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         if ((create || list_empty(&idev->tempaddr_list)) &&
2565             idev->cnf.use_tempaddr > 0) {
2566                 /* When a new public address is created as described
2567                  * in [ADDRCONF], also create a new temporary address.
2568                  * Also create a temporary address if it's enabled but
2569                  * no temporary address currently exists.
2570                  */
2571                 read_unlock_bh(&idev->lock);
2572                 ipv6_create_tempaddr(ifp, false);
2573         } else {
2574                 read_unlock_bh(&idev->lock);
2575         }
2576 }
2577
2578 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2579 {
2580         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2581                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2582 }
2583
2584 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2585                                  const struct prefix_info *pinfo,
2586                                  struct inet6_dev *in6_dev,
2587                                  const struct in6_addr *addr, int addr_type,
2588                                  u32 addr_flags, bool sllao, bool tokenized,
2589                                  __u32 valid_lft, u32 prefered_lft)
2590 {
2591         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2592         int create = 0, update_lft = 0;
2593
2594         if (!ifp && valid_lft) {
2595                 int max_addresses = in6_dev->cnf.max_addresses;
2596                 struct ifa6_config cfg = {
2597                         .pfx = addr,
2598                         .plen = pinfo->prefix_len,
2599                         .ifa_flags = addr_flags,
2600                         .valid_lft = valid_lft,
2601                         .preferred_lft = prefered_lft,
2602                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2603                         .ifa_proto = IFAPROT_KERNEL_RA
2604                 };
2605
2606 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2607                 if ((net->ipv6.devconf_all->optimistic_dad ||
2608                      in6_dev->cnf.optimistic_dad) &&
2609                     !net->ipv6.devconf_all->forwarding && sllao)
2610                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2611 #endif
2612
2613                 /* Do not allow to create too much of autoconfigured
2614                  * addresses; this would be too easy way to crash kernel.
2615                  */
2616                 if (!max_addresses ||
2617                     ipv6_count_addresses(in6_dev) < max_addresses)
2618                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2619
2620                 if (IS_ERR_OR_NULL(ifp))
2621                         return -1;
2622
2623                 create = 1;
2624                 spin_lock_bh(&ifp->lock);
2625                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2626                 ifp->cstamp = jiffies;
2627                 ifp->tokenized = tokenized;
2628                 spin_unlock_bh(&ifp->lock);
2629                 addrconf_dad_start(ifp);
2630         }
2631
2632         if (ifp) {
2633                 u32 flags;
2634                 unsigned long now;
2635                 u32 stored_lft;
2636
2637                 /* update lifetime (RFC2462 5.5.3 e) */
2638                 spin_lock_bh(&ifp->lock);
2639                 now = jiffies;
2640                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2641                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2642                 else
2643                         stored_lft = 0;
2644                 if (!create && stored_lft) {
2645                         const u32 minimum_lft = min_t(u32,
2646                                 stored_lft, MIN_VALID_LIFETIME);
2647                         valid_lft = max(valid_lft, minimum_lft);
2648
2649                         /* RFC4862 Section 5.5.3e:
2650                          * "Note that the preferred lifetime of the
2651                          *  corresponding address is always reset to
2652                          *  the Preferred Lifetime in the received
2653                          *  Prefix Information option, regardless of
2654                          *  whether the valid lifetime is also reset or
2655                          *  ignored."
2656                          *
2657                          * So we should always update prefered_lft here.
2658                          */
2659                         update_lft = 1;
2660                 }
2661
2662                 if (update_lft) {
2663                         ifp->valid_lft = valid_lft;
2664                         ifp->prefered_lft = prefered_lft;
2665                         ifp->tstamp = now;
2666                         flags = ifp->flags;
2667                         ifp->flags &= ~IFA_F_DEPRECATED;
2668                         spin_unlock_bh(&ifp->lock);
2669
2670                         if (!(flags&IFA_F_TENTATIVE))
2671                                 ipv6_ifa_notify(0, ifp);
2672                 } else
2673                         spin_unlock_bh(&ifp->lock);
2674
2675                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2676                                  create, now);
2677
2678                 in6_ifa_put(ifp);
2679                 addrconf_verify(net);
2680         }
2681
2682         return 0;
2683 }
2684 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2685
2686 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2687 {
2688         struct prefix_info *pinfo;
2689         __u32 valid_lft;
2690         __u32 prefered_lft;
2691         int addr_type, err;
2692         u32 addr_flags = 0;
2693         struct inet6_dev *in6_dev;
2694         struct net *net = dev_net(dev);
2695
2696         pinfo = (struct prefix_info *) opt;
2697
2698         if (len < sizeof(struct prefix_info)) {
2699                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2700                 return;
2701         }
2702
2703         /*
2704          *      Validation checks ([ADDRCONF], page 19)
2705          */
2706
2707         addr_type = ipv6_addr_type(&pinfo->prefix);
2708
2709         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2710                 return;
2711
2712         valid_lft = ntohl(pinfo->valid);
2713         prefered_lft = ntohl(pinfo->prefered);
2714
2715         if (prefered_lft > valid_lft) {
2716                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2717                 return;
2718         }
2719
2720         in6_dev = in6_dev_get(dev);
2721
2722         if (!in6_dev) {
2723                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2724                                     dev->name);
2725                 return;
2726         }
2727
2728         /*
2729          *      Two things going on here:
2730          *      1) Add routes for on-link prefixes
2731          *      2) Configure prefixes with the auto flag set
2732          */
2733
2734         if (pinfo->onlink) {
2735                 struct fib6_info *rt;
2736                 unsigned long rt_expires;
2737
2738                 /* Avoid arithmetic overflow. Really, we could
2739                  * save rt_expires in seconds, likely valid_lft,
2740                  * but it would require division in fib gc, that it
2741                  * not good.
2742                  */
2743                 if (HZ > USER_HZ)
2744                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2745                 else
2746                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2747
2748                 if (addrconf_finite_timeout(rt_expires))
2749                         rt_expires *= HZ;
2750
2751                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2752                                                pinfo->prefix_len,
2753                                                dev,
2754                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2755                                                RTF_DEFAULT, true);
2756
2757                 if (rt) {
2758                         /* Autoconf prefix route */
2759                         if (valid_lft == 0) {
2760                                 ip6_del_rt(net, rt, false);
2761                                 rt = NULL;
2762                         } else if (addrconf_finite_timeout(rt_expires)) {
2763                                 /* not infinity */
2764                                 fib6_set_expires(rt, jiffies + rt_expires);
2765                         } else {
2766                                 fib6_clean_expires(rt);
2767                         }
2768                 } else if (valid_lft) {
2769                         clock_t expires = 0;
2770                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2771                         if (addrconf_finite_timeout(rt_expires)) {
2772                                 /* not infinity */
2773                                 flags |= RTF_EXPIRES;
2774                                 expires = jiffies_to_clock_t(rt_expires);
2775                         }
2776                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2777                                               0, dev, expires, flags,
2778                                               GFP_ATOMIC);
2779                 }
2780                 fib6_info_release(rt);
2781         }
2782
2783         /* Try to figure out our local address for this prefix */
2784
2785         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2786                 struct in6_addr addr;
2787                 bool tokenized = false, dev_addr_generated = false;
2788
2789                 if (pinfo->prefix_len == 64) {
2790                         memcpy(&addr, &pinfo->prefix, 8);
2791
2792                         if (!ipv6_addr_any(&in6_dev->token)) {
2793                                 read_lock_bh(&in6_dev->lock);
2794                                 memcpy(addr.s6_addr + 8,
2795                                        in6_dev->token.s6_addr + 8, 8);
2796                                 read_unlock_bh(&in6_dev->lock);
2797                                 tokenized = true;
2798                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2799                                    !ipv6_generate_stable_address(&addr, 0,
2800                                                                  in6_dev)) {
2801                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2802                                 goto ok;
2803                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2804                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2805                                 goto put;
2806                         } else {
2807                                 dev_addr_generated = true;
2808                         }
2809                         goto ok;
2810                 }
2811                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2812                                     pinfo->prefix_len);
2813                 goto put;
2814
2815 ok:
2816                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2817                                                    &addr, addr_type,
2818                                                    addr_flags, sllao,
2819                                                    tokenized, valid_lft,
2820                                                    prefered_lft);
2821                 if (err)
2822                         goto put;
2823
2824                 /* Ignore error case here because previous prefix add addr was
2825                  * successful which will be notified.
2826                  */
2827                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2828                                               addr_type, addr_flags, sllao,
2829                                               tokenized, valid_lft,
2830                                               prefered_lft,
2831                                               dev_addr_generated);
2832         }
2833         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2834 put:
2835         in6_dev_put(in6_dev);
2836 }
2837
2838 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2839                 struct in6_ifreq *ireq)
2840 {
2841         struct ip_tunnel_parm p = { };
2842         int err;
2843
2844         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2845                 return -EADDRNOTAVAIL;
2846
2847         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2848         p.iph.version = 4;
2849         p.iph.ihl = 5;
2850         p.iph.protocol = IPPROTO_IPV6;
2851         p.iph.ttl = 64;
2852
2853         if (!dev->netdev_ops->ndo_tunnel_ctl)
2854                 return -EOPNOTSUPP;
2855         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2856         if (err)
2857                 return err;
2858
2859         dev = __dev_get_by_name(net, p.name);
2860         if (!dev)
2861                 return -ENOBUFS;
2862         return dev_open(dev, NULL);
2863 }
2864
2865 /*
2866  *      Set destination address.
2867  *      Special case for SIT interfaces where we create a new "virtual"
2868  *      device.
2869  */
2870 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2871 {
2872         struct net_device *dev;
2873         struct in6_ifreq ireq;
2874         int err = -ENODEV;
2875
2876         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2877                 return -ENODEV;
2878         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2879                 return -EFAULT;
2880
2881         rtnl_lock();
2882         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2883         if (dev && dev->type == ARPHRD_SIT)
2884                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2885         rtnl_unlock();
2886         return err;
2887 }
2888
2889 static int ipv6_mc_config(struct sock *sk, bool join,
2890                           const struct in6_addr *addr, int ifindex)
2891 {
2892         int ret;
2893
2894         ASSERT_RTNL();
2895
2896         lock_sock(sk);
2897         if (join)
2898                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2899         else
2900                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2901         release_sock(sk);
2902
2903         return ret;
2904 }
2905
2906 /*
2907  *      Manual configuration of address on an interface
2908  */
2909 static int inet6_addr_add(struct net *net, int ifindex,
2910                           struct ifa6_config *cfg,
2911                           struct netlink_ext_ack *extack)
2912 {
2913         struct inet6_ifaddr *ifp;
2914         struct inet6_dev *idev;
2915         struct net_device *dev;
2916         unsigned long timeout;
2917         clock_t expires;
2918         u32 flags;
2919
2920         ASSERT_RTNL();
2921
2922         if (cfg->plen > 128)
2923                 return -EINVAL;
2924
2925         /* check the lifetime */
2926         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2927                 return -EINVAL;
2928
2929         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2930                 return -EINVAL;
2931
2932         dev = __dev_get_by_index(net, ifindex);
2933         if (!dev)
2934                 return -ENODEV;
2935
2936         idev = addrconf_add_dev(dev);
2937         if (IS_ERR(idev))
2938                 return PTR_ERR(idev);
2939
2940         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2941                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2942                                          true, cfg->pfx, ifindex);
2943
2944                 if (ret < 0)
2945                         return ret;
2946         }
2947
2948         cfg->scope = ipv6_addr_scope(cfg->pfx);
2949
2950         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2951         if (addrconf_finite_timeout(timeout)) {
2952                 expires = jiffies_to_clock_t(timeout * HZ);
2953                 cfg->valid_lft = timeout;
2954                 flags = RTF_EXPIRES;
2955         } else {
2956                 expires = 0;
2957                 flags = 0;
2958                 cfg->ifa_flags |= IFA_F_PERMANENT;
2959         }
2960
2961         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2962         if (addrconf_finite_timeout(timeout)) {
2963                 if (timeout == 0)
2964                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2965                 cfg->preferred_lft = timeout;
2966         }
2967
2968         ifp = ipv6_add_addr(idev, cfg, true, extack);
2969         if (!IS_ERR(ifp)) {
2970                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2971                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2972                                               ifp->rt_priority, dev, expires,
2973                                               flags, GFP_KERNEL);
2974                 }
2975
2976                 /* Send a netlink notification if DAD is enabled and
2977                  * optimistic flag is not set
2978                  */
2979                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2980                         ipv6_ifa_notify(0, ifp);
2981                 /*
2982                  * Note that section 3.1 of RFC 4429 indicates
2983                  * that the Optimistic flag should not be set for
2984                  * manually configured addresses
2985                  */
2986                 addrconf_dad_start(ifp);
2987                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2988                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2989                                          cfg->preferred_lft, true, jiffies);
2990                 in6_ifa_put(ifp);
2991                 addrconf_verify_rtnl(net);
2992                 return 0;
2993         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2994                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2995                                cfg->pfx, ifindex);
2996         }
2997
2998         return PTR_ERR(ifp);
2999 }
3000
3001 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3002                           const struct in6_addr *pfx, unsigned int plen)
3003 {
3004         struct inet6_ifaddr *ifp;
3005         struct inet6_dev *idev;
3006         struct net_device *dev;
3007
3008         if (plen > 128)
3009                 return -EINVAL;
3010
3011         dev = __dev_get_by_index(net, ifindex);
3012         if (!dev)
3013                 return -ENODEV;
3014
3015         idev = __in6_dev_get(dev);
3016         if (!idev)
3017                 return -ENXIO;
3018
3019         read_lock_bh(&idev->lock);
3020         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3021                 if (ifp->prefix_len == plen &&
3022                     ipv6_addr_equal(pfx, &ifp->addr)) {
3023                         in6_ifa_hold(ifp);
3024                         read_unlock_bh(&idev->lock);
3025
3026                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3027                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3028                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3029                                                  jiffies);
3030                         ipv6_del_addr(ifp);
3031                         addrconf_verify_rtnl(net);
3032                         if (ipv6_addr_is_multicast(pfx)) {
3033                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3034                                                false, pfx, dev->ifindex);
3035                         }
3036                         return 0;
3037                 }
3038         }
3039         read_unlock_bh(&idev->lock);
3040         return -EADDRNOTAVAIL;
3041 }
3042
3043
3044 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3045 {
3046         struct ifa6_config cfg = {
3047                 .ifa_flags = IFA_F_PERMANENT,
3048                 .preferred_lft = INFINITY_LIFE_TIME,
3049                 .valid_lft = INFINITY_LIFE_TIME,
3050         };
3051         struct in6_ifreq ireq;
3052         int err;
3053
3054         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3055                 return -EPERM;
3056
3057         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3058                 return -EFAULT;
3059
3060         cfg.pfx = &ireq.ifr6_addr;
3061         cfg.plen = ireq.ifr6_prefixlen;
3062
3063         rtnl_lock();
3064         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3065         rtnl_unlock();
3066         return err;
3067 }
3068
3069 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3070 {
3071         struct in6_ifreq ireq;
3072         int err;
3073
3074         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3075                 return -EPERM;
3076
3077         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3078                 return -EFAULT;
3079
3080         rtnl_lock();
3081         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3082                              ireq.ifr6_prefixlen);
3083         rtnl_unlock();
3084         return err;
3085 }
3086
3087 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3088                      int plen, int scope, u8 proto)
3089 {
3090         struct inet6_ifaddr *ifp;
3091         struct ifa6_config cfg = {
3092                 .pfx = addr,
3093                 .plen = plen,
3094                 .ifa_flags = IFA_F_PERMANENT,
3095                 .valid_lft = INFINITY_LIFE_TIME,
3096                 .preferred_lft = INFINITY_LIFE_TIME,
3097                 .scope = scope,
3098                 .ifa_proto = proto
3099         };
3100
3101         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3102         if (!IS_ERR(ifp)) {
3103                 spin_lock_bh(&ifp->lock);
3104                 ifp->flags &= ~IFA_F_TENTATIVE;
3105                 spin_unlock_bh(&ifp->lock);
3106                 rt_genid_bump_ipv6(dev_net(idev->dev));
3107                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3108                 in6_ifa_put(ifp);
3109         }
3110 }
3111
3112 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3113 static void add_v4_addrs(struct inet6_dev *idev)
3114 {
3115         struct in6_addr addr;
3116         struct net_device *dev;
3117         struct net *net = dev_net(idev->dev);
3118         int scope, plen, offset = 0;
3119         u32 pflags = 0;
3120
3121         ASSERT_RTNL();
3122
3123         memset(&addr, 0, sizeof(struct in6_addr));
3124         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3125         if (idev->dev->addr_len == sizeof(struct in6_addr))
3126                 offset = sizeof(struct in6_addr) - 4;
3127         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3128
3129         if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3130                 scope = IPV6_ADDR_COMPATv4;
3131                 plen = 96;
3132                 pflags |= RTF_NONEXTHOP;
3133         } else {
3134                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3135                         return;
3136
3137                 addr.s6_addr32[0] = htonl(0xfe800000);
3138                 scope = IFA_LINK;
3139                 plen = 64;
3140         }
3141
3142         if (addr.s6_addr32[3]) {
3143                 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3144                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3145                                       GFP_KERNEL);
3146                 return;
3147         }
3148
3149         for_each_netdev(net, dev) {
3150                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3151                 if (in_dev && (dev->flags & IFF_UP)) {
3152                         struct in_ifaddr *ifa;
3153                         int flag = scope;
3154
3155                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3156                                 addr.s6_addr32[3] = ifa->ifa_local;
3157
3158                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3159                                         continue;
3160                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3161                                         if (idev->dev->flags&IFF_POINTOPOINT)
3162                                                 continue;
3163                                         flag |= IFA_HOST;
3164                                 }
3165
3166                                 add_addr(idev, &addr, plen, flag,
3167                                          IFAPROT_UNSPEC);
3168                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3169                                                       0, pflags, GFP_KERNEL);
3170                         }
3171                 }
3172         }
3173 }
3174 #endif
3175
3176 static void init_loopback(struct net_device *dev)
3177 {
3178         struct inet6_dev  *idev;
3179
3180         /* ::1 */
3181
3182         ASSERT_RTNL();
3183
3184         idev = ipv6_find_idev(dev);
3185         if (IS_ERR(idev)) {
3186                 pr_debug("%s: add_dev failed\n", __func__);
3187                 return;
3188         }
3189
3190         add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3191 }
3192
3193 void addrconf_add_linklocal(struct inet6_dev *idev,
3194                             const struct in6_addr *addr, u32 flags)
3195 {
3196         struct ifa6_config cfg = {
3197                 .pfx = addr,
3198                 .plen = 64,
3199                 .ifa_flags = flags | IFA_F_PERMANENT,
3200                 .valid_lft = INFINITY_LIFE_TIME,
3201                 .preferred_lft = INFINITY_LIFE_TIME,
3202                 .scope = IFA_LINK,
3203                 .ifa_proto = IFAPROT_KERNEL_LL
3204         };
3205         struct inet6_ifaddr *ifp;
3206
3207 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3208         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3209              idev->cnf.optimistic_dad) &&
3210             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3211                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3212 #endif
3213
3214         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3215         if (!IS_ERR(ifp)) {
3216                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3217                                       0, 0, GFP_ATOMIC);
3218                 addrconf_dad_start(ifp);
3219                 in6_ifa_put(ifp);
3220         }
3221 }
3222 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3223
3224 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3225 {
3226         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3227                 return true;
3228
3229         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3230             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3231                 return true;
3232
3233         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3234             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3235                 return true;
3236
3237         return false;
3238 }
3239
3240 static int ipv6_generate_stable_address(struct in6_addr *address,
3241                                         u8 dad_count,
3242                                         const struct inet6_dev *idev)
3243 {
3244         static DEFINE_SPINLOCK(lock);
3245         static __u32 digest[SHA1_DIGEST_WORDS];
3246         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3247
3248         static union {
3249                 char __data[SHA1_BLOCK_SIZE];
3250                 struct {
3251                         struct in6_addr secret;
3252                         __be32 prefix[2];
3253                         unsigned char hwaddr[MAX_ADDR_LEN];
3254                         u8 dad_count;
3255                 } __packed;
3256         } data;
3257
3258         struct in6_addr secret;
3259         struct in6_addr temp;
3260         struct net *net = dev_net(idev->dev);
3261
3262         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3263
3264         if (idev->cnf.stable_secret.initialized)
3265                 secret = idev->cnf.stable_secret.secret;
3266         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3267                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3268         else
3269                 return -1;
3270
3271 retry:
3272         spin_lock_bh(&lock);
3273
3274         sha1_init(digest);
3275         memset(&data, 0, sizeof(data));
3276         memset(workspace, 0, sizeof(workspace));
3277         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3278         data.prefix[0] = address->s6_addr32[0];
3279         data.prefix[1] = address->s6_addr32[1];
3280         data.secret = secret;
3281         data.dad_count = dad_count;
3282
3283         sha1_transform(digest, data.__data, workspace);
3284
3285         temp = *address;
3286         temp.s6_addr32[2] = (__force __be32)digest[0];
3287         temp.s6_addr32[3] = (__force __be32)digest[1];
3288
3289         spin_unlock_bh(&lock);
3290
3291         if (ipv6_reserved_interfaceid(temp)) {
3292                 dad_count++;
3293                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3294                         return -1;
3295                 goto retry;
3296         }
3297
3298         *address = temp;
3299         return 0;
3300 }
3301
3302 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3303 {
3304         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3305
3306         if (s->initialized)
3307                 return;
3308         s = &idev->cnf.stable_secret;
3309         get_random_bytes(&s->secret, sizeof(s->secret));
3310         s->initialized = true;
3311 }
3312
3313 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3314 {
3315         struct in6_addr addr;
3316
3317         /* no link local addresses on L3 master devices */
3318         if (netif_is_l3_master(idev->dev))
3319                 return;
3320
3321         /* no link local addresses on devices flagged as slaves */
3322         if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3323                 return;
3324
3325         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3326
3327         switch (idev->cnf.addr_gen_mode) {
3328         case IN6_ADDR_GEN_MODE_RANDOM:
3329                 ipv6_gen_mode_random_init(idev);
3330                 fallthrough;
3331         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3332                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3333                         addrconf_add_linklocal(idev, &addr,
3334                                                IFA_F_STABLE_PRIVACY);
3335                 else if (prefix_route)
3336                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3337                                               0, 0, GFP_KERNEL);
3338                 break;
3339         case IN6_ADDR_GEN_MODE_EUI64:
3340                 /* addrconf_add_linklocal also adds a prefix_route and we
3341                  * only need to care about prefix routes if ipv6_generate_eui64
3342                  * couldn't generate one.
