix_len == 64) {
2796                         memcpy(&addr, &pinfo->prefix, 8);
2797
2798                         if (!ipv6_addr_any(&in6_dev->token)) {
2799                                 read_lock_bh(&in6_dev->lock);
2800                                 memcpy(addr.s6_addr + 8,
2801                                        in6_dev->token.s6_addr + 8, 8);
2802                                 read_unlock_bh(&in6_dev->lock);
2803                                 tokenized = true;
2804                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2805                                    !ipv6_generate_stable_address(&addr, 0,
2806                                                                  in6_dev)) {
2807                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2808                                 goto ok;
2809                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2810                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2811                                 goto put;
2812                         } else {
2813                                 dev_addr_generated = true;
2814                         }
2815                         goto ok;
2816                 }
2817                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2818                                     pinfo->prefix_len);
2819                 goto put;
2820
2821 ok:
2822                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2823                                                    &addr, addr_type,
2824                                                    addr_flags, sllao,
2825                                                    tokenized, valid_lft,
2826                                                    prefered_lft);
2827                 if (err)
2828                         goto put;
2829
2830                 /* Ignore error case here because previous prefix add addr was
2831                  * successful which will be notified.
2832                  */
2833                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2834                                               addr_type, addr_flags, sllao,
2835                                               tokenized, valid_lft,
2836                                               prefered_lft,
2837                                               dev_addr_generated);
2838         }
2839         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2840 put:
2841         in6_dev_put(in6_dev);
2842 }
2843
2844 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2845                 struct in6_ifreq *ireq)
2846 {
2847         struct ip_tunnel_parm p = { };
2848         int err;
2849
2850         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2851                 return -EADDRNOTAVAIL;
2852
2853         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2854         p.iph.version = 4;
2855         p.iph.ihl = 5;
2856         p.iph.protocol = IPPROTO_IPV6;
2857         p.iph.ttl = 64;
2858
2859         if (!dev->netdev_ops->ndo_tunnel_ctl)
2860                 return -EOPNOTSUPP;
2861         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2862         if (err)
2863                 return err;
2864
2865         dev = __dev_get_by_name(net, p.name);
2866         if (!dev)
2867                 return -ENOBUFS;
2868         return dev_open(dev, NULL);
2869 }
2870
2871 /*
2872  *      Set destination address.
2873  *      Special case for SIT interfaces where we create a new "virtual"
2874  *      device.
2875  */
2876 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2877 {
2878         struct net_device *dev;
2879         struct in6_ifreq ireq;
2880         int err = -ENODEV;
2881
2882         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2883                 return -ENODEV;
2884         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2885                 return -EFAULT;
2886
2887         rtnl_lock();
2888         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2889         if (dev && dev->type == ARPHRD_SIT)
2890                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2891         rtnl_unlock();
2892         return err;
2893 }
2894
2895 static int ipv6_mc_config(struct sock *sk, bool join,
2896                           const struct in6_addr *addr, int ifindex)
2897 {
2898         int ret;
2899
2900         ASSERT_RTNL();
2901
2902         lock_sock(sk);
2903         if (join)
2904                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2905         else
2906                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2907         release_sock(sk);
2908
2909         return ret;
2910 }
2911
2912 /*
2913  *      Manual configuration of address on an interface
2914  */
2915 static int inet6_addr_add(struct net *net, int ifindex,
2916                           struct ifa6_config *cfg,
2917                           struct netlink_ext_ack *extack)
2918 {
2919         struct inet6_ifaddr *ifp;
2920         struct inet6_dev *idev;
2921         struct net_device *dev;
2922         unsigned long timeout;
2923         clock_t expires;
2924         u32 flags;
2925
2926         ASSERT_RTNL();
2927
2928         if (cfg->plen > 128)
2929                 return -EINVAL;
2930
2931         /* check the lifetime */
2932         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2933                 return -EINVAL;
2934
2935         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2936                 return -EINVAL;
2937
2938         dev = __dev_get_by_index(net, ifindex);
2939         if (!dev)
2940                 return -ENODEV;
2941
2942         idev = addrconf_add_dev(dev);
2943         if (IS_ERR(idev))
2944                 return PTR_ERR(idev);
2945
2946         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2947                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2948                                          true, cfg->pfx, ifindex);
2949
2950                 if (ret < 0)
2951                         return ret;
2952         }
2953
2954         cfg->scope = ipv6_addr_scope(cfg->pfx);
2955
2956         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2957         if (addrconf_finite_timeout(timeout)) {
2958                 expires = jiffies_to_clock_t(timeout * HZ);
2959                 cfg->valid_lft = timeout;
2960                 flags = RTF_EXPIRES;
2961         } else {
2962                 expires = 0;
2963                 flags = 0;
2964                 cfg->ifa_flags |= IFA_F_PERMANENT;
2965         }
2966
2967         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2968         if (addrconf_finite_timeout(timeout)) {
2969                 if (timeout == 0)
2970                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2971                 cfg->preferred_lft = timeout;
2972         }
2973
2974         ifp = ipv6_add_addr(idev, cfg, true, extack);
2975         if (!IS_ERR(ifp)) {
2976                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2977                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2978                                               ifp->rt_priority, dev, expires,
2979                                               flags, GFP_KERNEL);
2980                 }
2981
2982                 /* Send a netlink notification if DAD is enabled and
2983                  * optimistic flag is not set
2984                  */
2985                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2986                         ipv6_ifa_notify(0, ifp);
2987                 /*
2988                  * Note that section 3.1 of RFC 4429 indicates
2989                  * that the Optimistic flag should not be set for
2990                  * manually configured addresses
2991                  */
2992                 addrconf_dad_start(ifp);
2993                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2994                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2995                                          cfg->preferred_lft, true, jiffies);
2996                 in6_ifa_put(ifp);
2997                 addrconf_verify_rtnl(net);
2998                 return 0;
2999         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3000                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3001                                cfg->pfx, ifindex);
3002         }
3003
3004         return PTR_ERR(ifp);
3005 }
3006
3007 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3008                           const struct in6_addr *pfx, unsigned int plen)
3009 {
3010         struct inet6_ifaddr *ifp;
3011         struct inet6_dev *idev;
3012         struct net_device *dev;
3013
3014         if (plen > 128)
3015                 return -EINVAL;
3016
3017         dev = __dev_get_by_index(net, ifindex);
3018         if (!dev)
3019                 return -ENODEV;
3020
3021         idev = __in6_dev_get(dev);
3022         if (!idev)
3023                 return -ENXIO;
3024
3025         read_lock_bh(&idev->lock);
3026         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3027                 if (ifp->prefix_len == plen &&
3028                     ipv6_addr_equal(pfx, &ifp->addr)) {
3029                         in6_ifa_hold(ifp);
3030                         read_unlock_bh(&idev->lock);
3031
3032                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3033                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3034                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3035                                                  jiffies);
3036                         ipv6_del_addr(ifp);
3037                         addrconf_verify_rtnl(net);
3038                         if (ipv6_addr_is_multicast(pfx)) {
3039                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3040                                                false, pfx, dev->ifindex);
3041                         }
3042                         return 0;
3043                 }
3044         }
3045         read_unlock_bh(&idev->lock);
3046         return -EADDRNOTAVAIL;
3047 }
3048
3049
3050 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3051 {
3052         struct ifa6_config cfg = {
3053                 .ifa_flags = IFA_F_PERMANENT,
3054                 .preferred_lft = INFINITY_LIFE_TIME,
3055                 .valid_lft = INFINITY_LIFE_TIME,
3056         };
3057         struct in6_ifreq ireq;
3058         int err;
3059
3060         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3061                 return -EPERM;
3062
3063         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3064                 return -EFAULT;
3065
3066         cfg.pfx = &ireq.ifr6_addr;
3067         cfg.plen = ireq.ifr6_prefixlen;
3068
3069         rtnl_lock();
3070         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3071         rtnl_unlock();
3072         return err;
3073 }
3074
3075 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3076 {
3077         struct in6_ifreq ireq;
3078         int err;
3079
3080         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3081                 return -EPERM;
3082
3083         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3084                 return -EFAULT;
3085
3086         rtnl_lock();
3087         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3088                              ireq.ifr6_prefixlen);
3089         rtnl_unlock();
3090         return err;
3091 }
3092
3093 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3094                      int plen, int scope, u8 proto)
3095 {
3096         struct inet6_ifaddr *ifp;
3097         struct ifa6_config cfg = {
3098                 .pfx = addr,
3099                 .plen = plen,
3100                 .ifa_flags = IFA_F_PERMANENT,
3101                 .valid_lft = INFINITY_LIFE_TIME,
3102                 .preferred_lft = INFINITY_LIFE_TIME,
3103                 .scope = scope,
3104                 .ifa_proto = proto
3105         };
3106
3107         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3108         if (!IS_ERR(ifp)) {
3109                 spin_lock_bh(&ifp->lock);
3110                 ifp->flags &= ~IFA_F_TENTATIVE;
3111                 spin_unlock_bh(&ifp->lock);
3112                 rt_genid_bump_ipv6(dev_net(idev->dev));
3113                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3114                 in6_ifa_put(ifp);
3115         }
3116 }
3117
3118 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3119 static void add_v4_addrs(struct inet6_dev *idev)
3120 {
3121         struct in6_addr addr;
3122         struct net_device *dev;
3123         struct net *net = dev_net(idev->dev);
3124         int scope, plen, offset = 0;
3125         u32 pflags = 0;
3126
3127         ASSERT_RTNL();
3128
3129         memset(&addr, 0, sizeof(struct in6_addr));
3130         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3131         if (idev->dev->addr_len == sizeof(struct in6_addr))
3132                 offset = sizeof(struct in6_addr) - 4;
3133         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3134
3135         if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3136                 scope = IPV6_ADDR_COMPATv4;
3137                 plen = 96;
3138                 pflags |= RTF_NONEXTHOP;
3139         } else {
3140                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3141                         return;
3142
3143                 addr.s6_addr32[0] = htonl(0xfe800000);
3144                 scope = IFA_LINK;
3145                 plen = 64;
3146         }
3147
3148         if (addr.s6_addr32[3]) {
3149                 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3150                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3151                                       GFP_KERNEL);
3152                 return;
3153         }
3154
3155         for_each_netdev(net, dev) {
3156                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3157                 if (in_dev && (dev->flags & IFF_UP)) {
3158                         struct in_ifaddr *ifa;
3159                         int flag = scope;
3160
3161                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3162                                 addr.s6_addr32[3] = ifa->ifa_local;
3163
3164                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3165                                         continue;
3166                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3167                                         if (idev->dev->flags&IFF_POINTOPOINT)
3168                                                 continue;
3169                                         flag |= IFA_HOST;
3170                                 }
3171
3172                                 add_addr(idev, &addr, plen, flag,
3173                                          IFAPROT_UNSPEC);
3174                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3175                                                       0, pflags, GFP_KERNEL);
3176                         }
3177                 }
3178         }
3179 }
3180 #endif
3181
3182 static void init_loopback(struct net_device *dev)
3183 {
3184         struct inet6_dev  *idev;
3185
3186         /* ::1 */
3187
3188         ASSERT_RTNL();
3189
3190         idev = ipv6_find_idev(dev);
3191         if (IS_ERR(idev)) {
3192                 pr_debug("%s: add_dev failed\n", __func__);
3193                 return;
3194         }
3195
3196         add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3197 }
3198
3199 void addrconf_add_linklocal(struct inet6_dev *idev,
3200                             const struct in6_addr *addr, u32 flags)
3201 {
3202         struct ifa6_config cfg = {
3203                 .pfx = addr,
3204                 .plen = 64,
3205                 .ifa_flags = flags | IFA_F_PERMANENT,
3206                 .valid_lft = INFINITY_LIFE_TIME,
3207                 .preferred_lft = INFINITY_LIFE_TIME,
3208                 .scope = IFA_LINK,
3209                 .ifa_proto = IFAPROT_KERNEL_LL
3210         };
3211         struct inet6_ifaddr *ifp;
3212
3213 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3214         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3215              idev->cnf.optimistic_dad) &&
3216             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3217                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3218 #endif
3219
3220         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3221         if (!IS_ERR(ifp)) {
3222                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3223                                       0, 0, GFP_ATOMIC);
3224                 addrconf_dad_start(ifp);
3225                 in6_ifa_put(ifp);
3226         }
3227 }
3228 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3229
3230 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3231 {
3232         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3233                 return true;
3234
3235         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3236             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3237                 return true;
3238
3239         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3240             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3241                 return true;
3242
3243         return false;
3244 }
3245
3246 static int ipv6_generate_stable_address(struct in6_addr *address,
3247                                         u8 dad_count,
3248                                         const struct inet6_dev *idev)
3249 {
3250         static DEFINE_SPINLOCK(lock);
3251         static __u32 digest[SHA1_DIGEST_WORDS];
3252         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3253
3254         static union {
3255                 char __data[SHA1_BLOCK_SIZE];
3256                 struct {
3257                         struct in6_addr secret;
3258                         __be32 prefix[2];
3259                         unsigned char hwaddr[MAX_ADDR_LEN];
3260                         u8 dad_count;
3261                 } __packed;
3262         } data;
3263
3264         struct in6_addr secret;
3265         struct in6_addr temp;
3266         struct net *net = dev_net(idev->dev);
3267
3268         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3269
3270         if (idev->cnf.stable_secret.initialized)
3271                 secret = idev->cnf.stable_secret.secret;
3272         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3273                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3274         else
3275                 return -1;
3276
3277 retry:
3278         spin_lock_bh(&lock);
3279
3280         sha1_init(digest);
3281         memset(&data, 0, sizeof(data));
3282         memset(workspace, 0, sizeof(workspace));
3283         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3284         data.prefix[0] = address->s6_addr32[0];
3285         data.prefix[1] = address->s6_addr32[1];
3286         data.secret = secret;
3287         data.dad_count = dad_count;
3288
3289         sha1_transform(digest, data.__data, workspace);
3290
3291         temp = *address;
3292         temp.s6_addr32[2] = (__force __be32)digest[0];
3293         temp.s6_addr32[3] = (__force __be32)digest[1];
3294
3295         spin_unlock_bh(&lock);
3296
3297         if (ipv6_reserved_interfaceid(temp)) {
3298                 dad_count++;
3299                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3300                         return -1;
3301                 goto retry;
3302         }
3303
3304         *address = temp;
3305         return 0;
3306 }
3307
3308 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3309 {
3310         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3311
3312         if (s->initialized)
3313                 return;
3314         s = &idev->cnf.stable_secret;
3315         get_random_bytes(&s->secret, sizeof(s->secret));
3316         s->initialized = true;
3317 }
3318
3319 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3320 {
3321         struct in6_addr addr;
3322
3323         /* no link local addresses on L3 master devices */
3324         if (netif_is_l3_master(idev->dev))
3325                 return;
3326
3327         /* no link local addresses on devices flagged as slaves */
3328         if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3329                 return;
3330
3331         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3332
3333         switch (idev->cnf.addr_gen_mode) {
3334         case IN6_ADDR_GEN_MODE_RANDOM:
3335                 ipv6_gen_mode_random_init(idev);
3336                 fallthrough;
3337         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3338                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3339                         addrconf_add_linklocal(idev, &addr,
3340                                                IFA_F_STABLE_PRIVACY);
3341                 else if (prefix_route)
3342                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3343                                               0, 0, GFP_KERNEL);
3344                 break;
3345         case IN6_ADDR_GEN_MODE_EUI64:
3346                 /* addrconf_add_linklocal also adds a prefix_route and we
3347                  * only need to care about prefix routes if ipv6_generate_eui64
3348                  * couldn't generate one.
