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
  93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  94
  95 #define IPV6_MAX_STRLEN \
  96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  97
  98 static inline u32 cstamp_delta(unsigned long cstamp)
  99 {
 100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 101 }
 102
 103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 104 {
 105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 107         do_div(tmp, 1000000);
 108         return (s32)tmp;
 109 }
 110
 111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 112 {
 113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 119                 do_div(tmp, 1000000);
 120         }
 121         return (s32)tmp;
 122 }
 123
 124 #ifdef CONFIG_SYSCTL
 125 static int addrconf_sysctl_register(struct inet6_dev *idev);
 126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 127 #else
 128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 129 {
 130         return 0;
 131 }
 132
 133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 134 {
 135 }
 136 #endif
 137
 138 static void ipv6_regen_rndid(struct inet6_dev *idev);
 139 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 140
 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 142 static int ipv6_count_addresses(const struct inet6_dev *idev);
 143 static int ipv6_generate_stable_address(struct in6_addr *addr,
 144                                         u8 dad_count,
 145                                         const struct inet6_dev *idev);
 146
 147 #define IN6_ADDR_HSIZE_SHIFT    8
 148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 149 /*
 150  *      Configured unicast address hash table
 151  */
 152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 153 static DEFINE_SPINLOCK(addrconf_hash_lock);
 154
 155 static void addrconf_verify(void);
 156 static void addrconf_verify_rtnl(void);
 157 static void addrconf_verify_work(struct work_struct *);
 158
 159 static struct workqueue_struct *addrconf_wq;
 160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 161
 162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 164
 165 static void addrconf_type_change(struct net_device *dev,
 166                                  unsigned long event);
 167 static int addrconf_ifdown(struct net_device *dev, int how);
 168
 169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 170                                                   int plen,
 171                                                   const struct net_device *dev,
 172                                                   u32 flags, u32 noflags,
 173                                                   bool no_gw);
 174
 175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 176 static void addrconf_dad_work(struct work_struct *w);
 177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 178                                    bool send_na);
 179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 180 static void addrconf_rs_timer(struct timer_list *t);
 181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 183
 184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 185                                 struct prefix_info *pinfo);
 186
 187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 188         .forwarding             = 0,
 189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 190         .mtu6                   = IPV6_MIN_MTU,
 191         .accept_ra              = 1,
 192         .accept_redirects       = 1,
 193         .autoconf               = 1,
 194         .force_mld_version      = 0,
 195         .mldv1_unsolicited_report_interval = 10 * HZ,
 196         .mldv2_unsolicited_report_interval = HZ,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .accept_ra_from_local   = 0,
 210         .accept_ra_min_hop_limit= 1,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_min_plen = 0,
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 0,
 224         .suppress_frag_ndisc    = 1,
 225         .accept_ra_mtu          = 1,
 226         .stable_secret          = {
 227                 .initialized = false,
 228         },
 229         .use_oif_addrs_only     = 0,
 230         .ignore_routes_with_linkdown = 0,
 231         .keep_addr_on_down      = 0,
 232         .seg6_enabled           = 0,
 233 #ifdef CONFIG_IPV6_SEG6_HMAC
 234         .seg6_require_hmac      = 0,
 235 #endif
 236         .enhanced_dad           = 1,
 237         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 238         .disable_policy         = 0,
 239 };
 240
 241 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 242         .forwarding             = 0,
 243         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 244         .mtu6                   = IPV6_MIN_MTU,
 245         .accept_ra              = 1,
 246         .accept_redirects       = 1,
 247         .autoconf               = 1,
 248         .force_mld_version      = 0,
 249         .mldv1_unsolicited_report_interval = 10 * HZ,
 250         .mldv2_unsolicited_report_interval = HZ,
 251         .dad_transmits          = 1,
 252         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 253         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 254         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 255         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 256         .use_tempaddr           = 0,
 257         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 258         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 259         .regen_max_retry        = REGEN_MAX_RETRY,
 260         .max_desync_factor      = MAX_DESYNC_FACTOR,
 261         .max_addresses          = IPV6_MAX_ADDRESSES,
 262         .accept_ra_defrtr       = 1,
 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 };
 294
 295 /* Check if link is ready: is it up and is a valid qdisc available */
 296 static inline bool addrconf_link_ready(const struct net_device *dev)
 297 {
 298         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 299 }
 300
 301 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 302 {
 303         if (del_timer(&idev->rs_timer))
 304                 __in6_dev_put(idev);
 305 }
 306
 307 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 308 {
 309         if (cancel_delayed_work(&ifp->dad_work))
 310                 __in6_ifa_put(ifp);
 311 }
 312
 313 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 314                                   unsigned long when)
 315 {
 316         if (!timer_pending(&idev->rs_timer))
 317                 in6_dev_hold(idev);
 318         mod_timer(&idev->rs_timer, jiffies + when);
 319 }
 320
 321 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 322                                    unsigned long delay)
 323 {
 324         in6_ifa_hold(ifp);
 325         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 326                 in6_ifa_put(ifp);
 327 }
 328
 329 static int snmp6_alloc_dev(struct inet6_dev *idev)
 330 {
 331         int i;
 332
 333         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 334         if (!idev->stats.ipv6)
 335                 goto err_ip;
 336
 337         for_each_possible_cpu(i) {
 338                 struct ipstats_mib *addrconf_stats;
 339                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 340                 u64_stats_init(&addrconf_stats->syncp);
 341         }
 342
 343
 344         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 345                                         GFP_KERNEL);
 346         if (!idev->stats.icmpv6dev)
 347                 goto err_icmp;
 348         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 349                                            GFP_KERNEL);
 350         if (!idev->stats.icmpv6msgdev)
 351                 goto err_icmpmsg;
 352
 353         return 0;
 354
 355 err_icmpmsg:
 356         kfree(idev->stats.icmpv6dev);
 357 err_icmp:
 358         free_percpu(idev->stats.ipv6);
 359 err_ip:
 360         return -ENOMEM;
 361 }
 362
 363 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 364 {
 365         struct inet6_dev *ndev;
 366         int err = -ENOMEM;
 367
 368         ASSERT_RTNL();
 369
 370         if (dev->mtu < IPV6_MIN_MTU)
 371                 return ERR_PTR(-EINVAL);
 372
 373         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 374         if (!ndev)
 375                 return ERR_PTR(err);
 376
 377         rwlock_init(&ndev->lock);
 378         ndev->dev = dev;
 379         INIT_LIST_HEAD(&ndev->addr_list);
 380         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 381         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 382
 383         if (ndev->cnf.stable_secret.initialized)
 384                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 385
 386         ndev->cnf.mtu6 = dev->mtu;
 387         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 388         if (!ndev->nd_parms) {
 389                 kfree(ndev);
 390                 return ERR_PTR(err);
 391         }
 392         if (ndev->cnf.forwarding)
 393                 dev_disable_lro(dev);
 394         /* We refer to the device */
 395         dev_hold(dev);
 396
 397         if (snmp6_alloc_dev(ndev) < 0) {
 398                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 399                            __func__);
 400                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 401                 dev_put(dev);
 402                 kfree(ndev);
 403                 return ERR_PTR(err);
 404         }
 405
 406         if (snmp6_register_dev(ndev) < 0) {
 407                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 408                            __func__, dev->name);
 409                 goto err_release;
 410         }
 411
 412         /* One reference from device. */
 413         refcount_set(&ndev->refcnt, 1);
 414
 415         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 416                 ndev->cnf.accept_dad = -1;
 417
 418 #if IS_ENABLED(CONFIG_IPV6_SIT)
 419         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 420                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 421                 ndev->cnf.rtr_solicits = 0;
 422         }
 423 #endif
 424
 425         INIT_LIST_HEAD(&ndev->tempaddr_list);
 426         ndev->desync_factor = U32_MAX;
 427         if ((dev->flags&IFF_LOOPBACK) ||
 428             dev->type == ARPHRD_TUNNEL ||
 429             dev->type == ARPHRD_TUNNEL6 ||
 430             dev->type == ARPHRD_SIT ||
 431             dev->type == ARPHRD_NONE) {
 432                 ndev->cnf.use_tempaddr = -1;
 433         } else
 434                 ipv6_regen_rndid(ndev);
 435
 436         ndev->token = in6addr_any;
 437
 438         if (netif_running(dev) && addrconf_link_ready(dev))
 439                 ndev->if_flags |= IF_READY;
 440
 441         ipv6_mc_init_dev(ndev);
 442         ndev->tstamp = jiffies;
 443         err = addrconf_sysctl_register(ndev);
 444         if (err) {
 445                 ipv6_mc_destroy_dev(ndev);
 446                 snmp6_unregister_dev(ndev);
 447                 goto err_release;
 448         }
 449         /* protected by rtnl_lock */
 450         rcu_assign_pointer(dev->ip6_ptr, ndev);
 451
 452         /* Join interface-local all-node multicast group */
 453         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 454
 455         /* Join all-node multicast group */
 456         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 457
 458         /* Join all-router multicast group if forwarding is set */
 459         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 460                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 461
 462         return ndev;
 463
 464 err_release:
 465         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 466         ndev->dead = 1;
 467         in6_dev_finish_destroy(ndev);
 468         return ERR_PTR(err);
 469 }
 470
 471 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 472 {
 473         struct inet6_dev *idev;
 474
 475         ASSERT_RTNL();
 476
 477         idev = __in6_dev_get(dev);
 478         if (!idev) {
 479                 idev = ipv6_add_dev(dev);
 480                 if (IS_ERR(idev))
 481                         return NULL;
 482         }
 483
 484         if (dev->flags&IFF_UP)
 485                 ipv6_mc_up(idev);
 486         return idev;
 487 }
 488
 489 static int inet6_netconf_msgsize_devconf(int type)
 490 {
 491         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 492                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 493         bool all = false;
 494
 495         if (type == NETCONFA_ALL)
 496                 all = true;
 497
 498         if (all || type == NETCONFA_FORWARDING)
 499                 size += nla_total_size(4);
 500 #ifdef CONFIG_IPV6_MROUTE
 501         if (all || type == NETCONFA_MC_FORWARDING)
 502                 size += nla_total_size(4);
 503 #endif
 504         if (all || type == NETCONFA_PROXY_NEIGH)
 505                 size += nla_total_size(4);
 506
 507         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 508                 size += nla_total_size(4);
 509
 510         return size;
 511 }
 512
 513 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 514                                       struct ipv6_devconf *devconf, u32 portid,
 515                                       u32 seq, int event, unsigned int flags,
 516                                       int type)
 517 {
 518         struct nlmsghdr  *nlh;
 519         struct netconfmsg *ncm;
 520         bool all = false;
 521
 522         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 523                         flags);
 524         if (!nlh)
 525                 return -EMSGSIZE;
 526
 527         if (type == NETCONFA_ALL)
 528                 all = true;
 529
 530         ncm = nlmsg_data(nlh);
 531         ncm->ncm_family = AF_INET6;
 532
 533         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 534                 goto nla_put_failure;
 535
 536         if (!devconf)
 537                 goto out;
 538
 539         if ((all || type == NETCONFA_FORWARDING) &&
 540             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 541                 goto nla_put_failure;
 542 #ifdef CONFIG_IPV6_MROUTE
 543         if ((all || type == NETCONFA_MC_FORWARDING) &&
 544             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 545                         devconf->mc_forwarding) < 0)
 546                 goto nla_put_failure;
 547 #endif
 548         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 549             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 550                 goto nla_put_failure;
 551
 552         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 553             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 554                         devconf->ignore_routes_with_linkdown) < 0)
 555                 goto nla_put_failure;
 556
 557 out:
 558         nlmsg_end(skb, nlh);
 559         return 0;
 560
 561 nla_put_failure:
 562         nlmsg_cancel(skb, nlh);
 563         return -EMSGSIZE;
 564 }
 565
 566 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 567                                   int ifindex, struct ipv6_devconf *devconf)
 568 {
 569         struct sk_buff *skb;
 570         int err = -ENOBUFS;
 571
 572         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 573         if (!skb)
 574                 goto errout;
 575
 576         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 577                                          event, 0, type);
 578         if (err < 0) {
 579                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 580                 WARN_ON(err == -EMSGSIZE);
 581                 kfree_skb(skb);
 582                 goto errout;
 583         }