3343                  */
3344                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3345                         addrconf_add_linklocal(idev, &addr, 0);
3346                 else if (prefix_route)
3347                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3348                                               0, 0, GFP_KERNEL);
3349                 break;
3350         case IN6_ADDR_GEN_MODE_NONE:
3351         default:
3352                 /* will not add any link local address */
3353                 break;
3354         }
3355 }
3356
3357 static void addrconf_dev_config(struct net_device *dev)
3358 {
3359         struct inet6_dev *idev;
3360
3361         ASSERT_RTNL();
3362
3363         if ((dev->type != ARPHRD_ETHER) &&
3364             (dev->type != ARPHRD_FDDI) &&
3365             (dev->type != ARPHRD_ARCNET) &&
3366             (dev->type != ARPHRD_INFINIBAND) &&
3367             (dev->type != ARPHRD_IEEE1394) &&
3368             (dev->type != ARPHRD_TUNNEL6) &&
3369             (dev->type != ARPHRD_6LOWPAN) &&
3370             (dev->type != ARPHRD_TUNNEL) &&
3371             (dev->type != ARPHRD_NONE) &&
3372             (dev->type != ARPHRD_RAWIP)) {
3373                 /* Alas, we support only Ethernet autoconfiguration. */
3374                 idev = __in6_dev_get(dev);
3375                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3376                     dev->flags & IFF_MULTICAST)
3377                         ipv6_mc_up(idev);
3378                 return;
3379         }
3380
3381         idev = addrconf_add_dev(dev);
3382         if (IS_ERR(idev))
3383                 return;
3384
3385         /* this device type has no EUI support */
3386         if (dev->type == ARPHRD_NONE &&
3387             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3388                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3389
3390         addrconf_addr_gen(idev, false);
3391 }
3392
3393 #if IS_ENABLED(CONFIG_IPV6_SIT)
3394 static void addrconf_sit_config(struct net_device *dev)
3395 {
3396         struct inet6_dev *idev;
3397
3398         ASSERT_RTNL();
3399
3400         /*
3401          * Configure the tunnel with one of our IPv4
3402          * addresses... we should configure all of
3403          * our v4 addrs in the tunnel
3404          */
3405
3406         idev = ipv6_find_idev(dev);
3407         if (IS_ERR(idev)) {
3408                 pr_debug("%s: add_dev failed\n", __func__);
3409                 return;
3410         }
3411
3412         if (dev->priv_flags & IFF_ISATAP) {
3413                 addrconf_addr_gen(idev, false);
3414                 return;
3415         }
3416
3417         add_v4_addrs(idev);
3418
3419         if (dev->flags&IFF_POINTOPOINT)
3420                 addrconf_add_mroute(dev);
3421 }
3422 #endif
3423
3424 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3425 static void addrconf_gre_config(struct net_device *dev)
3426 {
3427         struct inet6_dev *idev;
3428
3429         ASSERT_RTNL();
3430
3431         idev = ipv6_find_idev(dev);
3432         if (IS_ERR(idev)) {
3433                 pr_debug("%s: add_dev failed\n", __func__);
3434                 return;
3435         }
3436
3437         if (dev->type == ARPHRD_ETHER) {
3438                 addrconf_addr_gen(idev, true);
3439                 return;
3440         }
3441
3442         add_v4_addrs(idev);
3443
3444         if (dev->flags & IFF_POINTOPOINT)
3445                 addrconf_add_mroute(dev);
3446 }
3447 #endif
3448
3449 static void addrconf_init_auto_addrs(struct net_device *dev)
3450 {
3451         switch (dev->type) {
3452 #if IS_ENABLED(CONFIG_IPV6_SIT)
3453         case ARPHRD_SIT:
3454                 addrconf_sit_config(dev);
3455                 break;
3456 #endif
3457 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3458         case ARPHRD_IP6GRE:
3459         case ARPHRD_IPGRE:
3460                 addrconf_gre_config(dev);
3461                 break;
3462 #endif
3463         case ARPHRD_LOOPBACK:
3464                 init_loopback(dev);
3465                 break;
3466
3467         default:
3468                 addrconf_dev_config(dev);
3469                 break;
3470         }
3471 }
3472
3473 static int fixup_permanent_addr(struct net *net,
3474                                 struct inet6_dev *idev,
3475                                 struct inet6_ifaddr *ifp)
3476 {
3477         /* !fib6_node means the host route was removed from the
3478          * FIB, for example, if 'lo' device is taken down. In that
3479          * case regenerate the host route.
3480          */
3481         if (!ifp->rt || !ifp->rt->fib6_node) {
3482                 struct fib6_info *f6i, *prev;
3483
3484                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3485                                          GFP_ATOMIC);
3486                 if (IS_ERR(f6i))
3487                         return PTR_ERR(f6i);
3488
3489                 /* ifp->rt can be accessed outside of rtnl */
3490                 spin_lock(&ifp->lock);
3491                 prev = ifp->rt;
3492                 ifp->rt = f6i;
3493                 spin_unlock(&ifp->lock);
3494
3495                 fib6_info_release(prev);
3496         }
3497
3498         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3499                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3500                                       ifp->rt_priority, idev->dev, 0, 0,
3501                                       GFP_ATOMIC);
3502         }
3503
3504         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3505                 addrconf_dad_start(ifp);
3506
3507         return 0;
3508 }
3509
3510 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3511 {
3512         struct inet6_ifaddr *ifp, *tmp;
3513         struct inet6_dev *idev;
3514
3515         idev = __in6_dev_get(dev);
3516         if (!idev)
3517                 return;
3518
3519         write_lock_bh(&idev->lock);
3520
3521         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3522                 if ((ifp->flags & IFA_F_PERMANENT) &&
3523                     fixup_permanent_addr(net, idev, ifp) < 0) {
3524                         write_unlock_bh(&idev->lock);
3525                         in6_ifa_hold(ifp);
3526                         ipv6_del_addr(ifp);
3527                         write_lock_bh(&idev->lock);
3528
3529                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3530                                              idev->dev->name, &ifp->addr);
3531                 }
3532         }
3533
3534         write_unlock_bh(&idev->lock);
3535 }
3536
3537 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3538                            void *ptr)
3539 {
3540         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3541         struct netdev_notifier_change_info *change_info;
3542         struct netdev_notifier_changeupper_info *info;
3543         struct inet6_dev *idev = __in6_dev_get(dev);
3544         struct net *net = dev_net(dev);
3545         int run_pending = 0;
3546         int err;
3547
3548         switch (event) {
3549         case NETDEV_REGISTER:
3550                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3551                         idev = ipv6_add_dev(dev);
3552                         if (IS_ERR(idev))
3553                                 return notifier_from_errno(PTR_ERR(idev));
3554                 }
3555                 break;
3556
3557         case NETDEV_CHANGEMTU:
3558                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3559                 if (dev->mtu < IPV6_MIN_MTU) {
3560                         addrconf_ifdown(dev, dev != net->loopback_dev);
3561                         break;
3562                 }
3563
3564                 if (idev) {
3565                         rt6_mtu_change(dev, dev->mtu);
3566                         idev->cnf.mtu6 = dev->mtu;
3567                         break;
3568                 }
3569
3570                 /* allocate new idev */
3571                 idev = ipv6_add_dev(dev);
3572                 if (IS_ERR(idev))
3573                         break;
3574
3575                 /* device is still not ready */
3576                 if (!(idev->if_flags & IF_READY))
3577                         break;
3578
3579                 run_pending = 1;
3580                 fallthrough;
3581         case NETDEV_UP:
3582         case NETDEV_CHANGE:
3583                 if (idev && idev->cnf.disable_ipv6)
3584                         break;
3585
3586                 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3587                         if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3588                             dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3589                                 ipv6_mc_up(idev);
3590                         break;
3591                 }
3592
3593                 if (event == NETDEV_UP) {
3594                         /* restore routes for permanent addresses */
3595                         addrconf_permanent_addr(net, dev);
3596
3597                         if (!addrconf_link_ready(dev)) {
3598                                 /* device is not ready yet. */
3599                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3600                                          dev->name);
3601                                 break;
3602                         }
3603
3604                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3605                                 idev = ipv6_add_dev(dev);
3606
3607                         if (!IS_ERR_OR_NULL(idev)) {
3608                                 idev->if_flags |= IF_READY;
3609                                 run_pending = 1;
3610                         }
3611                 } else if (event == NETDEV_CHANGE) {
3612                         if (!addrconf_link_ready(dev)) {
3613                                 /* device is still not ready. */
3614                                 rt6_sync_down_dev(dev, event);
3615                                 break;
3616                         }
3617
3618                         if (!IS_ERR_OR_NULL(idev)) {
3619                                 if (idev->if_flags & IF_READY) {
3620                                         /* device is already configured -
3621                                          * but resend MLD reports, we might
3622                                          * have roamed and need to update
3623                                          * multicast snooping switches
3624                                          */
3625                                         ipv6_mc_up(idev);
3626                                         change_info = ptr;
3627                                         if (change_info->flags_changed & IFF_NOARP)
3628                                                 addrconf_dad_run(idev, true);
3629                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3630                                         break;
3631                                 }
3632                                 idev->if_flags |= IF_READY;
3633                         }
3634
3635                         pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3636                                  dev->name);
3637
3638                         run_pending = 1;
3639                 }
3640
3641                 addrconf_init_auto_addrs(dev);
3642
3643                 if (!IS_ERR_OR_NULL(idev)) {
3644                         if (run_pending)
3645                                 addrconf_dad_run(idev, false);
3646
3647                         /* Device has an address by now */
3648                         rt6_sync_up(dev, RTNH_F_DEAD);
3649
3650                         /*
3651                          * If the MTU changed during the interface down,
3652                          * when the interface up, the changed MTU must be
3653                          * reflected in the idev as well as routers.
3654                          */
3655                         if (idev->cnf.mtu6 != dev->mtu &&
3656                             dev->mtu >= IPV6_MIN_MTU) {
3657                                 rt6_mtu_change(dev, dev->mtu);
3658                                 idev->cnf.mtu6 = dev->mtu;
3659                         }
3660                         idev->tstamp = jiffies;
3661                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3662
3663                         /*
3664                          * If the changed mtu during down is lower than
3665                          * IPV6_MIN_MTU stop IPv6 on this interface.
3666                          */
3667                         if (dev->mtu < IPV6_MIN_MTU)
3668                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3669                 }
3670                 break;
3671
3672         case NETDEV_DOWN:
3673         case NETDEV_UNREGISTER:
3674                 /*
3675                  *      Remove all addresses from this interface.
3676                  */
3677                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3678                 break;
3679
3680         case NETDEV_CHANGENAME:
3681                 if (idev) {
3682                         snmp6_unregister_dev(idev);
3683                         addrconf_sysctl_unregister(idev);
3684                         err = addrconf_sysctl_register(idev);
3685                         if (err)
3686                                 return notifier_from_errno(err);
3687                         err = snmp6_register_dev(idev);
3688                         if (err) {
3689                                 addrconf_sysctl_unregister(idev);
3690                                 return notifier_from_errno(err);
3691                         }
3692                 }
3693                 break;
3694
3695         case NETDEV_PRE_TYPE_CHANGE:
3696         case NETDEV_POST_TYPE_CHANGE:
3697                 if (idev)
3698                         addrconf_type_change(dev, event);
3699                 break;
3700
3701         case NETDEV_CHANGEUPPER:
3702                 info = ptr;
3703
3704                 /* flush all routes if dev is linked to or unlinked from
3705                  * an L3 master device (e.g., VRF)
3706                  */
3707                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3708                         addrconf_ifdown(dev, false);
3709         }
3710
3711         return NOTIFY_OK;
3712 }
3713
3714 /*
3715  *      addrconf module should be notified of a device going up
3716  */
3717 static struct notifier_block ipv6_dev_notf = {
3718         .notifier_call = addrconf_notify,
3719         .priority = ADDRCONF_NOTIFY_PRIORITY,
3720 };
3721
3722 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3723 {
3724         struct inet6_dev *idev;
3725         ASSERT_RTNL();
3726
3727         idev = __in6_dev_get(dev);
3728
3729         if (event == NETDEV_POST_TYPE_CHANGE)
3730                 ipv6_mc_remap(idev);
3731         else if (event == NETDEV_PRE_TYPE_CHANGE)
3732                 ipv6_mc_unmap(idev);
3733 }
3734
3735 static bool addr_is_local(const struct in6_addr *addr)
3736 {
3737         return ipv6_addr_type(addr) &
3738                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3739 }
3740
3741 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3742 {
3743         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3744         struct net *net = dev_net(dev);
3745         struct inet6_dev *idev;
3746         struct inet6_ifaddr *ifa;
3747         LIST_HEAD(tmp_addr_list);
3748         bool keep_addr = false;
3749         bool was_ready;
3750         int state, i;
3751
3752         ASSERT_RTNL();
3753
3754         rt6_disable_ip(dev, event);
3755
3756         idev = __in6_dev_get(dev);
3757         if (!idev)
3758                 return -ENODEV;
3759
3760         /*
3761          * Step 1: remove reference to ipv6 device from parent device.