3349                  */
3350                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3351                         addrconf_add_linklocal(idev, &addr, 0);
3352                 else if (prefix_route)
3353                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3354                                               0, 0, GFP_KERNEL);
3355                 break;
3356         case IN6_ADDR_GEN_MODE_NONE:
3357         default:
3358                 /* will not add any link local address */
3359                 break;
3360         }
3361 }
3362
3363 static void addrconf_dev_config(struct net_device *dev)
3364 {
3365         struct inet6_dev *idev;
3366
3367         ASSERT_RTNL();
3368
3369         if ((dev->type != ARPHRD_ETHER) &&
3370             (dev->type != ARPHRD_FDDI) &&
3371             (dev->type != ARPHRD_ARCNET) &&
3372             (dev->type != ARPHRD_INFINIBAND) &&
3373             (dev->type != ARPHRD_IEEE1394) &&
3374             (dev->type != ARPHRD_TUNNEL6) &&
3375             (dev->type != ARPHRD_6LOWPAN) &&
3376             (dev->type != ARPHRD_TUNNEL) &&
3377             (dev->type != ARPHRD_NONE) &&
3378             (dev->type != ARPHRD_RAWIP)) {
3379                 /* Alas, we support only Ethernet autoconfiguration. */
3380                 idev = __in6_dev_get(dev);
3381                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3382                     dev->flags & IFF_MULTICAST)
3383                         ipv6_mc_up(idev);
3384                 return;
3385         }
3386
3387         idev = addrconf_add_dev(dev);
3388         if (IS_ERR(idev))
3389                 return;
3390
3391         /* this device type has no EUI support */
3392         if (dev->type == ARPHRD_NONE &&
3393             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3394                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3395
3396         addrconf_addr_gen(idev, false);
3397 }
3398
3399 #if IS_ENABLED(CONFIG_IPV6_SIT)
3400 static void addrconf_sit_config(struct net_device *dev)
3401 {
3402         struct inet6_dev *idev;
3403
3404         ASSERT_RTNL();
3405
3406         /*
3407          * Configure the tunnel with one of our IPv4
3408          * addresses... we should configure all of
3409          * our v4 addrs in the tunnel
3410          */
3411
3412         idev = ipv6_find_idev(dev);
3413         if (IS_ERR(idev)) {
3414                 pr_debug("%s: add_dev failed\n", __func__);
3415                 return;
3416         }
3417
3418         if (dev->priv_flags & IFF_ISATAP) {
3419                 addrconf_addr_gen(idev, false);
3420                 return;
3421         }
3422
3423         add_v4_addrs(idev);
3424
3425         if (dev->flags&IFF_POINTOPOINT)
3426                 addrconf_add_mroute(dev);
3427 }
3428 #endif
3429
3430 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3431 static void addrconf_gre_config(struct net_device *dev)
3432 {
3433         struct inet6_dev *idev;
3434
3435         ASSERT_RTNL();
3436
3437         idev = ipv6_find_idev(dev);
3438         if (IS_ERR(idev)) {
3439                 pr_debug("%s: add_dev failed\n", __func__);
3440                 return;
3441         }
3442
3443         if (dev->type == ARPHRD_ETHER) {
3444                 addrconf_addr_gen(idev, true);
3445                 return;
3446         }
3447
3448         add_v4_addrs(idev);
3449
3450         if (dev->flags & IFF_POINTOPOINT)
3451                 addrconf_add_mroute(dev);
3452 }
3453 #endif
3454
3455 static void addrconf_init_auto_addrs(struct net_device *dev)
3456 {
3457         switch (dev->type) {
3458 #if IS_ENABLED(CONFIG_IPV6_SIT)
3459         case ARPHRD_SIT:
3460                 addrconf_sit_config(dev);
3461                 break;
3462 #endif
3463 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3464         case ARPHRD_IP6GRE:
3465         case ARPHRD_IPGRE:
3466                 addrconf_gre_config(dev);
3467                 break;
3468 #endif
3469         case ARPHRD_LOOPBACK:
3470                 init_loopback(dev);
3471                 break;
3472
3473         default:
3474                 addrconf_dev_config(dev);
3475                 break;
3476         }
3477 }
3478
3479 static int fixup_permanent_addr(struct net *net,
3480                                 struct inet6_dev *idev,
3481                                 struct inet6_ifaddr *ifp)
3482 {
3483         /* !fib6_node means the host route was removed from the
3484          * FIB, for example, if 'lo' device is taken down. In that
3485          * case regenerate the host route.
3486          */
3487         if (!ifp->rt || !ifp->rt->fib6_node) {
3488                 struct fib6_info *f6i, *prev;
3489
3490                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3491                                          GFP_ATOMIC);
3492                 if (IS_ERR(f6i))
3493                         return PTR_ERR(f6i);
3494
3495                 /* ifp->rt can be accessed outside of rtnl */
3496                 spin_lock(&ifp->lock);
3497                 prev = ifp->rt;
3498                 ifp->rt = f6i;
3499                 spin_unlock(&ifp->lock);
3500
3501                 fib6_info_release(prev);
3502         }
3503
3504         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3505                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3506                                       ifp->rt_priority, idev->dev, 0, 0,
3507                                       GFP_ATOMIC);
3508         }
3509
3510         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3511                 addrconf_dad_start(ifp);
3512
3513         return 0;
3514 }
3515
3516 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3517 {
3518         struct inet6_ifaddr *ifp, *tmp;
3519         struct inet6_dev *idev;
3520
3521         idev = __in6_dev_get(dev);
3522         if (!idev)
3523                 return;
3524
3525         write_lock_bh(&idev->lock);
3526
3527         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3528                 if ((ifp->flags & IFA_F_PERMANENT) &&
3529                     fixup_permanent_addr(net, idev, ifp) < 0) {
3530                         write_unlock_bh(&idev->lock);
3531                         in6_ifa_hold(ifp);
3532                         ipv6_del_addr(ifp);
3533                         write_lock_bh(&idev->lock);
3534
3535                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3536                                              idev->dev->name, &ifp->addr);
3537                 }
3538         }
3539
3540         write_unlock_bh(&idev->lock);
3541 }
3542
3543 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3544                            void *ptr)
3545 {
3546         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3547         struct netdev_notifier_change_info *change_info;
3548         struct netdev_notifier_changeupper_info *info;
3549         struct inet6_dev *idev = __in6_dev_get(dev);
3550         struct net *net = dev_net(dev);
3551         int run_pending = 0;
3552         int err;
3553
3554         switch (event) {
3555         case NETDEV_REGISTER:
3556                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3557                         idev = ipv6_add_dev(dev);
3558                         if (IS_ERR(idev))
3559                                 return notifier_from_errno(PTR_ERR(idev));
3560                 }
3561                 break;
3562
3563         case NETDEV_CHANGEMTU:
3564                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3565                 if (dev->mtu < IPV6_MIN_MTU) {
3566                         addrconf_ifdown(dev, dev != net->loopback_dev);
3567                         break;
3568                 }
3569
3570                 if (idev) {
3571                         rt6_mtu_change(dev, dev->mtu);
3572                         idev->cnf.mtu6 = dev->mtu;
3573                         break;
3574                 }
3575
3576                 /* allocate new idev */
3577                 idev = ipv6_add_dev(dev);
3578                 if (IS_ERR(idev))
3579                         break;
3580
3581                 /* device is still not ready */
3582                 if (!(idev->if_flags & IF_READY))
3583                         break;
3584
3585                 run_pending = 1;
3586                 fallthrough;
3587         case NETDEV_UP:
3588         case NETDEV_CHANGE:
3589                 if (idev && idev->cnf.disable_ipv6)
3590                         break;
3591
3592                 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3593                         if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3594                             dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3595                                 ipv6_mc_up(idev);
3596                         break;
3597                 }
3598
3599                 if (event == NETDEV_UP) {
3600                         /* restore routes for permanent addresses */
3601                         addrconf_permanent_addr(net, dev);
3602
3603                         if (!addrconf_link_ready(dev)) {
3604                                 /* device is not ready yet. */
3605                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3606                                          dev->name);
3607                                 break;
3608                         }
3609
3610                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3611                                 idev = ipv6_add_dev(dev);
3612
3613                         if (!IS_ERR_OR_NULL(idev)) {
3614                                 idev->if_flags |= IF_READY;
3615                                 run_pending = 1;
3616                         }
3617                 } else if (event == NETDEV_CHANGE) {
3618                         if (!addrconf_link_ready(dev)) {
3619                                 /* device is still not ready. */
3620                                 rt6_sync_down_dev(dev, event);
3621                                 break;
3622                         }
3623
3624                         if (!IS_ERR_OR_NULL(idev)) {
3625                                 if (idev->if_flags & IF_READY) {
3626                                         /* device is already configured -
3627                                          * but resend MLD reports, we might
3628                                          * have roamed and need to update
3629                                          * multicast snooping switches
3630                                          */
3631                                         ipv6_mc_up(idev);
3632                                         change_info = ptr;
3633                                         if (change_info->flags_changed & IFF_NOARP)
3634                                                 addrconf_dad_run(idev, true);
3635                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3636                                         break;
3637                                 }
3638                                 idev->if_flags |= IF_READY;
3639                         }
3640
3641                         pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3642                                  dev->name);
3643
3644                         run_pending = 1;
3645                 }
3646
3647                 addrconf_init_auto_addrs(dev);
3648
3649                 if (!IS_ERR_OR_NULL(idev)) {
3650                         if (run_pending)
3651                                 addrconf_dad_run(idev, false);
3652
3653                         /* Device has an address by now */
3654                         rt6_sync_up(dev, RTNH_F_DEAD);
3655
3656                         /*
3657                          * If the MTU changed during the interface down,
3658                          * when the interface up, the changed MTU must be
3659                          * reflected in the idev as well as routers.
3660                          */
3661                         if (idev->cnf.mtu6 != dev->mtu &&
3662                             dev->mtu >= IPV6_MIN_MTU) {
3663                                 rt6_mtu_change(dev, dev->mtu);
3664                                 idev->cnf.mtu6 = dev->mtu;
3665                         }
3666                         idev->tstamp = jiffies;
3667                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3668
3669                         /*
3670                          * If the changed mtu during down is lower than
3671                          * IPV6_MIN_MTU stop IPv6 on this interface.
3672                          */
3673                         if (dev->mtu < IPV6_MIN_MTU)
3674                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3675                 }
3676                 break;
3677
3678         case NETDEV_DOWN:
3679         case NETDEV_UNREGISTER:
3680                 /*
3681                  *      Remove all addresses from this interface.
3682                  */
3683                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3684                 break;
3685
3686         case NETDEV_CHANGENAME:
3687                 if (idev) {
3688                         snmp6_unregister_dev(idev);
3689                         addrconf_sysctl_unregister(idev);
3690                         err = addrconf_sysctl_register(idev);
3691                         if (err)
3692                                 return notifier_from_errno(err);
3693                         err = snmp6_register_dev(idev);
3694                         if (err) {
3695                                 addrconf_sysctl_unregister(idev);
3696                                 return notifier_from_errno(err);
3697                         }
3698                 }
3699                 break;
3700
3701         case NETDEV_PRE_TYPE_CHANGE:
3702         case NETDEV_POST_TYPE_CHANGE:
3703                 if (idev)
3704                         addrconf_type_change(dev, event);
3705                 break;
3706
3707         case NETDEV_CHANGEUPPER:
3708                 info = ptr;
3709
3710                 /* flush all routes if dev is linked to or unlinked from
3711                  * an L3 master device (e.g., VRF)
3712                  */
3713                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3714                         addrconf_ifdown(dev, false);
3715         }
3716
3717         return NOTIFY_OK;
3718 }
3719
3720 /*
3721  *      addrconf module should be notified of a device going up
3722  */
3723 static struct notifier_block ipv6_dev_notf = {
3724         .notifier_call = addrconf_notify,
3725         .priority = ADDRCONF_NOTIFY_PRIORITY,
3726 };
3727
3728 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3729 {
3730         struct inet6_dev *idev;
3731         ASSERT_RTNL();
3732
3733         idev = __in6_dev_get(dev);
3734
3735         if (event == NETDEV_POST_TYPE_CHANGE)
3736                 ipv6_mc_remap(idev);
3737         else if (event == NETDEV_PRE_TYPE_CHANGE)
3738                 ipv6_mc_unmap(idev);
3739 }
3740
3741 static bool addr_is_local(const struct in6_addr *addr)
3742 {
3743         return ipv6_addr_type(addr) &
3744                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3745 }
3746
3747 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3748 {
3749         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3750         struct net *net = dev_net(dev);
3751         struct inet6_dev *idev;
3752         struct inet6_ifaddr *ifa;
3753         LIST_HEAD(tmp_addr_list);
3754         bool keep_addr = false;
3755         bool was_ready;
3756         int state, i;
3757
3758         ASSERT_RTNL();
3759
3760         rt6_disable_ip(dev, event);
3761
3762         idev = __in6_dev_get(dev);
3763         if (!idev)
3764                 return -ENODEV;
3765
3766         /*
3767          * Step 1: remove reference to ipv6 device from parent device.
3768          *         Do not dev_put!
3769          */
3770         if (unregister) {
3771                 idev->dead = 1;
3772
3773                 /* protected by rtnl_lock */
3774                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3775
3776                 /* Step 1.5: remove snmp6 entry */
3777                 snmp6_unregister_dev(idev);
3778
3779         }
3780
3781         /* combine the user config with event to determine if permanent
3782          * addresses are to be removed from address hash table
3783          */
3784         if (!unregister && !idev->cnf.disable_ipv6) {
3785                 /* aggregate the system setting and interface setting */
3786                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3787
3788                 if (!_keep_addr)
3789                         _keep_addr = idev->cnf.keep_addr_on_down;
3790
3791                 keep_addr = (_keep_addr > 0);
3792         }
3793
3794         /* Step 2: clear hash table */
3795         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3796                 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3797
3798                 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3799 restart:
3800                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3801                         if (ifa->idev == idev) {
3802                                 addrconf_del_dad_work(ifa);
3803                                 /* combined flag + permanent flag decide if
3804                                  * address is retained on a down event
3805                                  */
3806                                 if (!keep_addr ||
3807                                     !(ifa->flags & IFA_F_PERMANENT) ||
3808                                     addr_is_local(&ifa->addr)) {
3809                                         hlist_del_init_rcu(&ifa->addr_lst);
3810                                         goto restart;
3811                                 }
3812                         }
3813                 }
3814                 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3815         }
3816
3817         write_lock_bh(&idev->lock);
3818
3819         addrconf_del_rs_timer(idev);
3820
3821         /* Step 2: clear flags for stateless addrconf, repeated down
3822          *         detection
3823          */
3824         was_ready = idev->if_flags & IF_READY;
3825         if (!unregister)
3826                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3827
3828         /* Step 3: clear tempaddr list */
3829         while (!list_empty(&idev->tempaddr_list)) {
3830                 ifa = list_first_entry(&idev->tempaddr_list,
3831                                        struct inet6_ifaddr, tmp_list);
3832                 list_del(&ifa->tmp_list);
3833                 write_unlock_bh(&idev->lock);
3834                 spin_lock_bh(&ifa->lock);
3835
3836                 if (ifa->ifpub) {
3837                         in6_ifa_put(ifa->ifpub);
3838                         ifa->ifpub = NULL;
3839                 }
3840                 spin_unlock_bh(&ifa->lock);
3841                 in6_ifa_put(ifa);
3842                 write_lock_bh(&idev->lock);
3843         }
3844
3845         list_for_each_entry(ifa, &idev->addr_list, if_list)
3846                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3847         write_unlock_bh(&idev->lock);
3848
3849         while (!list_empty(&tmp_addr_list)) {
3850                 struct fib6_info *rt = NULL;
3851                 bool keep;
3852
3853                 ifa = list_first_entry(&tmp_addr_list,
3854                                        struct inet6_ifaddr, if_list_aux);
3855                 list_del(&ifa->if_list_aux);
3856
3857                 addrconf_del_dad_work(ifa);
3858
3859                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3860                         !addr_is_local(&ifa->addr);
3861
3862                 spin_lock_bh(&ifa->lock);
3863
3864                 if (keep) {
3865                         /* set state to skip the notifier below */
3866                         state = INET6_IFADDR_STATE_DEAD;
3867                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3868                         if (!(ifa->flags & IFA_F_NODAD))
3869                                 ifa->flags |= IFA_F_TENTATIVE;
3870
3871                         rt = ifa->rt;
3872                         ifa->rt = NULL;
3873                 } else {
3874                         state = ifa->state;
3875                         ifa->state = INET6_IFADDR_STATE_DEAD;
3876                 }
3877
3878                 spin_unlock_bh(&ifa->lock);
3879
3880                 if (rt)
3881                         ip6_del_rt(net, rt, false);
3882
3883                 if (state != INET6_IFADDR_STATE_DEAD) {
3884                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3885                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3886                 } else {
3887                         if (idev->cnf.forwarding)
3888                                 addrconf_leave_anycast(ifa);
3889                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3890                 }
3891
3892                 if (!keep) {
3893                         write_lock_bh(&idev->lock);
3894                         list_del_rcu(&ifa->if_list);
3895                         write_unlock_bh(&idev->lock);
3896                         in6_ifa_put(ifa);
3897                 }
3898         }
3899
3900         /* Step 5: Discard anycast and multicast list */
3901         if (unregister) {
3902                 ipv6_ac_destroy_dev(idev);
3903                 ipv6_mc_destroy_dev(idev);
3904         } else if (was_ready) {
3905                 ipv6_mc_down(idev);
3906         }
3907
3908         idev->tstamp = jiffies;
3909         idev->ra_mtu = 0;
3910
3911         /* Last: Shot the device (if unregistered) */
3912         if (unregister) {
3913                 addrconf_sysctl_unregister(idev);
3914                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3915                 neigh_ifdown(&nd_tbl, dev);
3916                 in6_dev_put(idev);
3917         }
3918         return 0;
3919 }
3920
3921 static void addrconf_rs_timer(struct timer_list *t)
3922 {
3923         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3924         struct net_device *dev = idev->dev;
3925         struct in6_addr lladdr;
3926
3927         write_lock(&idev->lock);
3928         if (idev->dead || !(idev->if_flags & IF_READY))
3929                 goto out;
3930
3931         if (!ipv6_accept_ra(idev))
3932                 goto out;
3933
3934         /* Announcement received after solicitation was sent */
3935         if (idev->if_flags & IF_RA_RCVD)
3936                 goto out;
3937
3938         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3939                 write_unlock(&idev->lock);
3940                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3941                         ndisc_send_rs(dev, &lladdr,
3942                                       &in6addr_linklocal_allrouters);
3943                 else
3944                         goto put;
3945
3946                 write_lock(&idev->lock);
3947                 idev->rs_interval = rfc3315_s14_backoff_update(
3948                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3949                 /* The wait after the last probe can be shorter */
3950                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3951                                              idev->cnf.rtr_solicits) ?