 584         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 585         return;
 586 errout:
 587         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 588 }
 589
 590 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 591         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 592         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 593         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 594         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 595 };
 596
 597 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 598                                        const struct nlmsghdr *nlh,
 599                                        struct nlattr **tb,
 600                                        struct netlink_ext_ack *extack)
 601 {
 602         int i, err;
 603
 604         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 605                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 606                 return -EINVAL;
 607         }
 608
 609         if (!netlink_strict_get_check(skb))
 610                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 611                                               tb, NETCONFA_MAX,
 612                                               devconf_ipv6_policy, extack);
 613
 614         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 615                                             tb, NETCONFA_MAX,
 616                                             devconf_ipv6_policy, extack);
 617         if (err)
 618                 return err;
 619
 620         for (i = 0; i <= NETCONFA_MAX; i++) {
 621                 if (!tb[i])
 622                         continue;
 623
 624                 switch (i) {
 625                 case NETCONFA_IFINDEX:
 626                         break;
 627                 default:
 628                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 629                         return -EINVAL;
 630                 }
 631         }
 632
 633         return 0;
 634 }
 635
 636 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 637                                      struct nlmsghdr *nlh,
 638                                      struct netlink_ext_ack *extack)
 639 {
 640         struct net *net = sock_net(in_skb->sk);
 641         struct nlattr *tb[NETCONFA_MAX+1];
 642         struct inet6_dev *in6_dev = NULL;
 643         struct net_device *dev = NULL;
 644         struct sk_buff *skb;
 645         struct ipv6_devconf *devconf;
 646         int ifindex;
 647         int err;
 648
 649         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 650         if (err < 0)
 651                 return err;
 652
 653         if (!tb[NETCONFA_IFINDEX])
 654                 return -EINVAL;
 655
 656         err = -EINVAL;
 657         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 658         switch (ifindex) {
 659         case NETCONFA_IFINDEX_ALL:
 660                 devconf = net->ipv6.devconf_all;
 661                 break;
 662         case NETCONFA_IFINDEX_DEFAULT:
 663                 devconf = net->ipv6.devconf_dflt;
 664                 break;
 665         default:
 666                 dev = dev_get_by_index(net, ifindex);
 667                 if (!dev)
 668                         return -EINVAL;
 669                 in6_dev = in6_dev_get(dev);
 670                 if (!in6_dev)
 671                         goto errout;
 672                 devconf = &in6_dev->cnf;
 673                 break;
 674         }
 675
 676         err = -ENOBUFS;
 677         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 678         if (!skb)
 679                 goto errout;
 680
 681         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 682                                          NETLINK_CB(in_skb).portid,
 683                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 684                                          NETCONFA_ALL);
 685         if (err < 0) {
 686                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 687                 WARN_ON(err == -EMSGSIZE);
 688                 kfree_skb(skb);
 689                 goto errout;
 690         }
 691         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 692 errout:
 693         if (in6_dev)
 694                 in6_dev_put(in6_dev);
 695         if (dev)
 696                 dev_put(dev);
 697         return err;
 698 }
 699
 700 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 701                                       struct netlink_callback *cb)
 702 {
 703         const struct nlmsghdr *nlh = cb->nlh;
 704         struct net *net = sock_net(skb->sk);
 705         int h, s_h;
 706         int idx, s_idx;
 707         struct net_device *dev;
 708         struct inet6_dev *idev;
 709         struct hlist_head *head;
 710
 711         if (cb->strict_check) {
 712                 struct netlink_ext_ack *extack = cb->extack;
 713                 struct netconfmsg *ncm;
 714
 715                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 716                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 717                         return -EINVAL;
 718                 }
 719
 720                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 721                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 722                         return -EINVAL;
 723                 }
 724         }
 725
 726         s_h = cb->args[0];
 727         s_idx = idx = cb->args[1];
 728
 729         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 730                 idx = 0;
 731                 head = &net->dev_index_head[h];
 732                 rcu_read_lock();
 733                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 734                           net->dev_base_seq;
 735                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 736                         if (idx < s_idx)
 737                                 goto cont;
 738                         idev = __in6_dev_get(dev);
 739                         if (!idev)
 740                                 goto cont;
 741
 742                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 743                                                        &idev->cnf,
 744                                                        NETLINK_CB(cb->skb).portid,
 745                                                        nlh->nlmsg_seq,
 746                                                        RTM_NEWNETCONF,
 747                                                        NLM_F_MULTI,
 748                                                        NETCONFA_ALL) < 0) {
 749                                 rcu_read_unlock();
 750                                 goto done;
 751                         }
 752                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 753 cont:
 754                         idx++;
 755                 }
 756                 rcu_read_unlock();
 757         }
 758         if (h == NETDEV_HASHENTRIES) {
 759                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 760                                                net->ipv6.devconf_all,
 761                                                NETLINK_CB(cb->skb).portid,
 762                                                nlh->nlmsg_seq,
 763                                                RTM_NEWNETCONF, NLM_F_MULTI,
 764                                                NETCONFA_ALL) < 0)
 765                         goto done;
 766                 else
 767                         h++;
 768         }
 769         if (h == NETDEV_HASHENTRIES + 1) {
 770                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 771                                                net->ipv6.devconf_dflt,
 772                                                NETLINK_CB(cb->skb).portid,
 773                                                nlh->nlmsg_seq,
 774                                                RTM_NEWNETCONF, NLM_F_MULTI,
 775                                                NETCONFA_ALL) < 0)
 776                         goto done;
 777                 else
 778                         h++;
 779         }
 780 done:
 781         cb->args[0] = h;
 782         cb->args[1] = idx;
 783
 784         return skb->len;
 785 }
 786
 787 #ifdef CONFIG_SYSCTL
 788 static void dev_forward_change(struct inet6_dev *idev)
 789 {
 790         struct net_device *dev;
 791         struct inet6_ifaddr *ifa;
 792
 793         if (!idev)
 794                 return;
 795         dev = idev->dev;
 796         if (idev->cnf.forwarding)
 797                 dev_disable_lro(dev);
 798         if (dev->flags & IFF_MULTICAST) {
 799                 if (idev->cnf.forwarding) {
 800                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 801                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 802                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 803                 } else {
 804                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 805                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 806                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 807                 }
 808         }
 809
 810         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 811                 if (ifa->flags&IFA_F_TENTATIVE)
 812                         continue;
 813                 if (idev->cnf.forwarding)
 814                         addrconf_join_anycast(ifa);
 815                 else
 816                         addrconf_leave_anycast(ifa);
 817         }
 818         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 819                                      NETCONFA_FORWARDING,
 820                                      dev->ifindex, &idev->cnf);
 821 }
 822
 823
 824 static void addrconf_forward_change(struct net *net, __s32 newf)
 825 {
 826         struct net_device *dev;
 827         struct inet6_dev *idev;
 828
 829         for_each_netdev(net, dev) {
 830                 idev = __in6_dev_get(dev);
 831                 if (idev) {
 832                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 833                         idev->cnf.forwarding = newf;
 834                         if (changed)
 835                                 dev_forward_change(idev);
 836                 }
 837         }
 838 }
 839
 840 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 841 {
 842         struct net *net;
 843         int old;
 844
 845         if (!rtnl_trylock())
 846                 return restart_syscall();
 847
 848         net = (struct net *)table->extra2;
 849         old = *p;
 850         *p = newf;
 851
 852         if (p == &net->ipv6.devconf_dflt->forwarding) {
 853                 if ((!newf) ^ (!old))
 854                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 855                                                      NETCONFA_FORWARDING,
 856                                                      NETCONFA_IFINDEX_DEFAULT,
 857                                                      net->ipv6.devconf_dflt);
 858                 rtnl_unlock();
 859                 return 0;
 860         }
 861
 862         if (p == &net->ipv6.devconf_all->forwarding) {
 863                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 864
 865                 net->ipv6.devconf_dflt->forwarding = newf;
 866                 if ((!newf) ^ (!old_dflt))
 867                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 868                                                      NETCONFA_FORWARDING,
 869                                                      NETCONFA_IFINDEX_DEFAULT,
 870                                                      net->ipv6.devconf_dflt);
 871
 872                 addrconf_forward_change(net, newf);
 873                 if ((!newf) ^ (!old))
 874                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 875                                                      NETCONFA_FORWARDING,
 876                                                      NETCONFA_IFINDEX_ALL,
 877                                                      net->ipv6.devconf_all);
 878         } else if ((!newf) ^ (!old))
 879                 dev_forward_change((struct inet6_dev *)table->extra1);
 880         rtnl_unlock();
 881
 882         if (newf)
 883                 rt6_purge_dflt_routers(net);
 884         return 1;
 885 }
 886
 887 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 888 {
 889         struct net_device *dev;
 890         struct inet6_dev *idev;
 891
 892         for_each_netdev(net, dev) {
 893                 idev = __in6_dev_get(dev);
 894                 if (idev) {
 895                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 896
 897                         idev->cnf.ignore_routes_with_linkdown = newf;
 898                         if (changed)
 899                                 inet6_netconf_notify_devconf(dev_net(dev),
 900                                                              RTM_NEWNETCONF,
 901                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 902                                                              dev->ifindex,
 903                                                              &idev->cnf);
 904                 }
 905         }
 906 }
 907
 908 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 909 {
 910         struct net *net;
 911         int old;
 912
 913         if (!rtnl_trylock())
 914                 return restart_syscall();
 915
 916         net = (struct net *)table->extra2;
 917         old = *p;
 918         *p = newf;
 919
 920         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 921                 if ((!newf) ^ (!old))
 922                         inet6_netconf_notify_devconf(net,
 923                                                      RTM_NEWNETCONF,
 924                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 925                                                      NETCONFA_IFINDEX_DEFAULT,
 926                                                      net->ipv6.devconf_dflt);
 927                 rtnl_unlock();
 928                 return 0;
 929         }
 930
 931         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 932                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 933                 addrconf_linkdown_change(net, newf);
 934                 if ((!newf) ^ (!old))
 935                         inet6_netconf_notify_devconf(net,
 936                                                      RTM_NEWNETCONF,
 937                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 938                                                      NETCONFA_IFINDEX_ALL,
 939                                                      net->ipv6.devconf_all);
 940         }
 941         rtnl_unlock();
 942
 943         return 1;
 944 }
 945
 946 #endif
 947
 948 /* Nobody refers to this ifaddr, destroy it */
 949 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 950 {
 951         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 952
 953 #ifdef NET_REFCNT_DEBUG
 954         pr_debug("%s\n", __func__);
 955 #endif
 956
 957         in6_dev_put(ifp->idev);
 958
 959         if (cancel_delayed_work(&ifp->dad_work))
 960                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 961                           ifp);
 962
 963         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 964                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 965                 return;
 966         }
 967
 968         kfree_rcu(ifp, rcu);
 969 }
 970
 971 static void
 972 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 973 {
 974         struct list_head *p;
 975         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 976
 977         /*
 978          * Each device address list is sorted in order of scope -
 979          * global before linklocal.