3762          *         Do not dev_put!
3763          */
3764         if (unregister) {
3765                 idev->dead = 1;
3766
3767                 /* protected by rtnl_lock */
3768                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3769
3770                 /* Step 1.5: remove snmp6 entry */
3771                 snmp6_unregister_dev(idev);
3772
3773         }
3774
3775         /* combine the user config with event to determine if permanent
3776          * addresses are to be removed from address hash table
3777          */
3778         if (!unregister && !idev->cnf.disable_ipv6) {
3779                 /* aggregate the system setting and interface setting */
3780                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3781
3782                 if (!_keep_addr)
3783                         _keep_addr = idev->cnf.keep_addr_on_down;
3784
3785                 keep_addr = (_keep_addr > 0);
3786         }
3787
3788         /* Step 2: clear hash table */
3789         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3790                 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3791
3792                 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3793 restart:
3794                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3795                         if (ifa->idev == idev) {
3796                                 addrconf_del_dad_work(ifa);
3797                                 /* combined flag + permanent flag decide if
3798                                  * address is retained on a down event
3799                                  */
3800                                 if (!keep_addr ||
3801                                     !(ifa->flags & IFA_F_PERMANENT) ||
3802                                     addr_is_local(&ifa->addr)) {
3803                                         hlist_del_init_rcu(&ifa->addr_lst);
3804                                         goto restart;
3805                                 }
3806                         }
3807                 }
3808                 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3809         }
3810
3811         write_lock_bh(&idev->lock);
3812
3813         addrconf_del_rs_timer(idev);
3814
3815         /* Step 2: clear flags for stateless addrconf, repeated down
3816          *         detection
3817          */
3818         was_ready = idev->if_flags & IF_READY;
3819         if (!unregister)
3820                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3821
3822         /* Step 3: clear tempaddr list */
3823         while (!list_empty(&idev->tempaddr_list)) {
3824                 ifa = list_first_entry(&idev->tempaddr_list,
3825                                        struct inet6_ifaddr, tmp_list);
3826                 list_del(&ifa->tmp_list);
3827                 write_unlock_bh(&idev->lock);
3828                 spin_lock_bh(&ifa->lock);
3829
3830                 if (ifa->ifpub) {
3831                         in6_ifa_put(ifa->ifpub);
3832                         ifa->ifpub = NULL;
3833                 }
3834                 spin_unlock_bh(&ifa->lock);
3835                 in6_ifa_put(ifa);
3836                 write_lock_bh(&idev->lock);
3837         }
3838
3839         list_for_each_entry(ifa, &idev->addr_list, if_list)
3840                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3841         write_unlock_bh(&idev->lock);
3842
3843         while (!list_empty(&tmp_addr_list)) {
3844                 struct fib6_info *rt = NULL;
3845                 bool keep;
3846
3847                 ifa = list_first_entry(&tmp_addr_list,
3848                                        struct inet6_ifaddr, if_list_aux);
3849                 list_del(&ifa->if_list_aux);
3850
3851                 addrconf_del_dad_work(ifa);
3852
3853                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3854                         !addr_is_local(&ifa->addr);
3855
3856                 spin_lock_bh(&ifa->lock);
3857
3858                 if (keep) {
3859                         /* set state to skip the notifier below */
3860                         state = INET6_IFADDR_STATE_DEAD;
3861                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3862                         if (!(ifa->flags & IFA_F_NODAD))
3863                                 ifa->flags |= IFA_F_TENTATIVE;
3864
3865                         rt = ifa->rt;
3866                         ifa->rt = NULL;
3867                 } else {
3868                         state = ifa->state;
3869                         ifa->state = INET6_IFADDR_STATE_DEAD;
3870                 }
3871
3872                 spin_unlock_bh(&ifa->lock);
3873
3874                 if (rt)
3875                         ip6_del_rt(net, rt, false);
3876
3877                 if (state != INET6_IFADDR_STATE_DEAD) {
3878                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3879                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3880                 } else {
3881                         if (idev->cnf.forwarding)
3882                                 addrconf_leave_anycast(ifa);
3883                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3884                 }
3885
3886                 if (!keep) {
3887                         write_lock_bh(&idev->lock);
3888                         list_del_rcu(&ifa->if_list);
3889                         write_unlock_bh(&idev->lock);
3890                         in6_ifa_put(ifa);
3891                 }
3892         }
3893
3894         /* Step 5: Discard anycast and multicast list */
3895         if (unregister) {
3896                 ipv6_ac_destroy_dev(idev);
3897                 ipv6_mc_destroy_dev(idev);
3898         } else if (was_ready) {
3899                 ipv6_mc_down(idev);
3900         }
3901
3902         idev->tstamp = jiffies;
3903         idev->ra_mtu = 0;
3904
3905         /* Last: Shot the device (if unregistered) */
3906         if (unregister) {
3907                 addrconf_sysctl_unregister(idev);
3908                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3909                 neigh_ifdown(&nd_tbl, dev);
3910                 in6_dev_put(idev);
3911         }
3912         return 0;
3913 }
3914
3915 static void addrconf_rs_timer(struct timer_list *t)
3916 {
3917         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3918         struct net_device *dev = idev->dev;
3919         struct in6_addr lladdr;
3920
3921         write_lock(&idev->lock);
3922         if (idev->dead || !(idev->if_flags & IF_READY))
3923                 goto out;
3924
3925         if (!ipv6_accept_ra(idev))
3926                 goto out;
3927
3928         /* Announcement received after solicitation was sent */
3929         if (idev->if_flags & IF_RA_RCVD)
3930                 goto out;
3931
3932         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3933                 write_unlock(&idev->lock);
3934                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3935                         ndisc_send_rs(dev, &lladdr,
3936                                       &in6addr_linklocal_allrouters);
3937                 else
3938                         goto put;
3939
3940                 write_lock(&idev->lock);
3941                 idev->rs_interval = rfc3315_s14_backoff_update(
3942                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3943                 /* The wait after the last probe can be shorter */
3944                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3945                                              idev->cnf.rtr_solicits) ?
3946                                       idev->cnf.rtr_solicit_delay :
3947                                       idev->rs_interval);
3948         } else {
3949                 /*
3950                  * Note: we do not support deprecated "all on-link"
3951                  * assumption any longer.
3952                  */
3953                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3954         }
3955
3956 out:
3957         write_unlock(&idev->lock);
3958 put:
3959         in6_dev_put(idev);
3960 }
3961
3962 /*
3963  *      Duplicate Address Detection
3964  */
3965 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3966 {
3967         unsigned long rand_num;
3968         struct inet6_dev *idev = ifp->idev;
3969         u64 nonce;
3970
3971         if (ifp->flags & IFA_F_OPTIMISTIC)
3972                 rand_num = 0;
3973         else
3974                 rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1);
3975
3976         nonce = 0;
3977         if (idev->cnf.enhanced_dad ||
3978             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3979                 do
3980                         get_random_bytes(&nonce, 6);
3981                 while (nonce == 0);
3982         }
3983         ifp->dad_nonce = nonce;
3984         ifp->dad_probes = idev->cnf.dad_transmits;
3985         addrconf_mod_dad_work(ifp, rand_num);
3986 }
3987
3988 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3989 {
3990         struct inet6_dev *idev = ifp->idev;
3991         struct net_device *dev = idev->dev;
3992         bool bump_id, notify = false;
3993         struct net *net;
3994
3995         addrconf_join_solict(dev, &ifp->addr);
3996
3997         read_lock_bh(&idev->lock);
3998         spin_lock(&ifp->lock);
3999         if (ifp->state == INET6_IFADDR_STATE_DEAD)
4000                 goto out;
4001
4002         net = dev_net(dev);
4003         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4004             (net->ipv6.devconf_all->accept_dad < 1 &&
4005              idev->cnf.accept_dad < 1) ||
4006             !(ifp->flags&IFA_F_TENTATIVE) ||
4007             ifp->flags & IFA_F_NODAD) {
4008                 bool send_na = false;
4009
4010                 if (ifp->flags & IFA_F_TENTATIVE &&
4011                     !(ifp->flags & IFA_F_OPTIMISTIC))
4012                         send_na = true;
4013                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4014                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4015                 spin_unlock(&ifp->lock);
4016                 read_unlock_bh(&idev->lock);
4017
4018                 addrconf_dad_completed(ifp, bump_id, send_na);
4019                 return;
4020         }
4021
4022         if (!(idev->if_flags & IF_READY)) {
4023                 spin_unlock(&ifp->lock);
4024                 read_unlock_bh(&idev->lock);
4025                 /*
4026                  * If the device is not ready:
4027                  * - keep it tentative if it is a permanent address.
4028                  * - otherwise, kill it.
4029                  */
4030                 in6_ifa_hold(ifp);
4031                 addrconf_dad_stop(ifp, 0);
4032                 return;
4033         }
4034
4035         /*
4036          * Optimistic nodes can start receiving
4037          * Frames right away
4038          */
4039         if (ifp->flags & IFA_F_OPTIMISTIC) {
4040                 ip6_ins_rt(net, ifp->rt);
4041                 if (ipv6_use_optimistic_addr(net, idev)) {
4042                         /* Because optimistic nodes can use this address,
4043                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4044                          */
4045                         notify = true;
4046                 }
4047         }
4048
4049         addrconf_dad_kick(ifp);
4050 out:
4051         spin_unlock(&ifp->lock);
4052         read_unlock_bh(&idev->lock);
4053         if (notify)
4054                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4055 }
4056
4057 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4058 {
4059         bool begin_dad = false;
4060
4061         spin_lock_bh(&ifp->lock);
4062         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4063                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4064                 begin_dad = true;
4065         }
4066         spin_unlock_bh(&ifp->lock);
4067
4068         if (begin_dad)
4069                 addrconf_mod_dad_work(ifp, 0);
4070 }
4071
4072 static void addrconf_dad_work(struct work_struct *w)
4073 {
4074         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4075                                                 struct inet6_ifaddr,
4076                                                 dad_work);
4077         struct inet6_dev *idev = ifp->idev;
4078         bool bump_id, disable_ipv6 = false;
4079         struct in6_addr mcaddr;
4080
4081         enum {
4082                 DAD_PROCESS,
4083                 DAD_BEGIN,
4084                 DAD_ABORT,
4085         } action = DAD_PROCESS;
4086
4087         rtnl_lock();
4088
4089         spin_lock_bh(&ifp->lock);
4090         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4091                 action = DAD_BEGIN;
4092                 ifp->state = INET6_IFADDR_STATE_DAD;
4093         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4094                 action = DAD_ABORT;
4095                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4096
4097                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4098                      idev->cnf.accept_dad > 1) &&
4099                     !idev->cnf.disable_ipv6 &&
4100                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4101                         struct in6_addr addr;
4102
4103                         addr.s6_addr32[0] = htonl(0xfe800000);
4104                         addr.s6_addr32[1] = 0;
4105
4106                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4107                             ipv6_addr_equal(&ifp->addr, &addr)) {
4108                                 /* DAD failed for link-local based on MAC */
4109                                 idev->cnf.disable_ipv6 = 1;
4110
4111                                 pr_info("%s: IPv6 being disabled!\n",
4112                                         ifp->idev->dev->name);
4113                                 disable_ipv6 = true;
4114                         }
4115                 }
4116         }
4117         spin_unlock_bh(&ifp->lock);
4118
4119         if (action == DAD_BEGIN) {
4120                 addrconf_dad_begin(ifp);
4121                 goto out;
4122         } else if (action == DAD_ABORT) {
4123                 in6_ifa_hold(ifp);
4124                 addrconf_dad_stop(ifp, 1);
4125                 if (disable_ipv6)
4126                         addrconf_ifdown(idev->dev, false);
4127                 goto out;
4128         }
4129
4130         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4131                 goto out;
4132
4133         write_lock_bh(&idev->lock);
4134         if (idev->dead || !(idev->if_flags & IF_READY)) {
4135                 write_unlock_bh(&idev->lock);
4136                 goto out;
4137         }
4138
4139         spin_lock(&ifp->lock);
4140         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4141                 spin_unlock(&ifp->lock);
4142                 write_unlock_bh(&idev->lock);
4143                 goto out;
4144         }
4145
4146         if (ifp->dad_probes == 0) {
4147                 bool send_na = false;
4148
4149                 /*
4150                  * DAD was successful
4151                  */
4152
4153                 if (ifp->flags & IFA_F_TENTATIVE &&
4154                     !(ifp->flags & IFA_F_OPTIMISTIC))
4155                         send_na = true;
4156                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4157                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4158                 spin_unlock(&ifp->lock);
4159                 write_unlock_bh(&idev->lock);
4160
4161                 addrconf_dad_completed(ifp, bump_id, send_na);
4162
4163                 goto out;
4164         }
4165
4166         ifp->dad_probes--;
4167         addrconf_mod_dad_work(ifp,
4168                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4169                                   HZ/100));
4170         spin_unlock(&ifp->lock);
4171         write_unlock_bh(&idev->lock);
4172
4173         /* send a neighbour solicitation for our addr */
4174         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4175         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4176                       ifp->dad_nonce);
4177 out:
4178         in6_ifa_put(ifp);
4179         rtnl_unlock();
4180 }
4181
4182 /* ifp->idev must be at least read locked */
4183 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4184 {
4185         struct inet6_ifaddr *ifpiter;
4186         struct inet6_dev *idev = ifp->idev;
4187
4188         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4189                 if (ifpiter->scope > IFA_LINK)
4190                         break;
4191                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4192                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4193                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4194                     IFA_F_PERMANENT)
4195                         return false;
4196         }
4197         return true;
4198 }
4199
4200 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4201                                    bool send_na)
4202 {
4203         struct net_device *dev = ifp->idev->dev;
4204         struct in6_addr lladdr;
4205         bool send_rs, send_mld;
4206
4207         addrconf_del_dad_work(ifp);
4208
4209         /*
4210          *      Configure the address for reception. Now it is valid.
4211          */
4212
4213         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4214
4215         /* If added prefix is link local and we are prepared to process
4216            router advertisements, start sending router solicitations.
4217          */
4218
4219         read_lock_bh(&ifp->idev->lock);
4220         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4221         send_rs = send_mld &&
4222                   ipv6_accept_ra(ifp->idev) &&
4223                   ifp->idev->cnf.rtr_solicits != 0 &&
4224                   (dev->flags & IFF_LOOPBACK) == 0 &&
4225                   (dev->type != ARPHRD_TUNNEL) &&
4226                   !netif_is_team_port(dev);
4227         read_unlock_bh(&ifp->idev->lock);
4228
4229         /* While dad is in progress mld report's source address is in6_addrany.
4230          * Resend with proper ll now.
4231          */
4232         if (send_mld)
4233                 ipv6_mc_dad_complete(ifp->idev);
4234
4235         /* send unsolicited NA if enabled */
4236         if (send_na &&
4237             (ifp->idev->cnf.ndisc_notify ||
4238              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4239                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4240                               /*router=*/ !!ifp->idev->cnf.forwarding,
4241                               /*solicited=*/ false, /*override=*/ true,
4242                               /*inc_opt=*/ true);
4243         }
4244
4245         if (send_rs) {
4246                 /*
4247                  *      If a host as already performed a random delay
4248                  *      [...] as part of DAD [...] there is no need
4249                  *      to delay again before sending the first RS
4250                  */
4251                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4252                         return;
4253                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4254
4255                 write_lock_bh(&ifp->idev->lock);
4256                 spin_lock(&ifp->lock);
4257                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4258                         ifp->idev->cnf.rtr_solicit_interval);
4259                 ifp->idev->rs_probes = 1;
4260                 ifp->idev->if_flags |= IF_RS_SENT;
4261                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4262                 spin_unlock(&ifp->lock);
4263                 write_unlock_bh(&ifp->idev->lock);
4264         }
4265
4266         if (bump_id)
4267                 rt_genid_bump_ipv6(dev_net(dev));
4268
4269         /* Make sure that a new temporary address will be created
4270          * before this temporary address becomes deprecated.