3952                                       idev->cnf.rtr_solicit_delay :
3953                                       idev->rs_interval);
3954         } else {
3955                 /*
3956                  * Note: we do not support deprecated "all on-link"
3957                  * assumption any longer.
3958                  */
3959                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3960         }
3961
3962 out:
3963         write_unlock(&idev->lock);
3964 put:
3965         in6_dev_put(idev);
3966 }
3967
3968 /*
3969  *      Duplicate Address Detection
3970  */
3971 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3972 {
3973         unsigned long rand_num;
3974         struct inet6_dev *idev = ifp->idev;
3975         u64 nonce;
3976
3977         if (ifp->flags & IFA_F_OPTIMISTIC)
3978                 rand_num = 0;
3979         else
3980                 rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1);
3981
3982         nonce = 0;
3983         if (idev->cnf.enhanced_dad ||
3984             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3985                 do
3986                         get_random_bytes(&nonce, 6);
3987                 while (nonce == 0);
3988         }
3989         ifp->dad_nonce = nonce;
3990         ifp->dad_probes = idev->cnf.dad_transmits;
3991         addrconf_mod_dad_work(ifp, rand_num);
3992 }
3993
3994 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3995 {
3996         struct inet6_dev *idev = ifp->idev;
3997         struct net_device *dev = idev->dev;
3998         bool bump_id, notify = false;
3999         struct net *net;
4000
4001         addrconf_join_solict(dev, &ifp->addr);
4002
4003         read_lock_bh(&idev->lock);
4004         spin_lock(&ifp->lock);
4005         if (ifp->state == INET6_IFADDR_STATE_DEAD)
4006                 goto out;
4007
4008         net = dev_net(dev);
4009         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4010             (net->ipv6.devconf_all->accept_dad < 1 &&
4011              idev->cnf.accept_dad < 1) ||
4012             !(ifp->flags&IFA_F_TENTATIVE) ||
4013             ifp->flags & IFA_F_NODAD) {
4014                 bool send_na = false;
4015
4016                 if (ifp->flags & IFA_F_TENTATIVE &&
4017                     !(ifp->flags & IFA_F_OPTIMISTIC))
4018                         send_na = true;
4019                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4020                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4021                 spin_unlock(&ifp->lock);
4022                 read_unlock_bh(&idev->lock);
4023
4024                 addrconf_dad_completed(ifp, bump_id, send_na);
4025                 return;
4026         }
4027
4028         if (!(idev->if_flags & IF_READY)) {
4029                 spin_unlock(&ifp->lock);
4030                 read_unlock_bh(&idev->lock);
4031                 /*
4032                  * If the device is not ready:
4033                  * - keep it tentative if it is a permanent address.
4034                  * - otherwise, kill it.
4035                  */
4036                 in6_ifa_hold(ifp);
4037                 addrconf_dad_stop(ifp, 0);
4038                 return;
4039         }
4040
4041         /*
4042          * Optimistic nodes can start receiving
4043          * Frames right away
4044          */
4045         if (ifp->flags & IFA_F_OPTIMISTIC) {
4046                 ip6_ins_rt(net, ifp->rt);
4047                 if (ipv6_use_optimistic_addr(net, idev)) {
4048                         /* Because optimistic nodes can use this address,
4049                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4050                          */
4051                         notify = true;
4052                 }
4053         }
4054
4055         addrconf_dad_kick(ifp);
4056 out:
4057         spin_unlock(&ifp->lock);
4058         read_unlock_bh(&idev->lock);
4059         if (notify)
4060                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4061 }
4062
4063 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4064 {
4065         bool begin_dad = false;
4066
4067         spin_lock_bh(&ifp->lock);
4068         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4069                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4070                 begin_dad = true;
4071         }
4072         spin_unlock_bh(&ifp->lock);
4073
4074         if (begin_dad)
4075                 addrconf_mod_dad_work(ifp, 0);
4076 }
4077
4078 static void addrconf_dad_work(struct work_struct *w)
4079 {
4080         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4081                                                 struct inet6_ifaddr,
4082                                                 dad_work);
4083         struct inet6_dev *idev = ifp->idev;
4084         bool bump_id, disable_ipv6 = false;
4085         struct in6_addr mcaddr;
4086
4087         enum {
4088                 DAD_PROCESS,
4089                 DAD_BEGIN,
4090                 DAD_ABORT,
4091         } action = DAD_PROCESS;
4092
4093         rtnl_lock();
4094
4095         spin_lock_bh(&ifp->lock);
4096         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4097                 action = DAD_BEGIN;
4098                 ifp->state = INET6_IFADDR_STATE_DAD;
4099         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4100                 action = DAD_ABORT;
4101                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4102
4103                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4104                      idev->cnf.accept_dad > 1) &&
4105                     !idev->cnf.disable_ipv6 &&
4106                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4107                         struct in6_addr addr;
4108
4109                         addr.s6_addr32[0] = htonl(0xfe800000);
4110                         addr.s6_addr32[1] = 0;
4111
4112                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4113                             ipv6_addr_equal(&ifp->addr, &addr)) {
4114                                 /* DAD failed for link-local based on MAC */
4115                                 idev->cnf.disable_ipv6 = 1;
4116
4117                                 pr_info("%s: IPv6 being disabled!\n",
4118                                         ifp->idev->dev->name);
4119                                 disable_ipv6 = true;
4120                         }
4121                 }
4122         }
4123         spin_unlock_bh(&ifp->lock);
4124
4125         if (action == DAD_BEGIN) {
4126                 addrconf_dad_begin(ifp);
4127                 goto out;
4128         } else if (action == DAD_ABORT) {
4129                 in6_ifa_hold(ifp);
4130                 addrconf_dad_stop(ifp, 1);
4131                 if (disable_ipv6)
4132                         addrconf_ifdown(idev->dev, false);
4133                 goto out;
4134         }
4135
4136         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4137                 goto out;
4138
4139         write_lock_bh(&idev->lock);
4140         if (idev->dead || !(idev->if_flags & IF_READY)) {
4141                 write_unlock_bh(&idev->lock);
4142                 goto out;
4143         }
4144
4145         spin_lock(&ifp->lock);
4146         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4147                 spin_unlock(&ifp->lock);
4148                 write_unlock_bh(&idev->lock);
4149                 goto out;
4150         }
4151
4152         if (ifp->dad_probes == 0) {
4153                 bool send_na = false;
4154
4155                 /*
4156                  * DAD was successful
4157                  */
4158
4159                 if (ifp->flags & IFA_F_TENTATIVE &&
4160                     !(ifp->flags & IFA_F_OPTIMISTIC))
4161                         send_na = true;
4162                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4163                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4164                 spin_unlock(&ifp->lock);
4165                 write_unlock_bh(&idev->lock);
4166
4167                 addrconf_dad_completed(ifp, bump_id, send_na);
4168
4169                 goto out;
4170         }
4171
4172         ifp->dad_probes--;
4173         addrconf_mod_dad_work(ifp,
4174                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4175                                   HZ/100));
4176         spin_unlock(&ifp->lock);
4177         write_unlock_bh(&idev->lock);
4178
4179         /* send a neighbour solicitation for our addr */
4180         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4181         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4182                       ifp->dad_nonce);
4183 out:
4184         in6_ifa_put(ifp);
4185         rtnl_unlock();
4186 }
4187
4188 /* ifp->idev must be at least read locked */
4189 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4190 {
4191         struct inet6_ifaddr *ifpiter;
4192         struct inet6_dev *idev = ifp->idev;
4193
4194         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4195                 if (ifpiter->scope > IFA_LINK)
4196                         break;
4197                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4198                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4199                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4200                     IFA_F_PERMANENT)
4201                         return false;
4202         }
4203         return true;
4204 }
4205
4206 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4207                                    bool send_na)
4208 {
4209         struct net_device *dev = ifp->idev->dev;
4210         struct in6_addr lladdr;
4211         bool send_rs, send_mld;
4212
4213         addrconf_del_dad_work(ifp);
4214
4215         /*
4216          *      Configure the address for reception. Now it is valid.
4217          */
4218
4219         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4220
4221         /* If added prefix is link local and we are prepared to process
4222            router advertisements, start sending router solicitations.
4223          */
4224
4225         read_lock_bh(&ifp->idev->lock);
4226         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4227         send_rs = send_mld &&
4228                   ipv6_accept_ra(ifp->idev) &&
4229                   ifp->idev->cnf.rtr_solicits != 0 &&
4230                   (dev->flags & IFF_LOOPBACK) == 0 &&
4231                   (dev->type != ARPHRD_TUNNEL) &&
4232                   !netif_is_team_port(dev);
4233         read_unlock_bh(&ifp->idev->lock);
4234
4235         /* While dad is in progress mld report's source address is in6_addrany.
4236          * Resend with proper ll now.
4237          */
4238         if (send_mld)
4239                 ipv6_mc_dad_complete(ifp->idev);
4240
4241         /* send unsolicited NA if enabled */
4242         if (send_na &&
4243             (ifp->idev->cnf.ndisc_notify ||
4244              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4245                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4246                               /*router=*/ !!ifp->idev->cnf.forwarding,
4247                               /*solicited=*/ false, /*override=*/ true,
4248                               /*inc_opt=*/ true);
4249         }
4250
4251         if (send_rs) {
4252                 /*
4253                  *      If a host as already performed a random delay
4254                  *      [...] as part of DAD [...] there is no need
4255                  *      to delay again before sending the first RS
4256                  */
4257                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4258                         return;
4259                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4260
4261                 write_lock_bh(&ifp->idev->lock);
4262                 spin_lock(&ifp->lock);
4263                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4264                         ifp->idev->cnf.rtr_solicit_interval);
4265                 ifp->idev->rs_probes = 1;
4266                 ifp->idev->if_flags |= IF_RS_SENT;
4267                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4268                 spin_unlock(&ifp->lock);
4269                 write_unlock_bh(&ifp->idev->lock);
4270         }
4271
4272         if (bump_id)
4273                 rt_genid_bump_ipv6(dev_net(dev));
4274
4275         /* Make sure that a new temporary address will be created
4276          * before this temporary address becomes deprecated.
4277          */
4278         if (ifp->flags & IFA_F_TEMPORARY)
4279                 addrconf_verify_rtnl(dev_net(dev));
4280 }
4281
4282 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4283 {
4284         struct inet6_ifaddr *ifp;
4285
4286         read_lock_bh(&idev->lock);
4287         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4288                 spin_lock(&ifp->lock);
4289                 if ((ifp->flags & IFA_F_TENTATIVE &&
4290                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4291                         if (restart)
4292                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4293                         addrconf_dad_kick(ifp);
4294                 }
4295                 spin_unlock(&ifp->lock);
4296         }
4297         read_unlock_bh(&idev->lock);
4298 }
4299
4300 #ifdef CONFIG_PROC_FS
4301 struct if6_iter_state {
4302         struct seq_net_private p;
4303         int bucket;
4304         int offset;
4305 };
4306
4307 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4308 {
4309         struct if6_iter_state *state = seq->private;
4310         struct net *net = seq_file_net(seq);
4311         struct inet6_ifaddr *ifa = NULL;
4312         int p = 0;
4313
4314         /* initial bucket if pos is 0 */
4315         if (pos == 0) {
4316                 state->bucket = 0;
4317                 state->offset = 0;
4318         }
4319
4320         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4321                 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4322                                          addr_lst) {
4323                         /* sync with offset */
4324                         if (p < state->offset) {
4325                                 p++;
4326                                 continue;
4327                         }
4328                         return ifa;
4329                 }
4330
4331                 /* prepare for next bucket */
4332                 state->offset = 0;
4333                 p = 0;
4334         }
4335         return NULL;
4336 }
4337
4338 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4339                                          struct inet6_ifaddr *ifa)
4340 {
4341         struct if6_iter_state *state = seq->private;
4342         struct net *net = seq_file_net(seq);
4343
4344         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4345                 state->offset++;
4346                 return ifa;
4347         }
4348
4349         state->offset = 0;
4350         while (++state->bucket < IN6_ADDR_HSIZE) {
4351                 hlist_for_each_entry_rcu(ifa,
4352                                      &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4353                         return ifa;
4354                 }
4355         }
4356
4357         return NULL;
4358 }
4359
4360 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4361         __acquires(rcu)
4362 {
4363         rcu_read_lock();
4364         return if6_get_first(seq, *pos);
4365 }
4366
4367 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4368 {
4369         struct inet6_ifaddr *ifa;
4370
4371         ifa = if6_get_next(seq, v);
4372         ++*pos;
4373         return ifa;
4374 }
4375
4376 static void if6_seq_stop(struct seq_file *seq, void *v)
4377         __releases(rcu)
4378 {
4379         rcu_read_unlock();
4380 }
4381
4382 static int if6_seq_show(struct seq_file *seq, void *v)
4383 {
4384         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4385         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4386                    &ifp->addr,
4387                    ifp->idev->dev->ifindex,
4388                    ifp->prefix_len,
4389                    ifp->scope,
4390                    (u8) ifp->flags,
4391                    ifp->idev->dev->name);
4392         return 0;
4393 }
4394
4395 static const struct seq_operations if6_seq_ops = {
4396         .start  = if6_seq_start,
4397         .next   = if6_seq_next,
4398         .show   = if6_seq_show,
4399         .stop   = if6_seq_stop,
4400 };
4401
4402 static int __net_init if6_proc_net_init(struct net *net)
4403 {
4404         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4405                         sizeof(struct if6_iter_state)))
4406                 return -ENOMEM;
4407         return 0;
4408 }
4409
4410 static void __net_exit if6_proc_net_exit(struct net *net)
4411 {
4412         remove_proc_entry("if_inet6", net->proc_net);
4413 }
4414
4415 static struct pernet_operations if6_proc_net_ops = {
4416         .init = if6_proc_net_init,
4417         .exit = if6_proc_net_exit,
4418 };
4419
4420 int __init if6_proc_init(void)
4421 {
4422         return register_pernet_subsys(&if6_proc_net_ops);
4423 }
4424
4425 void if6_proc_exit(void)
4426 {
4427         unregister_pernet_subsys(&if6_proc_net_ops);
4428 }
4429 #endif  /* CONFIG_PROC_FS */
4430
4431 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4432 /* Check if address is a home address configured on any interface. */
4433 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4434 {
4435         unsigned int hash = inet6_addr_hash(net, addr);
4436         struct inet6_ifaddr *ifp = NULL;
4437         int ret = 0;
4438
4439         rcu_read_lock();
4440         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4441                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4442                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4443                         ret = 1;
4444                         break;
4445                 }
4446         }
4447         rcu_read_unlock();
4448         return ret;
4449 }
4450 #endif
4451
4452 /* RFC6554 has some algorithm to avoid loops in segment routing by
4453  * checking if the segments contains any of a local interface address.
4454  *
4455  * Quote:
4456  *
4457  * To detect loops in the SRH, a router MUST determine if the SRH
4458  * includes multiple addresses assigned to any interface on that router.
4459  * If such addresses appear more than once and are separated by at least
4460  * one address not assigned to that router.