 980          */
 981         list_for_each(p, &idev->addr_list) {
 982                 struct inet6_ifaddr *ifa
 983                         = list_entry(p, struct inet6_ifaddr, if_list);
 984                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 985                         break;
 986         }
 987
 988         list_add_tail_rcu(&ifp->if_list, p);
 989 }
 990
 991 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 992 {
 993         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 994
 995         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 996 }
 997
 998 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 999                                struct net_device *dev, unsigned int hash)
1000 {
1001         struct inet6_ifaddr *ifp;
1002
1003         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1004                 if (!net_eq(dev_net(ifp->idev->dev), net))
1005                         continue;
1006                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1007                         if (!dev || ifp->idev->dev == dev)
1008                                 return true;
1009                 }
1010         }
1011         return false;
1012 }
1013
1014 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1015 {
1016         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1017         int err = 0;
1018
1019         spin_lock(&addrconf_hash_lock);
1020
1021         /* Ignore adding duplicate addresses on an interface */
1022         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1023                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1024                 err = -EEXIST;
1025         } else {
1026                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1027         }
1028
1029         spin_unlock(&addrconf_hash_lock);
1030
1031         return err;
1032 }
1033
1034 /* On success it returns ifp with increased reference count */
1035
1036 static struct inet6_ifaddr *
1037 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1038               bool can_block, struct netlink_ext_ack *extack)
1039 {
1040         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1041         int addr_type = ipv6_addr_type(cfg->pfx);
1042         struct net *net = dev_net(idev->dev);
1043         struct inet6_ifaddr *ifa = NULL;
1044         struct fib6_info *f6i = NULL;
1045         int err = 0;
1046
1047         if (addr_type == IPV6_ADDR_ANY ||
1048             addr_type & IPV6_ADDR_MULTICAST ||
1049             (!(idev->dev->flags & IFF_LOOPBACK) &&
1050              !netif_is_l3_master(idev->dev) &&
1051              addr_type & IPV6_ADDR_LOOPBACK))
1052                 return ERR_PTR(-EADDRNOTAVAIL);
1053
1054         if (idev->dead) {
1055                 err = -ENODEV;                  /*XXX*/
1056                 goto out;
1057         }
1058
1059         if (idev->cnf.disable_ipv6) {
1060                 err = -EACCES;
1061                 goto out;
1062         }
1063
1064         /* validator notifier needs to be blocking;
1065          * do not call in atomic context
1066          */
1067         if (can_block) {
1068                 struct in6_validator_info i6vi = {
1069                         .i6vi_addr = *cfg->pfx,
1070                         .i6vi_dev = idev,
1071                         .extack = extack,
1072                 };
1073
1074                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1075                 err = notifier_to_errno(err);
1076                 if (err < 0)
1077                         goto out;
1078         }
1079
1080         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1081         if (!ifa) {
1082                 err = -ENOBUFS;
1083                 goto out;
1084         }
1085
1086         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1087         if (IS_ERR(f6i)) {
1088                 err = PTR_ERR(f6i);
1089                 f6i = NULL;
1090                 goto out;
1091         }
1092
1093         if (net->ipv6.devconf_all->disable_policy ||
1094             idev->cnf.disable_policy)
1095                 f6i->dst_nopolicy = true;
1096
1097         neigh_parms_data_state_setall(idev->nd_parms);
1098
1099         ifa->addr = *cfg->pfx;
1100         if (cfg->peer_pfx)
1101                 ifa->peer_addr = *cfg->peer_pfx;
1102
1103         spin_lock_init(&ifa->lock);
1104         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1105         INIT_HLIST_NODE(&ifa->addr_lst);
1106         ifa->scope = cfg->scope;
1107         ifa->prefix_len = cfg->plen;
1108         ifa->rt_priority = cfg->rt_priority;
1109         ifa->flags = cfg->ifa_flags;
1110         /* No need to add the TENTATIVE flag for addresses with NODAD */
1111         if (!(cfg->ifa_flags & IFA_F_NODAD))
1112                 ifa->flags |= IFA_F_TENTATIVE;
1113         ifa->valid_lft = cfg->valid_lft;
1114         ifa->prefered_lft = cfg->preferred_lft;
1115         ifa->cstamp = ifa->tstamp = jiffies;
1116         ifa->tokenized = false;
1117
1118         ifa->rt = f6i;
1119
1120         ifa->idev = idev;
1121         in6_dev_hold(idev);
1122
1123         /* For caller */
1124         refcount_set(&ifa->refcnt, 1);
1125
1126         rcu_read_lock_bh();
1127
1128         err = ipv6_add_addr_hash(idev->dev, ifa);
1129         if (err < 0) {
1130                 rcu_read_unlock_bh();
1131                 goto out;
1132         }
1133
1134         write_lock(&idev->lock);
1135
1136         /* Add to inet6_dev unicast addr list. */
1137         ipv6_link_dev_addr(idev, ifa);
1138
1139         if (ifa->flags&IFA_F_TEMPORARY) {
1140                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1141                 in6_ifa_hold(ifa);
1142         }
1143
1144         in6_ifa_hold(ifa);
1145         write_unlock(&idev->lock);
1146
1147         rcu_read_unlock_bh();
1148
1149         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1150 out:
1151         if (unlikely(err < 0)) {
1152                 fib6_info_release(f6i);
1153
1154                 if (ifa) {
1155                         if (ifa->idev)
1156                                 in6_dev_put(ifa->idev);
1157                         kfree(ifa);
1158                 }
1159                 ifa = ERR_PTR(err);
1160         }
1161
1162         return ifa;
1163 }
1164
1165 enum cleanup_prefix_rt_t {
1166         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1167         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1168         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1169 };
1170
1171 /*
1172  * Check, whether the prefix for ifp would still need a prefix route
1173  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1174  * constants.
1175  *
1176  * 1) we don't purge prefix if address was not permanent.
1177  *    prefix is managed by its own lifetime.
1178  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1179  * 3) if there are no addresses, delete prefix.
1180  * 4) if there are still other permanent address(es),
1181  *    corresponding prefix is still permanent.
1182  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1183  *    don't purge the prefix, assume user space is managing it.
1184  * 6) otherwise, update prefix lifetime to the
1185  *    longest valid lifetime among the corresponding
1186  *    addresses on the device.
1187  *    Note: subsequent RA will update lifetime.
1188  **/
1189 static enum cleanup_prefix_rt_t
1190 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1191 {
1192         struct inet6_ifaddr *ifa;
1193         struct inet6_dev *idev = ifp->idev;
1194         unsigned long lifetime;
1195         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1196
1197         *expires = jiffies;
1198
1199         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1200                 if (ifa == ifp)
1201                         continue;
1202                 if (ifa->prefix_len != ifp->prefix_len ||
1203                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1204                                        ifp->prefix_len))
1205                         continue;
1206                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1207                         return CLEANUP_PREFIX_RT_NOP;
1208
1209                 action = CLEANUP_PREFIX_RT_EXPIRE;
1210
1211                 spin_lock(&ifa->lock);
1212
1213                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1214                 /*
1215                  * Note: Because this address is
1216                  * not permanent, lifetime <
1217                  * LONG_MAX / HZ here.
1218                  */
1219                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1220                         *expires = ifa->tstamp + lifetime * HZ;
1221                 spin_unlock(&ifa->lock);
1222         }
1223
1224         return action;
1225 }
1226
1227 static void
1228 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1229 {
1230         struct fib6_info *f6i;
1231
1232         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
1233                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1234         if (f6i) {
1235                 if (del_rt)
1236                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1237                 else {
1238                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1239                                 fib6_set_expires(f6i, expires);
1240                         fib6_info_release(f6i);
1241                 }
1242         }
1243 }
1244
1245
1246 /* This function wants to get referenced ifp and releases it before return */
1247
1248 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1249 {
1250         int state;
1251         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1252         unsigned long expires;
1253
1254         ASSERT_RTNL();
1255
1256         spin_lock_bh(&ifp->lock);
1257         state = ifp->state;
1258         ifp->state = INET6_IFADDR_STATE_DEAD;
1259         spin_unlock_bh(&ifp->lock);
1260
1261         if (state == INET6_IFADDR_STATE_DEAD)
1262                 goto out;
1263
1264         spin_lock_bh(&addrconf_hash_lock);
1265         hlist_del_init_rcu(&ifp->addr_lst);
1266         spin_unlock_bh(&addrconf_hash_lock);
1267
1268         write_lock_bh(&ifp->idev->lock);
1269
1270         if (ifp->flags&IFA_F_TEMPORARY) {
1271                 list_del(&ifp->tmp_list);
1272                 if (ifp->ifpub) {
1273                         in6_ifa_put(ifp->ifpub);
1274                         ifp->ifpub = NULL;
1275                 }
1276                 __in6_ifa_put(ifp);
1277         }
1278
1279         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1280                 action = check_cleanup_prefix_route(ifp, &expires);
1281
1282         list_del_rcu(&ifp->if_list);
1283         __in6_ifa_put(ifp);
1284
1285         write_unlock_bh(&ifp->idev->lock);
1286
1287         addrconf_del_dad_work(ifp);
1288
1289         ipv6_ifa_notify(RTM_DELADDR, ifp);
1290
1291         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1292
1293         if (action != CLEANUP_PREFIX_RT_NOP) {
1294                 cleanup_prefix_route(ifp, expires,
1295                         action == CLEANUP_PREFIX_RT_DEL);
1296         }
1297
1298         /* clean up prefsrc entries */
1299         rt6_remove_prefsrc(ifp);
1300 out:
1301         in6_ifa_put(ifp);
1302 }
1303
1304 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1305                                 struct inet6_ifaddr *ift,
1306                                 bool block)
1307 {
1308         struct inet6_dev *idev = ifp->idev;
1309         struct in6_addr addr, *tmpaddr;
1310         unsigned long tmp_tstamp, age;
1311         unsigned long regen_advance;
1312         struct ifa6_config cfg;
1313         int ret = 0;
1314         unsigned long now = jiffies;
1315         long max_desync_factor;
1316         s32 cnf_temp_preferred_lft;
1317
1318         write_lock_bh(&idev->lock);
1319         if (ift) {
1320                 spin_lock_bh(&ift->lock);
1321                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1322                 spin_unlock_bh(&ift->lock);
1323                 tmpaddr = &addr;
1324         } else {
1325                 tmpaddr = NULL;
1326         }
1327 retry:
1328         in6_dev_hold(idev);
1329         if (idev->cnf.use_tempaddr <= 0) {
1330                 write_unlock_bh(&idev->lock);
1331                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1332                 in6_dev_put(idev);
1333                 ret = -1;
1334                 goto out;
1335         }
1336         spin_lock_bh(&ifp->lock);
1337         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1338                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1339                 spin_unlock_bh(&ifp->lock);
1340                 write_unlock_bh(&idev->lock);
1341                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1342                         __func__);
1343                 in6_dev_put(idev);
1344                 ret = -1;
1345                 goto out;
1346         }
1347         in6_ifa_hold(ifp);
1348         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1349         ipv6_try_regen_rndid(idev, tmpaddr);
1350         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1351         age = (now - ifp->tstamp) / HZ;
1352
1353         regen_advance = idev->cnf.regen_max_retry *
1354                         idev->cnf.dad_transmits *
1355                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1356
1357         /* recalculate max_desync_factor each time and update
1358          * idev->desync_factor if it's larger
1359          */
1360         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1361         max_desync_factor = min_t(__u32,
1362                                   idev->cnf.max_desync_factor,
1363                                   cnf_temp_preferred_lft - regen_advance);
1364
1365         if (unlikely(idev->desync_factor > max_desync_factor)) {
1366                 if (max_desync_factor > 0) {
1367                         get_random_bytes(&idev->desync_factor,
1368                                          sizeof(idev->desync_factor));
1369                         idev->desync_factor %= max_desync_factor;
1370                 } else {
1371                         idev->desync_factor = 0;
1372                 }
1373         }
1374
1375         memset(&cfg, 0, sizeof(cfg));
1376         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1377                               idev->cnf.temp_valid_lft + age);
1378         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1379         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1380
1381         cfg.plen = ifp->prefix_len;
1382         tmp_tstamp = ifp->tstamp;
1383         spin_unlock_bh(&ifp->lock);
1384
1385         write_unlock_bh(&idev->lock);
1386
1387         /* A temporary address is created only if this calculated Preferred
1388          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1389          * an implementation must not create a temporary address with a zero
1390          * Preferred Lifetime.
1391          * Use age calculation as in addrconf_verify to avoid unnecessary
1392          * temporary addresses being generated.
1393          */
1394         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1395         if (cfg.preferred_lft <= regen_advance + age) {
1396                 in6_ifa_put(ifp);
1397                 in6_dev_put(idev);
1398                 ret = -1;
1399                 goto out;
1400         }
1401
1402         cfg.ifa_flags = IFA_F_TEMPORARY;
1403         /* set in addrconf_prefix_rcv() */
1404         if (ifp->flags & IFA_F_OPTIMISTIC)
1405                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1406
1407         cfg.pfx = &addr;
1408         cfg.scope = ipv6_addr_scope(cfg.pfx);
1409
1410         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1411         if (IS_ERR(ift)) {
1412                 in6_ifa_put(ifp);
1413                 in6_dev_put(idev);
1414                 pr_info("%s: retry temporary address regeneration\n", __func__);
1415                 tmpaddr = &addr;
1416                 write_lock_bh(&idev->lock);
1417                 goto retry;
1418         }
1419
1420         spin_lock_bh(&ift->lock);
1421         ift->ifpub = ifp;
1422         ift->cstamp = now;
1423         ift->tstamp = tmp_tstamp;
1424         spin_unlock_bh(&ift->lock);
1425
1426         addrconf_dad_start(ift);
1427         in6_ifa_put(ift);
1428         in6_dev_put(idev);
1429 out:
1430         return ret;
1431 }
1432
1433 /*
1434  *      Choose an appropriate source address (RFC3484)
1435  */
1436 enum {
1437         IPV6_SADDR_RULE_INIT = 0,
1438         IPV6_SADDR_RULE_LOCAL,
1439         IPV6_SADDR_RULE_SCOPE,
1440         IPV6_SADDR_RULE_PREFERRED,
1441 #ifdef CONFIG_IPV6_MIP6
1442         IPV6_SADDR_RULE_HOA,
1443 #endif
1444         IPV6_SADDR_RULE_OIF,
1445         IPV6_SADDR_RULE_LABEL,
1446         IPV6_SADDR_RULE_PRIVACY,
1447         IPV6_SADDR_RULE_ORCHID,
1448         IPV6_SADDR_RULE_PREFIX,
1449 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1450         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1451 #endif
1452         IPV6_SADDR_RULE_MAX
1453 };
1454
1455 struct ipv6_saddr_score {
1456         int                     rule;
1457         int                     addr_type;
1458         struct inet6_ifaddr     *ifa;
1459         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1460         int                     scopedist;
1461         int                     matchlen;
1462 };
1463
1464 struct ipv6_saddr_dst {
1465         const struct in6_addr *addr;
1466         int ifindex;
1467         int scope;
1468         int label;
1469         unsigned int prefs;
1470 };
1471
1472 static inline int ipv6_saddr_preferred(int type)
1473 {
1474         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1475                 return 1;
1476         return 0;
1477 }
1478
1479 static bool ipv6_use_optimistic_addr(struct net *net,
1480                                      struct inet6_dev *idev)
1481 {
1482 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1483         if (!idev)
1484                 return false;
1485         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1486                 return false;
1487         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1488                 return false;
1489
1490         return true;
1491 #else
1492         return false;
1493 #endif
1494 }
1495
1496 static bool ipv6_allow_optimistic_dad(struct net *net,
1497                                       struct inet6_dev *idev)
1498 {
1499 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1500         if (!idev)
1501                 return false;
1502         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1503                 return false;
1504
1505         return true;
1506 #else
1507         return false;
1508 #endif
1509 }
1510
1511 static int ipv6_get_saddr_eval(struct net *net,
1512                                struct ipv6_saddr_score *score,
1513                                struct ipv6_saddr_dst *dst,
1514                                int i)
1515 {
1516         int ret;
1517
1518         if (i <= score->rule) {
1519                 switch (i) {
1520                 case IPV6_SADDR_RULE_SCOPE:
1521                         ret = score->scopedist;
1522                         break;
1523                 case IPV6_SADDR_RULE_PREFIX:
1524                         ret = score->matchlen;
1525                         break;
1526                 default:
1527                         ret = !!test_bit(i, score->scorebits);
1528                 }
1529                 goto out;
1530         }
1531
1532         switch (i) {
1533         case IPV6_SADDR_RULE_INIT:
1534                 /* Rule 0: remember if hiscore is not ready yet */
1535                 ret = !!score->ifa;
1536                 break;
1537         case IPV6_SADDR_RULE_LOCAL:
1538                 /* Rule 1: Prefer same address */
1539                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1540                 break;
1541         case IPV6_SADDR_RULE_SCOPE:
1542                 /* Rule 2: Prefer appropriate scope
1543                  *
1544                  *      ret
1545                  *       ^
1546                  *    -1 |  d 15
1547                  *    ---+--+-+---> scope
1548                  *       |
1549                  *       |             d is scope of the destination.