4271          */
4272         if (ifp->flags & IFA_F_TEMPORARY)
4273                 addrconf_verify_rtnl(dev_net(dev));
4274 }
4275
4276 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4277 {
4278         struct inet6_ifaddr *ifp;
4279
4280         read_lock_bh(&idev->lock);
4281         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4282                 spin_lock(&ifp->lock);
4283                 if ((ifp->flags & IFA_F_TENTATIVE &&
4284                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4285                         if (restart)
4286                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4287                         addrconf_dad_kick(ifp);
4288                 }
4289                 spin_unlock(&ifp->lock);
4290         }
4291         read_unlock_bh(&idev->lock);
4292 }
4293
4294 #ifdef CONFIG_PROC_FS
4295 struct if6_iter_state {
4296         struct seq_net_private p;
4297         int bucket;
4298         int offset;
4299 };
4300
4301 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4302 {
4303         struct if6_iter_state *state = seq->private;
4304         struct net *net = seq_file_net(seq);
4305         struct inet6_ifaddr *ifa = NULL;
4306         int p = 0;
4307
4308         /* initial bucket if pos is 0 */
4309         if (pos == 0) {
4310                 state->bucket = 0;
4311                 state->offset = 0;
4312         }
4313
4314         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4315                 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4316                                          addr_lst) {
4317                         /* sync with offset */
4318                         if (p < state->offset) {
4319                                 p++;
4320                                 continue;
4321                         }
4322                         return ifa;
4323                 }
4324
4325                 /* prepare for next bucket */
4326                 state->offset = 0;
4327                 p = 0;
4328         }
4329         return NULL;
4330 }
4331
4332 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4333                                          struct inet6_ifaddr *ifa)
4334 {
4335         struct if6_iter_state *state = seq->private;
4336         struct net *net = seq_file_net(seq);
4337
4338         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4339                 state->offset++;
4340                 return ifa;
4341         }
4342
4343         state->offset = 0;
4344         while (++state->bucket < IN6_ADDR_HSIZE) {
4345                 hlist_for_each_entry_rcu(ifa,
4346                                      &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4347                         return ifa;
4348                 }
4349         }
4350
4351         return NULL;
4352 }
4353
4354 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4355         __acquires(rcu)
4356 {
4357         rcu_read_lock();
4358         return if6_get_first(seq, *pos);
4359 }
4360
4361 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4362 {
4363         struct inet6_ifaddr *ifa;
4364
4365         ifa = if6_get_next(seq, v);
4366         ++*pos;
4367         return ifa;
4368 }
4369
4370 static void if6_seq_stop(struct seq_file *seq, void *v)
4371         __releases(rcu)
4372 {
4373         rcu_read_unlock();
4374 }
4375
4376 static int if6_seq_show(struct seq_file *seq, void *v)
4377 {
4378         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4379         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4380                    &ifp->addr,
4381                    ifp->idev->dev->ifindex,
4382                    ifp->prefix_len,
4383                    ifp->scope,
4384                    (u8) ifp->flags,
4385                    ifp->idev->dev->name);
4386         return 0;
4387 }
4388
4389 static const struct seq_operations if6_seq_ops = {
4390         .start  = if6_seq_start,
4391         .next   = if6_seq_next,
4392         .show   = if6_seq_show,
4393         .stop   = if6_seq_stop,
4394 };
4395
4396 static int __net_init if6_proc_net_init(struct net *net)
4397 {
4398         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4399                         sizeof(struct if6_iter_state)))
4400                 return -ENOMEM;
4401         return 0;
4402 }
4403
4404 static void __net_exit if6_proc_net_exit(struct net *net)
4405 {
4406         remove_proc_entry("if_inet6", net->proc_net);
4407 }
4408
4409 static struct pernet_operations if6_proc_net_ops = {
4410         .init = if6_proc_net_init,
4411         .exit = if6_proc_net_exit,
4412 };
4413
4414 int __init if6_proc_init(void)
4415 {
4416         return register_pernet_subsys(&if6_proc_net_ops);
4417 }
4418
4419 void if6_proc_exit(void)
4420 {
4421         unregister_pernet_subsys(&if6_proc_net_ops);
4422 }
4423 #endif  /* CONFIG_PROC_FS */
4424
4425 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4426 /* Check if address is a home address configured on any interface. */
4427 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4428 {
4429         unsigned int hash = inet6_addr_hash(net, addr);
4430         struct inet6_ifaddr *ifp = NULL;
4431         int ret = 0;
4432
4433         rcu_read_lock();
4434         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4435                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4436                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4437                         ret = 1;
4438                         break;
4439                 }
4440         }
4441         rcu_read_unlock();
4442         return ret;
4443 }
4444 #endif
4445
4446 /* RFC6554 has some algorithm to avoid loops in segment routing by
4447  * checking if the segments contains any of a local interface address.
4448  *
4449  * Quote:
4450  *
4451  * To detect loops in the SRH, a router MUST determine if the SRH
4452  * includes multiple addresses assigned to any interface on that router.
4453  * If such addresses appear more than once and are separated by at least
4454  * one address not assigned to that router.
4455  */
4456 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4457                           unsigned char nsegs)
4458 {
4459         const struct in6_addr *addr;
4460         int i, ret = 0, found = 0;
4461         struct inet6_ifaddr *ifp;
4462         bool separated = false;
4463         unsigned int hash;
4464         bool hash_found;
4465
4466         rcu_read_lock();
4467         for (i = 0; i < nsegs; i++) {
4468                 addr = &segs[i];
4469                 hash = inet6_addr_hash(net, addr);
4470
4471                 hash_found = false;
4472                 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4473
4474                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4475                                 hash_found = true;
4476                                 break;
4477                         }
4478                 }
4479
4480                 if (hash_found) {
4481                         if (found > 1 && separated) {
4482                                 ret = 1;
4483                                 break;
4484                         }
4485
4486                         separated = false;
4487                         found++;
4488                 } else {
4489                         separated = true;
4490                 }
4491         }
4492         rcu_read_unlock();
4493
4494         return ret;
4495 }
4496
4497 /*
4498  *      Periodic address status verification
4499  */
4500
4501 static void addrconf_verify_rtnl(struct net *net)
4502 {
4503         unsigned long now, next, next_sec, next_sched;
4504         struct inet6_ifaddr *ifp;
4505         int i;
4506
4507         ASSERT_RTNL();
4508
4509         rcu_read_lock_bh();
4510         now = jiffies;
4511         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4512
4513         cancel_delayed_work(&net->ipv6.addr_chk_work);
4514
4515         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4516 restart:
4517                 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4518                         unsigned long age;
4519
4520                         /* When setting preferred_lft to a value not zero or
4521                          * infinity, while valid_lft is infinity
4522                          * IFA_F_PERMANENT has a non-infinity life time.
4523                          */
4524                         if ((ifp->flags & IFA_F_PERMANENT) &&
4525                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4526                                 continue;
4527
4528                         spin_lock(&ifp->lock);
4529                         /* We try to batch several events at once. */
4530                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4531
4532                         if ((ifp->flags&IFA_F_TEMPORARY) &&
4533                             !(ifp->flags&IFA_F_TENTATIVE) &&
4534                             ifp->prefered_lft != INFINITY_LIFE_TIME &&
4535                             !ifp->regen_count && ifp->ifpub) {
4536                                 /* This is a non-regenerated temporary addr. */
4537
4538                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4539                                         ifp->idev->cnf.dad_transmits *
4540                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4541
4542                                 if (age + regen_advance >= ifp->prefered_lft) {
4543                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4544                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4545                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4546
4547                                         ifp->regen_count++;
4548                                         in6_ifa_hold(ifp);
4549                                         in6_ifa_hold(ifpub);
4550                                         spin_unlock(&ifp->lock);
4551
4552                                         spin_lock(&ifpub->lock);
4553                                         ifpub->regen_count = 0;
4554                                         spin_unlock(&ifpub->lock);
4555                                         rcu_read_unlock_bh();
4556                                         ipv6_create_tempaddr(ifpub, true);
4557                                         in6_ifa_put(ifpub);
4558                                         in6_ifa_put(ifp);
4559                                         rcu_read_lock_bh();
4560                                         goto restart;
4561                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4562                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4563                         }
4564
4565                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4566                             age >= ifp->valid_lft) {
4567                                 spin_unlock(&ifp->lock);
4568                                 in6_ifa_hold(ifp);
4569                                 rcu_read_unlock_bh();
4570                                 ipv6_del_addr(ifp);
4571                                 rcu_read_lock_bh();
4572                                 goto restart;
4573                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4574                                 spin_unlock(&ifp->lock);
4575                                 continue;
4576                         } else if (age >= ifp->prefered_lft) {
4577                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4578                                 int deprecate = 0;
4579
4580                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4581                                         deprecate = 1;
4582                                         ifp->flags |= IFA_F_DEPRECATED;
4583                                 }
4584
4585                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4586                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4587                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4588
4589                                 spin_unlock(&ifp->lock);
4590
4591                                 if (deprecate) {
4592                                         in6_ifa_hold(ifp);
4593
4594                                         ipv6_ifa_notify(0, ifp);
4595                                         in6_ifa_put(ifp);
4596                                         goto restart;
4597                                 }
4598                         } else {
4599                                 /* ifp->prefered_lft <= ifp->valid_lft */
4600                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4601                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4602                                 spin_unlock(&ifp->lock);
4603                         }
4604                 }
4605         }
4606
4607         next_sec = round_jiffies_up(next);
4608         next_sched = next;
4609
4610         /* If rounded timeout is accurate enough, accept it. */
4611         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4612                 next_sched = next_sec;
4613
4614         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4615         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4616                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4617
4618         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4619                  now, next, next_sec, next_sched);
4620         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4621         rcu_read_unlock_bh();
4622 }
4623
4624 static void addrconf_verify_work(struct work_struct *w)
4625 {
4626         struct net *net = container_of(to_delayed_work(w), struct net,
4627                                        ipv6.addr_chk_work);
4628
4629         rtnl_lock();
4630         addrconf_verify_rtnl(net);
4631         rtnl_unlock();
4632 }
4633
4634 static void addrconf_verify(struct net *net)
4635 {
4636         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4637 }
4638
4639 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4640                                      struct in6_addr **peer_pfx)
4641 {
4642         struct in6_addr *pfx = NULL;
4643
4644         *peer_pfx = NULL;
4645
4646         if (addr)
4647                 pfx = nla_data(addr);
4648
4649         if (local) {
4650                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4651                         *peer_pfx = pfx;
4652                 pfx = nla_data(local);
4653         }
4654
4655         return pfx;
4656 }
4657
4658 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4659         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4660         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4661         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4662         [IFA_FLAGS]             = { .len = sizeof(u32) },
4663         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4664         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4665         [IFA_PROTO]             = { .type = NLA_U8 },
4666 };
4667
4668 static int
4669 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4670                   struct netlink_ext_ack *extack)
4671 {
4672         struct net *net = sock_net(skb->sk);
4673         struct ifaddrmsg *ifm;
4674         struct nlattr *tb[IFA_MAX+1];
4675         struct in6_addr *pfx, *peer_pfx;
4676         u32 ifa_flags;
4677         int err;
4678
4679         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4680                                      ifa_ipv6_policy, extack);
4681         if (err < 0)
4682                 return err;
4683
4684         ifm = nlmsg_data(nlh);
4685         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4686         if (!pfx)
4687                 return -EINVAL;
4688
4689         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4690
4691         /* We ignore other flags so far. */
4692         ifa_flags &= IFA_F_MANAGETEMPADDR;
4693
4694         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4695                               ifm->ifa_prefixlen);
4696 }
4697
4698 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4699                                unsigned long expires, u32 flags,
4700                                bool modify_peer)
4701 {
4702         struct fib6_info *f6i;
4703         u32 prio;
4704
4705         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4706                                         ifp->prefix_len,
4707                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4708         if (!f6i)
4709                 return -ENOENT;
4710
4711         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4712         if (f6i->fib6_metric != prio) {
4713                 /* delete old one */
4714                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4715
4716                 /* add new one */
4717                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4718                                       ifp->prefix_len,
4719                                       ifp->rt_priority, ifp->idev->dev,
4720                                       expires, flags, GFP_KERNEL);
4721         } else {
4722                 if (!expires)
4723                         fib6_clean_expires(f6i);
4724                 else
4725                         fib6_set_expires(f6i, expires);
4726
4727                 fib6_info_release(f6i);
4728         }
4729
4730         return 0;
4731 }
4732
4733 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4734                              struct ifa6_config *cfg)
4735 {
4736         u32 flags;
4737         clock_t expires;
4738         unsigned long timeout;
4739         bool was_managetempaddr;
4740         bool had_prefixroute;
4741         bool new_peer = false;
4742
4743         ASSERT_RTNL();
4744
4745         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4746                 return -EINVAL;
4747
4748         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4749             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4750                 return -EINVAL;
4751
4752         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4753                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4754
4755         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4756         if (addrconf_finite_timeout(timeout)) {
4757                 expires = jiffies_to_clock_t(timeout * HZ);
4758                 cfg->valid_lft = timeout;
4759                 flags = RTF_EXPIRES;
4760         } else {
4761                 expires = 0;
4762                 flags = 0;
4763                 cfg->ifa_flags |= IFA_F_PERMANENT;
4764         }
4765
4766         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4767         if (addrconf_finite_timeout(timeout)) {
4768                 if (timeout == 0)
4769                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4770                 cfg->preferred_lft = timeout;
4771         }
4772
4773         if (cfg->peer_pfx &&
4774             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4775                 if (!ipv6_addr_any(&ifp->peer_addr))
4776                         cleanup_prefix_route(ifp, expires, true, true);
4777                 new_peer = true;
4778         }
4779
4780         spin_lock_bh(&ifp->lock);
4781         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4782         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4783                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4784         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4785                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4786                         IFA_F_NOPREFIXROUTE);
4787         ifp->flags |= cfg->ifa_flags;
4788         ifp->tstamp = jiffies;
4789         ifp->valid_lft = cfg->valid_lft;
4790         ifp->prefered_lft = cfg->preferred_lft;
4791         ifp->ifa_proto = cfg->ifa_proto;
4792
4793         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4794                 ifp->rt_priority = cfg->rt_priority;
4795
4796         if (new_peer)
4797                 ifp->peer_addr = *cfg->peer_pfx;
4798
4799         spin_unlock_bh(&ifp->lock);
4800         if (!(ifp->flags&IFA_F_TENTATIVE))
4801                 ipv6_ifa_notify(0, ifp);
4802
4803         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4804                 int rc = -ENOENT;
4805
4806                 if (had_prefixroute)
4807                         rc = modify_prefix_route(ifp, expires, flags, false);
4808
4809                 /* prefix route could have been deleted; if so restore it */
4810                 if (rc == -ENOENT) {
4811                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4812                                               ifp->rt_priority, ifp->idev->dev,
4813                                               expires, flags, GFP_KERNEL);
4814                 }
4815
4816                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4817                         rc = modify_prefix_route(ifp, expires, flags, true);
4818
4819                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4820                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4821                                               ifp->rt_priority, ifp->idev->dev,
4822                                               expires, flags, GFP_KERNEL);
4823                 }
4824         } else if (had_prefixroute) {
4825                 enum cleanup_prefix_rt_t action;
4826                 unsigned long rt_expires;
4827
4828                 write_lock_bh(&ifp->idev->lock);
4829                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4830                 write_unlock_bh(&ifp->idev->lock);
4831
4832                 if (action != CLEANUP_PREFIX_RT_NOP) {
4833                         cleanup_prefix_route(ifp, rt_expires,
4834                                 action == CLEANUP_PREFIX_RT_DEL, false);
4835                 }
4836         }
4837
4838         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4839                 if (was_managetempaddr &&
4840                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4841                         cfg->valid_lft = 0;
4842                         cfg->preferred_lft = 0;
4843                 }
4844                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4845                                  cfg->preferred_lft, !was_managetempaddr,
4846                                  jiffies);
4847         }
4848
4849         addrconf_verify_rtnl(net);
4850
4851         return 0;
4852 }
4853
4854 static int
4855 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4856                   struct netlink_ext_ack *extack)
4857 {
4858         struct net *net = sock_net(skb->sk);
4859         struct ifaddrmsg *ifm;
4860         struct nlattr *tb[IFA_MAX+1];
4861         struct in6_addr *peer_pfx;
4862         struct inet6_ifaddr *ifa;
4863         struct net_device *dev;
4864         struct inet6_dev *idev;
4865         struct ifa6_config cfg;
4866         int err;
4867
4868         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4869                                      ifa_ipv6_policy, extack);
4870         if (err < 0)
4871                 return err;
4872
4873         memset(&cfg, 0, sizeof(cfg));
4874
4875         ifm = nlmsg_data(nlh);
4876         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4877         if (!cfg.pfx)
4878                 return -EINVAL;
4879
4880         cfg.peer_pfx = peer_pfx;
4881         cfg.plen = ifm->ifa_prefixlen;
4882         if (tb[IFA_RT_PRIORITY])
4883                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4884
4885         if (tb[IFA_PROTO])
4886                 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4887
4888         cfg.valid_lft = INFINITY_LIFE_TIME;
4889         cfg.preferred_lft = INFINITY_LIFE_TIME;
4890
4891         if (tb[IFA_CACHEINFO]) {
4892                 struct ifa_cacheinfo *ci;
4893
4894                 ci = nla_data(tb[IFA_CACHEINFO]);
4895                 cfg.valid_lft = ci->ifa_valid;
4896                 cfg.preferred_lft = ci->ifa_prefered;
4897         }
4898
4899         dev =  __dev_get_by_index(net, ifm->ifa_index);
4900         if (!dev)
4901                 return -ENODEV;
4902
4903         if (tb[IFA_FLAGS])
4904                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4905         else
4906                 cfg.ifa_flags = ifm->ifa_flags;
4907
4908         /* We ignore other flags so far. */
4909         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4910                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4911                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4912
4913         idev = ipv6_find_idev(dev);
4914         if (IS_ERR(idev))
4915                 return PTR_ERR(idev);
4916
4917         if (!ipv6_allow_optimistic_dad(net, idev))
4918                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4919
4920         if (cfg.ifa_flags & IFA_F_NODAD &&
4921             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4922                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4923                 return -EINVAL;
4924         }
4925
4926         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4927         if (!ifa) {
4928                 /*
4929                  * It would be best to check for !NLM_F_CREATE here but
4930                  * userspace already relies on not having to provide this.