4461  */
4462 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4463                           unsigned char nsegs)
4464 {
4465         const struct in6_addr *addr;
4466         int i, ret = 0, found = 0;
4467         struct inet6_ifaddr *ifp;
4468         bool separated = false;
4469         unsigned int hash;
4470         bool hash_found;
4471
4472         rcu_read_lock();
4473         for (i = 0; i < nsegs; i++) {
4474                 addr = &segs[i];
4475                 hash = inet6_addr_hash(net, addr);
4476
4477                 hash_found = false;
4478                 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4479
4480                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4481                                 hash_found = true;
4482                                 break;
4483                         }
4484                 }
4485
4486                 if (hash_found) {
4487                         if (found > 1 && separated) {
4488                                 ret = 1;
4489                                 break;
4490                         }
4491
4492                         separated = false;
4493                         found++;
4494                 } else {
4495                         separated = true;
4496                 }
4497         }
4498         rcu_read_unlock();
4499
4500         return ret;
4501 }
4502
4503 /*
4504  *      Periodic address status verification
4505  */
4506
4507 static void addrconf_verify_rtnl(struct net *net)
4508 {
4509         unsigned long now, next, next_sec, next_sched;
4510         struct inet6_ifaddr *ifp;
4511         int i;
4512
4513         ASSERT_RTNL();
4514
4515         rcu_read_lock_bh();
4516         now = jiffies;
4517         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4518
4519         cancel_delayed_work(&net->ipv6.addr_chk_work);
4520
4521         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4522 restart:
4523                 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4524                         unsigned long age;
4525
4526                         /* When setting preferred_lft to a value not zero or
4527                          * infinity, while valid_lft is infinity
4528                          * IFA_F_PERMANENT has a non-infinity life time.
4529                          */
4530                         if ((ifp->flags & IFA_F_PERMANENT) &&
4531                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4532                                 continue;
4533
4534                         spin_lock(&ifp->lock);
4535                         /* We try to batch several events at once. */
4536                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4537
4538                         if ((ifp->flags&IFA_F_TEMPORARY) &&
4539                             !(ifp->flags&IFA_F_TENTATIVE) &&
4540                             ifp->prefered_lft != INFINITY_LIFE_TIME &&
4541                             !ifp->regen_count && ifp->ifpub) {
4542                                 /* This is a non-regenerated temporary addr. */
4543
4544                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4545                                         ifp->idev->cnf.dad_transmits *
4546                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4547
4548                                 if (age + regen_advance >= ifp->prefered_lft) {
4549                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4550                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4551                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4552
4553                                         ifp->regen_count++;
4554                                         in6_ifa_hold(ifp);
4555                                         in6_ifa_hold(ifpub);
4556                                         spin_unlock(&ifp->lock);
4557
4558                                         spin_lock(&ifpub->lock);
4559                                         ifpub->regen_count = 0;
4560                                         spin_unlock(&ifpub->lock);
4561                                         rcu_read_unlock_bh();
4562                                         ipv6_create_tempaddr(ifpub, true);
4563                                         in6_ifa_put(ifpub);
4564                                         in6_ifa_put(ifp);
4565                                         rcu_read_lock_bh();
4566                                         goto restart;
4567                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4568                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4569                         }
4570
4571                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4572                             age >= ifp->valid_lft) {
4573                                 spin_unlock(&ifp->lock);
4574                                 in6_ifa_hold(ifp);
4575                                 rcu_read_unlock_bh();
4576                                 ipv6_del_addr(ifp);
4577                                 rcu_read_lock_bh();
4578                                 goto restart;
4579                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4580                                 spin_unlock(&ifp->lock);
4581                                 continue;
4582                         } else if (age >= ifp->prefered_lft) {
4583                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4584                                 int deprecate = 0;
4585
4586                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4587                                         deprecate = 1;
4588                                         ifp->flags |= IFA_F_DEPRECATED;
4589                                 }
4590
4591                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4592                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4593                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4594
4595                                 spin_unlock(&ifp->lock);
4596
4597                                 if (deprecate) {
4598                                         in6_ifa_hold(ifp);
4599
4600                                         ipv6_ifa_notify(0, ifp);
4601                                         in6_ifa_put(ifp);
4602                                         goto restart;
4603                                 }
4604                         } else {
4605                                 /* ifp->prefered_lft <= ifp->valid_lft */
4606                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4607                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4608                                 spin_unlock(&ifp->lock);
4609                         }
4610                 }
4611         }
4612
4613         next_sec = round_jiffies_up(next);
4614         next_sched = next;
4615
4616         /* If rounded timeout is accurate enough, accept it. */
4617         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4618                 next_sched = next_sec;
4619
4620         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4621         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4622                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4623
4624         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4625                  now, next, next_sec, next_sched);
4626         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4627         rcu_read_unlock_bh();
4628 }
4629
4630 static void addrconf_verify_work(struct work_struct *w)
4631 {
4632         struct net *net = container_of(to_delayed_work(w), struct net,
4633                                        ipv6.addr_chk_work);
4634
4635         rtnl_lock();
4636         addrconf_verify_rtnl(net);
4637         rtnl_unlock();
4638 }
4639
4640 static void addrconf_verify(struct net *net)
4641 {
4642         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4643 }
4644
4645 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4646                                      struct in6_addr **peer_pfx)
4647 {
4648         struct in6_addr *pfx = NULL;
4649
4650         *peer_pfx = NULL;
4651
4652         if (addr)
4653                 pfx = nla_data(addr);
4654
4655         if (local) {
4656                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4657                         *peer_pfx = pfx;
4658                 pfx = nla_data(local);
4659         }
4660
4661         return pfx;
4662 }
4663
4664 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4665         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4666         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4667         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4668         [IFA_FLAGS]             = { .len = sizeof(u32) },
4669         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4670         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4671         [IFA_PROTO]             = { .type = NLA_U8 },
4672 };
4673
4674 static int
4675 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4676                   struct netlink_ext_ack *extack)
4677 {
4678         struct net *net = sock_net(skb->sk);
4679         struct ifaddrmsg *ifm;
4680         struct nlattr *tb[IFA_MAX+1];
4681         struct in6_addr *pfx, *peer_pfx;
4682         u32 ifa_flags;
4683         int err;
4684
4685         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4686                                      ifa_ipv6_policy, extack);
4687         if (err < 0)
4688                 return err;
4689
4690         ifm = nlmsg_data(nlh);
4691         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4692         if (!pfx)
4693                 return -EINVAL;
4694
4695         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4696
4697         /* We ignore other flags so far. */
4698         ifa_flags &= IFA_F_MANAGETEMPADDR;
4699
4700         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4701                               ifm->ifa_prefixlen);
4702 }
4703
4704 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4705                                unsigned long expires, u32 flags,
4706                                bool modify_peer)
4707 {
4708         struct fib6_info *f6i;
4709         u32 prio;
4710
4711         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4712                                         ifp->prefix_len,
4713                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4714         if (!f6i)
4715                 return -ENOENT;
4716
4717         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4718         if (f6i->fib6_metric != prio) {
4719                 /* delete old one */
4720                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4721
4722                 /* add new one */
4723                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4724                                       ifp->prefix_len,
4725                                       ifp->rt_priority, ifp->idev->dev,
4726                                       expires, flags, GFP_KERNEL);
4727         } else {
4728                 if (!expires)
4729                         fib6_clean_expires(f6i);
4730                 else
4731                         fib6_set_expires(f6i, expires);
4732
4733                 fib6_info_release(f6i);
4734         }
4735
4736         return 0;
4737 }
4738
4739 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4740                              struct ifa6_config *cfg)
4741 {
4742         u32 flags;
4743         clock_t expires;
4744         unsigned long timeout;
4745         bool was_managetempaddr;
4746         bool had_prefixroute;
4747         bool new_peer = false;
4748
4749         ASSERT_RTNL();
4750
4751         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4752                 return -EINVAL;
4753
4754         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4755             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4756                 return -EINVAL;
4757
4758         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4759                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4760
4761         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4762         if (addrconf_finite_timeout(timeout)) {
4763                 expires = jiffies_to_clock_t(timeout * HZ);
4764                 cfg->valid_lft = timeout;
4765                 flags = RTF_EXPIRES;
4766         } else {
4767                 expires = 0;
4768                 flags = 0;
4769                 cfg->ifa_flags |= IFA_F_PERMANENT;
4770         }
4771
4772         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4773         if (addrconf_finite_timeout(timeout)) {
4774                 if (timeout == 0)
4775                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4776                 cfg->preferred_lft = timeout;
4777         }
4778
4779         if (cfg->peer_pfx &&
4780             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4781                 if (!ipv6_addr_any(&ifp->peer_addr))
4782                         cleanup_prefix_route(ifp, expires, true, true);
4783                 new_peer = true;
4784         }
4785
4786         spin_lock_bh(&ifp->lock);
4787         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4788         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4789                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4790         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4791                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4792                         IFA_F_NOPREFIXROUTE);
4793         ifp->flags |= cfg->ifa_flags;
4794         ifp->tstamp = jiffies;
4795         ifp->valid_lft = cfg->valid_lft;
4796         ifp->prefered_lft = cfg->preferred_lft;
4797         ifp->ifa_proto = cfg->ifa_proto;
4798
4799         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4800                 ifp->rt_priority = cfg->rt_priority;
4801
4802         if (new_peer)
4803                 ifp->peer_addr = *cfg->peer_pfx;
4804
4805         spin_unlock_bh(&ifp->lock);
4806         if (!(ifp->flags&IFA_F_TENTATIVE))
4807                 ipv6_ifa_notify(0, ifp);
4808
4809         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4810                 int rc = -ENOENT;
4811
4812                 if (had_prefixroute)
4813                         rc = modify_prefix_route(ifp, expires, flags, false);
4814
4815                 /* prefix route could have been deleted; if so restore it */
4816                 if (rc == -ENOENT) {
4817                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4818                                               ifp->rt_priority, ifp->idev->dev,
4819                                               expires, flags, GFP_KERNEL);
4820                 }
4821
4822                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4823                         rc = modify_prefix_route(ifp, expires, flags, true);
4824
4825                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4826                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4827                                               ifp->rt_priority, ifp->idev->dev,
4828                                               expires, flags, GFP_KERNEL);
4829                 }
4830         } else if (had_prefixroute) {
4831                 enum cleanup_prefix_rt_t action;
4832                 unsigned long rt_expires;
4833
4834                 write_lock_bh(&ifp->idev->lock);
4835                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4836                 write_unlock_bh(&ifp->idev->lock);
4837
4838                 if (action != CLEANUP_PREFIX_RT_NOP) {
4839                         cleanup_prefix_route(ifp, rt_expires,
4840                                 action == CLEANUP_PREFIX_RT_DEL, false);
4841                 }
4842         }
4843
4844         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4845                 if (was_managetempaddr &&
4846                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4847                         cfg->valid_lft = 0;
4848                         cfg->preferred_lft = 0;
4849                 }
4850                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4851                                  cfg->preferred_lft, !was_managetempaddr,
4852                                  jiffies);
4853         }
4854
4855         addrconf_verify_rtnl(net);
4856
4857         return 0;
4858 }
4859
4860 static int
4861 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4862                   struct netlink_ext_ack *extack)
4863 {
4864         struct net *net = sock_net(skb->sk);
4865         struct ifaddrmsg *ifm;
4866         struct nlattr *tb[IFA_MAX+1];
4867         struct in6_addr *peer_pfx;
4868         struct inet6_ifaddr *ifa;
4869         struct net_device *dev;
4870         struct inet6_dev *idev;
4871         struct ifa6_config cfg;
4872         int err;
4873
4874         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4875                                      ifa_ipv6_policy, extack);
4876         if (err < 0)
4877                 return err;
4878
4879         memset(&cfg, 0, sizeof(cfg));
4880
4881         ifm = nlmsg_data(nlh);
4882         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4883         if (!cfg.pfx)
4884                 return -EINVAL;
4885
4886         cfg.peer_pfx = peer_pfx;
4887         cfg.plen = ifm->ifa_prefixlen;
4888         if (tb[IFA_RT_PRIORITY])
4889                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4890
4891         if (tb[IFA_PROTO])
4892                 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4893
4894         cfg.valid_lft = INFINITY_LIFE_TIME;
4895         cfg.preferred_lft = INFINITY_LIFE_TIME;
4896
4897         if (tb[IFA_CACHEINFO]) {
4898                 struct ifa_cacheinfo *ci;
4899
4900                 ci = nla_data(tb[IFA_CACHEINFO]);
4901                 cfg.valid_lft = ci->ifa_valid;
4902                 cfg.preferred_lft = ci->ifa_prefered;
4903         }
4904
4905         dev =  __dev_get_by_index(net, ifm->ifa_index);
4906         if (!dev)
4907                 return -ENODEV;
4908
4909         if (tb[IFA_FLAGS])
4910                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4911         else
4912                 cfg.ifa_flags = ifm->ifa_flags;
4913
4914         /* We ignore other flags so far. */
4915         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4916                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4917                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4918
4919         idev = ipv6_find_idev(dev);
4920         if (IS_ERR(idev))
4921                 return PTR_ERR(idev);
4922
4923         if (!ipv6_allow_optimistic_dad(net, idev))
4924                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4925
4926         if (cfg.ifa_flags & IFA_F_NODAD &&
4927             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4928                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4929                 return -EINVAL;
4930         }
4931
4932         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4933         if (!ifa) {
4934                 /*
4935                  * It would be best to check for !NLM_F_CREATE here but
4936                  * userspace already relies on not having to provide this.