1550                  *  B-d  |  \
1551                  *       |   \      <- smaller scope is better if
1552                  *  B-15 |    \        if scope is enough for destination.
1553                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1554                  * d-C-1 | /
1555                  *       |/         <- greater is better
1556                  *   -C  /             if scope is not enough for destination.
1557                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1558                  *
1559                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1560                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1561                  * Assume B = 0 and we get C > 29.
1562                  */
1563                 ret = __ipv6_addr_src_scope(score->addr_type);
1564                 if (ret >= dst->scope)
1565                         ret = -ret;
1566                 else
1567                         ret -= 128;     /* 30 is enough */
1568                 score->scopedist = ret;
1569                 break;
1570         case IPV6_SADDR_RULE_PREFERRED:
1571             {
1572                 /* Rule 3: Avoid deprecated and optimistic addresses */
1573                 u8 avoid = IFA_F_DEPRECATED;
1574
1575                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1576                         avoid |= IFA_F_OPTIMISTIC;
1577                 ret = ipv6_saddr_preferred(score->addr_type) ||
1578                       !(score->ifa->flags & avoid);
1579                 break;
1580             }
1581 #ifdef CONFIG_IPV6_MIP6
1582         case IPV6_SADDR_RULE_HOA:
1583             {
1584                 /* Rule 4: Prefer home address */
1585                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1586                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1587                 break;
1588             }
1589 #endif
1590         case IPV6_SADDR_RULE_OIF:
1591                 /* Rule 5: Prefer outgoing interface */
1592                 ret = (!dst->ifindex ||
1593                        dst->ifindex == score->ifa->idev->dev->ifindex);
1594                 break;
1595         case IPV6_SADDR_RULE_LABEL:
1596                 /* Rule 6: Prefer matching label */
1597                 ret = ipv6_addr_label(net,
1598                                       &score->ifa->addr, score->addr_type,
1599                                       score->ifa->idev->dev->ifindex) == dst->label;
1600                 break;
1601         case IPV6_SADDR_RULE_PRIVACY:
1602             {
1603                 /* Rule 7: Prefer public address
1604                  * Note: prefer temporary address if use_tempaddr >= 2
1605                  */
1606                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1607                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1608                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1609                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1610                 break;
1611             }
1612         case IPV6_SADDR_RULE_ORCHID:
1613                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1614                  *          non-ORCHID vs non-ORCHID
1615                  */
1616                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1617                         ipv6_addr_orchid(dst->addr));
1618                 break;
1619         case IPV6_SADDR_RULE_PREFIX:
1620                 /* Rule 8: Use longest matching prefix */
1621                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1622                 if (ret > score->ifa->prefix_len)
1623                         ret = score->ifa->prefix_len;
1624                 score->matchlen = ret;
1625                 break;
1626 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1627         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1628                 /* Optimistic addresses still have lower precedence than other
1629                  * preferred addresses.
1630                  */
1631                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1632                 break;
1633 #endif
1634         default:
1635                 ret = 0;
1636         }
1637
1638         if (ret)
1639                 __set_bit(i, score->scorebits);
1640         score->rule = i;
1641 out:
1642         return ret;
1643 }
1644
1645 static int __ipv6_dev_get_saddr(struct net *net,
1646                                 struct ipv6_saddr_dst *dst,
1647                                 struct inet6_dev *idev,
1648                                 struct ipv6_saddr_score *scores,
1649                                 int hiscore_idx)
1650 {
1651         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1652
1653         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1654                 int i;
1655
1656                 /*
1657                  * - Tentative Address (RFC2462 section 5.4)
1658                  *  - A tentative address is not considered
1659                  *    "assigned to an interface" in the traditional
1660                  *    sense, unless it is also flagged as optimistic.
1661                  * - Candidate Source Address (section 4)
1662                  *  - In any case, anycast addresses, multicast
1663                  *    addresses, and the unspecified address MUST
1664                  *    NOT be included in a candidate set.
1665                  */
1666                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1667                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1668                         continue;
1669
1670                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1671
1672                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1673                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1674                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1675                                             idev->dev->name);
1676                         continue;
1677                 }
1678
1679                 score->rule = -1;
1680                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1681
1682                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1683                         int minihiscore, miniscore;
1684
1685                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1686                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1687
1688                         if (minihiscore > miniscore) {
1689                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1690                                     score->scopedist > 0) {
1691                                         /*
1692                                          * special case:
1693                                          * each remaining entry
1694                                          * has too small (not enough)
1695                                          * scope, because ifa entries
1696                                          * are sorted by their scope
1697                                          * values.
1698                                          */
1699                                         goto out;
1700                                 }
1701                                 break;
1702                         } else if (minihiscore < miniscore) {
1703                                 swap(hiscore, score);
1704                                 hiscore_idx = 1 - hiscore_idx;
1705
1706                                 /* restore our iterator */
1707                                 score->ifa = hiscore->ifa;
1708
1709                                 break;
1710                         }
1711                 }
1712         }
1713 out:
1714         return hiscore_idx;
1715 }
1716
1717 static int ipv6_get_saddr_master(struct net *net,
1718                                  const struct net_device *dst_dev,
1719                                  const struct net_device *master,
1720                                  struct ipv6_saddr_dst *dst,
1721                                  struct ipv6_saddr_score *scores,
1722                                  int hiscore_idx)
1723 {
1724         struct inet6_dev *idev;
1725
1726         idev = __in6_dev_get(dst_dev);
1727         if (idev)
1728                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1729                                                    scores, hiscore_idx);
1730
1731         idev = __in6_dev_get(master);
1732         if (idev)
1733                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1734                                                    scores, hiscore_idx);
1735
1736         return hiscore_idx;
1737 }
1738
1739 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1740                        const struct in6_addr *daddr, unsigned int prefs,
1741                        struct in6_addr *saddr)
1742 {
1743         struct ipv6_saddr_score scores[2], *hiscore;
1744         struct ipv6_saddr_dst dst;
1745         struct inet6_dev *idev;
1746         struct net_device *dev;
1747         int dst_type;
1748         bool use_oif_addr = false;
1749         int hiscore_idx = 0;
1750         int ret = 0;
1751
1752         dst_type = __ipv6_addr_type(daddr);
1753         dst.addr = daddr;
1754         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1755         dst.scope = __ipv6_addr_src_scope(dst_type);
1756         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1757         dst.prefs = prefs;
1758
1759         scores[hiscore_idx].rule = -1;
1760         scores[hiscore_idx].ifa = NULL;
1761
1762         rcu_read_lock();
1763
1764         /* Candidate Source Address (section 4)
1765          *  - multicast and link-local destination address,
1766          *    the set of candidate source address MUST only
1767          *    include addresses assigned to interfaces
1768          *    belonging to the same link as the outgoing
1769          *    interface.
1770          * (- For site-local destination addresses, the
1771          *    set of candidate source addresses MUST only
1772          *    include addresses assigned to interfaces
1773          *    belonging to the same site as the outgoing
1774          *    interface.)
1775          *  - "It is RECOMMENDED that the candidate source addresses
1776          *    be the set of unicast addresses assigned to the
1777          *    interface that will be used to send to the destination
1778          *    (the 'outgoing' interface)." (RFC 6724)
1779          */
1780         if (dst_dev) {
1781                 idev = __in6_dev_get(dst_dev);
1782                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1783                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1784                     (idev && idev->cnf.use_oif_addrs_only)) {
1785                         use_oif_addr = true;
1786                 }
1787         }
1788
1789         if (use_oif_addr) {
1790                 if (idev)
1791                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1792         } else {
1793                 const struct net_device *master;
1794                 int master_idx = 0;
1795
1796                 /* if dst_dev exists and is enslaved to an L3 device, then
1797                  * prefer addresses from dst_dev and then the master over
1798                  * any other enslaved devices in the L3 domain.
1799                  */
1800                 master = l3mdev_master_dev_rcu(dst_dev);
1801                 if (master) {
1802                         master_idx = master->ifindex;
1803
1804                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1805                                                             master, &dst,
1806                                                             scores, hiscore_idx);
1807
1808                         if (scores[hiscore_idx].ifa)
1809                                 goto out;
1810                 }
1811
1812                 for_each_netdev_rcu(net, dev) {
1813                         /* only consider addresses on devices in the
1814                          * same L3 domain
1815                          */
1816                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1817                                 continue;
1818                         idev = __in6_dev_get(dev);
1819                         if (!idev)
1820                                 continue;
1821                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1822                 }
1823         }
1824
1825 out:
1826         hiscore = &scores[hiscore_idx];
1827         if (!hiscore->ifa)
1828                 ret = -EADDRNOTAVAIL;
1829         else
1830                 *saddr = hiscore->ifa->addr;
1831
1832         rcu_read_unlock();
1833         return ret;
1834 }
1835 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1836
1837 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1838                       u32 banned_flags)
1839 {
1840         struct inet6_ifaddr *ifp;
1841         int err = -EADDRNOTAVAIL;
1842
1843         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1844                 if (ifp->scope > IFA_LINK)
1845                         break;
1846                 if (ifp->scope == IFA_LINK &&
1847                     !(ifp->flags & banned_flags)) {
1848                         *addr = ifp->addr;
1849                         err = 0;
1850                         break;
1851                 }
1852         }
1853         return err;
1854 }
1855
1856 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1857                     u32 banned_flags)
1858 {
1859         struct inet6_dev *idev;
1860         int err = -EADDRNOTAVAIL;
1861
1862         rcu_read_lock();
1863         idev = __in6_dev_get(dev);
1864         if (idev) {
1865                 read_lock_bh(&idev->lock);
1866                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1867                 read_unlock_bh(&idev->lock);
1868         }
1869         rcu_read_unlock();
1870         return err;
1871 }
1872
1873 static int ipv6_count_addresses(const struct inet6_dev *idev)
1874 {
1875         const struct inet6_ifaddr *ifp;
1876         int cnt = 0;
1877
1878         rcu_read_lock();
1879         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1880                 cnt++;
1881         rcu_read_unlock();
1882         return cnt;
1883 }
1884
1885 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1886                   const struct net_device *dev, int strict)
1887 {
1888         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1889                                        strict, IFA_F_TENTATIVE);
1890 }
1891 EXPORT_SYMBOL(ipv6_chk_addr);
1892
1893 /* device argument is used to find the L3 domain of interest. If
1894  * skip_dev_check is set, then the ifp device is not checked against
1895  * the passed in dev argument. So the 2 cases for addresses checks are:
1896  *   1. does the address exist in the L3 domain that dev is part of
1897  *      (skip_dev_check = true), or
1898  *
1899  *   2. does the address exist on the specific device
1900  *      (skip_dev_check = false)
1901  */
1902 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1903                             const struct net_device *dev, bool skip_dev_check,
1904                             int strict, u32 banned_flags)
1905 {
1906         unsigned int hash = inet6_addr_hash(net, addr);
1907         const struct net_device *l3mdev;
1908         struct inet6_ifaddr *ifp;
1909         u32 ifp_flags;
1910
1911         rcu_read_lock();
1912
1913         l3mdev = l3mdev_master_dev_rcu(dev);
1914         if (skip_dev_check)
1915                 dev = NULL;
1916
1917         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1918                 if (!net_eq(dev_net(ifp->idev->dev), net))
1919                         continue;
1920
1921                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1922                         continue;
1923
1924                 /* Decouple optimistic from tentative for evaluation here.