4931                  */
4932                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4933         }
4934
4935         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4936             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4937                 err = -EEXIST;
4938         else
4939                 err = inet6_addr_modify(net, ifa, &cfg);
4940
4941         in6_ifa_put(ifa);
4942
4943         return err;
4944 }
4945
4946 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4947                           u8 scope, int ifindex)
4948 {
4949         struct ifaddrmsg *ifm;
4950
4951         ifm = nlmsg_data(nlh);
4952         ifm->ifa_family = AF_INET6;
4953         ifm->ifa_prefixlen = prefixlen;
4954         ifm->ifa_flags = flags;
4955         ifm->ifa_scope = scope;
4956         ifm->ifa_index = ifindex;
4957 }
4958
4959 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4960                          unsigned long tstamp, u32 preferred, u32 valid)
4961 {
4962         struct ifa_cacheinfo ci;
4963
4964         ci.cstamp = cstamp_delta(cstamp);
4965         ci.tstamp = cstamp_delta(tstamp);
4966         ci.ifa_prefered = preferred;
4967         ci.ifa_valid = valid;
4968
4969         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4970 }
4971
4972 static inline int rt_scope(int ifa_scope)
4973 {
4974         if (ifa_scope & IFA_HOST)
4975                 return RT_SCOPE_HOST;
4976         else if (ifa_scope & IFA_LINK)
4977                 return RT_SCOPE_LINK;
4978         else if (ifa_scope & IFA_SITE)
4979                 return RT_SCOPE_SITE;
4980         else
4981                 return RT_SCOPE_UNIVERSE;
4982 }
4983
4984 static inline int inet6_ifaddr_msgsize(void)
4985 {
4986         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4987                + nla_total_size(16) /* IFA_LOCAL */
4988                + nla_total_size(16) /* IFA_ADDRESS */
4989                + nla_total_size(sizeof(struct ifa_cacheinfo))
4990                + nla_total_size(4)  /* IFA_FLAGS */
4991                + nla_total_size(1)  /* IFA_PROTO */
4992                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4993 }
4994
4995 enum addr_type_t {
4996         UNICAST_ADDR,
4997         MULTICAST_ADDR,
4998         ANYCAST_ADDR,
4999 };
5000
5001 struct inet6_fill_args {
5002         u32 portid;
5003         u32 seq;
5004         int event;
5005         unsigned int flags;
5006         int netnsid;
5007         int ifindex;
5008         enum addr_type_t type;
5009 };
5010
5011 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5012                              struct inet6_fill_args *args)
5013 {
5014         struct nlmsghdr  *nlh;
5015         u32 preferred, valid;
5016
5017         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5018                         sizeof(struct ifaddrmsg), args->flags);
5019         if (!nlh)
5020                 return -EMSGSIZE;
5021
5022         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5023                       ifa->idev->dev->ifindex);
5024
5025         if (args->netnsid >= 0 &&
5026             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5027                 goto error;
5028
5029         spin_lock_bh(&ifa->lock);
5030         if (!((ifa->flags&IFA_F_PERMANENT) &&
5031               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5032                 preferred = ifa->prefered_lft;
5033                 valid = ifa->valid_lft;
5034                 if (preferred != INFINITY_LIFE_TIME) {
5035                         long tval = (jiffies - ifa->tstamp)/HZ;
5036                         if (preferred > tval)
5037                                 preferred -= tval;
5038                         else
5039                                 preferred = 0;
5040                         if (valid != INFINITY_LIFE_TIME) {
5041                                 if (valid > tval)
5042                                         valid -= tval;
5043                                 else
5044                                         valid = 0;
5045                         }
5046                 }
5047         } else {
5048                 preferred = INFINITY_LIFE_TIME;
5049                 valid = INFINITY_LIFE_TIME;
5050         }
5051         spin_unlock_bh(&ifa->lock);
5052
5053         if (!ipv6_addr_any(&ifa->peer_addr)) {
5054                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5055                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5056                         goto error;
5057         } else
5058                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5059                         goto error;
5060
5061         if (ifa->rt_priority &&
5062             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5063                 goto error;
5064
5065         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5066                 goto error;
5067
5068         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5069                 goto error;
5070
5071         if (ifa->ifa_proto &&
5072             nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5073                 goto error;
5074
5075         nlmsg_end(skb, nlh);
5076         return 0;
5077
5078 error:
5079         nlmsg_cancel(skb, nlh);
5080         return -EMSGSIZE;
5081 }
5082
5083 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5084                                struct inet6_fill_args *args)
5085 {
5086         struct nlmsghdr  *nlh;
5087         u8 scope = RT_SCOPE_UNIVERSE;
5088         int ifindex = ifmca->idev->dev->ifindex;
5089
5090         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5091                 scope = RT_SCOPE_SITE;
5092
5093         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5094                         sizeof(struct ifaddrmsg), args->flags);
5095         if (!nlh)
5096                 return -EMSGSIZE;
5097
5098         if (args->netnsid >= 0 &&
5099             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5100                 nlmsg_cancel(skb, nlh);
5101                 return -EMSGSIZE;
5102         }
5103
5104         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5105         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5106             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5107                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5108                 nlmsg_cancel(skb, nlh);
5109                 return -EMSGSIZE;
5110         }
5111
5112         nlmsg_end(skb, nlh);
5113         return 0;
5114 }
5115
5116 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5117                                struct inet6_fill_args *args)
5118 {
5119         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5120         int ifindex = dev ? dev->ifindex : 1;
5121         struct nlmsghdr  *nlh;
5122         u8 scope = RT_SCOPE_UNIVERSE;
5123
5124         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5125                 scope = RT_SCOPE_SITE;
5126
5127         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5128                         sizeof(struct ifaddrmsg), args->flags);
5129         if (!nlh)
5130                 return -EMSGSIZE;
5131
5132         if (args->netnsid >= 0 &&
5133             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5134                 nlmsg_cancel(skb, nlh);
5135                 return -EMSGSIZE;
5136         }
5137
5138         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5139         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5140             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5141                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5142                 nlmsg_cancel(skb, nlh);
5143                 return -EMSGSIZE;
5144         }
5145
5146         nlmsg_end(skb, nlh);
5147         return 0;
5148 }
5149
5150 /* called with rcu_read_lock() */
5151 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5152                           struct netlink_callback *cb, int s_ip_idx,
5153                           struct inet6_fill_args *fillargs)
5154 {
5155         struct ifmcaddr6 *ifmca;
5156         struct ifacaddr6 *ifaca;
5157         int ip_idx = 0;
5158         int err = 1;
5159
5160         read_lock_bh(&idev->lock);
5161         switch (fillargs->type) {
5162         case UNICAST_ADDR: {
5163                 struct inet6_ifaddr *ifa;
5164                 fillargs->event = RTM_NEWADDR;
5165
5166                 /* unicast address incl. temp addr */
5167                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5168                         if (ip_idx < s_ip_idx)
5169                                 goto next;
5170                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5171                         if (err < 0)
5172                                 break;
5173                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5174 next:
5175                         ip_idx++;
5176                 }
5177                 break;
5178         }
5179         case MULTICAST_ADDR:
5180                 read_unlock_bh(&idev->lock);
5181                 fillargs->event = RTM_GETMULTICAST;
5182
5183                 /* multicast address */
5184                 for (ifmca = rtnl_dereference(idev->mc_list);
5185                      ifmca;
5186                      ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5187                         if (ip_idx < s_ip_idx)
5188                                 continue;
5189                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5190                         if (err < 0)
5191                                 break;
5192                 }
5193                 read_lock_bh(&idev->lock);
5194                 break;
5195         case ANYCAST_ADDR:
5196                 fillargs->event = RTM_GETANYCAST;
5197                 /* anycast address */
5198                 for (ifaca = idev->ac_list; ifaca;
5199                      ifaca = ifaca->aca_next, ip_idx++) {
5200                         if (ip_idx < s_ip_idx)
5201                                 continue;
5202                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5203                         if (err < 0)
5204                                 break;
5205                 }
5206                 break;
5207         default:
5208                 break;
5209         }
5210         read_unlock_bh(&idev->lock);
5211         cb->args[2] = ip_idx;
5212         return err;
5213 }
5214
5215 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5216                                        struct inet6_fill_args *fillargs,
5217                                        struct net **tgt_net, struct sock *sk,
5218                                        struct netlink_callback *cb)
5219 {
5220         struct netlink_ext_ack *extack = cb->extack;
5221         struct nlattr *tb[IFA_MAX+1];
5222         struct ifaddrmsg *ifm;
5223         int err, i;
5224
5225         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5226                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5227                 return -EINVAL;
5228         }
5229
5230         ifm = nlmsg_data(nlh);
5231         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5232                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5233                 return -EINVAL;
5234         }
5235
5236         fillargs->ifindex = ifm->ifa_index;
5237         if (fillargs->ifindex) {
5238                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5239                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5240         }
5241
5242         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5243                                             ifa_ipv6_policy, extack);
5244         if (err < 0)
5245                 return err;
5246
5247         for (i = 0; i <= IFA_MAX; ++i) {
5248                 if (!tb[i])
5249                         continue;
5250
5251                 if (i == IFA_TARGET_NETNSID) {
5252                         struct net *net;
5253
5254                         fillargs->netnsid = nla_get_s32(tb[i]);
5255                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5256                         if (IS_ERR(net)) {
5257                                 fillargs->netnsid = -1;
5258                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5259                                 return PTR_ERR(net);
5260                         }
5261                         *tgt_net = net;
5262                 } else {
5263                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5264                         return -EINVAL;
5265                 }
5266         }
5267
5268         return 0;
5269 }
5270
5271 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5272                            enum addr_type_t type)
5273 {
5274         const struct nlmsghdr *nlh = cb->nlh;
5275         struct inet6_fill_args fillargs = {
5276                 .portid = NETLINK_CB(cb->skb).portid,
5277                 .seq = cb->nlh->nlmsg_seq,
5278                 .flags = NLM_F_MULTI,
5279                 .netnsid = -1,
5280                 .type = type,
5281         };
5282         struct net *tgt_net = sock_net(skb->sk);
5283         int idx, s_idx, s_ip_idx;
5284         int h, s_h;
5285         struct net_device *dev;
5286         struct inet6_dev *idev;
5287         struct hlist_head *head;
5288         int err = 0;
5289
5290         s_h = cb->args[0];
5291         s_idx = idx = cb->args[1];
5292         s_ip_idx = cb->args[2];
5293
5294         if (cb->strict_check) {
5295                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5296                                                   skb->sk, cb);
5297                 if (err < 0)
5298                         goto put_tgt_net;
5299
5300                 err = 0;
5301                 if (fillargs.ifindex) {
5302                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5303                         if (!dev) {
5304                                 err = -ENODEV;
5305                                 goto put_tgt_net;
5306                         }
5307                         idev = __in6_dev_get(dev);
5308                         if (idev) {
5309                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5310                                                      &fillargs);
5311                                 if (err > 0)
5312                                         err = 0;
5313                         }
5314                         goto put_tgt_net;
5315                 }
5316         }
5317
5318         rcu_read_lock();
5319         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5320         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5321                 idx = 0;
5322                 head = &tgt_net->dev_index_head[h];
5323                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5324                         if (idx < s_idx)
5325                                 goto cont;
5326                         if (h > s_h || idx > s_idx)
5327                                 s_ip_idx = 0;
5328                         idev = __in6_dev_get(dev);
5329                         if (!idev)
5330                                 goto cont;
5331
5332                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5333                                            &fillargs) < 0)
5334                                 goto done;
5335 cont:
5336                         idx++;
5337                 }
5338         }
5339 done:
5340         rcu_read_unlock();
5341         cb->args[0] = h;
5342         cb->args[1] = idx;
5343 put_tgt_net:
5344         if (fillargs.netnsid >= 0)
5345                 put_net(tgt_net);
5346
5347         return skb->len ? : err;
5348 }
5349
5350 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5351 {
5352         enum addr_type_t type = UNICAST_ADDR;
5353
5354         return inet6_dump_addr(skb, cb, type);
5355 }
5356
5357 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5358 {
5359         enum addr_type_t type = MULTICAST_ADDR;
5360
5361         return inet6_dump_addr(skb, cb, type);
5362 }
5363
5364
5365 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5366 {
5367         enum addr_type_t type = ANYCAST_ADDR;
5368
5369         return inet6_dump_addr(skb, cb, type);
5370 }
5371
5372 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5373                                        const struct nlmsghdr *nlh,
5374                                        struct nlattr **tb,
5375                                        struct netlink_ext_ack *extack)
5376 {
5377         struct ifaddrmsg *ifm;
5378         int i, err;
5379
5380         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5381                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5382                 return -EINVAL;
5383         }
5384
5385         if (!netlink_strict_get_check(skb))
5386                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5387                                               ifa_ipv6_policy, extack);
5388
5389         ifm = nlmsg_data(nlh);
5390         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5391                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5392                 return -EINVAL;
5393         }
5394
5395         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5396                                             ifa_ipv6_policy, extack);
5397         if (err)
5398                 return err;
5399
5400         for (i = 0; i <= IFA_MAX; i++) {
5401                 if (!tb[i])
5402                         continue;
5403
5404                 switch (i) {
5405                 case IFA_TARGET_NETNSID:
5406                 case IFA_ADDRESS:
5407                 case IFA_LOCAL:
5408                         break;
5409                 default:
5410                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5411                         return -EINVAL;
5412                 }
5413         }
5414
5415         return 0;
5416 }
5417
5418 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5419                              struct netlink_ext_ack *extack)
5420 {
5421         struct net *tgt_net = sock_net(in_skb->sk);
5422         struct inet6_fill_args fillargs = {
5423                 .portid = NETLINK_CB(in_skb).portid,
5424                 .seq = nlh->nlmsg_seq,
5425                 .event = RTM_NEWADDR,
5426                 .flags = 0,
5427                 .netnsid = -1,
5428         };
5429         struct ifaddrmsg *ifm;
5430         struct nlattr *tb[IFA_MAX+1];
5431         struct in6_addr *addr = NULL, *peer;
5432         struct net_device *dev = NULL;
5433         struct inet6_ifaddr *ifa;
5434         struct sk_buff *skb;
5435         int err;
5436
5437         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5438         if (err < 0)
5439                 return err;
5440
5441         if (tb[IFA_TARGET_NETNSID]) {
5442                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5443
5444                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5445                                                   fillargs.netnsid);
5446                 if (IS_ERR(tgt_net))
5447                         return PTR_ERR(tgt_net);
5448         }
5449
5450         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5451         if (!addr)
5452                 return -EINVAL;
5453
5454         ifm = nlmsg_data(nlh);
5455         if (ifm->ifa_index)
5456                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5457
5458         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5459         if (!ifa) {
5460                 err = -EADDRNOTAVAIL;
5461                 goto errout;
5462         }
5463
5464         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5465         if (!skb) {
5466                 err = -ENOBUFS;
5467                 goto errout_ifa;
5468         }
5469
5470         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5471         if (err < 0) {
5472                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5473                 WARN_ON(err == -EMSGSIZE);
5474                 kfree_skb(skb);
5475                 goto errout_ifa;
5476         }
5477         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5478 errout_ifa:
5479         in6_ifa_put(ifa);
5480 errout:
5481         dev_put(dev);
5482         if (fillargs.netnsid >= 0)
5483                 put_net(tgt_net);
5484
5485         return err;
5486 }
5487
5488 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5489 {
5490         struct sk_buff *skb;
5491         struct net *net = dev_net(ifa->idev->dev);
5492         struct inet6_fill_args fillargs = {
5493                 .portid = 0,
5494                 .seq = 0,
5495                 .event = event,
5496                 .flags = 0,
5497                 .netnsid = -1,
5498         };
5499         int err = -ENOBUFS;
5500
5501         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5502         if (!skb)
5503                 goto errout;
5504
5505         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5506         if (err < 0) {
5507                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5508                 WARN_ON(err == -EMSGSIZE);
5509                 kfree_skb(skb);
5510                 goto errout;
5511         }
5512         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5513         return;
5514 errout:
5515         if (err < 0)
5516                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5517 }
5518
5519 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5520                                 __s32 *array, int bytes)
5521 {
5522         BUG_ON(bytes < (DEVCONF_MAX * 4));
5523
5524         memset(array, 0, bytes);
5525         array[DEVCONF_FORWARDING] = cnf->forwarding;
5526         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5527         array[DEVCONF_MTU6] = cnf->mtu6;
5528         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5529         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5530         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5531         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5532         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5533         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5534                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5535         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5536                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5537         array[DEVCONF_RTR_SOLICIT_DELAY] =
5538                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5539         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5540         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5541                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5542         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5543                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5544         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5545         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5546         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5547         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5548         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5549         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5550         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5551         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5552         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5553         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5554 #ifdef CONFIG_IPV6_ROUTER_PREF
5555         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5556         array[DEVCONF_RTR_PROBE_INTERVAL] =
5557                 jiffies_to_msecs(cnf->rtr_probe_interval);
5558 #ifdef CONFIG_IPV6_ROUTE_INFO
5559         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5560         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5561 #endif
5562 #endif
5563         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5564         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5565 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5566         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5567         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5568 #endif
5569 #ifdef CONFIG_IPV6_MROUTE
5570         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5571 #endif
5572         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5573         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5574         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5575         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5576         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5577         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5578         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5579         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5580         /* we omit DEVCONF_STABLE_SECRET for now */
5581         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5582         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5583         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5584         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5585         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5586 #ifdef CONFIG_IPV6_SEG6_HMAC
5587         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5588 #endif
5589         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5590         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5591         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5592         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5593         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5594         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5595         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5596         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5597         array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5598         array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5599 }
5600
5601 static inline size_t inet6_ifla6_size(void)
5602 {
5603         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5604              + nla_total_size(sizeof(struct ifla_cacheinfo))
5605              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5606              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5607              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5608              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5609              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5610              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5611              + 0;
5612 }
5613
5614 static inline size_t inet6_if_nlmsg_size(void)
5615 {
5616         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5617                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5618                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5619                + nla_total_size(4) /* IFLA_MTU */
5620                + nla_total_size(4) /* IFLA_LINK */
5621                + nla_total_size(1) /* IFLA_OPERSTATE */
5622                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5623 }
5624
5625 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5626                                         int bytes)
5627 {
5628         int i;
5629         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5630         BUG_ON(pad < 0);
5631
5632         /* Use put_unaligned() because stats may not be aligned for u64. */
5633         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5634         for (i = 1; i < ICMP6_MIB_MAX; i++)
5635                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5636
5637         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5638 }
5639
5640 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5641                                         int bytes, size_t syncpoff)
5642 {
5643         int i, c;
5644         u64 buff[IPSTATS_MIB_MAX];
5645         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5646
5647         BUG_ON(pad < 0);
5648
5649         memset(buff, 0, sizeof(buff));
5650         buff[0] = IPSTATS_MIB_MAX;
5651
5652         for_each_possible_cpu(c) {
5653                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5654                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5655         }
5656
5657         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5658         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5659 }
5660
5661 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5662                              int bytes)
5663 {
5664         switch (attrtype) {
5665         case IFLA_INET6_STATS:
5666                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5667                                      offsetof(struct ipstats_mib, syncp));
5668                 break;
5669         case IFLA_INET6_ICMP6STATS:
5670                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5671                 break;
5672         }
5673 }
5674
5675 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5676                                   u32 ext_filter_mask)
5677 {
5678         struct nlattr *nla;
5679         struct ifla_cacheinfo ci;
5680
5681         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5682                 goto nla_put_failure;
5683         ci.max_reasm_len = IPV6_MAXPLEN;
5684         ci.tstamp = cstamp_delta(idev->tstamp);
5685         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5686         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5687         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5688                 goto nla_put_failure;
5689         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5690         if (!nla)
5691                 goto nla_put_failure;
5692         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5693
5694         /* XXX - MC not implemented */
5695
5696         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5697                 return 0;
5698
5699         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5700         if (!nla)
5701                 goto nla_put_failure;
5702         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5703
5704         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5705         if (!nla)
5706                 goto nla_put_failure;
5707         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5708
5709         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5710         if (!nla)
5711                 goto nla_put_failure;
5712         read_lock_bh(&idev->lock);
5713         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5714         read_unlock_bh(&idev->lock);
5715
5716         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5717                 goto nla_put_failure;
5718
5719         if (idev->ra_mtu &&
5720             nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5721                 goto nla_put_failure;
5722
5723         return 0;
5724
5725 nla_put_failure:
5726         return -EMSGSIZE;
5727 }
5728
5729 static size_t inet6_get_link_af_size(const struct net_device *dev,
5730                                      u32 ext_filter_mask)
5731 {
5732         if (!__in6_dev_get(dev))
5733                 return 0;
5734
5735         return inet6_ifla6_size();
5736 }
5737
5738 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5739                               u32 ext_filter_mask)
5740 {
5741         struct inet6_dev *idev = __in6_dev_get(dev);
5742
5743         if (!idev)
5744                 return -ENODATA;
5745
5746         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5747                 return -EMSGSIZE;
5748
5749         return 0;
5750 }
5751
5752 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5753                              struct netlink_ext_ack *extack)
5754 {
5755         struct inet6_ifaddr *ifp;
5756         struct net_device *dev = idev->dev;
5757         bool clear_token, update_rs = false;
5758         struct in6_addr ll_addr;
5759
5760         ASSERT_RTNL();
5761
5762         if (!token)
5763                 return -EINVAL;
5764
5765         if (dev->flags & IFF_LOOPBACK) {
5766                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5767                 return -EINVAL;
5768         }
5769
5770         if (dev->flags & IFF_NOARP) {
5771                 NL_SET_ERR_MSG_MOD(extack,
5772                                    "Device does not do neighbour discovery");
5773                 return -EINVAL;
5774         }
5775
5776         if (!ipv6_accept_ra(idev)) {
5777                 NL_SET_ERR_MSG_MOD(extack,
5778                                    "Router advertisement is disabled on device");
5779                 return -EINVAL;
5780         }
5781
5782         if (idev->cnf.rtr_solicits == 0) {
5783                 NL_SET_ERR_MSG(extack,
5784                                "Router solicitation is disabled on device");
5785                 return -EINVAL;
5786         }
5787
5788         write_lock_bh(&idev->lock);
5789
5790         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5791         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5792
5793         write_unlock_bh(&idev->lock);
5794
5795         clear_token = ipv6_addr_any(token);
5796         if (clear_token)
5797                 goto update_lft;
5798
5799         if (!idev->dead && (idev->if_flags & IF_READY) &&
5800             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5801                              IFA_F_OPTIMISTIC)) {
5802                 /* If we're not ready, then normal ifup will take care
5803                  * of this. Otherwise, we need to request our rs here.