4937                  */
4938                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4939         }
4940
4941         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4942             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4943                 err = -EEXIST;
4944         else
4945                 err = inet6_addr_modify(net, ifa, &cfg);
4946
4947         in6_ifa_put(ifa);
4948
4949         return err;
4950 }
4951
4952 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4953                           u8 scope, int ifindex)
4954 {
4955         struct ifaddrmsg *ifm;
4956
4957         ifm = nlmsg_data(nlh);
4958         ifm->ifa_family = AF_INET6;
4959         ifm->ifa_prefixlen = prefixlen;
4960         ifm->ifa_flags = flags;
4961         ifm->ifa_scope = scope;
4962         ifm->ifa_index = ifindex;
4963 }
4964
4965 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4966                          unsigned long tstamp, u32 preferred, u32 valid)
4967 {
4968         struct ifa_cacheinfo ci;
4969
4970         ci.cstamp = cstamp_delta(cstamp);
4971         ci.tstamp = cstamp_delta(tstamp);
4972         ci.ifa_prefered = preferred;
4973         ci.ifa_valid = valid;
4974
4975         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4976 }
4977
4978 static inline int rt_scope(int ifa_scope)
4979 {
4980         if (ifa_scope & IFA_HOST)
4981                 return RT_SCOPE_HOST;
4982         else if (ifa_scope & IFA_LINK)
4983                 return RT_SCOPE_LINK;
4984         else if (ifa_scope & IFA_SITE)
4985                 return RT_SCOPE_SITE;
4986         else
4987                 return RT_SCOPE_UNIVERSE;
4988 }
4989
4990 static inline int inet6_ifaddr_msgsize(void)
4991 {
4992         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4993                + nla_total_size(16) /* IFA_LOCAL */
4994                + nla_total_size(16) /* IFA_ADDRESS */
4995                + nla_total_size(sizeof(struct ifa_cacheinfo))
4996                + nla_total_size(4)  /* IFA_FLAGS */
4997                + nla_total_size(1)  /* IFA_PROTO */
4998                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4999 }
5000
5001 enum addr_type_t {
5002         UNICAST_ADDR,
5003         MULTICAST_ADDR,
5004         ANYCAST_ADDR,
5005 };
5006
5007 struct inet6_fill_args {
5008         u32 portid;
5009         u32 seq;
5010         int event;
5011         unsigned int flags;
5012         int netnsid;
5013         int ifindex;
5014         enum addr_type_t type;
5015 };
5016
5017 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5018                              struct inet6_fill_args *args)
5019 {
5020         struct nlmsghdr  *nlh;
5021         u32 preferred, valid;
5022
5023         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5024                         sizeof(struct ifaddrmsg), args->flags);
5025         if (!nlh)
5026                 return -EMSGSIZE;
5027
5028         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5029                       ifa->idev->dev->ifindex);
5030
5031         if (args->netnsid >= 0 &&
5032             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5033                 goto error;
5034
5035         spin_lock_bh(&ifa->lock);
5036         if (!((ifa->flags&IFA_F_PERMANENT) &&
5037               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5038                 preferred = ifa->prefered_lft;
5039                 valid = ifa->valid_lft;
5040                 if (preferred != INFINITY_LIFE_TIME) {
5041                         long tval = (jiffies - ifa->tstamp)/HZ;
5042                         if (preferred > tval)
5043                                 preferred -= tval;
5044                         else
5045                                 preferred = 0;
5046                         if (valid != INFINITY_LIFE_TIME) {
5047                                 if (valid > tval)
5048                                         valid -= tval;
5049                                 else
5050                                         valid = 0;
5051                         }
5052                 }
5053         } else {
5054                 preferred = INFINITY_LIFE_TIME;
5055                 valid = INFINITY_LIFE_TIME;
5056         }
5057         spin_unlock_bh(&ifa->lock);
5058
5059         if (!ipv6_addr_any(&ifa->peer_addr)) {
5060                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5061                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5062                         goto error;
5063         } else
5064                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5065                         goto error;
5066
5067         if (ifa->rt_priority &&
5068             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5069                 goto error;
5070
5071         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5072                 goto error;
5073
5074         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5075                 goto error;
5076
5077         if (ifa->ifa_proto &&
5078             nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5079                 goto error;
5080
5081         nlmsg_end(skb, nlh);
5082         return 0;
5083
5084 error:
5085         nlmsg_cancel(skb, nlh);
5086         return -EMSGSIZE;
5087 }
5088
5089 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5090                                struct inet6_fill_args *args)
5091 {
5092         struct nlmsghdr  *nlh;
5093         u8 scope = RT_SCOPE_UNIVERSE;
5094         int ifindex = ifmca->idev->dev->ifindex;
5095
5096         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5097                 scope = RT_SCOPE_SITE;
5098
5099         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5100                         sizeof(struct ifaddrmsg), args->flags);
5101         if (!nlh)
5102                 return -EMSGSIZE;
5103
5104         if (args->netnsid >= 0 &&
5105             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5106                 nlmsg_cancel(skb, nlh);
5107                 return -EMSGSIZE;
5108         }
5109
5110         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5111         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5112             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5113                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5114                 nlmsg_cancel(skb, nlh);
5115                 return -EMSGSIZE;
5116         }
5117
5118         nlmsg_end(skb, nlh);
5119         return 0;
5120 }
5121
5122 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5123                                struct inet6_fill_args *args)
5124 {
5125         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5126         int ifindex = dev ? dev->ifindex : 1;
5127         struct nlmsghdr  *nlh;
5128         u8 scope = RT_SCOPE_UNIVERSE;
5129
5130         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5131                 scope = RT_SCOPE_SITE;
5132
5133         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5134                         sizeof(struct ifaddrmsg), args->flags);
5135         if (!nlh)
5136                 return -EMSGSIZE;
5137
5138         if (args->netnsid >= 0 &&
5139             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5140                 nlmsg_cancel(skb, nlh);
5141                 return -EMSGSIZE;
5142         }
5143
5144         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5145         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5146             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5147                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5148                 nlmsg_cancel(skb, nlh);
5149                 return -EMSGSIZE;
5150         }
5151
5152         nlmsg_end(skb, nlh);
5153         return 0;
5154 }
5155
5156 /* called with rcu_read_lock() */
5157 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5158                           struct netlink_callback *cb, int s_ip_idx,
5159                           struct inet6_fill_args *fillargs)
5160 {
5161         struct ifmcaddr6 *ifmca;
5162         struct ifacaddr6 *ifaca;
5163         int ip_idx = 0;
5164         int err = 1;
5165
5166         read_lock_bh(&idev->lock);
5167         switch (fillargs->type) {
5168         case UNICAST_ADDR: {
5169                 struct inet6_ifaddr *ifa;
5170                 fillargs->event = RTM_NEWADDR;
5171
5172                 /* unicast address incl. temp addr */
5173                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5174                         if (ip_idx < s_ip_idx)
5175                                 goto next;
5176                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5177                         if (err < 0)
5178                                 break;
5179                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5180 next:
5181                         ip_idx++;
5182                 }
5183                 break;
5184         }
5185         case MULTICAST_ADDR:
5186                 read_unlock_bh(&idev->lock);
5187                 fillargs->event = RTM_GETMULTICAST;
5188
5189                 /* multicast address */
5190                 for (ifmca = rtnl_dereference(idev->mc_list);
5191                      ifmca;
5192                      ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5193                         if (ip_idx < s_ip_idx)
5194                                 continue;
5195                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5196                         if (err < 0)
5197                                 break;
5198                 }
5199                 read_lock_bh(&idev->lock);
5200                 break;
5201         case ANYCAST_ADDR:
5202                 fillargs->event = RTM_GETANYCAST;
5203                 /* anycast address */
5204                 for (ifaca = idev->ac_list; ifaca;
5205                      ifaca = ifaca->aca_next, ip_idx++) {
5206                         if (ip_idx < s_ip_idx)
5207                                 continue;
5208                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5209                         if (err < 0)
5210                                 break;
5211                 }
5212                 break;
5213         default:
5214                 break;
5215         }
5216         read_unlock_bh(&idev->lock);
5217         cb->args[2] = ip_idx;
5218         return err;
5219 }
5220
5221 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5222                                        struct inet6_fill_args *fillargs,
5223                                        struct net **tgt_net, struct sock *sk,
5224                                        struct netlink_callback *cb)
5225 {
5226         struct netlink_ext_ack *extack = cb->extack;
5227         struct nlattr *tb[IFA_MAX+1];
5228         struct ifaddrmsg *ifm;
5229         int err, i;
5230
5231         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5232                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5233                 return -EINVAL;
5234         }
5235
5236         ifm = nlmsg_data(nlh);
5237         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5238                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5239                 return -EINVAL;
5240         }
5241
5242         fillargs->ifindex = ifm->ifa_index;
5243         if (fillargs->ifindex) {
5244                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5245                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5246         }
5247
5248         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5249                                             ifa_ipv6_policy, extack);
5250         if (err < 0)
5251                 return err;
5252
5253         for (i = 0; i <= IFA_MAX; ++i) {
5254                 if (!tb[i])
5255                         continue;
5256
5257                 if (i == IFA_TARGET_NETNSID) {
5258                         struct net *net;
5259
5260                         fillargs->netnsid = nla_get_s32(tb[i]);
5261                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5262                         if (IS_ERR(net)) {
5263                                 fillargs->netnsid = -1;
5264                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5265                                 return PTR_ERR(net);
5266                         }
5267                         *tgt_net = net;
5268                 } else {
5269                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5270                         return -EINVAL;
5271                 }
5272         }
5273
5274         return 0;
5275 }
5276
5277 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5278                            enum addr_type_t type)
5279 {
5280         const struct nlmsghdr *nlh = cb->nlh;
5281         struct inet6_fill_args fillargs = {
5282                 .portid = NETLINK_CB(cb->skb).portid,
5283                 .seq = cb->nlh->nlmsg_seq,
5284                 .flags = NLM_F_MULTI,
5285                 .netnsid = -1,
5286                 .type = type,
5287         };
5288         struct net *tgt_net = sock_net(skb->sk);
5289         int idx, s_idx, s_ip_idx;
5290         int h, s_h;
5291         struct net_device *dev;
5292         struct inet6_dev *idev;
5293         struct hlist_head *head;
5294         int err = 0;
5295
5296         s_h = cb->args[0];
5297         s_idx = idx = cb->args[1];
5298         s_ip_idx = cb->args[2];
5299
5300         if (cb->strict_check) {
5301                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5302                                                   skb->sk, cb);
5303                 if (err < 0)
5304                         goto put_tgt_net;
5305
5306                 err = 0;
5307                 if (fillargs.ifindex) {
5308                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5309                         if (!dev) {
5310                                 err = -ENODEV;
5311                                 goto put_tgt_net;
5312                         }
5313                         idev = __in6_dev_get(dev);
5314                         if (idev) {
5315                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5316                                                      &fillargs);
5317                                 if (err > 0)
5318                                         err = 0;
5319                         }
5320                         goto put_tgt_net;
5321                 }
5322         }
5323
5324         rcu_read_lock();
5325         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5326         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5327                 idx = 0;
5328                 head = &tgt_net->dev_index_head[h];
5329                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5330                         if (idx < s_idx)
5331                                 goto cont;
5332                         if (h > s_h || idx > s_idx)
5333                                 s_ip_idx = 0;
5334                         idev = __in6_dev_get(dev);
5335                         if (!idev)
5336                                 goto cont;
5337
5338                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5339                                            &fillargs) < 0)
5340                                 goto done;
5341 cont:
5342                         idx++;
5343                 }
5344         }
5345 done:
5346         rcu_read_unlock();
5347         cb->args[0] = h;
5348         cb->args[1] = idx;
5349 put_tgt_net:
5350         if (fillargs.netnsid >= 0)
5351                 put_net(tgt_net);
5352
5353         return skb->len ? : err;
5354 }
5355
5356 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5357 {
5358         enum addr_type_t type = UNICAST_ADDR;
5359
5360         return inet6_dump_addr(skb, cb, type);
5361 }
5362
5363 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5364 {
5365         enum addr_type_t type = MULTICAST_ADDR;
5366
5367         return inet6_dump_addr(skb, cb, type);
5368 }
5369
5370
5371 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5372 {
5373         enum addr_type_t type = ANYCAST_ADDR;
5374
5375         return inet6_dump_addr(skb, cb, type);
5376 }
5377
5378 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5379                                        const struct nlmsghdr *nlh,
5380                                        struct nlattr **tb,
5381                                        struct netlink_ext_ack *extack)
5382 {
5383         struct ifaddrmsg *ifm;
5384         int i, err;
5385
5386         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5387                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5388                 return -EINVAL;
5389         }
5390
5391         if (!netlink_strict_get_check(skb))
5392                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5393                                               ifa_ipv6_policy, extack);
5394
5395         ifm = nlmsg_data(nlh);
5396         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5397                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5398                 return -EINVAL;
5399         }
5400
5401         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5402                                             ifa_ipv6_policy, extack);
5403         if (err)
5404                 return err;
5405
5406         for (i = 0; i <= IFA_MAX; i++) {
5407                 if (!tb[i])
5408                         continue;
5409
5410                 switch (i) {
5411                 case IFA_TARGET_NETNSID:
5412                 case IFA_ADDRESS:
5413                 case IFA_LOCAL:
5414                         break;
5415                 default:
5416                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5417                         return -EINVAL;
5418                 }
5419         }
5420
5421         return 0;
5422 }
5423
5424 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5425                              struct netlink_ext_ack *extack)
5426 {
5427         struct net *tgt_net = sock_net(in_skb->sk);
5428         struct inet6_fill_args fillargs = {
5429                 .portid = NETLINK_CB(in_skb).portid,
5430                 .seq = nlh->nlmsg_seq,
5431                 .event = RTM_NEWADDR,
5432                 .flags = 0,
5433                 .netnsid = -1,
5434         };
5435         struct ifaddrmsg *ifm;
5436         struct nlattr *tb[IFA_MAX+1];
5437         struct in6_addr *addr = NULL, *peer;
5438         struct net_device *dev = NULL;
5439         struct inet6_ifaddr *ifa;
5440         struct sk_buff *skb;
5441         int err;
5442
5443         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5444         if (err < 0)
5445                 return err;
5446
5447         if (tb[IFA_TARGET_NETNSID]) {
5448                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5449
5450                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5451                                                   fillargs.netnsid);
5452                 if (IS_ERR(tgt_net))
5453                         return PTR_ERR(tgt_net);
5454         }
5455
5456         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5457         if (!addr)
5458                 return -EINVAL;
5459
5460         ifm = nlmsg_data(nlh);
5461         if (ifm->ifa_index)
5462                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5463
5464         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5465         if (!ifa) {
5466                 err = -EADDRNOTAVAIL;
5467                 goto errout;
5468         }
5469
5470         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5471         if (!skb) {
5472                 err = -ENOBUFS;
5473                 goto errout_ifa;
5474         }
5475
5476         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5477         if (err < 0) {
5478                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5479                 WARN_ON(err == -EMSGSIZE);
5480                 kfree_skb(skb);
5481                 goto errout_ifa;
5482         }
5483         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5484 errout_ifa:
5485         in6_ifa_put(ifa);
5486 errout:
5487         dev_put(dev);
5488         if (fillargs.netnsid >= 0)
5489                 put_net(tgt_net);
5490
5491         return err;
5492 }
5493
5494 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5495 {
5496         struct sk_buff *skb;
5497         struct net *net = dev_net(ifa->idev->dev);
5498         struct inet6_fill_args fillargs = {
5499                 .portid = 0,
5500                 .seq = 0,
5501                 .event = event,
5502                 .flags = 0,
5503                 .netnsid = -1,
5504         };
5505         int err = -ENOBUFS;
5506
5507         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5508         if (!skb)
5509                 goto errout;
5510
5511         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5512         if (err < 0) {
5513                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5514                 WARN_ON(err == -EMSGSIZE);
5515                 kfree_skb(skb);
5516                 goto errout;
5517         }
5518         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5519         return;
5520 errout:
5521         if (err < 0)
5522                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5523 }
5524
5525 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5526                                 __s32 *array, int bytes)
5527 {
5528         BUG_ON(bytes < (DEVCONF_MAX * 4));
5529
5530         memset(array, 0, bytes);
5531         array[DEVCONF_FORWARDING] = cnf->forwarding;
5532         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5533         array[DEVCONF_MTU6] = cnf->mtu6;
5534         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5535         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5536         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5537         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5538         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5539         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5540                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5541         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5542                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5543         array[DEVCONF_RTR_SOLICIT_DELAY] =
5544                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5545         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5546         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5547                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5548         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5549                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5550         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5551         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5552         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5553         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5554         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5555         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5556         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5557         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5558         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5559         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5560 #ifdef CONFIG_IPV6_ROUTER_PREF
5561         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5562         array[DEVCONF_RTR_PROBE_INTERVAL] =
5563                 jiffies_to_msecs(cnf->rtr_probe_interval);
5564 #ifdef CONFIG_IPV6_ROUTE_INFO
5565         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5566         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5567 #endif
5568 #endif
5569         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5570         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5571 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5572         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5573         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5574 #endif
5575 #ifdef CONFIG_IPV6_MROUTE
5576         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5577 #endif
5578         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5579         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5580         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5581         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5582         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5583         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5584         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5585         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5586         /* we omit DEVCONF_STABLE_SECRET for now */
5587         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5588         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5589         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5590         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5591         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5592 #ifdef CONFIG_IPV6_SEG6_HMAC
5593         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5594 #endif
5595         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5596         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5597         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5598         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5599         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5600         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5601         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5602         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5603         array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5604         array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5605 }
5606
5607 static inline size_t inet6_ifla6_size(void)
5608 {
5609         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5610              + nla_total_size(sizeof(struct ifla_cacheinfo))
5611              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5612              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5613              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5614              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5615              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5616              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5617              + 0;
5618 }
5619
5620 static inline size_t inet6_if_nlmsg_size(void)
5621 {
5622         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5623                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5624                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5625                + nla_total_size(4) /* IFLA_MTU */
5626                + nla_total_size(4) /* IFLA_LINK */
5627                + nla_total_size(1) /* IFLA_OPERSTATE */
5628                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5629 }
5630
5631 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5632                                         int bytes)
5633 {
5634         int i;
5635         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5636         BUG_ON(pad < 0);
5637
5638         /* Use put_unaligned() because stats may not be aligned for u64. */
5639         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5640         for (i = 1; i < ICMP6_MIB_MAX; i++)
5641                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5642
5643         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5644 }
5645
5646 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5647                                         int bytes, size_t syncpoff)
5648 {
5649         int i, c;
5650         u64 buff[IPSTATS_MIB_MAX];
5651         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5652
5653         BUG_ON(pad < 0);
5654
5655         memset(buff, 0, sizeof(buff));
5656         buff[0] = IPSTATS_MIB_MAX;
5657
5658         for_each_possible_cpu(c) {
5659                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5660                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5661         }
5662
5663         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5664         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5665 }
5666
5667 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5668                              int bytes)
5669 {
5670         switch (attrtype) {
5671         case IFLA_INET6_STATS:
5672                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5673                                      offsetof(struct ipstats_mib, syncp));
5674                 break;
5675         case IFLA_INET6_ICMP6STATS:
5676                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5677                 break;
5678         }
5679 }
5680
5681 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5682                                   u32 ext_filter_mask)
5683 {
5684         struct nlattr *nla;
5685         struct ifla_cacheinfo ci;
5686
5687         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5688                 goto nla_put_failure;
5689         ci.max_reasm_len = IPV6_MAXPLEN;
5690         ci.tstamp = cstamp_delta(idev->tstamp);
5691         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5692         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5693         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5694                 goto nla_put_failure;
5695         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5696         if (!nla)
5697                 goto nla_put_failure;
5698         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5699
5700         /* XXX - MC not implemented */
5701
5702         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5703                 return 0;
5704
5705         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5706         if (!nla)
5707                 goto nla_put_failure;
5708         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5709
5710         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5711         if (!nla)
5712                 goto nla_put_failure;
5713         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5714
5715         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5716         if (!nla)
5717                 goto nla_put_failure;
5718         read_lock_bh(&idev->lock);
5719         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5720         read_unlock_bh(&idev->lock);
5721
5722         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5723                 goto nla_put_failure;
5724
5725         if (idev->ra_mtu &&
5726             nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5727                 goto nla_put_failure;
5728
5729         return 0;
5730
5731 nla_put_failure:
5732         return -EMSGSIZE;
5733 }
5734
5735 static size_t inet6_get_link_af_size(const struct net_device *dev,
5736                                      u32 ext_filter_mask)
5737 {
5738         if (!__in6_dev_get(dev))
5739                 return 0;
5740
5741         return inet6_ifla6_size();
5742 }
5743
5744 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5745                               u32 ext_filter_mask)
5746 {
5747         struct inet6_dev *idev = __in6_dev_get(dev);
5748
5749         if (!idev)
5750                 return -ENODATA;
5751
5752         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5753                 return -EMSGSIZE;
5754
5755         return 0;
5756 }
5757
5758 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5759                              struct netlink_ext_ack *extack)
5760 {
5761         struct inet6_ifaddr *ifp;
5762         struct net_device *dev = idev->dev;
5763         bool clear_token, update_rs = false;
5764         struct in6_addr ll_addr;
5765
5766         ASSERT_RTNL();
5767
5768         if (!token)
5769                 return -EINVAL;
5770
5771         if (dev->flags & IFF_LOOPBACK) {
5772                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5773                 return -EINVAL;
5774         }
5775
5776         if (dev->flags & IFF_NOARP) {
5777                 NL_SET_ERR_MSG_MOD(extack,
5778                                    "Device does not do neighbour discovery");
5779                 return -EINVAL;
5780         }
5781
5782         if (!ipv6_accept_ra(idev)) {
5783                 NL_SET_ERR_MSG_MOD(extack,
5784                                    "Router advertisement is disabled on device");
5785                 return -EINVAL;
5786         }
5787
5788         if (idev->cnf.rtr_solicits == 0) {
5789                 NL_SET_ERR_MSG(extack,
5790                                "Router solicitation is disabled on device");
5791                 return -EINVAL;
5792         }
5793
5794         write_lock_bh(&idev->lock);
5795
5796         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5797         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5798
5799         write_unlock_bh(&idev->lock);
5800
5801         clear_token = ipv6_addr_any(token);
5802         if (clear_token)
5803                 goto update_lft;
5804
5805         if (!idev->dead && (idev->if_flags & IF_READY) &&
5806             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5807                              IFA_F_OPTIMISTIC)) {
5808                 /* If we're not ready, then normal ifup will take care
5809                  * of this. Otherwise, we need to request our rs here.