1925                  * Ban optimistic addresses explicitly, when required.
1926                  */
1927                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1928                             ? (ifp->flags&~IFA_F_TENTATIVE)
1929                             : ifp->flags;
1930                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1931                     !(ifp_flags&banned_flags) &&
1932                     (!dev || ifp->idev->dev == dev ||
1933                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1934                         rcu_read_unlock();
1935                         return 1;
1936                 }
1937         }
1938
1939         rcu_read_unlock();
1940         return 0;
1941 }
1942 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1943
1944
1945 /* Compares an address/prefix_len with addresses on device @dev.
1946  * If one is found it returns true.
1947  */
1948 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1949         const unsigned int prefix_len, struct net_device *dev)
1950 {
1951         const struct inet6_ifaddr *ifa;
1952         const struct inet6_dev *idev;
1953         bool ret = false;
1954
1955         rcu_read_lock();
1956         idev = __in6_dev_get(dev);
1957         if (idev) {
1958                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1959                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1960                         if (ret)
1961                                 break;
1962                 }
1963         }
1964         rcu_read_unlock();
1965
1966         return ret;
1967 }
1968 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1969
1970 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1971 {
1972         const struct inet6_ifaddr *ifa;
1973         const struct inet6_dev *idev;
1974         int     onlink;
1975
1976         onlink = 0;
1977         rcu_read_lock();
1978         idev = __in6_dev_get(dev);
1979         if (idev) {
1980                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1981                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1982                                                    ifa->prefix_len);
1983                         if (onlink)
1984                                 break;
1985                 }
1986         }
1987         rcu_read_unlock();
1988         return onlink;
1989 }
1990 EXPORT_SYMBOL(ipv6_chk_prefix);
1991
1992 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1993                                      struct net_device *dev, int strict)
1994 {
1995         unsigned int hash = inet6_addr_hash(net, addr);
1996         struct inet6_ifaddr *ifp, *result = NULL;
1997
1998         rcu_read_lock();
1999         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2000                 if (!net_eq(dev_net(ifp->idev->dev), net))
2001                         continue;
2002                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2003                         if (!dev || ifp->idev->dev == dev ||
2004                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2005                                 result = ifp;
2006                                 in6_ifa_hold(ifp);
2007                                 break;
2008                         }
2009                 }
2010         }
2011         rcu_read_unlock();
2012
2013         return result;
2014 }
2015
2016 /* Gets referenced address, destroys ifaddr */
2017
2018 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2019 {
2020         if (dad_failed)
2021                 ifp->flags |= IFA_F_DADFAILED;
2022
2023         if (ifp->flags&IFA_F_TEMPORARY) {
2024                 struct inet6_ifaddr *ifpub;
2025                 spin_lock_bh(&ifp->lock);
2026                 ifpub = ifp->ifpub;
2027                 if (ifpub) {
2028                         in6_ifa_hold(ifpub);
2029                         spin_unlock_bh(&ifp->lock);
2030                         ipv6_create_tempaddr(ifpub, ifp, true);
2031                         in6_ifa_put(ifpub);
2032                 } else {
2033                         spin_unlock_bh(&ifp->lock);
2034                 }
2035                 ipv6_del_addr(ifp);
2036         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2037                 spin_lock_bh(&ifp->lock);
2038                 addrconf_del_dad_work(ifp);
2039                 ifp->flags |= IFA_F_TENTATIVE;
2040                 if (dad_failed)
2041                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2042                 spin_unlock_bh(&ifp->lock);
2043                 if (dad_failed)
2044                         ipv6_ifa_notify(0, ifp);
2045                 in6_ifa_put(ifp);
2046         } else {
2047                 ipv6_del_addr(ifp);
2048         }
2049 }
2050
2051 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2052 {
2053         int err = -ENOENT;
2054
2055         spin_lock_bh(&ifp->lock);
2056         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2057                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2058                 err = 0;
2059         }
2060         spin_unlock_bh(&ifp->lock);
2061
2062         return err;
2063 }
2064
2065 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2066 {
2067         struct inet6_dev *idev = ifp->idev;
2068         struct net *net = dev_net(ifp->idev->dev);
2069
2070         if (addrconf_dad_end(ifp)) {
2071                 in6_ifa_put(ifp);
2072                 return;
2073         }
2074
2075         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2076                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2077
2078         spin_lock_bh(&ifp->lock);
2079
2080         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2081                 struct in6_addr new_addr;
2082                 struct inet6_ifaddr *ifp2;
2083                 int retries = ifp->stable_privacy_retry + 1;
2084                 struct ifa6_config cfg = {
2085                         .pfx = &new_addr,
2086                         .plen = ifp->prefix_len,
2087                         .ifa_flags = ifp->flags,
2088                         .valid_lft = ifp->valid_lft,
2089                         .preferred_lft = ifp->prefered_lft,
2090                         .scope = ifp->scope,
2091                 };
2092
2093                 if (retries > net->ipv6.sysctl.idgen_retries) {
2094                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2095                                              ifp->idev->dev->name);
2096                         goto errdad;
2097                 }
2098
2099                 new_addr = ifp->addr;
2100                 if (ipv6_generate_stable_address(&new_addr, retries,
2101                                                  idev))
2102                         goto errdad;
2103
2104                 spin_unlock_bh(&ifp->lock);
2105
2106                 if (idev->cnf.max_addresses &&
2107                     ipv6_count_addresses(idev) >=
2108                     idev->cnf.max_addresses)
2109                         goto lock_errdad;
2110
2111                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2112                                      ifp->idev->dev->name);
2113
2114                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2115                 if (IS_ERR(ifp2))
2116                         goto lock_errdad;
2117
2118                 spin_lock_bh(&ifp2->lock);
2119                 ifp2->stable_privacy_retry = retries;
2120                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2121                 spin_unlock_bh(&ifp2->lock);
2122
2123                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2124                 in6_ifa_put(ifp2);
2125 lock_errdad:
2126                 spin_lock_bh(&ifp->lock);
2127         }
2128
2129 errdad:
2130         /* transition from _POSTDAD to _ERRDAD */
2131         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2132         spin_unlock_bh(&ifp->lock);
2133
2134         addrconf_mod_dad_work(ifp, 0);
2135         in6_ifa_put(ifp);
2136 }
2137
2138 /* Join to solicited addr multicast group.
2139  * caller must hold RTNL */
2140 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2141 {
2142         struct in6_addr maddr;
2143
2144         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2145                 return;
2146
2147         addrconf_addr_solict_mult(addr, &maddr);
2148         ipv6_dev_mc_inc(dev, &maddr);
2149 }
2150
2151 /* caller must hold RTNL */
2152 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2153 {
2154         struct in6_addr maddr;
2155
2156         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2157                 return;
2158
2159         addrconf_addr_solict_mult(addr, &maddr);
2160         __ipv6_dev_mc_dec(idev, &maddr);
2161 }
2162
2163 /* caller must hold RTNL */
2164 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2165 {
2166         struct in6_addr addr;
2167
2168         if (ifp->prefix_len >= 127) /* RFC 6164 */
2169                 return;
2170         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2171         if (ipv6_addr_any(&addr))
2172                 return;
2173         __ipv6_dev_ac_inc(ifp->idev, &addr);
2174 }
2175
2176 /* caller must hold RTNL */
2177 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2178 {
2179         struct in6_addr addr;
2180
2181         if (ifp->prefix_len >= 127) /* RFC 6164 */
2182                 return;
2183         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2184         if (ipv6_addr_any(&addr))
2185                 return;
2186         __ipv6_dev_ac_dec(ifp->idev, &addr);
2187 }
2188
2189 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2190 {
2191         switch (dev->addr_len) {
2192         case ETH_ALEN:
2193                 memcpy(eui, dev->dev_addr, 3);
2194                 eui[3] = 0xFF;
2195                 eui[4] = 0xFE;
2196                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2197                 break;
2198         case EUI64_ADDR_LEN:
2199                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2200                 eui[0] ^= 2;
2201                 break;
2202         default:
2203                 return -1;
2204         }
2205
2206         return 0;
2207 }
2208
2209 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2210 {
2211         union fwnet_hwaddr *ha;
2212
2213         if (dev->addr_len != FWNET_ALEN)
2214                 return -1;
2215
2216         ha = (union fwnet_hwaddr *)dev->dev_addr;
2217
2218         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2219         eui[0] ^= 2;
2220         return 0;
2221 }
2222
2223 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2224 {
2225         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2226         if (dev->addr_len != ARCNET_ALEN)
2227                 return -1;
2228         memset(eui, 0, 7);
2229         eui[7] = *(u8 *)dev->dev_addr;
2230         return 0;
2231 }
2232
2233 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2234 {
2235         if (dev->addr_len != INFINIBAND_ALEN)
2236                 return -1;
2237         memcpy(eui, dev->dev_addr + 12, 8);
2238         eui[0] |= 2;
2239         return 0;
2240 }
2241
2242 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2243 {
2244         if (addr == 0)
2245                 return -1;
2246         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2247                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2248                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2249                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2250                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2251                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2252         eui[1] = 0;
2253         eui[2] = 0x5E;
2254         eui[3] = 0xFE;
2255         memcpy(eui + 4, &addr, 4);
2256         return 0;
2257 }
2258
2259 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2260 {
2261         if (dev->priv_flags & IFF_ISATAP)
2262                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2263         return -1;
2264 }
2265
2266 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2267 {
2268         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2269 }
2270
2271 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2272 {
2273         memcpy(eui, dev->perm_addr, 3);
2274         memcpy(eui + 5, dev->perm_addr + 3, 3);
2275         eui[3] = 0xFF;
2276         eui[4] = 0xFE;
2277         eui[0] ^= 2;
2278         return 0;
2279 }
2280
2281 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2282 {
2283         switch (dev->type) {
2284         case ARPHRD_ETHER:
2285         case ARPHRD_FDDI:
2286                 return addrconf_ifid_eui48(eui, dev);
2287         case ARPHRD_ARCNET:
2288                 return addrconf_ifid_arcnet(eui, dev);
2289         case ARPHRD_INFINIBAND:
2290                 return addrconf_ifid_infiniband(eui, dev);
2291         case ARPHRD_SIT:
2292                 return addrconf_ifid_sit(eui, dev);
2293         case ARPHRD_IPGRE:
2294         case ARPHRD_TUNNEL:
2295                 return addrconf_ifid_gre(eui, dev);
2296         case ARPHRD_6LOWPAN:
2297                 return addrconf_ifid_6lowpan(eui, dev);
2298         case ARPHRD_IEEE1394:
2299                 return addrconf_ifid_ieee1394(eui, dev);
2300         case ARPHRD_TUNNEL6:
2301         case ARPHRD_IP6GRE:
2302         case ARPHRD_RAWIP:
2303                 return addrconf_ifid_ip6tnl(eui, dev);
2304         }
2305         return -1;
2306 }
2307
2308 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2309 {
2310         int err = -1;
2311         struct inet6_ifaddr *ifp;
2312
2313         read_lock_bh(&idev->lock);
2314         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2315                 if (ifp->scope > IFA_LINK)
2316                         break;
2317                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2318                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2319                         err = 0;
2320                         break;
2321                 }
2322         }
2323         read_unlock_bh(&idev->lock);
2324         return err;
2325 }
2326
2327 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2328 static void ipv6_regen_rndid(struct inet6_dev *idev)
2329 {
2330 regen:
2331         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2332         idev->rndid[0] &= ~0x02;
2333
2334         /*
2335          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2336          * check if generated address is not inappropriate
2337          *
2338          *  - Reserved subnet anycast (RFC 2526)
2339          *      11111101 11....11 1xxxxxxx
2340          *  - ISATAP (RFC4214) 6.1
2341          *      00-00-5E-FE-xx-xx-xx-xx
2342          *  - value 0
2343          *  - XXX: already assigned to an address on the device
2344          */
2345         if (idev->rndid[0] == 0xfd &&
2346             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2347             (idev->rndid[7]&0x80))
2348                 goto regen;
2349         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2350                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2351                         goto regen;
2352                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2353                         goto regen;
2354         }
2355 }
2356
2357 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2358 {
2359         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2360                 ipv6_regen_rndid(idev);
2361 }
2362
2363 /*
2364  *      Add prefix route.