5804                  */
5805                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5806                 update_rs = true;
5807         }
5808
5809 update_lft:
5810         write_lock_bh(&idev->lock);
5811
5812         if (update_rs) {
5813                 idev->if_flags |= IF_RS_SENT;
5814                 idev->rs_interval = rfc3315_s14_backoff_init(
5815                         idev->cnf.rtr_solicit_interval);
5816                 idev->rs_probes = 1;
5817                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5818         }
5819
5820         /* Well, that's kinda nasty ... */
5821         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5822                 spin_lock(&ifp->lock);
5823                 if (ifp->tokenized) {
5824                         ifp->valid_lft = 0;
5825                         ifp->prefered_lft = 0;
5826                 }
5827                 spin_unlock(&ifp->lock);
5828         }
5829
5830         write_unlock_bh(&idev->lock);
5831         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5832         addrconf_verify_rtnl(dev_net(dev));
5833         return 0;
5834 }
5835
5836 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5837         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5838         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5839         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5840                                             .reject_message =
5841                                                 "IFLA_INET6_RA_MTU can not be set" },
5842 };
5843
5844 static int check_addr_gen_mode(int mode)
5845 {
5846         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5847             mode != IN6_ADDR_GEN_MODE_NONE &&
5848             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5849             mode != IN6_ADDR_GEN_MODE_RANDOM)
5850                 return -EINVAL;
5851         return 1;
5852 }
5853
5854 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5855                                 int mode)
5856 {
5857         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5858             !idev->cnf.stable_secret.initialized &&
5859             !net->ipv6.devconf_dflt->stable_secret.initialized)
5860                 return -EINVAL;
5861         return 1;
5862 }
5863
5864 static int inet6_validate_link_af(const struct net_device *dev,
5865                                   const struct nlattr *nla,
5866                                   struct netlink_ext_ack *extack)
5867 {
5868         struct nlattr *tb[IFLA_INET6_MAX + 1];
5869         struct inet6_dev *idev = NULL;
5870         int err;
5871
5872         if (dev) {
5873                 idev = __in6_dev_get(dev);
5874                 if (!idev)
5875                         return -EAFNOSUPPORT;
5876         }
5877
5878         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5879                                           inet6_af_policy, extack);
5880         if (err)
5881                 return err;
5882
5883         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5884                 return -EINVAL;
5885
5886         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5887                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5888
5889                 if (check_addr_gen_mode(mode) < 0)
5890                         return -EINVAL;
5891                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5892                         return -EINVAL;
5893         }
5894
5895         return 0;
5896 }
5897
5898 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5899                              struct netlink_ext_ack *extack)
5900 {
5901         struct inet6_dev *idev = __in6_dev_get(dev);
5902         struct nlattr *tb[IFLA_INET6_MAX + 1];
5903         int err;
5904
5905         if (!idev)
5906                 return -EAFNOSUPPORT;
5907
5908         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5909                 return -EINVAL;
5910
5911         if (tb[IFLA_INET6_TOKEN]) {
5912                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5913                                         extack);
5914                 if (err)
5915                         return err;
5916         }
5917
5918         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5919                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5920
5921                 idev->cnf.addr_gen_mode = mode;
5922         }
5923
5924         return 0;
5925 }
5926
5927 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5928                              u32 portid, u32 seq, int event, unsigned int flags)
5929 {
5930         struct net_device *dev = idev->dev;
5931         struct ifinfomsg *hdr;
5932         struct nlmsghdr *nlh;
5933         void *protoinfo;
5934
5935         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5936         if (!nlh)
5937                 return -EMSGSIZE;
5938
5939         hdr = nlmsg_data(nlh);
5940         hdr->ifi_family = AF_INET6;
5941         hdr->__ifi_pad = 0;
5942         hdr->ifi_type = dev->type;
5943         hdr->ifi_index = dev->ifindex;
5944         hdr->ifi_flags = dev_get_flags(dev);
5945         hdr->ifi_change = 0;
5946
5947         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5948             (dev->addr_len &&
5949              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5950             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5951             (dev->ifindex != dev_get_iflink(dev) &&
5952              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5953             nla_put_u8(skb, IFLA_OPERSTATE,
5954                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5955                 goto nla_put_failure;
5956         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5957         if (!protoinfo)
5958                 goto nla_put_failure;
5959
5960         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5961                 goto nla_put_failure;
5962
5963         nla_nest_end(skb, protoinfo);
5964         nlmsg_end(skb, nlh);
5965         return 0;
5966
5967 nla_put_failure:
5968         nlmsg_cancel(skb, nlh);
5969         return -EMSGSIZE;
5970 }
5971
5972 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5973                                    struct netlink_ext_ack *extack)
5974 {
5975         struct ifinfomsg *ifm;
5976
5977         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5978                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5979                 return -EINVAL;
5980         }
5981
5982         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5983                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5984                 return -EINVAL;
5985         }
5986
5987         ifm = nlmsg_data(nlh);
5988         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5989             ifm->ifi_change || ifm->ifi_index) {
5990                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5991                 return -EINVAL;
5992         }
5993
5994         return 0;
5995 }
5996
5997 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5998 {
5999         struct net *net = sock_net(skb->sk);
6000         int h, s_h;
6001         int idx = 0, s_idx;
6002         struct net_device *dev;
6003         struct inet6_dev *idev;
6004         struct hlist_head *head;
6005
6006         /* only requests using strict checking can pass data to
6007          * influence the dump
6008          */
6009         if (cb->strict_check) {
6010                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6011
6012                 if (err < 0)
6013                         return err;
6014         }
6015
6016         s_h = cb->args[0];
6017         s_idx = cb->args[1];
6018
6019         rcu_read_lock();
6020         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6021                 idx = 0;
6022                 head = &net->dev_index_head[h];
6023                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6024                         if (idx < s_idx)
6025                                 goto cont;
6026                         idev = __in6_dev_get(dev);
6027                         if (!idev)
6028                                 goto cont;
6029                         if (inet6_fill_ifinfo(skb, idev,
6030                                               NETLINK_CB(cb->skb).portid,
6031                                               cb->nlh->nlmsg_seq,
6032                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
6033                                 goto out;
6034 cont:
6035                         idx++;
6036                 }
6037         }
6038 out:
6039         rcu_read_unlock();
6040         cb->args[1] = idx;
6041         cb->args[0] = h;
6042
6043         return skb->len;
6044 }
6045
6046 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6047 {
6048         struct sk_buff *skb;
6049         struct net *net = dev_net(idev->dev);
6050         int err = -ENOBUFS;
6051
6052         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6053         if (!skb)
6054                 goto errout;
6055
6056         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6057         if (err < 0) {
6058                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6059                 WARN_ON(err == -EMSGSIZE);
6060                 kfree_skb(skb);
6061                 goto errout;
6062         }
6063         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6064         return;
6065 errout:
6066         if (err < 0)
6067                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6068 }
6069
6070 static inline size_t inet6_prefix_nlmsg_size(void)
6071 {
6072         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6073                + nla_total_size(sizeof(struct in6_addr))
6074                + nla_total_size(sizeof(struct prefix_cacheinfo));
6075 }
6076
6077 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6078                              struct prefix_info *pinfo, u32 portid, u32 seq,
6079                              int event, unsigned int flags)
6080 {
6081         struct prefixmsg *pmsg;
6082         struct nlmsghdr *nlh;
6083         struct prefix_cacheinfo ci;
6084
6085         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6086         if (!nlh)
6087                 return -EMSGSIZE;
6088
6089         pmsg = nlmsg_data(nlh);
6090         pmsg->prefix_family = AF_INET6;
6091         pmsg->prefix_pad1 = 0;
6092         pmsg->prefix_pad2 = 0;
6093         pmsg->prefix_ifindex = idev->dev->ifindex;
6094         pmsg->prefix_len = pinfo->prefix_len;
6095         pmsg->prefix_type = pinfo->type;
6096         pmsg->prefix_pad3 = 0;
6097         pmsg->prefix_flags = 0;
6098         if (pinfo->onlink)
6099                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6100         if (pinfo->autoconf)
6101                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6102
6103         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6104                 goto nla_put_failure;
6105         ci.preferred_time = ntohl(pinfo->prefered);
6106         ci.valid_time = ntohl(pinfo->valid);
6107         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6108                 goto nla_put_failure;
6109         nlmsg_end(skb, nlh);
6110         return 0;
6111
6112 nla_put_failure:
6113         nlmsg_cancel(skb, nlh);
6114         return -EMSGSIZE;
6115 }
6116
6117 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6118                          struct prefix_info *pinfo)
6119 {
6120         struct sk_buff *skb;
6121         struct net *net = dev_net(idev->dev);
6122         int err = -ENOBUFS;
6123
6124         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6125         if (!skb)
6126                 goto errout;
6127
6128         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6129         if (err < 0) {
6130                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6131                 WARN_ON(err == -EMSGSIZE);
6132                 kfree_skb(skb);
6133                 goto errout;
6134         }
6135         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6136         return;
6137 errout:
6138         if (err < 0)
6139                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6140 }
6141
6142 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6143 {
6144         struct net *net = dev_net(ifp->idev->dev);
6145
6146         if (event)
6147                 ASSERT_RTNL();
6148
6149         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6150
6151         switch (event) {
6152         case RTM_NEWADDR:
6153                 /*
6154                  * If the address was optimistic we inserted the route at the
6155                  * start of our DAD process, so we don't need to do it again.
6156                  * If the device was taken down in the middle of the DAD
6157                  * cycle there is a race where we could get here without a
6158                  * host route, so nothing to insert. That will be fixed when
6159                  * the device is brought up.
6160                  */
6161                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6162                         ip6_ins_rt(net, ifp->rt);
6163                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6164                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6165                                 &ifp->addr, ifp->idev->dev->name);
6166                 }
6167
6168                 if (ifp->idev->cnf.forwarding)
6169                         addrconf_join_anycast(ifp);
6170                 if (!ipv6_addr_any(&ifp->peer_addr))
6171                         addrconf_prefix_route(&ifp->peer_addr, 128,
6172                                               ifp->rt_priority, ifp->idev->dev,
6173                                               0, 0, GFP_ATOMIC);
6174                 break;
6175         case RTM_DELADDR:
6176                 if (ifp->idev->cnf.forwarding)
6177                         addrconf_leave_anycast(ifp);
6178                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6179                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6180                         struct fib6_info *rt;
6181
6182                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6183                                                        ifp->idev->dev, 0, 0,
6184                                                        false);
6185                         if (rt)
6186                                 ip6_del_rt(net, rt, false);
6187                 }
6188                 if (ifp->rt) {
6189                         ip6_del_rt(net, ifp->rt, false);
6190                         ifp->rt = NULL;
6191                 }
6192                 rt_genid_bump_ipv6(net);
6193                 break;
6194         }
6195         atomic_inc(&net->ipv6.dev_addr_genid);
6196 }
6197
6198 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6199 {
6200         if (likely(ifp->idev->dead == 0))
6201                 __ipv6_ifa_notify(event, ifp);
6202 }
6203
6204 #ifdef CONFIG_SYSCTL
6205
6206 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6207                 void *buffer, size_t *lenp, loff_t *ppos)
6208 {
6209         int *valp = ctl->data;
6210         int val = *valp;
6211         loff_t pos = *ppos;
6212         struct ctl_table lctl;
6213         int ret;
6214
6215         /*
6216          * ctl->data points to idev->cnf.forwarding, we should
6217          * not modify it until we get the rtnl lock.
6218          */
6219         lctl = *ctl;
6220         lctl.data = &val;
6221
6222         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6223
6224         if (write)
6225                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6226         if (ret)
6227                 *ppos = pos;
6228         return ret;
6229 }
6230
6231 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6232                 void *buffer, size_t *lenp, loff_t *ppos)
6233 {
6234         struct inet6_dev *idev = ctl->extra1;
6235         int min_mtu = IPV6_MIN_MTU;
6236         struct ctl_table lctl;
6237
6238         lctl = *ctl;
6239         lctl.extra1 = &min_mtu;
6240         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6241
6242         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6243 }
6244
6245 static void dev_disable_change(struct inet6_dev *idev)
6246 {
6247         struct netdev_notifier_info info;
6248
6249         if (!idev || !idev->dev)
6250                 return;
6251
6252         netdev_notifier_info_init(&info, idev->dev);
6253         if (idev->cnf.disable_ipv6)
6254                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6255         else
6256                 addrconf_notify(NULL, NETDEV_UP, &info);
6257 }
6258
6259 static void addrconf_disable_change(struct net *net, __s32 newf)
6260 {
6261         struct net_device *dev;
6262         struct inet6_dev *idev;
6263
6264         for_each_netdev(net, dev) {
6265                 idev = __in6_dev_get(dev);
6266                 if (idev) {
6267                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6268                         idev->cnf.disable_ipv6 = newf;
6269                         if (changed)
6270                                 dev_disable_change(idev);
6271                 }
6272         }
6273 }
6274
6275 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6276 {
6277         struct net *net;
6278         int old;
6279
6280         if (!rtnl_trylock())
6281                 return restart_syscall();
6282
6283         net = (struct net *)table->extra2;
6284         old = *p;
6285         *p = newf;
6286
6287         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6288                 rtnl_unlock();
6289                 return 0;
6290         }
6291
6292         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6293                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6294                 addrconf_disable_change(net, newf);
6295         } else if ((!newf) ^ (!old))
6296                 dev_disable_change((struct inet6_dev *)table->extra1);
6297
6298         rtnl_unlock();
6299         return 0;
6300 }
6301
6302 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6303                 void *buffer, size_t *lenp, loff_t *ppos)
6304 {
6305         int *valp = ctl->data;
6306         int val = *valp;
6307         loff_t pos = *ppos;
6308         struct ctl_table lctl;
6309         int ret;
6310
6311         /*
6312          * ctl->data points to idev->cnf.disable_ipv6, we should
6313          * not modify it until we get the rtnl lock.