5810                  */
5811                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5812                 update_rs = true;
5813         }
5814
5815 update_lft:
5816         write_lock_bh(&idev->lock);
5817
5818         if (update_rs) {
5819                 idev->if_flags |= IF_RS_SENT;
5820                 idev->rs_interval = rfc3315_s14_backoff_init(
5821                         idev->cnf.rtr_solicit_interval);
5822                 idev->rs_probes = 1;
5823                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5824         }
5825
5826         /* Well, that's kinda nasty ... */
5827         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5828                 spin_lock(&ifp->lock);
5829                 if (ifp->tokenized) {
5830                         ifp->valid_lft = 0;
5831                         ifp->prefered_lft = 0;
5832                 }
5833                 spin_unlock(&ifp->lock);
5834         }
5835
5836         write_unlock_bh(&idev->lock);
5837         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5838         addrconf_verify_rtnl(dev_net(dev));
5839         return 0;
5840 }
5841
5842 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5843         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5844         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5845         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5846                                             .reject_message =
5847                                                 "IFLA_INET6_RA_MTU can not be set" },
5848 };
5849
5850 static int check_addr_gen_mode(int mode)
5851 {
5852         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5853             mode != IN6_ADDR_GEN_MODE_NONE &&
5854             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5855             mode != IN6_ADDR_GEN_MODE_RANDOM)
5856                 return -EINVAL;
5857         return 1;
5858 }
5859
5860 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5861                                 int mode)
5862 {
5863         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5864             !idev->cnf.stable_secret.initialized &&
5865             !net->ipv6.devconf_dflt->stable_secret.initialized)
5866                 return -EINVAL;
5867         return 1;
5868 }
5869
5870 static int inet6_validate_link_af(const struct net_device *dev,
5871                                   const struct nlattr *nla,
5872                                   struct netlink_ext_ack *extack)
5873 {
5874         struct nlattr *tb[IFLA_INET6_MAX + 1];
5875         struct inet6_dev *idev = NULL;
5876         int err;
5877
5878         if (dev) {
5879                 idev = __in6_dev_get(dev);
5880                 if (!idev)
5881                         return -EAFNOSUPPORT;
5882         }
5883
5884         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5885                                           inet6_af_policy, extack);
5886         if (err)
5887                 return err;
5888
5889         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5890                 return -EINVAL;
5891
5892         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5893                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5894
5895                 if (check_addr_gen_mode(mode) < 0)
5896                         return -EINVAL;
5897                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5898                         return -EINVAL;
5899         }
5900
5901         return 0;
5902 }
5903
5904 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5905                              struct netlink_ext_ack *extack)
5906 {
5907         struct inet6_dev *idev = __in6_dev_get(dev);
5908         struct nlattr *tb[IFLA_INET6_MAX + 1];
5909         int err;
5910
5911         if (!idev)
5912                 return -EAFNOSUPPORT;
5913
5914         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5915                 return -EINVAL;
5916
5917         if (tb[IFLA_INET6_TOKEN]) {
5918                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5919                                         extack);
5920                 if (err)
5921                         return err;
5922         }
5923
5924         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5925                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5926
5927                 idev->cnf.addr_gen_mode = mode;
5928         }
5929
5930         return 0;
5931 }
5932
5933 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5934                              u32 portid, u32 seq, int event, unsigned int flags)
5935 {
5936         struct net_device *dev = idev->dev;
5937         struct ifinfomsg *hdr;
5938         struct nlmsghdr *nlh;
5939         void *protoinfo;
5940
5941         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5942         if (!nlh)
5943                 return -EMSGSIZE;
5944
5945         hdr = nlmsg_data(nlh);
5946         hdr->ifi_family = AF_INET6;
5947         hdr->__ifi_pad = 0;
5948         hdr->ifi_type = dev->type;
5949         hdr->ifi_index = dev->ifindex;
5950         hdr->ifi_flags = dev_get_flags(dev);
5951         hdr->ifi_change = 0;
5952
5953         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5954             (dev->addr_len &&
5955              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5956             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5957             (dev->ifindex != dev_get_iflink(dev) &&
5958              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5959             nla_put_u8(skb, IFLA_OPERSTATE,
5960                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5961                 goto nla_put_failure;
5962         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5963         if (!protoinfo)
5964                 goto nla_put_failure;
5965
5966         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5967                 goto nla_put_failure;
5968
5969         nla_nest_end(skb, protoinfo);
5970         nlmsg_end(skb, nlh);
5971         return 0;
5972
5973 nla_put_failure:
5974         nlmsg_cancel(skb, nlh);
5975         return -EMSGSIZE;
5976 }
5977
5978 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5979                                    struct netlink_ext_ack *extack)
5980 {
5981         struct ifinfomsg *ifm;
5982
5983         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5984                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5985                 return -EINVAL;
5986         }
5987
5988         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5989                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5990                 return -EINVAL;
5991         }
5992
5993         ifm = nlmsg_data(nlh);
5994         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5995             ifm->ifi_change || ifm->ifi_index) {
5996                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5997                 return -EINVAL;
5998         }
5999
6000         return 0;
6001 }
6002
6003 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6004 {
6005         struct net *net = sock_net(skb->sk);
6006         int h, s_h;
6007         int idx = 0, s_idx;
6008         struct net_device *dev;
6009         struct inet6_dev *idev;
6010         struct hlist_head *head;
6011
6012         /* only requests using strict checking can pass data to
6013          * influence the dump
6014          */
6015         if (cb->strict_check) {
6016                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6017
6018                 if (err < 0)
6019                         return err;
6020         }
6021
6022         s_h = cb->args[0];
6023         s_idx = cb->args[1];
6024
6025         rcu_read_lock();
6026         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6027                 idx = 0;
6028                 head = &net->dev_index_head[h];
6029                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6030                         if (idx < s_idx)
6031                                 goto cont;
6032                         idev = __in6_dev_get(dev);
6033                         if (!idev)
6034                                 goto cont;
6035                         if (inet6_fill_ifinfo(skb, idev,
6036                                               NETLINK_CB(cb->skb).portid,
6037                                               cb->nlh->nlmsg_seq,
6038                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
6039                                 goto out;
6040 cont:
6041                         idx++;
6042                 }
6043         }
6044 out:
6045         rcu_read_unlock();
6046         cb->args[1] = idx;
6047         cb->args[0] = h;
6048
6049         return skb->len;
6050 }
6051
6052 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6053 {
6054         struct sk_buff *skb;
6055         struct net *net = dev_net(idev->dev);
6056         int err = -ENOBUFS;
6057
6058         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6059         if (!skb)
6060                 goto errout;
6061
6062         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6063         if (err < 0) {
6064                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6065                 WARN_ON(err == -EMSGSIZE);
6066                 kfree_skb(skb);
6067                 goto errout;
6068         }
6069         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6070         return;
6071 errout:
6072         if (err < 0)
6073                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6074 }
6075
6076 static inline size_t inet6_prefix_nlmsg_size(void)
6077 {
6078         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6079                + nla_total_size(sizeof(struct in6_addr))
6080                + nla_total_size(sizeof(struct prefix_cacheinfo));
6081 }
6082
6083 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6084                              struct prefix_info *pinfo, u32 portid, u32 seq,
6085                              int event, unsigned int flags)
6086 {
6087         struct prefixmsg *pmsg;
6088         struct nlmsghdr *nlh;
6089         struct prefix_cacheinfo ci;
6090
6091         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6092         if (!nlh)
6093                 return -EMSGSIZE;
6094
6095         pmsg = nlmsg_data(nlh);
6096         pmsg->prefix_family = AF_INET6;
6097         pmsg->prefix_pad1 = 0;
6098         pmsg->prefix_pad2 = 0;
6099         pmsg->prefix_ifindex = idev->dev->ifindex;
6100         pmsg->prefix_len = pinfo->prefix_len;
6101         pmsg->prefix_type = pinfo->type;
6102         pmsg->prefix_pad3 = 0;
6103         pmsg->prefix_flags = 0;
6104         if (pinfo->onlink)
6105                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6106         if (pinfo->autoconf)
6107                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6108
6109         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6110                 goto nla_put_failure;
6111         ci.preferred_time = ntohl(pinfo->prefered);
6112         ci.valid_time = ntohl(pinfo->valid);
6113         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6114                 goto nla_put_failure;
6115         nlmsg_end(skb, nlh);
6116         return 0;
6117
6118 nla_put_failure:
6119         nlmsg_cancel(skb, nlh);
6120         return -EMSGSIZE;
6121 }
6122
6123 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6124                          struct prefix_info *pinfo)
6125 {
6126         struct sk_buff *skb;
6127         struct net *net = dev_net(idev->dev);
6128         int err = -ENOBUFS;
6129
6130         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6131         if (!skb)
6132                 goto errout;
6133
6134         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6135         if (err < 0) {
6136                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6137                 WARN_ON(err == -EMSGSIZE);
6138                 kfree_skb(skb);
6139                 goto errout;
6140         }
6141         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6142         return;
6143 errout:
6144         if (err < 0)
6145                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6146 }
6147
6148 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6149 {
6150         struct net *net = dev_net(ifp->idev->dev);
6151
6152         if (event)
6153                 ASSERT_RTNL();
6154
6155         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6156
6157         switch (event) {
6158         case RTM_NEWADDR:
6159                 /*
6160                  * If the address was optimistic we inserted the route at the
6161                  * start of our DAD process, so we don't need to do it again.
6162                  * If the device was taken down in the middle of the DAD
6163                  * cycle there is a race where we could get here without a
6164                  * host route, so nothing to insert. That will be fixed when
6165                  * the device is brought up.
6166                  */
6167                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6168                         ip6_ins_rt(net, ifp->rt);
6169                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6170                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6171                                 &ifp->addr, ifp->idev->dev->name);
6172                 }
6173
6174                 if (ifp->idev->cnf.forwarding)
6175                         addrconf_join_anycast(ifp);
6176                 if (!ipv6_addr_any(&ifp->peer_addr))
6177                         addrconf_prefix_route(&ifp->peer_addr, 128,
6178                                               ifp->rt_priority, ifp->idev->dev,
6179                                               0, 0, GFP_ATOMIC);
6180                 break;
6181         case RTM_DELADDR:
6182                 if (ifp->idev->cnf.forwarding)
6183                         addrconf_leave_anycast(ifp);
6184                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6185                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6186                         struct fib6_info *rt;
6187
6188                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6189                                                        ifp->idev->dev, 0, 0,
6190                                                        false);
6191                         if (rt)
6192                                 ip6_del_rt(net, rt, false);
6193                 }
6194                 if (ifp->rt) {
6195                         ip6_del_rt(net, ifp->rt, false);
6196                         ifp->rt = NULL;
6197                 }
6198                 rt_genid_bump_ipv6(net);
6199                 break;
6200         }
6201         atomic_inc(&net->ipv6.dev_addr_genid);
6202 }
6203
6204 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6205 {
6206         if (likely(ifp->idev->dead == 0))
6207                 __ipv6_ifa_notify(event, ifp);
6208 }
6209
6210 #ifdef CONFIG_SYSCTL
6211
6212 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6213                 void *buffer, size_t *lenp, loff_t *ppos)
6214 {
6215         int *valp = ctl->data;
6216         int val = *valp;
6217         loff_t pos = *ppos;
6218         struct ctl_table lctl;
6219         int ret;
6220
6221         /*
6222          * ctl->data points to idev->cnf.forwarding, we should
6223          * not modify it until we get the rtnl lock.
6224          */
6225         lctl = *ctl;
6226         lctl.data = &val;
6227
6228         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6229
6230         if (write)
6231                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6232         if (ret)
6233                 *ppos = pos;
6234         return ret;
6235 }
6236
6237 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6238                 void *buffer, size_t *lenp, loff_t *ppos)
6239 {
6240         struct inet6_dev *idev = ctl->extra1;
6241         int min_mtu = IPV6_MIN_MTU;
6242         struct ctl_table lctl;
6243
6244         lctl = *ctl;
6245         lctl.extra1 = &min_mtu;
6246         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6247
6248         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6249 }
6250
6251 static void dev_disable_change(struct inet6_dev *idev)
6252 {
6253         struct netdev_notifier_info info;
6254
6255         if (!idev || !idev->dev)
6256                 return;
6257
6258         netdev_notifier_info_init(&info, idev->dev);
6259         if (idev->cnf.disable_ipv6)
6260                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6261         else
6262                 addrconf_notify(NULL, NETDEV_UP, &info);
6263 }
6264
6265 static void addrconf_disable_change(struct net *net, __s32 newf)
6266 {
6267         struct net_device *dev;
6268         struct inet6_dev *idev;
6269
6270         for_each_netdev(net, dev) {
6271                 idev = __in6_dev_get(dev);
6272                 if (idev) {
6273                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6274                         idev->cnf.disable_ipv6 = newf;
6275                         if (changed)
6276                                 dev_disable_change(idev);
6277                 }
6278         }
6279 }
6280
6281 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6282 {
6283         struct net *net;
6284         int old;
6285
6286         if (!rtnl_trylock())
6287                 return restart_syscall();
6288
6289         net = (struct net *)table->extra2;
6290         old = *p;
6291         *p = newf;
6292
6293         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6294                 rtnl_unlock();
6295                 return 0;
6296         }
6297
6298         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6299                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6300                 addrconf_disable_change(net, newf);
6301         } else if ((!newf) ^ (!old))
6302                 dev_disable_change((struct inet6_dev *)table->extra1);
6303
6304         rtnl_unlock();
6305         return 0;
6306 }
6307
6308 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6309                 void *buffer, size_t *lenp, loff_t *ppos)
6310 {
6311         int *valp = ctl->data;
6312         int val = *valp;
6313         loff_t pos = *ppos;
6314         struct ctl_table lctl;
6315         int ret;
6316
6317         /*
6318          * ctl->data points to idev->cnf.disable_ipv6, we should
6319          * not modify it until we get the rtnl lock.