2365  */
2366
2367 static void
2368 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2369                       struct net_device *dev, unsigned long expires,
2370                       u32 flags, gfp_t gfp_flags)
2371 {
2372         struct fib6_config cfg = {
2373                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2374                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2375                 .fc_ifindex = dev->ifindex,
2376                 .fc_expires = expires,
2377                 .fc_dst_len = plen,
2378                 .fc_flags = RTF_UP | flags,
2379                 .fc_nlinfo.nl_net = dev_net(dev),
2380                 .fc_protocol = RTPROT_KERNEL,
2381                 .fc_type = RTN_UNICAST,
2382         };
2383
2384         cfg.fc_dst = *pfx;
2385
2386         /* Prevent useless cloning on PtP SIT.
2387            This thing is done here expecting that the whole
2388            class of non-broadcast devices need not cloning.
2389          */
2390 #if IS_ENABLED(CONFIG_IPV6_SIT)
2391         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2392                 cfg.fc_flags |= RTF_NONEXTHOP;
2393 #endif
2394
2395         ip6_route_add(&cfg, gfp_flags, NULL);
2396 }
2397
2398
2399 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2400                                                   int plen,
2401                                                   const struct net_device *dev,
2402                                                   u32 flags, u32 noflags,
2403                                                   bool no_gw)
2404 {
2405         struct fib6_node *fn;
2406         struct fib6_info *rt = NULL;
2407         struct fib6_table *table;
2408         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2409
2410         table = fib6_get_table(dev_net(dev), tb_id);
2411         if (!table)
2412                 return NULL;
2413
2414         rcu_read_lock();
2415         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2416         if (!fn)
2417                 goto out;
2418
2419         for_each_fib6_node_rt_rcu(fn) {
2420                 if (rt->fib6_nh.fib_nh_dev->ifindex != dev->ifindex)
2421                         continue;
2422                 if (no_gw && rt->fib6_nh.fib_nh_gw_family)
2423                         continue;
2424                 if ((rt->fib6_flags & flags) != flags)
2425                         continue;
2426                 if ((rt->fib6_flags & noflags) != 0)
2427                         continue;
2428                 if (!fib6_info_hold_safe(rt))
2429                         continue;
2430                 break;
2431         }
2432 out:
2433         rcu_read_unlock();
2434         return rt;
2435 }
2436
2437
2438 /* Create "default" multicast route to the interface */
2439
2440 static void addrconf_add_mroute(struct net_device *dev)
2441 {
2442         struct fib6_config cfg = {
2443                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2444                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2445                 .fc_ifindex = dev->ifindex,
2446                 .fc_dst_len = 8,
2447                 .fc_flags = RTF_UP,
2448                 .fc_type = RTN_UNICAST,
2449                 .fc_nlinfo.nl_net = dev_net(dev),
2450         };
2451
2452         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2453
2454         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2455 }
2456
2457 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2458 {
2459         struct inet6_dev *idev;
2460
2461         ASSERT_RTNL();
2462
2463         idev = ipv6_find_idev(dev);
2464         if (!idev)
2465                 return ERR_PTR(-ENOBUFS);
2466
2467         if (idev->cnf.disable_ipv6)
2468                 return ERR_PTR(-EACCES);
2469
2470         /* Add default multicast route */
2471         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2472                 addrconf_add_mroute(dev);
2473
2474         return idev;
2475 }
2476
2477 static void manage_tempaddrs(struct inet6_dev *idev,
2478                              struct inet6_ifaddr *ifp,
2479                              __u32 valid_lft, __u32 prefered_lft,
2480                              bool create, unsigned long now)
2481 {
2482         u32 flags;
2483         struct inet6_ifaddr *ift;
2484
2485         read_lock_bh(&idev->lock);
2486         /* update all temporary addresses in the list */
2487         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2488                 int age, max_valid, max_prefered;
2489
2490                 if (ifp != ift->ifpub)
2491                         continue;
2492
2493                 /* RFC 4941 section 3.3:
2494                  * If a received option will extend the lifetime of a public
2495                  * address, the lifetimes of temporary addresses should
2496                  * be extended, subject to the overall constraint that no
2497                  * temporary addresses should ever remain "valid" or "preferred"
2498                  * for a time longer than (TEMP_VALID_LIFETIME) or
2499                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2500                  */
2501                 age = (now - ift->cstamp) / HZ;
2502                 max_valid = idev->cnf.temp_valid_lft - age;
2503                 if (max_valid < 0)
2504                         max_valid = 0;
2505
2506                 max_prefered = idev->cnf.temp_prefered_lft -
2507                                idev->desync_factor - age;
2508                 if (max_prefered < 0)
2509                         max_prefered = 0;
2510
2511                 if (valid_lft > max_valid)
2512                         valid_lft = max_valid;
2513
2514                 if (prefered_lft > max_prefered)
2515                         prefered_lft = max_prefered;
2516
2517                 spin_lock(&ift->lock);
2518                 flags = ift->flags;
2519                 ift->valid_lft = valid_lft;
2520                 ift->prefered_lft = prefered_lft;
2521                 ift->tstamp = now;
2522                 if (prefered_lft > 0)
2523                         ift->flags &= ~IFA_F_DEPRECATED;
2524
2525                 spin_unlock(&ift->lock);
2526                 if (!(flags&IFA_F_TENTATIVE))
2527                         ipv6_ifa_notify(0, ift);
2528         }
2529
2530         if ((create || list_empty(&idev->tempaddr_list)) &&
2531             idev->cnf.use_tempaddr > 0) {
2532                 /* When a new public address is created as described
2533                  * in [ADDRCONF], also create a new temporary address.
2534                  * Also create a temporary address if it's enabled but
2535                  * no temporary address currently exists.
2536                  */
2537                 read_unlock_bh(&idev->lock);
2538                 ipv6_create_tempaddr(ifp, NULL, false);
2539         } else {
2540                 read_unlock_bh(&idev->lock);
2541         }
2542 }
2543
2544 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2545 {
2546         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2547                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2548 }
2549
2550 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2551                                  const struct prefix_info *pinfo,
2552                                  struct inet6_dev *in6_dev,
2553                                  const struct in6_addr *addr, int addr_type,
2554                                  u32 addr_flags, bool sllao, bool tokenized,
2555                                  __u32 valid_lft, u32 prefered_lft)
2556 {
2557         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2558         int create = 0, update_lft = 0;
2559
2560         if (!ifp && valid_lft) {
2561                 int max_addresses = in6_dev->cnf.max_addresses;
2562                 struct ifa6_config cfg = {
2563                         .pfx = addr,
2564                         .plen = pinfo->prefix_len,
2565                         .ifa_flags = addr_flags,
2566                         .valid_lft = valid_lft,
2567                         .preferred_lft = prefered_lft,
2568                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2569                 };
2570
2571 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2572                 if ((net->ipv6.devconf_all->optimistic_dad ||
2573                      in6_dev->cnf.optimistic_dad) &&
2574                     !net->ipv6.devconf_all->forwarding && sllao)
2575                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2576 #endif
2577
2578                 /* Do not allow to create too much of autoconfigured
2579                  * addresses; this would be too easy way to crash kernel.
2580                  */
2581                 if (!max_addresses ||
2582                     ipv6_count_addresses(in6_dev) < max_addresses)
2583                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2584
2585                 if (IS_ERR_OR_NULL(ifp))
2586                         return -1;
2587
2588                 create = 1;
2589                 spin_lock_bh(&ifp->lock);
2590                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2591                 ifp->cstamp = jiffies;
2592                 ifp->tokenized = tokenized;
2593                 spin_unlock_bh(&ifp->lock);
2594                 addrconf_dad_start(ifp);
2595         }
2596
2597         if (ifp) {
2598                 u32 flags;
2599                 unsigned long now;
2600                 u32 stored_lft;
2601
2602                 /* update lifetime (RFC2462 5.5.3 e) */
2603                 spin_lock_bh(&ifp->lock);
2604                 now = jiffies;
2605                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2606                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2607                 else
2608                         stored_lft = 0;
2609                 if (!create && stored_lft) {
2610                         const u32 minimum_lft = min_t(u32,
2611                                 stored_lft, MIN_VALID_LIFETIME);
2612                         valid_lft = max(valid_lft, minimum_lft);
2613
2614                         /* RFC4862 Section 5.5.3e:
2615                          * "Note that the preferred lifetime of the
2616                          *  corresponding address is always reset to
2617                          *  the Preferred Lifetime in the received
2618                          *  Prefix Information option, regardless of
2619                          *  whether the valid lifetime is also reset or
2620                          *  ignored."
2621                          *
2622                          * So we should always update prefered_lft here.
2623                          */
2624                         update_lft = 1;
2625                 }
2626
2627                 if (update_lft) {
2628                         ifp->valid_lft = valid_lft;
2629                         ifp->prefered_lft = prefered_lft;
2630                         ifp->tstamp = now;
2631                         flags = ifp->flags;
2632                         ifp->flags &= ~IFA_F_DEPRECATED;
2633                         spin_unlock_bh(&ifp->lock);
2634
2635                         if (!(flags&IFA_F_TENTATIVE))
2636                                 ipv6_ifa_notify(0, ifp);
2637                 } else
2638                         spin_unlock_bh(&ifp->lock);
2639
2640                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2641                                  create, now);
2642
2643                 in6_ifa_put(ifp);
2644                 addrconf_verify();
2645         }
2646
2647         return 0;
2648 }
2649 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2650
2651 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2652 {
2653         struct prefix_info *pinfo;
2654         __u32 valid_lft;
2655         __u32 prefered_lft;
2656         int addr_type, err;
2657         u32 addr_flags = 0;
2658         struct inet6_dev *in6_dev;
2659         struct net *net = dev_net(dev);
2660
2661         pinfo = (struct prefix_info *) opt;
2662
2663         if (len < sizeof(struct prefix_info)) {
2664                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2665                 return;
2666         }
2667
2668         /*
2669          *      Validation checks ([ADDRCONF], page 19)
2670          */
2671
2672         addr_type = ipv6_addr_type(&pinfo->prefix);
2673
2674         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2675                 return;
2676
2677         valid_lft = ntohl(pinfo->valid);
2678         prefered_lft = ntohl(pinfo->prefered);
2679
2680         if (prefered_lft > valid_lft) {
2681                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2682                 return;
2683         }
2684
2685         in6_dev = in6_dev_get(dev);
2686
2687         if (!in6_dev) {
2688                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2689                                     dev->name);
2690                 return;
2691         }
2692
2693         /*
2694          *      Two things going on here:
2695          *      1) Add routes for on-link prefixes
2696          *      2) Configure prefixes with the auto flag set
2697          */
2698
2699         if (pinfo->onlink) {
2700                 struct fib6_info *rt;
2701                 unsigned long rt_expires;
2702
2703                 /* Avoid arithmetic overflow. Really, we could
2704                  * save rt_expires in seconds, likely valid_lft,
2705                  * but it would require division in fib gc, that it
2706                  * not good.
2707                  */
2708                 if (HZ > USER_HZ)
2709                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2710                 else
2711                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2712
2713                 if (addrconf_finite_timeout(rt_expires))
2714                         rt_expires *= HZ;
2715
2716                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2717                                                pinfo->prefix_len,
2718                                                dev,
2719                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2720                                                RTF_DEFAULT, true);
2721
2722                 if (rt) {
2723                         /* Autoconf prefix route */
2724                         if (valid_lft == 0) {
2725                                 ip6_del_rt(net, rt);
2726                                 rt = NULL;
2727                         } else if (addrconf_finite_timeout(rt_expires)) {
2728                                 /* not infinity */
2729                                 fib6_set_expires(rt, jiffies + rt_expires);
2730                         } else {
2731                                 fib6_clean_expires(rt);
2732                         }
2733                 } else if (valid_lft) {
2734                         clock_t expires = 0;
2735                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2736                         if (addrconf_finite_timeout(rt_expires)) {
2737                                 /* not infinity */
2738                                 flags |= RTF_EXPIRES;
2739                                 expires = jiffies_to_clock_t(rt_expires);
2740                         }
2741                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2742                                               0, dev, expires, flags,
2743                                               GFP_ATOMIC);
2744                 }
2745                 fib6_info_release(rt);
2746         }
2747
2748         /* Try to figure out our local address for this prefix */
2749
2750         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2751                 struct in6_addr addr;
2752                 bool tokenized = false, dev_addr_generated = false;
2753
2754                 if (pinfo->prefix_len == 64) {
2755                         memcpy(&addr, &pinfo->prefix, 8);
2756
2757                         if (!ipv6_addr_any(&in6_dev->token)) {
2758                                 read_lock_bh(&in6_dev->lock);
2759                                 memcpy(addr.s6_addr + 8,
2760                                        in6_dev->token.s6_addr + 8, 8);
2761                                 read_unlock_bh(&in6_dev->lock);
2762                                 tokenized = true;
2763                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2764                                    !ipv6_generate_stable_address(&addr, 0,
2765                                                                  in6_dev)) {
2766                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2767                                 goto ok;
2768                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2769                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2770                                 goto put;
2771                         } else {
2772                                 dev_addr_generated = true;
2773                         }
2774                         goto ok;
2775                 }
2776                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2777                                     pinfo->prefix_len);
2778                 goto put;
2779
2780 ok:
2781                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2782                                                    &addr, addr_type,
2783                                                    addr_flags, sllao,
2784                                                    tokenized, valid_lft,
2785                                                    prefered_lft);
2786                 if (err)
2787                         goto put;
2788
2789                 /* Ignore error case here because previous prefix add addr was
2790                  * successful which will be notified.