6314          */
6315         lctl = *ctl;
6316         lctl.data = &val;
6317
6318         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6319
6320         if (write)
6321                 ret = addrconf_disable_ipv6(ctl, valp, val);
6322         if (ret)
6323                 *ppos = pos;
6324         return ret;
6325 }
6326
6327 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6328                 void *buffer, size_t *lenp, loff_t *ppos)
6329 {
6330         int *valp = ctl->data;
6331         int ret;
6332         int old, new;
6333
6334         old = *valp;
6335         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6336         new = *valp;
6337
6338         if (write && old != new) {
6339                 struct net *net = ctl->extra2;
6340
6341                 if (!rtnl_trylock())
6342                         return restart_syscall();
6343
6344                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6345                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6346                                                      NETCONFA_PROXY_NEIGH,
6347                                                      NETCONFA_IFINDEX_DEFAULT,
6348                                                      net->ipv6.devconf_dflt);
6349                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6350                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6351                                                      NETCONFA_PROXY_NEIGH,
6352                                                      NETCONFA_IFINDEX_ALL,
6353                                                      net->ipv6.devconf_all);
6354                 else {
6355                         struct inet6_dev *idev = ctl->extra1;
6356
6357                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6358                                                      NETCONFA_PROXY_NEIGH,
6359                                                      idev->dev->ifindex,
6360                                                      &idev->cnf);
6361                 }
6362                 rtnl_unlock();
6363         }
6364
6365         return ret;
6366 }
6367
6368 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6369                                          void *buffer, size_t *lenp,
6370                                          loff_t *ppos)
6371 {
6372         int ret = 0;
6373         u32 new_val;
6374         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6375         struct net *net = (struct net *)ctl->extra2;
6376         struct ctl_table tmp = {
6377                 .data = &new_val,
6378                 .maxlen = sizeof(new_val),
6379                 .mode = ctl->mode,
6380         };
6381
6382         if (!rtnl_trylock())
6383                 return restart_syscall();
6384
6385         new_val = *((u32 *)ctl->data);
6386
6387         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6388         if (ret != 0)
6389                 goto out;
6390
6391         if (write) {
6392                 if (check_addr_gen_mode(new_val) < 0) {
6393                         ret = -EINVAL;
6394                         goto out;
6395                 }
6396
6397                 if (idev) {
6398                         if (check_stable_privacy(idev, net, new_val) < 0) {
6399                                 ret = -EINVAL;
6400                                 goto out;
6401                         }
6402
6403                         if (idev->cnf.addr_gen_mode != new_val) {
6404                                 idev->cnf.addr_gen_mode = new_val;
6405                                 addrconf_init_auto_addrs(idev->dev);
6406                         }
6407                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6408                         struct net_device *dev;
6409
6410                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6411                         for_each_netdev(net, dev) {
6412                                 idev = __in6_dev_get(dev);
6413                                 if (idev &&
6414                                     idev->cnf.addr_gen_mode != new_val) {
6415                                         idev->cnf.addr_gen_mode = new_val;
6416                                         addrconf_init_auto_addrs(idev->dev);
6417                                 }
6418                         }
6419                 }
6420
6421                 *((u32 *)ctl->data) = new_val;
6422         }
6423
6424 out:
6425         rtnl_unlock();
6426
6427         return ret;
6428 }
6429
6430 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6431                                          void *buffer, size_t *lenp,
6432                                          loff_t *ppos)
6433 {
6434         int err;
6435         struct in6_addr addr;
6436         char str[IPV6_MAX_STRLEN];
6437         struct ctl_table lctl = *ctl;
6438         struct net *net = ctl->extra2;
6439         struct ipv6_stable_secret *secret = ctl->data;
6440
6441         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6442                 return -EIO;
6443
6444         lctl.maxlen = IPV6_MAX_STRLEN;
6445         lctl.data = str;
6446
6447         if (!rtnl_trylock())
6448                 return restart_syscall();
6449
6450         if (!write && !secret->initialized) {
6451                 err = -EIO;
6452                 goto out;
6453         }
6454
6455         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6456         if (err >= sizeof(str)) {
6457                 err = -EIO;
6458                 goto out;
6459         }
6460
6461         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6462         if (err || !write)
6463                 goto out;
6464
6465         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6466                 err = -EIO;
6467                 goto out;
6468         }
6469
6470         secret->initialized = true;
6471         secret->secret = addr;
6472
6473         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6474                 struct net_device *dev;
6475
6476                 for_each_netdev(net, dev) {
6477                         struct inet6_dev *idev = __in6_dev_get(dev);
6478
6479                         if (idev) {
6480                                 idev->cnf.addr_gen_mode =
6481                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6482                         }
6483                 }
6484         } else {
6485                 struct inet6_dev *idev = ctl->extra1;
6486
6487                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6488         }
6489
6490 out:
6491         rtnl_unlock();
6492
6493         return err;
6494 }
6495
6496 static
6497 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6498                                                 int write, void *buffer,
6499                                                 size_t *lenp,
6500                                                 loff_t *ppos)
6501 {
6502         int *valp = ctl->data;
6503         int val = *valp;
6504         loff_t pos = *ppos;
6505         struct ctl_table lctl;
6506         int ret;
6507
6508         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6509          * we should not modify it until we get the rtnl lock.
6510          */
6511         lctl = *ctl;
6512         lctl.data = &val;
6513
6514         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6515
6516         if (write)
6517                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6518         if (ret)
6519                 *ppos = pos;
6520         return ret;
6521 }
6522
6523 static
6524 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6525 {
6526         if (rt) {
6527                 if (action)
6528                         rt->dst.flags |= DST_NOPOLICY;
6529                 else
6530                         rt->dst.flags &= ~DST_NOPOLICY;
6531         }
6532 }
6533
6534 static
6535 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6536 {
6537         struct inet6_ifaddr *ifa;
6538
6539         read_lock_bh(&idev->lock);
6540         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6541                 spin_lock(&ifa->lock);
6542                 if (ifa->rt) {
6543                         /* host routes only use builtin fib6_nh */
6544                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6545                         int cpu;
6546
6547                         rcu_read_lock();
6548                         ifa->rt->dst_nopolicy = val ? true : false;
6549                         if (nh->rt6i_pcpu) {
6550                                 for_each_possible_cpu(cpu) {
6551                                         struct rt6_info **rtp;
6552
6553                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6554                                         addrconf_set_nopolicy(*rtp, val);
6555                                 }
6556                         }
6557                         rcu_read_unlock();
6558                 }
6559                 spin_unlock(&ifa->lock);
6560         }
6561         read_unlock_bh(&idev->lock);
6562 }
6563
6564 static
6565 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6566 {
6567         struct inet6_dev *idev;
6568         struct net *net;
6569
6570         if (!rtnl_trylock())
6571                 return restart_syscall();
6572
6573         *valp = val;
6574
6575         net = (struct net *)ctl->extra2;
6576         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6577                 rtnl_unlock();
6578                 return 0;
6579         }
6580
6581         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6582                 struct net_device *dev;
6583
6584                 for_each_netdev(net, dev) {
6585                         idev = __in6_dev_get(dev);
6586                         if (idev)
6587                                 addrconf_disable_policy_idev(idev, val);
6588                 }
6589         } else {
6590                 idev = (struct inet6_dev *)ctl->extra1;
6591                 addrconf_disable_policy_idev(idev, val);
6592         }
6593
6594         rtnl_unlock();
6595         return 0;
6596 }
6597
6598 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6599                                    void *buffer, size_t *lenp, loff_t *ppos)
6600 {
6601         int *valp = ctl->data;
6602         int val = *valp;
6603         loff_t pos = *ppos;
6604         struct ctl_table lctl;
6605         int ret;
6606
6607         lctl = *ctl;
6608         lctl.data = &val;
6609         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6610
6611         if (write && (*valp != val))
6612                 ret = addrconf_disable_policy(ctl, valp, val);
6613
6614         if (ret)
6615                 *ppos = pos;
6616
6617         return ret;
6618 }
6619
6620 static int minus_one = -1;
6621 static const int two_five_five = 255;
6622 static u32 ioam6_if_id_max = U16_MAX;
6623
6624 static const struct ctl_table addrconf_sysctl[] = {
6625         {
6626                 .procname       = "forwarding",
6627                 .data           = &ipv6_devconf.forwarding,
6628                 .maxlen         = sizeof(int),
6629                 .mode           = 0644,
6630                 .proc_handler   = addrconf_sysctl_forward,
6631         },
6632         {
6633                 .procname       = "hop_limit",
6634                 .data           = &ipv6_devconf.hop_limit,
6635                 .maxlen         = sizeof(int),
6636                 .mode           = 0644,
6637                 .proc_handler   = proc_dointvec_minmax,
6638                 .extra1         = (void *)SYSCTL_ONE,
6639                 .extra2         = (void *)&two_five_five,
6640         },
6641         {
6642                 .procname       = "mtu",
6643                 .data           = &ipv6_devconf.mtu6,
6644                 .maxlen         = sizeof(int),
6645                 .mode           = 0644,
6646                 .proc_handler   = addrconf_sysctl_mtu,
6647         },
6648         {
6649                 .procname       = "accept_ra",
6650                 .data           = &ipv6_devconf.accept_ra,
6651                 .maxlen         = sizeof(int),
6652                 .mode           = 0644,
6653                 .proc_handler   = proc_dointvec,
6654         },
6655         {
6656                 .procname       = "accept_redirects",
6657                 .data           = &ipv6_devconf.accept_redirects,
6658                 .maxlen         = sizeof(int),
6659                 .mode           = 0644,
6660                 .proc_handler   = proc_dointvec,
6661         },
6662         {
6663                 .procname       = "autoconf",
6664                 .data           = &ipv6_devconf.autoconf,
6665                 .maxlen         = sizeof(int),
6666                 .mode           = 0644,
6667                 .proc_handler   = proc_dointvec,
6668         },
6669         {
6670                 .procname       = "dad_transmits",
6671                 .data           = &ipv6_devconf.dad_transmits,
6672                 .maxlen         = sizeof(int),
6673                 .mode           = 0644,
6674                 .proc_handler   = proc_dointvec,
6675         },
6676         {
6677                 .procname       = "router_solicitations",
6678                 .data           = &ipv6_devconf.rtr_solicits,
6679                 .maxlen         = sizeof(int),
6680                 .mode           = 0644,
6681                 .proc_handler   = proc_dointvec_minmax,
6682                 .extra1         = &minus_one,
6683         },
6684         {
6685                 .procname       = "router_solicitation_interval",
6686                 .data           = &ipv6_devconf.rtr_solicit_interval,
6687                 .maxlen         = sizeof(int),
6688                 .mode           = 0644,
6689                 .proc_handler   = proc_dointvec_jiffies,
6690         },
6691         {
6692                 .procname       = "router_solicitation_max_interval",
6693                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6694                 .maxlen         = sizeof(int),
6695                 .mode           = 0644,
6696                 .proc_handler   = proc_dointvec_jiffies,
6697         },
6698         {
6699                 .procname       = "router_solicitation_delay",
6700                 .data           = &ipv6_devconf.rtr_solicit_delay,
6701                 .maxlen         = sizeof(int),
6702                 .mode           = 0644,
6703                 .proc_handler   = proc_dointvec_jiffies,
6704         },
6705         {
6706                 .procname       = "force_mld_version",
6707                 .data           = &ipv6_devconf.force_mld_version,
6708                 .maxlen         = sizeof(int),
6709                 .mode           = 0644,
6710                 .proc_handler   = proc_dointvec,
6711         },
6712         {
6713                 .procname       = "mldv1_unsolicited_report_interval",
6714                 .data           =
6715                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6716                 .maxlen         = sizeof(int),
6717                 .mode           = 0644,
6718                 .proc_handler   = proc_dointvec_ms_jiffies,
6719         },
6720         {
6721                 .procname       = "mldv2_unsolicited_report_interval",
6722                 .data           =
6723                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6724                 .maxlen         = sizeof(int),
6725                 .mode           = 0644,
6726                 .proc_handler   = proc_dointvec_ms_jiffies,
6727         },
6728         {
6729                 .procname       = "use_tempaddr",
6730                 .data           = &ipv6_devconf.use_tempaddr,
6731                 .maxlen         = sizeof(int),
6732                 .mode           = 0644,
6733                 .proc_handler   = proc_dointvec,
6734         },
6735         {
6736                 .procname       = "temp_valid_lft",
6737                 .data           = &ipv6_devconf.temp_valid_lft,
6738                 .maxlen         = sizeof(int),
6739                 .mode           = 0644,
6740                 .proc_handler   = proc_dointvec,
6741         },
6742         {
6743                 .procname       = "temp_prefered_lft",
6744                 .data           = &ipv6_devconf.temp_prefered_lft,
6745                 .maxlen         = sizeof(int),
6746                 .mode           = 0644,
6747                 .proc_handler   = proc_dointvec,
6748         },
6749         {
6750                 .procname       = "regen_max_retry",
6751                 .data           = &ipv6_devconf.regen_max_retry,
6752                 .maxlen         = sizeof(int),
6753                 .mode           = 0644,
6754                 .proc_handler   = proc_dointvec,
6755         },
6756         {
6757                 .procname       = "max_desync_factor",
6758                 .data           = &ipv6_devconf.max_desync_factor,
6759                 .maxlen         = sizeof(int),
6760                 .mode           = 0644,
6761                 .proc_handler   = proc_dointvec,
6762         },
6763         {
6764                 .procname       = "max_addresses",
6765                 .data           = &ipv6_devconf.max_addresses,
6766                 .maxlen         = sizeof(int),
6767                 .mode           = 0644,
6768                 .proc_handler   = proc_dointvec,
6769         },
6770         {
6771                 .procname       = "accept_ra_defrtr",
6772                 .data           = &ipv6_devconf.accept_ra_defrtr,
6773                 .maxlen         = sizeof(int),
6774                 .mode           = 0644,
6775                 .proc_handler   = proc_dointvec,
6776         },
6777         {
6778                 .procname       = "ra_defrtr_metric",
6779                 .data           = &ipv6_devconf.ra_defrtr_metric,
6780                 .maxlen         = sizeof(u32),
6781                 .mode           = 0644,
6782                 .proc_handler   = proc_douintvec_minmax,
6783                 .extra1         = (void *)SYSCTL_ONE,
6784         },
6785         {
6786                 .procname       = "accept_ra_min_hop_limit",
6787                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6788                 .maxlen         = sizeof(int),
6789                 .mode           = 0644,
6790                 .proc_handler   = proc_dointvec,
6791         },
6792         {
6793                 .procname       = "accept_ra_pinfo",
6794                 .data           = &ipv6_devconf.accept_ra_pinfo,
6795                 .maxlen         = sizeof(int),
6796                 .mode           = 0644,
6797                 .proc_handler   = proc_dointvec,
6798         },
6799 #ifdef CONFIG_IPV6_ROUTER_PREF
6800         {
6801                 .procname       = "accept_ra_rtr_pref",
6802                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6803                 .maxlen         = sizeof(int),
6804                 .mode           = 0644,
6805                 .proc_handler   = proc_dointvec,
6806         },
6807         {
6808                 .procname       = "router_probe_interval",
6809                 .data           = &ipv6_devconf.rtr_probe_interval,
6810                 .maxlen         = sizeof(int),
6811                 .mode           = 0644,
6812                 .proc_handler   = proc_dointvec_jiffies,
6813         },
6814 #ifdef CONFIG_IPV6_ROUTE_INFO
6815         {
6816                 .procname       = "accept_ra_rt_info_min_plen",
6817                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6818                 .maxlen         = sizeof(int),
6819                 .mode           = 0644,
6820                 .proc_handler   = proc_dointvec,
6821         },
6822         {
6823                 .procname       = "accept_ra_rt_info_max_plen",
6824                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6825                 .maxlen         = sizeof(int),
6826                 .mode           = 0644,
6827                 .proc_handler   = proc_dointvec,
6828         },
6829 #endif
6830 #endif
6831         {
6832                 .procname       = "proxy_ndp",
6833                 .data           = &ipv6_devconf.proxy_ndp,
6834                 .maxlen         = sizeof(int),
6835                 .mode           = 0644,
6836                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6837         },
6838         {
6839                 .procname       = "accept_source_route",
6840                 .data           = &ipv6_devconf.accept_source_route,
6841                 .maxlen         = sizeof(int),
6842                 .mode           = 0644,
6843                 .proc_handler   = proc_dointvec,
6844         },
6845 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6846         {
6847                 .procname       = "optimistic_dad",
6848                 .data           = &ipv6_devconf.optimistic_dad,