6320          */
6321         lctl = *ctl;
6322         lctl.data = &val;
6323
6324         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6325
6326         if (write)
6327                 ret = addrconf_disable_ipv6(ctl, valp, val);
6328         if (ret)
6329                 *ppos = pos;
6330         return ret;
6331 }
6332
6333 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6334                 void *buffer, size_t *lenp, loff_t *ppos)
6335 {
6336         int *valp = ctl->data;
6337         int ret;
6338         int old, new;
6339
6340         old = *valp;
6341         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6342         new = *valp;
6343
6344         if (write && old != new) {
6345                 struct net *net = ctl->extra2;
6346
6347                 if (!rtnl_trylock())
6348                         return restart_syscall();
6349
6350                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6351                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6352                                                      NETCONFA_PROXY_NEIGH,
6353                                                      NETCONFA_IFINDEX_DEFAULT,
6354                                                      net->ipv6.devconf_dflt);
6355                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6356                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6357                                                      NETCONFA_PROXY_NEIGH,
6358                                                      NETCONFA_IFINDEX_ALL,
6359                                                      net->ipv6.devconf_all);
6360                 else {
6361                         struct inet6_dev *idev = ctl->extra1;
6362
6363                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6364                                                      NETCONFA_PROXY_NEIGH,
6365                                                      idev->dev->ifindex,
6366                                                      &idev->cnf);
6367                 }
6368                 rtnl_unlock();
6369         }
6370
6371         return ret;
6372 }
6373
6374 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6375                                          void *buffer, size_t *lenp,
6376                                          loff_t *ppos)
6377 {
6378         int ret = 0;
6379         u32 new_val;
6380         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6381         struct net *net = (struct net *)ctl->extra2;
6382         struct ctl_table tmp = {
6383                 .data = &new_val,
6384                 .maxlen = sizeof(new_val),
6385                 .mode = ctl->mode,
6386         };
6387
6388         if (!rtnl_trylock())
6389                 return restart_syscall();
6390
6391         new_val = *((u32 *)ctl->data);
6392
6393         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6394         if (ret != 0)
6395                 goto out;
6396
6397         if (write) {
6398                 if (check_addr_gen_mode(new_val) < 0) {
6399                         ret = -EINVAL;
6400                         goto out;
6401                 }
6402
6403                 if (idev) {
6404                         if (check_stable_privacy(idev, net, new_val) < 0) {
6405                                 ret = -EINVAL;
6406                                 goto out;
6407                         }
6408
6409                         if (idev->cnf.addr_gen_mode != new_val) {
6410                                 idev->cnf.addr_gen_mode = new_val;
6411                                 addrconf_init_auto_addrs(idev->dev);
6412                         }
6413                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6414                         struct net_device *dev;
6415
6416                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6417                         for_each_netdev(net, dev) {
6418                                 idev = __in6_dev_get(dev);
6419                                 if (idev &&
6420                                     idev->cnf.addr_gen_mode != new_val) {
6421                                         idev->cnf.addr_gen_mode = new_val;
6422                                         addrconf_init_auto_addrs(idev->dev);
6423                                 }
6424                         }
6425                 }
6426
6427                 *((u32 *)ctl->data) = new_val;
6428         }
6429
6430 out:
6431         rtnl_unlock();
6432
6433         return ret;
6434 }
6435
6436 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6437                                          void *buffer, size_t *lenp,
6438                                          loff_t *ppos)
6439 {
6440         int err;
6441         struct in6_addr addr;
6442         char str[IPV6_MAX_STRLEN];
6443         struct ctl_table lctl = *ctl;
6444         struct net *net = ctl->extra2;
6445         struct ipv6_stable_secret *secret = ctl->data;
6446
6447         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6448                 return -EIO;
6449
6450         lctl.maxlen = IPV6_MAX_STRLEN;
6451         lctl.data = str;
6452
6453         if (!rtnl_trylock())
6454                 return restart_syscall();
6455
6456         if (!write && !secret->initialized) {
6457                 err = -EIO;
6458                 goto out;
6459         }
6460
6461         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6462         if (err >= sizeof(str)) {
6463                 err = -EIO;
6464                 goto out;
6465         }
6466
6467         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6468         if (err || !write)
6469                 goto out;
6470
6471         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6472                 err = -EIO;
6473                 goto out;
6474         }
6475
6476         secret->initialized = true;
6477         secret->secret = addr;
6478
6479         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6480                 struct net_device *dev;
6481
6482                 for_each_netdev(net, dev) {
6483                         struct inet6_dev *idev = __in6_dev_get(dev);
6484
6485                         if (idev) {
6486                                 idev->cnf.addr_gen_mode =
6487                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6488                         }
6489                 }
6490         } else {
6491                 struct inet6_dev *idev = ctl->extra1;
6492
6493                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6494         }
6495
6496 out:
6497         rtnl_unlock();
6498
6499         return err;
6500 }
6501
6502 static
6503 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6504                                                 int write, void *buffer,
6505                                                 size_t *lenp,
6506                                                 loff_t *ppos)
6507 {
6508         int *valp = ctl->data;
6509         int val = *valp;
6510         loff_t pos = *ppos;
6511         struct ctl_table lctl;
6512         int ret;
6513
6514         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6515          * we should not modify it until we get the rtnl lock.
6516          */
6517         lctl = *ctl;
6518         lctl.data = &val;
6519
6520         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6521
6522         if (write)
6523                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6524         if (ret)
6525                 *ppos = pos;
6526         return ret;
6527 }
6528
6529 static
6530 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6531 {
6532         if (rt) {
6533                 if (action)
6534                         rt->dst.flags |= DST_NOPOLICY;
6535                 else
6536                         rt->dst.flags &= ~DST_NOPOLICY;
6537         }
6538 }
6539
6540 static
6541 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6542 {
6543         struct inet6_ifaddr *ifa;
6544
6545         read_lock_bh(&idev->lock);
6546         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6547                 spin_lock(&ifa->lock);
6548                 if (ifa->rt) {
6549                         /* host routes only use builtin fib6_nh */
6550                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6551                         int cpu;
6552
6553                         rcu_read_lock();
6554                         ifa->rt->dst_nopolicy = val ? true : false;
6555                         if (nh->rt6i_pcpu) {
6556                                 for_each_possible_cpu(cpu) {
6557                                         struct rt6_info **rtp;
6558
6559                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6560                                         addrconf_set_nopolicy(*rtp, val);
6561                                 }
6562                         }
6563                         rcu_read_unlock();
6564                 }
6565                 spin_unlock(&ifa->lock);
6566         }
6567         read_unlock_bh(&idev->lock);
6568 }
6569
6570 static
6571 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6572 {
6573         struct inet6_dev *idev;
6574         struct net *net;
6575
6576         if (!rtnl_trylock())
6577                 return restart_syscall();
6578
6579         *valp = val;
6580
6581         net = (struct net *)ctl->extra2;
6582         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6583                 rtnl_unlock();
6584                 return 0;
6585         }
6586
6587         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6588                 struct net_device *dev;
6589
6590                 for_each_netdev(net, dev) {
6591                         idev = __in6_dev_get(dev);
6592                         if (idev)
6593                                 addrconf_disable_policy_idev(idev, val);
6594                 }
6595         } else {
6596                 idev = (struct inet6_dev *)ctl->extra1;
6597                 addrconf_disable_policy_idev(idev, val);
6598         }
6599
6600         rtnl_unlock();
6601         return 0;
6602 }
6603
6604 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6605                                    void *buffer, size_t *lenp, loff_t *ppos)
6606 {
6607         int *valp = ctl->data;
6608         int val = *valp;
6609         loff_t pos = *ppos;
6610         struct ctl_table lctl;
6611         int ret;
6612
6613         lctl = *ctl;
6614         lctl.data = &val;
6615         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6616
6617         if (write && (*valp != val))
6618                 ret = addrconf_disable_policy(ctl, valp, val);
6619
6620         if (ret)
6621                 *ppos = pos;
6622
6623         return ret;
6624 }
6625
6626 static int minus_one = -1;
6627 static const int two_five_five = 255;
6628 static u32 ioam6_if_id_max = U16_MAX;
6629
6630 static const struct ctl_table addrconf_sysctl[] = {
6631         {
6632                 .procname       = "forwarding",
6633                 .data           = &ipv6_devconf.forwarding,
6634                 .maxlen         = sizeof(int),
6635                 .mode           = 0644,
6636                 .proc_handler   = addrconf_sysctl_forward,
6637         },
6638         {
6639                 .procname       = "hop_limit",
6640                 .data           = &ipv6_devconf.hop_limit,
6641                 .maxlen         = sizeof(int),
6642                 .mode           = 0644,
6643                 .proc_handler   = proc_dointvec_minmax,
6644                 .extra1         = (void *)SYSCTL_ONE,
6645                 .extra2         = (void *)&two_five_five,
6646         },
6647         {
6648                 .procname       = "mtu",
6649                 .data           = &ipv6_devconf.mtu6,
6650                 .maxlen         = sizeof(int),
6651                 .mode           = 0644,
6652                 .proc_handler   = addrconf_sysctl_mtu,
6653         },
6654         {
6655                 .procname       = "accept_ra",
6656                 .data           = &ipv6_devconf.accept_ra,
6657                 .maxlen         = sizeof(int),
6658                 .mode           = 0644,
6659                 .proc_handler   = proc_dointvec,
6660         },
6661         {
6662                 .procname       = "accept_redirects",
6663                 .data           = &ipv6_devconf.accept_redirects,
6664                 .maxlen         = sizeof(int),
6665                 .mode           = 0644,
6666                 .proc_handler   = proc_dointvec,
6667         },
6668         {
6669                 .procname       = "autoconf",
6670                 .data           = &ipv6_devconf.autoconf,
6671                 .maxlen         = sizeof(int),
6672                 .mode           = 0644,
6673                 .proc_handler   = proc_dointvec,
6674         },
6675         {
6676                 .procname       = "dad_transmits",
6677                 .data           = &ipv6_devconf.dad_transmits,
6678                 .maxlen         = sizeof(int),
6679                 .mode           = 0644,
6680                 .proc_handler   = proc_dointvec,
6681         },
6682         {
6683                 .procname       = "router_solicitations",
6684                 .data           = &ipv6_devconf.rtr_solicits,
6685                 .maxlen         = sizeof(int),
6686                 .mode           = 0644,
6687                 .proc_handler   = proc_dointvec_minmax,
6688                 .extra1         = &minus_one,
6689         },
6690         {
6691                 .procname       = "router_solicitation_interval",
6692                 .data           = &ipv6_devconf.rtr_solicit_interval,
6693                 .maxlen         = sizeof(int),
6694                 .mode           = 0644,
6695                 .proc_handler   = proc_dointvec_jiffies,
6696         },
6697         {
6698                 .procname       = "router_solicitation_max_interval",
6699                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6700                 .maxlen         = sizeof(int),
6701                 .mode           = 0644,
6702                 .proc_handler   = proc_dointvec_jiffies,
6703         },
6704         {
6705                 .procname       = "router_solicitation_delay",
6706                 .data           = &ipv6_devconf.rtr_solicit_delay,
6707                 .maxlen         = sizeof(int),
6708                 .mode           = 0644,
6709                 .proc_handler   = proc_dointvec_jiffies,
6710         },
6711         {
6712                 .procname       = "force_mld_version",
6713                 .data           = &ipv6_devconf.force_mld_version,
6714                 .maxlen         = sizeof(int),
6715                 .mode           = 0644,
6716                 .proc_handler   = proc_dointvec,
6717         },
6718         {
6719                 .procname       = "mldv1_unsolicited_report_interval",
6720                 .data           =
6721                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6722                 .maxlen         = sizeof(int),
6723                 .mode           = 0644,
6724                 .proc_handler   = proc_dointvec_ms_jiffies,
6725         },
6726         {
6727                 .procname       = "mldv2_unsolicited_report_interval",
6728                 .data           =
6729                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6730                 .maxlen         = sizeof(int),
6731                 .mode           = 0644,
6732                 .proc_handler   = proc_dointvec_ms_jiffies,
6733         },
6734         {
6735                 .procname       = "use_tempaddr",
6736                 .data           = &ipv6_devconf.use_tempaddr,
6737                 .maxlen         = sizeof(int),
6738                 .mode           = 0644,
6739                 .proc_handler   = proc_dointvec,
6740         },
6741         {
6742                 .procname       = "temp_valid_lft",
6743                 .data           = &ipv6_devconf.temp_valid_lft,
6744                 .maxlen         = sizeof(int),
6745                 .mode           = 0644,
6746                 .proc_handler   = proc_dointvec,
6747         },
6748         {
6749                 .procname       = "temp_prefered_lft",
6750                 .data           = &ipv6_devconf.temp_prefered_lft,
6751                 .maxlen         = sizeof(int),
6752                 .mode           = 0644,
6753                 .proc_handler   = proc_dointvec,
6754         },
6755         {
6756                 .procname       = "regen_max_retry",
6757                 .data           = &ipv6_devconf.regen_max_retry,
6758                 .maxlen         = sizeof(int),
6759                 .mode           = 0644,
6760                 .proc_handler   = proc_dointvec,
6761         },
6762         {
6763                 .procname       = "max_desync_factor",
6764                 .data           = &ipv6_devconf.max_desync_factor,
6765                 .maxlen         = sizeof(int),
6766                 .mode           = 0644,
6767                 .proc_handler   = proc_dointvec,
6768         },
6769         {
6770                 .procname       = "max_addresses",
6771                 .data           = &ipv6_devconf.max_addresses,
6772                 .maxlen         = sizeof(int),
6773                 .mode           = 0644,
6774                 .proc_handler   = proc_dointvec,
6775         },
6776         {
6777                 .procname       = "accept_ra_defrtr",
6778                 .data           = &ipv6_devconf.accept_ra_defrtr,
6779                 .maxlen         = sizeof(int),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_dointvec,
6782         },
6783         {
6784                 .procname       = "ra_defrtr_metric",
6785                 .data           = &ipv6_devconf.ra_defrtr_metric,
6786                 .maxlen         = sizeof(u32),
6787                 .mode           = 0644,
6788                 .proc_handler   = proc_douintvec_minmax,
6789                 .extra1         = (void *)SYSCTL_ONE,
6790         },
6791         {
6792                 .procname       = "accept_ra_min_hop_limit",
6793                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6794                 .maxlen         = sizeof(int),
6795                 .mode           = 0644,
6796                 .proc_handler   = proc_dointvec,
6797         },
6798         {
6799                 .procname       = "accept_ra_pinfo",
6800                 .data           = &ipv6_devconf.accept_ra_pinfo,
6801                 .maxlen         = sizeof(int),
6802                 .mode           = 0644,
6803                 .proc_handler   = proc_dointvec,
6804         },
6805 #ifdef CONFIG_IPV6_ROUTER_PREF
6806         {
6807                 .procname       = "accept_ra_rtr_pref",
6808                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6809                 .maxlen         = sizeof(int),
6810                 .mode           = 0644,
6811                 .proc_handler   = proc_dointvec,
6812         },
6813         {
6814                 .procname       = "router_probe_interval",
6815                 .data           = &ipv6_devconf.rtr_probe_interval,
6816                 .maxlen         = sizeof(int),
6817                 .mode           = 0644,
6818                 .proc_handler   = proc_dointvec_jiffies,
6819         },
6820 #ifdef CONFIG_IPV6_ROUTE_INFO
6821         {
6822                 .procname       = "accept_ra_rt_info_min_plen",
6823                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6824                 .maxlen         = sizeof(int),
6825                 .mode           = 0644,
6826                 .proc_handler   = proc_dointvec,
6827         },
6828         {
6829                 .procname       = "accept_ra_rt_info_max_plen",
6830                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6831                 .maxlen         = sizeof(int),
6832                 .mode           = 0644,
6833                 .proc_handler   = proc_dointvec,
6834         },
6835 #endif
6836 #endif
6837         {
6838                 .procname       = "proxy_ndp",
6839                 .data           = &ipv6_devconf.proxy_ndp,
6840                 .maxlen         = sizeof(int),
6841                 .mode           = 0644,
6842                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6843         },
6844         {
6845                 .procname       = "accept_source_route",
6846                 .data           = &ipv6_devconf.accept_source_route,
6847                 .maxlen         = sizeof(int),
6848                 .mode           = 0644,
6849                 .proc_handler   = proc_dointvec,
6850         },
6851 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6852         {
6853                 .procname       = "optimistic_dad",
6854                 .data           = &ipv6_devconf.optimistic_dad,