2791                  */
2792                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2793                                               addr_type, addr_flags, sllao,
2794                                               tokenized, valid_lft,
2795                                               prefered_lft,
2796                                               dev_addr_generated);
2797         }
2798         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2799 put:
2800         in6_dev_put(in6_dev);
2801 }
2802
2803 /*
2804  *      Set destination address.
2805  *      Special case for SIT interfaces where we create a new "virtual"
2806  *      device.
2807  */
2808 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2809 {
2810         struct in6_ifreq ireq;
2811         struct net_device *dev;
2812         int err = -EINVAL;
2813
2814         rtnl_lock();
2815
2816         err = -EFAULT;
2817         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2818                 goto err_exit;
2819
2820         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2821
2822         err = -ENODEV;
2823         if (!dev)
2824                 goto err_exit;
2825
2826 #if IS_ENABLED(CONFIG_IPV6_SIT)
2827         if (dev->type == ARPHRD_SIT) {
2828                 const struct net_device_ops *ops = dev->netdev_ops;
2829                 struct ifreq ifr;
2830                 struct ip_tunnel_parm p;
2831
2832                 err = -EADDRNOTAVAIL;
2833                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2834                         goto err_exit;
2835
2836                 memset(&p, 0, sizeof(p));
2837                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2838                 p.iph.saddr = 0;
2839                 p.iph.version = 4;
2840                 p.iph.ihl = 5;
2841                 p.iph.protocol = IPPROTO_IPV6;
2842                 p.iph.ttl = 64;
2843                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2844
2845                 if (ops->ndo_do_ioctl) {
2846                         mm_segment_t oldfs = get_fs();
2847
2848                         set_fs(KERNEL_DS);
2849                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2850                         set_fs(oldfs);
2851                 } else
2852                         err = -EOPNOTSUPP;
2853
2854                 if (err == 0) {
2855                         err = -ENOBUFS;
2856                         dev = __dev_get_by_name(net, p.name);
2857                         if (!dev)
2858                                 goto err_exit;
2859                         err = dev_open(dev, NULL);
2860                 }
2861         }
2862 #endif
2863
2864 err_exit:
2865         rtnl_unlock();
2866         return err;
2867 }
2868
2869 static int ipv6_mc_config(struct sock *sk, bool join,
2870                           const struct in6_addr *addr, int ifindex)
2871 {
2872         int ret;
2873
2874         ASSERT_RTNL();
2875
2876         lock_sock(sk);
2877         if (join)
2878                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2879         else
2880                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2881         release_sock(sk);
2882
2883         return ret;
2884 }
2885
2886 /*
2887  *      Manual configuration of address on an interface
2888  */
2889 static int inet6_addr_add(struct net *net, int ifindex,
2890                           struct ifa6_config *cfg,
2891                           struct netlink_ext_ack *extack)
2892 {
2893         struct inet6_ifaddr *ifp;
2894         struct inet6_dev *idev;
2895         struct net_device *dev;
2896         unsigned long timeout;
2897         clock_t expires;
2898         u32 flags;
2899
2900         ASSERT_RTNL();
2901
2902         if (cfg->plen > 128)
2903                 return -EINVAL;
2904
2905         /* check the lifetime */
2906         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2907                 return -EINVAL;
2908
2909         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2910                 return -EINVAL;
2911
2912         dev = __dev_get_by_index(net, ifindex);
2913         if (!dev)
2914                 return -ENODEV;
2915
2916         idev = addrconf_add_dev(dev);
2917         if (IS_ERR(idev))
2918                 return PTR_ERR(idev);
2919
2920         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2921                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2922                                          true, cfg->pfx, ifindex);
2923
2924                 if (ret < 0)
2925                         return ret;
2926         }
2927
2928         cfg->scope = ipv6_addr_scope(cfg->pfx);
2929
2930         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2931         if (addrconf_finite_timeout(timeout)) {
2932                 expires = jiffies_to_clock_t(timeout * HZ);
2933                 cfg->valid_lft = timeout;
2934                 flags = RTF_EXPIRES;
2935         } else {
2936                 expires = 0;
2937                 flags = 0;
2938                 cfg->ifa_flags |= IFA_F_PERMANENT;
2939         }
2940
2941         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2942         if (addrconf_finite_timeout(timeout)) {
2943                 if (timeout == 0)
2944                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2945                 cfg->preferred_lft = timeout;
2946         }
2947
2948         ifp = ipv6_add_addr(idev, cfg, true, extack);
2949         if (!IS_ERR(ifp)) {
2950                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2951                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2952                                               ifp->rt_priority, dev, expires,
2953                                               flags, GFP_KERNEL);
2954                 }
2955
2956                 /* Send a netlink notification if DAD is enabled and
2957                  * optimistic flag is not set
2958                  */
2959                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2960                         ipv6_ifa_notify(0, ifp);
2961                 /*
2962                  * Note that section 3.1 of RFC 4429 indicates
2963                  * that the Optimistic flag should not be set for
2964                  * manually configured addresses
2965                  */
2966                 addrconf_dad_start(ifp);
2967                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2968                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2969                                          cfg->preferred_lft, true, jiffies);
2970                 in6_ifa_put(ifp);
2971                 addrconf_verify_rtnl();
2972                 return 0;
2973         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2974                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2975                                cfg->pfx, ifindex);
2976         }
2977
2978         return PTR_ERR(ifp);
2979 }
2980
2981 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2982                           const struct in6_addr *pfx, unsigned int plen)
2983 {
2984         struct inet6_ifaddr *ifp;
2985         struct inet6_dev *idev;
2986         struct net_device *dev;
2987
2988         if (plen > 128)
2989                 return -EINVAL;
2990
2991         dev = __dev_get_by_index(net, ifindex);
2992         if (!dev)
2993                 return -ENODEV;
2994
2995         idev = __in6_dev_get(dev);
2996         if (!idev)
2997                 return -ENXIO;
2998
2999         read_lock_bh(&idev->lock);
3000         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3001                 if (ifp->prefix_len == plen &&
3002                     ipv6_addr_equal(pfx, &ifp->addr)) {
3003                         in6_ifa_hold(ifp);
3004                         read_unlock_bh(&idev->lock);
3005
3006                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3007                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3008                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3009                                                  jiffies);
3010                         ipv6_del_addr(ifp);
3011                         addrconf_verify_rtnl();
3012                         if (ipv6_addr_is_multicast(pfx)) {
3013                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3014                                                false, pfx, dev->ifindex);
3015                         }
3016                         return 0;
3017                 }
3018         }
3019         read_unlock_bh(&idev->lock);
3020         return -EADDRNOTAVAIL;
3021 }
3022
3023
3024 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3025 {
3026         struct ifa6_config cfg = {
3027                 .ifa_flags = IFA_F_PERMANENT,
3028                 .preferred_lft = INFINITY_LIFE_TIME,
3029                 .valid_lft = INFINITY_LIFE_TIME,
3030         };
3031         struct in6_ifreq ireq;
3032         int err;
3033
3034         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3035                 return -EPERM;
3036
3037         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3038                 return -EFAULT;
3039
3040         cfg.pfx = &ireq.ifr6_addr;
3041         cfg.plen = ireq.ifr6_prefixlen;
3042
3043         rtnl_lock();
3044         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3045         rtnl_unlock();
3046         return err;
3047 }
3048
3049 int addrconf_del_ifaddr(struct net *net, void __user *arg)
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         rtnl_lock();
3061         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3062                              ireq.ifr6_prefixlen);
3063         rtnl_unlock();
3064         return err;
3065 }
3066
3067 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3068                      int plen, int scope)
3069 {
3070         struct inet6_ifaddr *ifp;
3071         struct ifa6_config cfg = {
3072                 .pfx = addr,
3073                 .plen = plen,
3074                 .ifa_flags = IFA_F_PERMANENT,
3075                 .valid_lft = INFINITY_LIFE_TIME,
3076                 .preferred_lft = INFINITY_LIFE_TIME,
3077                 .scope = scope
3078         };
3079
3080         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3081         if (!IS_ERR(ifp)) {
3082                 spin_lock_bh(&ifp->lock);
3083                 ifp->flags &= ~IFA_F_TENTATIVE;
3084                 spin_unlock_bh(&ifp->lock);
3085                 rt_genid_bump_ipv6(dev_net(idev->dev));
3086                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3087                 in6_ifa_put(ifp);
3088         }
3089 }
3090
3091 #if IS_ENABLED(CONFIG_IPV6_SIT)
3092 static void sit_add_v4_addrs(struct inet6_dev *idev)
3093 {
3094         struct in6_addr addr;
3095         struct net_device *dev;
3096         struct net *net = dev_net(idev->dev);
3097         int scope, plen;
3098         u32 pflags = 0;
3099
3100         ASSERT_RTNL();
3101
3102         memset(&addr, 0, sizeof(struct in6_addr));
3103         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3104
3105         if (idev->dev->flags&IFF_POINTOPOINT) {
3106                 addr.s6_addr32[0] = htonl(0xfe800000);
3107                 scope = IFA_LINK;
3108                 plen = 64;
3109         } else {
3110                 scope = IPV6_ADDR_COMPATv4;
3111                 plen = 96;
3112                 pflags |= RTF_NONEXTHOP;
3113         }
3114
3115         if (addr.s6_addr32[3]) {
3116                 add_addr(idev, &addr, plen, scope);
3117                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3118                                       GFP_KERNEL);
3119                 return;
3120         }
3121
3122         for_each_netdev(net, dev) {
3123                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3124                 if (in_dev && (dev->flags & IFF_UP)) {
3125                         struct in_ifaddr *ifa;
3126
3127                         int flag = scope;
3128
3129                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3130
3131                                 addr.s6_addr32[3] = ifa->ifa_local;
3132
3133                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3134                                         continue;
3135                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3136                                         if (idev->dev->flags&IFF_POINTOPOINT)
3137                                                 continue;
3138                                         flag |= IFA_HOST;
3139                                 }
3140
3141                                 add_addr(idev, &addr, plen, flag);
3142                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3143                                                       0, pflags, GFP_KERNEL);
3144                         }
3145                 }
3146         }
3147 }
3148 #endif
3149
3150 static void init_loopback(struct net_device *dev)
3151 {
3152         struct inet6_dev  *idev;
3153
3154         /* ::1 */
3155
3156         ASSERT_RTNL();
3157
3158         idev = ipv6_find_idev(dev);
3159         if (!idev) {
3160                 pr_debug("%s: add_dev failed\n", __func__);
3161                 return;
3162         }
3163
3164         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3165 }
3166
3167 void addrconf_add_linklocal(struct inet6_dev *idev,
3168                             const struct in6_addr *addr, u32 flags)
3169 {
3170         struct ifa6_config cfg = {
3171                 .pfx = addr,
3172                 .plen = 64,
3173                 .ifa_flags = flags | IFA_F_PERMANENT,
3174                 .valid_lft = INFINITY_LIFE_TIME,
3175                 .preferred_lft = INFINITY_LIFE_TIME,
3176                 .scope = IFA_LINK
3177         };
3178         struct inet6_ifaddr *ifp;
3179
3180 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3181         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3182              idev->cnf.optimistic_dad) &&
3183             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3184                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3185 #endif
3186
3187         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3188         if (!IS_ERR(ifp)) {
3189                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3190                                       0, 0, GFP_ATOMIC);
3191                 addrconf_dad_start(ifp);
3192                 in6_ifa_put(ifp);
3193         }
3194 }
3195 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3196
3197 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3198 {
3199         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3200                 return true;
3201
3202         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3203             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3204                 return true;
3205
3206         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3207             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3208                 return true;
3209
3210         return false;
3211 }
3212
3213 static int ipv6_generate_stable_address(struct in6_addr *address,
3214                                         u8 dad_count,
3215                                         const struct inet6_dev *idev)
3216 {
3217         static DEFINE_SPINLOCK(lock);
3218         static __u32 digest[SHA_DIGEST_WORDS];
3219         static __u32 workspace[SHA_WORKSPACE_WORDS];
3220
3221         static union {
3222                 char __data[SHA_MESSAGE_BYTES];
3223                 struct {
3224                         struct in6_addr secret;
3225                         __be32 prefix[2];
3226                         unsigned char hwaddr[MAX_ADDR_LEN];
3227                         u8 dad_count;
3228                 } __packed;
3229         } data;
3230
3231         struct in6_addr secret;
3232         struct in6_addr temp;
3233         struct net *net = dev_net(idev->dev);
3234
3235         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3236
3237         if (idev->cnf.stable_secret.initialized)
3238                 secret = idev->cnf.stable_secret.secret;
3239         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3240                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3241         else
3242                 return -1;
3243
3244 retry:
3245         spin_lock_bh(&lock);
3246
3247         sha_init(digest);
3248         memset(&data, 0, sizeof(data));
3249         memset(workspace, 0, sizeof(workspace));
3250         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3251         data.prefix[0] = address->s6_addr32[0];
3252         data.prefix[1] = address->s6_addr32[1];
3253         data.secret = secret;
3254         data.dad_count = dad_count;
3255
3256         sha_transform(digest, data.__data, workspace);
3257
3258         temp = *address;
3259         temp.s6_addr32[2] = (__force __be32)digest[0];
3260         temp.s6_addr32[3] = (__force __be32)digest[1];
3261
3262         spin_unlock_bh(&lock);
3263
3264         if (ipv6_reserved_interfaceid(temp)) {
3265                 dad_count++;
3266                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3267                         return -1;
3268                 goto retry;
3269         }
3270
3271         *address = temp;
3272         return 0;
3273 }
3274
3275 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3276 {
3277         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3278
3279         if (s->initialized)
3280                 return;
3281         s = &idev->cnf.stable_secret;
3282         get_random_bytes(&s->secret, sizeof(s->secret));
3283         s->initialized = true;
3284 }
3285
3286 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3287 {
3288         struct in6_addr addr;
3289
3290         /* no link local addresses on L3 master devices */
3291         if (netif_is_l3_master(idev->dev))
3292                 return;
3293
3294         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3295
3296         switch (idev->cnf.addr_gen_mode) {
3297         case IN6_ADDR_GEN_MODE_RANDOM:
3298                 ipv6_gen_mode_random_init(idev);
3299                 /* fallthrough */
3300         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3301                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3302                         addrconf_add_linklocal(idev, &addr,
3303                                                IFA_F_STABLE_PRIVACY);
3304                 else if (prefix_route)
3305                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3306                                               0, 0, GFP_KERNEL);
3307                 break;
3308         case IN6_ADDR_GEN_MODE_EUI64:
3309                 /* addrconf_add_linklocal also adds a prefix_route and we
3310                  * only need to care about prefix routes if ipv6_generate_eui64
3311                  * couldn't generate one.