6849                 .maxlen         = sizeof(int),
6850                 .mode           = 0644,
6851                 .proc_handler   = proc_dointvec,
6852         },
6853         {
6854                 .procname       = "use_optimistic",
6855                 .data           = &ipv6_devconf.use_optimistic,
6856                 .maxlen         = sizeof(int),
6857                 .mode           = 0644,
6858                 .proc_handler   = proc_dointvec,
6859         },
6860 #endif
6861 #ifdef CONFIG_IPV6_MROUTE
6862         {
6863                 .procname       = "mc_forwarding",
6864                 .data           = &ipv6_devconf.mc_forwarding,
6865                 .maxlen         = sizeof(int),
6866                 .mode           = 0444,
6867                 .proc_handler   = proc_dointvec,
6868         },
6869 #endif
6870         {
6871                 .procname       = "disable_ipv6",
6872                 .data           = &ipv6_devconf.disable_ipv6,
6873                 .maxlen         = sizeof(int),
6874                 .mode           = 0644,
6875                 .proc_handler   = addrconf_sysctl_disable,
6876         },
6877         {
6878                 .procname       = "accept_dad",
6879                 .data           = &ipv6_devconf.accept_dad,
6880                 .maxlen         = sizeof(int),
6881                 .mode           = 0644,
6882                 .proc_handler   = proc_dointvec,
6883         },
6884         {
6885                 .procname       = "force_tllao",
6886                 .data           = &ipv6_devconf.force_tllao,
6887                 .maxlen         = sizeof(int),
6888                 .mode           = 0644,
6889                 .proc_handler   = proc_dointvec
6890         },
6891         {
6892                 .procname       = "ndisc_notify",
6893                 .data           = &ipv6_devconf.ndisc_notify,
6894                 .maxlen         = sizeof(int),
6895                 .mode           = 0644,
6896                 .proc_handler   = proc_dointvec
6897         },
6898         {
6899                 .procname       = "suppress_frag_ndisc",
6900                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6901                 .maxlen         = sizeof(int),
6902                 .mode           = 0644,
6903                 .proc_handler   = proc_dointvec
6904         },
6905         {
6906                 .procname       = "accept_ra_from_local",
6907                 .data           = &ipv6_devconf.accept_ra_from_local,
6908                 .maxlen         = sizeof(int),
6909                 .mode           = 0644,
6910                 .proc_handler   = proc_dointvec,
6911         },
6912         {
6913                 .procname       = "accept_ra_mtu",
6914                 .data           = &ipv6_devconf.accept_ra_mtu,
6915                 .maxlen         = sizeof(int),
6916                 .mode           = 0644,
6917                 .proc_handler   = proc_dointvec,
6918         },
6919         {
6920                 .procname       = "stable_secret",
6921                 .data           = &ipv6_devconf.stable_secret,
6922                 .maxlen         = IPV6_MAX_STRLEN,
6923                 .mode           = 0600,
6924                 .proc_handler   = addrconf_sysctl_stable_secret,
6925         },
6926         {
6927                 .procname       = "use_oif_addrs_only",
6928                 .data           = &ipv6_devconf.use_oif_addrs_only,
6929                 .maxlen         = sizeof(int),
6930                 .mode           = 0644,
6931                 .proc_handler   = proc_dointvec,
6932         },
6933         {
6934                 .procname       = "ignore_routes_with_linkdown",
6935                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6936                 .maxlen         = sizeof(int),
6937                 .mode           = 0644,
6938                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6939         },
6940         {
6941                 .procname       = "drop_unicast_in_l2_multicast",
6942                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6943                 .maxlen         = sizeof(int),
6944                 .mode           = 0644,
6945                 .proc_handler   = proc_dointvec,
6946         },
6947         {
6948                 .procname       = "drop_unsolicited_na",
6949                 .data           = &ipv6_devconf.drop_unsolicited_na,
6950                 .maxlen         = sizeof(int),
6951                 .mode           = 0644,
6952                 .proc_handler   = proc_dointvec,
6953         },
6954         {
6955                 .procname       = "keep_addr_on_down",
6956                 .data           = &ipv6_devconf.keep_addr_on_down,
6957                 .maxlen         = sizeof(int),
6958                 .mode           = 0644,
6959                 .proc_handler   = proc_dointvec,
6960
6961         },
6962         {
6963                 .procname       = "seg6_enabled",
6964                 .data           = &ipv6_devconf.seg6_enabled,
6965                 .maxlen         = sizeof(int),
6966                 .mode           = 0644,
6967                 .proc_handler   = proc_dointvec,
6968         },
6969 #ifdef CONFIG_IPV6_SEG6_HMAC
6970         {
6971                 .procname       = "seg6_require_hmac",
6972                 .data           = &ipv6_devconf.seg6_require_hmac,
6973                 .maxlen         = sizeof(int),
6974                 .mode           = 0644,
6975                 .proc_handler   = proc_dointvec,
6976         },
6977 #endif
6978         {
6979                 .procname       = "enhanced_dad",
6980                 .data           = &ipv6_devconf.enhanced_dad,
6981                 .maxlen         = sizeof(int),
6982                 .mode           = 0644,
6983                 .proc_handler   = proc_dointvec,
6984         },
6985         {
6986                 .procname       = "addr_gen_mode",
6987                 .data           = &ipv6_devconf.addr_gen_mode,
6988                 .maxlen         = sizeof(int),
6989                 .mode           = 0644,
6990                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6991         },
6992         {
6993                 .procname       = "disable_policy",
6994                 .data           = &ipv6_devconf.disable_policy,
6995                 .maxlen         = sizeof(int),
6996                 .mode           = 0644,
6997                 .proc_handler   = addrconf_sysctl_disable_policy,
6998         },
6999         {
7000                 .procname       = "ndisc_tclass",
7001                 .data           = &ipv6_devconf.ndisc_tclass,
7002                 .maxlen         = sizeof(int),
7003                 .mode           = 0644,
7004                 .proc_handler   = proc_dointvec_minmax,
7005                 .extra1         = (void *)SYSCTL_ZERO,
7006                 .extra2         = (void *)&two_five_five,
7007         },
7008         {
7009                 .procname       = "rpl_seg_enabled",
7010                 .data           = &ipv6_devconf.rpl_seg_enabled,
7011                 .maxlen         = sizeof(int),
7012                 .mode           = 0644,
7013                 .proc_handler   = proc_dointvec,
7014         },
7015         {
7016                 .procname       = "ioam6_enabled",
7017                 .data           = &ipv6_devconf.ioam6_enabled,
7018                 .maxlen         = sizeof(u8),
7019                 .mode           = 0644,
7020                 .proc_handler   = proc_dou8vec_minmax,
7021                 .extra1         = (void *)SYSCTL_ZERO,
7022                 .extra2         = (void *)SYSCTL_ONE,
7023         },
7024         {
7025                 .procname       = "ioam6_id",
7026                 .data           = &ipv6_devconf.ioam6_id,
7027                 .maxlen         = sizeof(u32),
7028                 .mode           = 0644,
7029                 .proc_handler   = proc_douintvec_minmax,
7030                 .extra1         = (void *)SYSCTL_ZERO,
7031                 .extra2         = (void *)&ioam6_if_id_max,
7032         },
7033         {
7034                 .procname       = "ioam6_id_wide",
7035                 .data           = &ipv6_devconf.ioam6_id_wide,
7036                 .maxlen         = sizeof(u32),
7037                 .mode           = 0644,
7038                 .proc_handler   = proc_douintvec,
7039         },
7040         {
7041                 .procname       = "ndisc_evict_nocarrier",
7042                 .data           = &ipv6_devconf.ndisc_evict_nocarrier,
7043                 .maxlen         = sizeof(u8),
7044                 .mode           = 0644,
7045                 .proc_handler   = proc_dou8vec_minmax,
7046                 .extra1         = (void *)SYSCTL_ZERO,
7047                 .extra2         = (void *)SYSCTL_ONE,
7048         },
7049         {
7050                 .procname       = "accept_untracked_na",
7051                 .data           = &ipv6_devconf.accept_untracked_na,
7052                 .maxlen         = sizeof(int),
7053                 .mode           = 0644,
7054                 .proc_handler   = proc_dointvec_minmax,
7055                 .extra1         = SYSCTL_ZERO,
7056                 .extra2         = SYSCTL_TWO,
7057         },
7058         {
7059                 /* sentinel */
7060         }
7061 };
7062
7063 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7064                 struct inet6_dev *idev, struct ipv6_devconf *p)
7065 {
7066         int i, ifindex;
7067         struct ctl_table *table;
7068         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7069
7070         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7071         if (!table)
7072                 goto out;
7073
7074         for (i = 0; table[i].data; i++) {
7075                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7076                 /* If one of these is already set, then it is not safe to
7077                  * overwrite either of them: this makes proc_dointvec_minmax
7078                  * usable.
7079                  */
7080                 if (!table[i].extra1 && !table[i].extra2) {
7081                         table[i].extra1 = idev; /* embedded; no ref */
7082                         table[i].extra2 = net;
7083                 }
7084         }
7085
7086         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7087
7088         p->sysctl_header = register_net_sysctl(net, path, table);
7089         if (!p->sysctl_header)
7090                 goto free;
7091
7092         if (!strcmp(dev_name, "all"))
7093                 ifindex = NETCONFA_IFINDEX_ALL;
7094         else if (!strcmp(dev_name, "default"))
7095                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7096         else
7097                 ifindex = idev->dev->ifindex;
7098         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7099                                      ifindex, p);
7100         return 0;
7101
7102 free:
7103         kfree(table);
7104 out:
7105         return -ENOBUFS;
7106 }
7107
7108 static void __addrconf_sysctl_unregister(struct net *net,
7109                                          struct ipv6_devconf *p, int ifindex)
7110 {
7111         struct ctl_table *table;
7112
7113         if (!p->sysctl_header)
7114                 return;
7115
7116         table = p->sysctl_header->ctl_table_arg;
7117         unregister_net_sysctl_table(p->sysctl_header);
7118         p->sysctl_header = NULL;
7119         kfree(table);
7120
7121         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7122 }
7123
7124 static int addrconf_sysctl_register(struct inet6_dev *idev)
7125 {
7126         int err;
7127
7128         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7129                 return -EINVAL;
7130
7131         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7132                                     &ndisc_ifinfo_sysctl_change);
7133         if (err)
7134                 return err;
7135         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7136                                          idev, &idev->cnf);
7137         if (err)
7138                 neigh_sysctl_unregister(idev->nd_parms);
7139
7140         return err;
7141 }
7142
7143 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7144 {
7145         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7146                                      idev->dev->ifindex);
7147         neigh_sysctl_unregister(idev->nd_parms);
7148 }
7149
7150
7151 #endif
7152
7153 static int __net_init addrconf_init_net(struct net *net)
7154 {
7155         int err = -ENOMEM;
7156         struct ipv6_devconf *all, *dflt;
7157
7158         spin_lock_init(&net->ipv6.addrconf_hash_lock);
7159         INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7160         net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7161                                            sizeof(struct hlist_head),
7162                                            GFP_KERNEL);
7163         if (!net->ipv6.inet6_addr_lst)
7164                 goto err_alloc_addr;
7165
7166         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7167         if (!all)
7168                 goto err_alloc_all;
7169
7170         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7171         if (!dflt)
7172                 goto err_alloc_dflt;
7173
7174         if (!net_eq(net, &init_net)) {
7175                 switch (net_inherit_devconf()) {
7176                 case 1:  /* copy from init_net */
7177                         memcpy(all, init_net.ipv6.devconf_all,
7178                                sizeof(ipv6_devconf));
7179                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7180                                sizeof(ipv6_devconf_dflt));
7181                         break;
7182                 case 3: /* copy from the current netns */
7183                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7184                                sizeof(ipv6_devconf));
7185                         memcpy(dflt,
7186                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7187                                sizeof(ipv6_devconf_dflt));
7188                         break;
7189                 case 0:
7190                 case 2:
7191                         /* use compiled values */
7192                         break;
7193                 }
7194         }
7195
7196         /* these will be inherited by all namespaces */
7197         dflt->autoconf = ipv6_defaults.autoconf;
7198         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7199
7200         dflt->stable_secret.initialized = false;
7201         all->stable_secret.initialized = false;
7202
7203         net->ipv6.devconf_all = all;
7204         net->ipv6.devconf_dflt = dflt;
7205
7206 #ifdef CONFIG_SYSCTL
7207         err = __addrconf_sysctl_register(net, "all", NULL, all);
7208         if (err < 0)
7209                 goto err_reg_all;
7210
7211         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7212         if (err < 0)
7213                 goto err_reg_dflt;
7214 #endif
7215         return 0;
7216
7217 #ifdef CONFIG_SYSCTL
7218 err_reg_dflt:
7219         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7220 err_reg_all:
7221         kfree(dflt);
7222         net->ipv6.devconf_dflt = NULL;
7223 #endif
7224 err_alloc_dflt:
7225         kfree(all);
7226         net->ipv6.devconf_all = NULL;
7227 err_alloc_all:
7228         kfree(net->ipv6.inet6_addr_lst);
7229 err_alloc_addr:
7230         return err;
7231 }
7232
7233 static void __net_exit addrconf_exit_net(struct net *net)
7234 {
7235         int i;
7236
7237 #ifdef CONFIG_SYSCTL
7238         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7239                                      NETCONFA_IFINDEX_DEFAULT);
7240         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7241                                      NETCONFA_IFINDEX_ALL);
7242 #endif
7243         kfree(net->ipv6.devconf_dflt);
7244         net->ipv6.devconf_dflt = NULL;
7245         kfree(net->ipv6.devconf_all);
7246         net->ipv6.devconf_all = NULL;
7247
7248         cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7249         /*
7250          *      Check hash table, then free it.
7251          */
7252         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7253                 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7254
7255         kfree(net->ipv6.inet6_addr_lst);
7256         net->ipv6.inet6_addr_lst = NULL;
7257 }
7258
7259 static struct pernet_operations addrconf_ops = {
7260         .init = addrconf_init_net,
7261         .exit = addrconf_exit_net,
7262 };
7263
7264 static struct rtnl_af_ops inet6_ops __read_mostly = {
7265         .family           = AF_INET6,
7266         .fill_link_af     = inet6_fill_link_af,
7267         .get_link_af_size = inet6_get_link_af_size,
7268         .validate_link_af = inet6_validate_link_af,
7269         .set_link_af      = inet6_set_link_af,
7270 };
7271
7272 /*
7273  *      Init / cleanup code
7274  */
7275
7276 int __init addrconf_init(void)
7277 {
7278         struct inet6_dev *idev;
7279         int err;
7280
7281         err = ipv6_addr_label_init();
7282         if (err < 0) {
7283                 pr_crit("%s: cannot initialize default policy table: %d\n",
7284                         __func__, err);
7285                 goto out;
7286         }
7287
7288         err = register_pernet_subsys(&addrconf_ops);
7289         if (err < 0)
7290                 goto out_addrlabel;
7291
7292         addrconf_wq = create_workqueue("ipv6_addrconf");
7293         if (!addrconf_wq) {
7294                 err = -ENOMEM;
7295                 goto out_nowq;
7296         }
7297
7298         rtnl_lock();
7299         idev = ipv6_add_dev(blackhole_netdev);
7300         rtnl_unlock();
7301         if (IS_ERR(idev)) {
7302                 err = PTR_ERR(idev);
7303                 goto errlo;
7304         }
7305
7306         ip6_route_init_special_entries();
7307
7308         register_netdevice_notifier(&ipv6_dev_notf);
7309
7310         addrconf_verify(&init_net);
7311
7312         rtnl_af_register(&inet6_ops);
7313
7314         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7315                                    NULL, inet6_dump_ifinfo, 0);
7316         if (err < 0)
7317                 goto errout;
7318
7319         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7320                                    inet6_rtm_newaddr, NULL, 0);
7321         if (err < 0)
7322                 goto errout;
7323         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7324                                    inet6_rtm_deladdr, NULL, 0);
7325         if (err < 0)
7326                 goto errout;
7327         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7328                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7329                                    RTNL_FLAG_DOIT_UNLOCKED);
7330         if (err < 0)
7331                 goto errout;
7332         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7333                                    NULL, inet6_dump_ifmcaddr, 0);
7334         if (err < 0)
7335                 goto errout;
7336         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7337                                    NULL, inet6_dump_ifacaddr, 0);
7338         if (err < 0)
7339                 goto errout;
7340         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7341                                    inet6_netconf_get_devconf,
7342                                    inet6_netconf_dump_devconf,
7343                                    RTNL_FLAG_DOIT_UNLOCKED);
7344         if (err < 0)
7345                 goto errout;
7346         err = ipv6_addr_label_rtnl_register();
7347         if (err < 0)
7348                 goto errout;
7349
7350         return 0;
7351 errout:
7352         rtnl_unregister_all(PF_INET6);
7353         rtnl_af_unregister(&inet6_ops);
7354         unregister_netdevice_notifier(&ipv6_dev_notf);
7355 errlo:
7356         destroy_workqueue(addrconf_wq);
7357 out_nowq:
7358         unregister_pernet_subsys(&addrconf_ops);
7359 out_addrlabel:
7360         ipv6_addr_label_cleanup();
7361 out:
7362         return err;
7363 }
7364
7365 void addrconf_cleanup(void)
7366 {
7367         struct net_device *dev;
7368
7369         unregister_netdevice_notifier(&ipv6_dev_notf);
7370         unregister_pernet_subsys(&addrconf_ops);
7371         ipv6_addr_label_cleanup();
7372
7373         rtnl_af_unregister(&inet6_ops);
7374
7375         rtnl_lock();
7376
7377         /* clean dev list */
7378         for_each_netdev(&init_net, dev) {
7379                 if (__in6_dev_get(dev) == NULL)
7380                         continue;
7381                 addrconf_ifdown(dev, true);
7382         }
7383         addrconf_ifdown(init_net.loopback_dev, true);
7384
7385         rtnl_unlock();
7386
7387         destroy_workqueue(addrconf_wq);
7388 }