6855                 .maxlen         = sizeof(int),
6856                 .mode           = 0644,
6857                 .proc_handler   = proc_dointvec,
6858         },
6859         {
6860                 .procname       = "use_optimistic",
6861                 .data           = &ipv6_devconf.use_optimistic,
6862                 .maxlen         = sizeof(int),
6863                 .mode           = 0644,
6864                 .proc_handler   = proc_dointvec,
6865         },
6866 #endif
6867 #ifdef CONFIG_IPV6_MROUTE
6868         {
6869                 .procname       = "mc_forwarding",
6870                 .data           = &ipv6_devconf.mc_forwarding,
6871                 .maxlen         = sizeof(int),
6872                 .mode           = 0444,
6873                 .proc_handler   = proc_dointvec,
6874         },
6875 #endif
6876         {
6877                 .procname       = "disable_ipv6",
6878                 .data           = &ipv6_devconf.disable_ipv6,
6879                 .maxlen         = sizeof(int),
6880                 .mode           = 0644,
6881                 .proc_handler   = addrconf_sysctl_disable,
6882         },
6883         {
6884                 .procname       = "accept_dad",
6885                 .data           = &ipv6_devconf.accept_dad,
6886                 .maxlen         = sizeof(int),
6887                 .mode           = 0644,
6888                 .proc_handler   = proc_dointvec,
6889         },
6890         {
6891                 .procname       = "force_tllao",
6892                 .data           = &ipv6_devconf.force_tllao,
6893                 .maxlen         = sizeof(int),
6894                 .mode           = 0644,
6895                 .proc_handler   = proc_dointvec
6896         },
6897         {
6898                 .procname       = "ndisc_notify",
6899                 .data           = &ipv6_devconf.ndisc_notify,
6900                 .maxlen         = sizeof(int),
6901                 .mode           = 0644,
6902                 .proc_handler   = proc_dointvec
6903         },
6904         {
6905                 .procname       = "suppress_frag_ndisc",
6906                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6907                 .maxlen         = sizeof(int),
6908                 .mode           = 0644,
6909                 .proc_handler   = proc_dointvec
6910         },
6911         {
6912                 .procname       = "accept_ra_from_local",
6913                 .data           = &ipv6_devconf.accept_ra_from_local,
6914                 .maxlen         = sizeof(int),
6915                 .mode           = 0644,
6916                 .proc_handler   = proc_dointvec,
6917         },
6918         {
6919                 .procname       = "accept_ra_mtu",
6920                 .data           = &ipv6_devconf.accept_ra_mtu,
6921                 .maxlen         = sizeof(int),
6922                 .mode           = 0644,
6923                 .proc_handler   = proc_dointvec,
6924         },
6925         {
6926                 .procname       = "stable_secret",
6927                 .data           = &ipv6_devconf.stable_secret,
6928                 .maxlen         = IPV6_MAX_STRLEN,
6929                 .mode           = 0600,
6930                 .proc_handler   = addrconf_sysctl_stable_secret,
6931         },
6932         {
6933                 .procname       = "use_oif_addrs_only",
6934                 .data           = &ipv6_devconf.use_oif_addrs_only,
6935                 .maxlen         = sizeof(int),
6936                 .mode           = 0644,
6937                 .proc_handler   = proc_dointvec,
6938         },
6939         {
6940                 .procname       = "ignore_routes_with_linkdown",
6941                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6942                 .maxlen         = sizeof(int),
6943                 .mode           = 0644,
6944                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6945         },
6946         {
6947                 .procname       = "drop_unicast_in_l2_multicast",
6948                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6949                 .maxlen         = sizeof(int),
6950                 .mode           = 0644,
6951                 .proc_handler   = proc_dointvec,
6952         },
6953         {
6954                 .procname       = "drop_unsolicited_na",
6955                 .data           = &ipv6_devconf.drop_unsolicited_na,
6956                 .maxlen         = sizeof(int),
6957                 .mode           = 0644,
6958                 .proc_handler   = proc_dointvec,
6959         },
6960         {
6961                 .procname       = "keep_addr_on_down",
6962                 .data           = &ipv6_devconf.keep_addr_on_down,
6963                 .maxlen         = sizeof(int),
6964                 .mode           = 0644,
6965                 .proc_handler   = proc_dointvec,
6966
6967         },
6968         {
6969                 .procname       = "seg6_enabled",
6970                 .data           = &ipv6_devconf.seg6_enabled,
6971                 .maxlen         = sizeof(int),
6972                 .mode           = 0644,
6973                 .proc_handler   = proc_dointvec,
6974         },
6975 #ifdef CONFIG_IPV6_SEG6_HMAC
6976         {
6977                 .procname       = "seg6_require_hmac",
6978                 .data           = &ipv6_devconf.seg6_require_hmac,
6979                 .maxlen         = sizeof(int),
6980                 .mode           = 0644,
6981                 .proc_handler   = proc_dointvec,
6982         },
6983 #endif
6984         {
6985                 .procname       = "enhanced_dad",
6986                 .data           = &ipv6_devconf.enhanced_dad,
6987                 .maxlen         = sizeof(int),
6988                 .mode           = 0644,
6989                 .proc_handler   = proc_dointvec,
6990         },
6991         {
6992                 .procname       = "addr_gen_mode",
6993                 .data           = &ipv6_devconf.addr_gen_mode,
6994                 .maxlen         = sizeof(int),
6995                 .mode           = 0644,
6996                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6997         },
6998         {
6999                 .procname       = "disable_policy",
7000                 .data           = &ipv6_devconf.disable_policy,
7001                 .maxlen         = sizeof(int),
7002                 .mode           = 0644,
7003                 .proc_handler   = addrconf_sysctl_disable_policy,
7004         },
7005         {
7006                 .procname       = "ndisc_tclass",
7007                 .data           = &ipv6_devconf.ndisc_tclass,
7008                 .maxlen         = sizeof(int),
7009                 .mode           = 0644,
7010                 .proc_handler   = proc_dointvec_minmax,
7011                 .extra1         = (void *)SYSCTL_ZERO,
7012                 .extra2         = (void *)&two_five_five,
7013         },
7014         {
7015                 .procname       = "rpl_seg_enabled",
7016                 .data           = &ipv6_devconf.rpl_seg_enabled,
7017                 .maxlen         = sizeof(int),
7018                 .mode           = 0644,
7019                 .proc_handler   = proc_dointvec,
7020         },
7021         {
7022                 .procname       = "ioam6_enabled",
7023                 .data           = &ipv6_devconf.ioam6_enabled,
7024                 .maxlen         = sizeof(u8),
7025                 .mode           = 0644,
7026                 .proc_handler   = proc_dou8vec_minmax,
7027                 .extra1         = (void *)SYSCTL_ZERO,
7028                 .extra2         = (void *)SYSCTL_ONE,
7029         },
7030         {
7031                 .procname       = "ioam6_id",
7032                 .data           = &ipv6_devconf.ioam6_id,
7033                 .maxlen         = sizeof(u32),
7034                 .mode           = 0644,
7035                 .proc_handler   = proc_douintvec_minmax,
7036                 .extra1         = (void *)SYSCTL_ZERO,
7037                 .extra2         = (void *)&ioam6_if_id_max,
7038         },
7039         {
7040                 .procname       = "ioam6_id_wide",
7041                 .data           = &ipv6_devconf.ioam6_id_wide,
7042                 .maxlen         = sizeof(u32),
7043                 .mode           = 0644,
7044                 .proc_handler   = proc_douintvec,
7045         },
7046         {
7047                 .procname       = "ndisc_evict_nocarrier",
7048                 .data           = &ipv6_devconf.ndisc_evict_nocarrier,
7049                 .maxlen         = sizeof(u8),
7050                 .mode           = 0644,
7051                 .proc_handler   = proc_dou8vec_minmax,
7052                 .extra1         = (void *)SYSCTL_ZERO,
7053                 .extra2         = (void *)SYSCTL_ONE,
7054         },
7055         {
7056                 .procname       = "accept_untracked_na",
7057                 .data           = &ipv6_devconf.accept_untracked_na,
7058                 .maxlen         = sizeof(int),
7059                 .mode           = 0644,
7060                 .proc_handler   = proc_dointvec_minmax,
7061                 .extra1         = SYSCTL_ZERO,
7062                 .extra2         = SYSCTL_TWO,
7063         },
7064         {
7065                 /* sentinel */
7066         }
7067 };
7068
7069 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7070                 struct inet6_dev *idev, struct ipv6_devconf *p)
7071 {
7072         int i, ifindex;
7073         struct ctl_table *table;
7074         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7075
7076         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7077         if (!table)
7078                 goto out;
7079
7080         for (i = 0; table[i].data; i++) {
7081                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7082                 /* If one of these is already set, then it is not safe to
7083                  * overwrite either of them: this makes proc_dointvec_minmax
7084                  * usable.
7085                  */
7086                 if (!table[i].extra1 && !table[i].extra2) {
7087                         table[i].extra1 = idev; /* embedded; no ref */
7088                         table[i].extra2 = net;
7089                 }
7090         }
7091
7092         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7093
7094         p->sysctl_header = register_net_sysctl(net, path, table);
7095         if (!p->sysctl_header)
7096                 goto free;
7097
7098         if (!strcmp(dev_name, "all"))
7099                 ifindex = NETCONFA_IFINDEX_ALL;
7100         else if (!strcmp(dev_name, "default"))
7101                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7102         else
7103                 ifindex = idev->dev->ifindex;
7104         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7105                                      ifindex, p);
7106         return 0;
7107
7108 free:
7109         kfree(table);
7110 out:
7111         return -ENOBUFS;
7112 }
7113
7114 static void __addrconf_sysctl_unregister(struct net *net,
7115                                          struct ipv6_devconf *p, int ifindex)
7116 {
7117         struct ctl_table *table;
7118
7119         if (!p->sysctl_header)
7120                 return;
7121
7122         table = p->sysctl_header->ctl_table_arg;
7123         unregister_net_sysctl_table(p->sysctl_header);
7124         p->sysctl_header = NULL;
7125         kfree(table);
7126
7127         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7128 }
7129
7130 static int addrconf_sysctl_register(struct inet6_dev *idev)
7131 {
7132         int err;
7133
7134         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7135                 return -EINVAL;
7136
7137         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7138                                     &ndisc_ifinfo_sysctl_change);
7139         if (err)
7140                 return err;
7141         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7142                                          idev, &idev->cnf);
7143         if (err)
7144                 neigh_sysctl_unregister(idev->nd_parms);
7145
7146         return err;
7147 }
7148
7149 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7150 {
7151         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7152                                      idev->dev->ifindex);
7153         neigh_sysctl_unregister(idev->nd_parms);
7154 }
7155
7156
7157 #endif
7158
7159 static int __net_init addrconf_init_net(struct net *net)
7160 {
7161         int err = -ENOMEM;
7162         struct ipv6_devconf *all, *dflt;
7163
7164         spin_lock_init(&net->ipv6.addrconf_hash_lock);
7165         INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7166         net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7167                                            sizeof(struct hlist_head),
7168                                            GFP_KERNEL);
7169         if (!net->ipv6.inet6_addr_lst)
7170                 goto err_alloc_addr;
7171
7172         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7173         if (!all)
7174                 goto err_alloc_all;
7175
7176         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7177         if (!dflt)
7178                 goto err_alloc_dflt;
7179
7180         if (!net_eq(net, &init_net)) {
7181                 switch (net_inherit_devconf()) {
7182                 case 1:  /* copy from init_net */
7183                         memcpy(all, init_net.ipv6.devconf_all,
7184                                sizeof(ipv6_devconf));
7185                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7186                                sizeof(ipv6_devconf_dflt));
7187                         break;
7188                 case 3: /* copy from the current netns */
7189                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7190                                sizeof(ipv6_devconf));
7191                         memcpy(dflt,
7192                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7193                                sizeof(ipv6_devconf_dflt));
7194                         break;
7195                 case 0:
7196                 case 2:
7197                         /* use compiled values */
7198                         break;
7199                 }
7200         }
7201
7202         /* these will be inherited by all namespaces */
7203         dflt->autoconf = ipv6_defaults.autoconf;
7204         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7205
7206         dflt->stable_secret.initialized = false;
7207         all->stable_secret.initialized = false;
7208
7209         net->ipv6.devconf_all = all;
7210         net->ipv6.devconf_dflt = dflt;
7211
7212 #ifdef CONFIG_SYSCTL
7213         err = __addrconf_sysctl_register(net, "all", NULL, all);
7214         if (err < 0)
7215                 goto err_reg_all;
7216
7217         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7218         if (err < 0)
7219                 goto err_reg_dflt;
7220 #endif
7221         return 0;
7222
7223 #ifdef CONFIG_SYSCTL
7224 err_reg_dflt:
7225         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7226 err_reg_all:
7227         kfree(dflt);
7228         net->ipv6.devconf_dflt = NULL;
7229 #endif
7230 err_alloc_dflt:
7231         kfree(all);
7232         net->ipv6.devconf_all = NULL;
7233 err_alloc_all:
7234         kfree(net->ipv6.inet6_addr_lst);
7235 err_alloc_addr:
7236         return err;
7237 }
7238
7239 static void __net_exit addrconf_exit_net(struct net *net)
7240 {
7241         int i;
7242
7243 #ifdef CONFIG_SYSCTL
7244         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7245                                      NETCONFA_IFINDEX_DEFAULT);
7246         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7247                                      NETCONFA_IFINDEX_ALL);
7248 #endif
7249         kfree(net->ipv6.devconf_dflt);
7250         net->ipv6.devconf_dflt = NULL;
7251         kfree(net->ipv6.devconf_all);
7252         net->ipv6.devconf_all = NULL;
7253
7254         cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7255         /*
7256          *      Check hash table, then free it.
7257          */
7258         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7259                 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7260
7261         kfree(net->ipv6.inet6_addr_lst);
7262         net->ipv6.inet6_addr_lst = NULL;
7263 }
7264
7265 static struct pernet_operations addrconf_ops = {
7266         .init = addrconf_init_net,
7267         .exit = addrconf_exit_net,
7268 };
7269
7270 static struct rtnl_af_ops inet6_ops __read_mostly = {
7271         .family           = AF_INET6,
7272         .fill_link_af     = inet6_fill_link_af,
7273         .get_link_af_size = inet6_get_link_af_size,
7274         .validate_link_af = inet6_validate_link_af,
7275         .set_link_af      = inet6_set_link_af,
7276 };
7277
7278 /*
7279  *      Init / cleanup code
7280  */
7281
7282 int __init addrconf_init(void)
7283 {
7284         struct inet6_dev *idev;
7285         int err;
7286
7287         err = ipv6_addr_label_init();
7288         if (err < 0) {
7289                 pr_crit("%s: cannot initialize default policy table: %d\n",
7290                         __func__, err);
7291                 goto out;
7292         }
7293
7294         err = register_pernet_subsys(&addrconf_ops);
7295         if (err < 0)
7296                 goto out_addrlabel;
7297
7298         addrconf_wq = create_workqueue("ipv6_addrconf");
7299         if (!addrconf_wq) {
7300                 err = -ENOMEM;
7301                 goto out_nowq;
7302         }
7303
7304         rtnl_lock();
7305         idev = ipv6_add_dev(blackhole_netdev);
7306         rtnl_unlock();
7307         if (IS_ERR(idev)) {
7308                 err = PTR_ERR(idev);
7309                 goto errlo;
7310         }
7311
7312         ip6_route_init_special_entries();
7313
7314         register_netdevice_notifier(&ipv6_dev_notf);
7315
7316         addrconf_verify(&init_net);
7317
7318         rtnl_af_register(&inet6_ops);
7319
7320         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7321                                    NULL, inet6_dump_ifinfo, 0);
7322         if (err < 0)
7323                 goto errout;
7324
7325         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7326                                    inet6_rtm_newaddr, NULL, 0);
7327         if (err < 0)
7328                 goto errout;
7329         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7330                                    inet6_rtm_deladdr, NULL, 0);
7331         if (err < 0)
7332                 goto errout;
7333         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7334                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7335                                    RTNL_FLAG_DOIT_UNLOCKED);
7336         if (err < 0)
7337                 goto errout;
7338         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7339                                    NULL, inet6_dump_ifmcaddr, 0);
7340         if (err < 0)
7341                 goto errout;
7342         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7343                                    NULL, inet6_dump_ifacaddr, 0);
7344         if (err < 0)
7345                 goto errout;
7346         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7347                                    inet6_netconf_get_devconf,
7348                                    inet6_netconf_dump_devconf,
7349                                    RTNL_FLAG_DOIT_UNLOCKED);
7350         if (err < 0)
7351                 goto errout;
7352         err = ipv6_addr_label_rtnl_register();
7353         if (err < 0)
7354                 goto errout;
7355
7356         return 0;
7357 errout:
7358         rtnl_unregister_all(PF_INET6);
7359         rtnl_af_unregister(&inet6_ops);
7360         unregister_netdevice_notifier(&ipv6_dev_notf);
7361 errlo:
7362         destroy_workqueue(addrconf_wq);
7363 out_nowq:
7364         unregister_pernet_subsys(&addrconf_ops);
7365 out_addrlabel:
7366         ipv6_addr_label_cleanup();
7367 out:
7368         return err;
7369 }
7370
7371 void addrconf_cleanup(void)
7372 {
7373         struct net_device *dev;
7374
7375         unregister_netdevice_notifier(&ipv6_dev_notf);
7376         unregister_pernet_subsys(&addrconf_ops);
7377         ipv6_addr_label_cleanup();
7378
7379         rtnl_af_unregister(&inet6_ops);
7380
7381         rtnl_lock();
7382
7383         /* clean dev list */
7384         for_each_netdev(&init_net, dev) {
7385                 if (__in6_dev_get(dev) == NULL)
7386                         continue;
7387                 addrconf_ifdown(dev, true);
7388         }
7389         addrconf_ifdown(init_net.loopback_dev, true);
7390
7391         rtnl_unlock();
7392
7393         destroy_workqueue(addrconf_wq);
7394 }