3312                  */
3313                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3314                         addrconf_add_linklocal(idev, &addr, 0);
3315                 else if (prefix_route)
3316                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3317                                               0, 0, GFP_KERNEL);
3318                 break;
3319         case IN6_ADDR_GEN_MODE_NONE:
3320         default:
3321                 /* will not add any link local address */
3322                 break;
3323         }
3324 }
3325
3326 static void addrconf_dev_config(struct net_device *dev)
3327 {
3328         struct inet6_dev *idev;
3329
3330         ASSERT_RTNL();
3331
3332         if ((dev->type != ARPHRD_ETHER) &&
3333             (dev->type != ARPHRD_FDDI) &&
3334             (dev->type != ARPHRD_ARCNET) &&
3335             (dev->type != ARPHRD_INFINIBAND) &&
3336             (dev->type != ARPHRD_IEEE1394) &&
3337             (dev->type != ARPHRD_TUNNEL6) &&
3338             (dev->type != ARPHRD_6LOWPAN) &&
3339             (dev->type != ARPHRD_IP6GRE) &&
3340             (dev->type != ARPHRD_IPGRE) &&
3341             (dev->type != ARPHRD_TUNNEL) &&
3342             (dev->type != ARPHRD_NONE) &&
3343             (dev->type != ARPHRD_RAWIP)) {
3344                 /* Alas, we support only Ethernet autoconfiguration. */
3345                 return;
3346         }
3347
3348         idev = addrconf_add_dev(dev);
3349         if (IS_ERR(idev))
3350                 return;
3351
3352         /* this device type has no EUI support */
3353         if (dev->type == ARPHRD_NONE &&
3354             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3355                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3356
3357         addrconf_addr_gen(idev, false);
3358 }
3359
3360 #if IS_ENABLED(CONFIG_IPV6_SIT)
3361 static void addrconf_sit_config(struct net_device *dev)
3362 {
3363         struct inet6_dev *idev;
3364
3365         ASSERT_RTNL();
3366
3367         /*
3368          * Configure the tunnel with one of our IPv4
3369          * addresses... we should configure all of
3370          * our v4 addrs in the tunnel
3371          */
3372
3373         idev = ipv6_find_idev(dev);
3374         if (!idev) {
3375                 pr_debug("%s: add_dev failed\n", __func__);
3376                 return;
3377         }
3378
3379         if (dev->priv_flags & IFF_ISATAP) {
3380                 addrconf_addr_gen(idev, false);
3381                 return;
3382         }
3383
3384         sit_add_v4_addrs(idev);
3385
3386         if (dev->flags&IFF_POINTOPOINT)
3387                 addrconf_add_mroute(dev);
3388 }
3389 #endif
3390
3391 #if IS_ENABLED(CONFIG_NET_IPGRE)
3392 static void addrconf_gre_config(struct net_device *dev)
3393 {
3394         struct inet6_dev *idev;
3395
3396         ASSERT_RTNL();
3397
3398         idev = ipv6_find_idev(dev);
3399         if (!idev) {
3400                 pr_debug("%s: add_dev failed\n", __func__);
3401                 return;
3402         }
3403
3404         addrconf_addr_gen(idev, true);
3405         if (dev->flags & IFF_POINTOPOINT)
3406                 addrconf_add_mroute(dev);
3407 }
3408 #endif
3409
3410 static int fixup_permanent_addr(struct net *net,
3411                                 struct inet6_dev *idev,
3412                                 struct inet6_ifaddr *ifp)
3413 {
3414         /* !fib6_node means the host route was removed from the
3415          * FIB, for example, if 'lo' device is taken down. In that
3416          * case regenerate the host route.
3417          */
3418         if (!ifp->rt || !ifp->rt->fib6_node) {
3419                 struct fib6_info *f6i, *prev;
3420
3421                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3422                                          GFP_ATOMIC);
3423                 if (IS_ERR(f6i))
3424                         return PTR_ERR(f6i);
3425
3426                 /* ifp->rt can be accessed outside of rtnl */
3427                 spin_lock(&ifp->lock);
3428                 prev = ifp->rt;
3429                 ifp->rt = f6i;
3430                 spin_unlock(&ifp->lock);
3431
3432                 fib6_info_release(prev);
3433         }
3434
3435         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3436                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3437                                       ifp->rt_priority, idev->dev, 0, 0,
3438                                       GFP_ATOMIC);
3439         }
3440
3441         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3442                 addrconf_dad_start(ifp);
3443
3444         return 0;
3445 }
3446
3447 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3448 {
3449         struct inet6_ifaddr *ifp, *tmp;
3450         struct inet6_dev *idev;
3451
3452         idev = __in6_dev_get(dev);
3453         if (!idev)
3454                 return;
3455
3456         write_lock_bh(&idev->lock);
3457
3458         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3459                 if ((ifp->flags & IFA_F_PERMANENT) &&
3460                     fixup_permanent_addr(net, idev, ifp) < 0) {
3461                         write_unlock_bh(&idev->lock);
3462                         in6_ifa_hold(ifp);
3463                         ipv6_del_addr(ifp);
3464                         write_lock_bh(&idev->lock);
3465
3466                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3467                                              idev->dev->name, &ifp->addr);
3468                 }
3469         }
3470
3471         write_unlock_bh(&idev->lock);
3472 }
3473
3474 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3475                            void *ptr)
3476 {
3477         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3478         struct netdev_notifier_change_info *change_info;
3479         struct netdev_notifier_changeupper_info *info;
3480         struct inet6_dev *idev = __in6_dev_get(dev);
3481         struct net *net = dev_net(dev);
3482         int run_pending = 0;
3483         int err;
3484
3485         switch (event) {
3486         case NETDEV_REGISTER:
3487                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3488                         idev = ipv6_add_dev(dev);
3489                         if (IS_ERR(idev))
3490                                 return notifier_from_errno(PTR_ERR(idev));
3491                 }
3492                 break;
3493
3494         case NETDEV_CHANGEMTU:
3495                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3496                 if (dev->mtu < IPV6_MIN_MTU) {
3497                         addrconf_ifdown(dev, dev != net->loopback_dev);
3498                         break;
3499                 }
3500
3501                 if (idev) {
3502                         rt6_mtu_change(dev, dev->mtu);
3503                         idev->cnf.mtu6 = dev->mtu;
3504                         break;
3505                 }
3506
3507                 /* allocate new idev */
3508                 idev = ipv6_add_dev(dev);
3509                 if (IS_ERR(idev))
3510                         break;
3511
3512                 /* device is still not ready */
3513                 if (!(idev->if_flags & IF_READY))
3514                         break;
3515
3516                 run_pending = 1;
3517
3518                 /* fall through */
3519
3520         case NETDEV_UP:
3521         case NETDEV_CHANGE:
3522                 if (dev->flags & IFF_SLAVE)
3523                         break;
3524
3525                 if (idev && idev->cnf.disable_ipv6)
3526                         break;
3527
3528                 if (event == NETDEV_UP) {
3529                         /* restore routes for permanent addresses */
3530                         addrconf_permanent_addr(net, dev);
3531
3532                         if (!addrconf_link_ready(dev)) {
3533                                 /* device is not ready yet. */
3534                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3535                                          dev->name);
3536                                 break;
3537                         }
3538
3539                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3540                                 idev = ipv6_add_dev(dev);
3541
3542                         if (!IS_ERR_OR_NULL(idev)) {
3543                                 idev->if_flags |= IF_READY;
3544                                 run_pending = 1;
3545                         }
3546                 } else if (event == NETDEV_CHANGE) {
3547                         if (!addrconf_link_ready(dev)) {
3548                                 /* device is still not ready. */
3549                                 rt6_sync_down_dev(dev, event);
3550                                 break;
3551                         }
3552
3553                         if (!IS_ERR_OR_NULL(idev)) {
3554                                 if (idev->if_flags & IF_READY) {
3555                                         /* device is already configured -
3556                                          * but resend MLD reports, we might
3557                                          * have roamed and need to update
3558                                          * multicast snooping switches
3559                                          */
3560                                         ipv6_mc_up(idev);
3561                                         change_info = ptr;
3562                                         if (change_info->flags_changed & IFF_NOARP)
3563                                                 addrconf_dad_run(idev, true);
3564                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3565                                         break;
3566                                 }
3567                                 idev->if_flags |= IF_READY;
3568                         }
3569
3570                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3571                                 dev->name);
3572
3573                         run_pending = 1;
3574                 }
3575
3576                 switch (dev->type) {
3577 #if IS_ENABLED(CONFIG_IPV6_SIT)
3578                 case ARPHRD_SIT:
3579                         addrconf_sit_config(dev);
3580                         break;
3581 #endif
3582 #if IS_ENABLED(CONFIG_NET_IPGRE)
3583                 case ARPHRD_IPGRE:
3584                         addrconf_gre_config(dev);
3585                         break;
3586 #endif
3587                 case ARPHRD_LOOPBACK:
3588                         init_loopback(dev);
3589                         break;
3590
3591                 default:
3592                         addrconf_dev_config(dev);
3593                         break;
3594                 }
3595
3596                 if (!IS_ERR_OR_NULL(idev)) {
3597                         if (run_pending)
3598                                 addrconf_dad_run(idev, false);
3599
3600                         /* Device has an address by now */
3601                         rt6_sync_up(dev, RTNH_F_DEAD);
3602
3603                         /*
3604                          * If the MTU changed during the interface down,
3605                          * when the interface up, the changed MTU must be
3606                          * reflected in the idev as well as routers.
3607                          */
3608                         if (idev->cnf.mtu6 != dev->mtu &&
3609                             dev->mtu >= IPV6_MIN_MTU) {
3610                                 rt6_mtu_change(dev, dev->mtu);
3611                                 idev->cnf.mtu6 = dev->mtu;
3612                         }
3613                         idev->tstamp = jiffies;
3614                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3615
3616                         /*
3617                          * If the changed mtu during down is lower than
3618                          * IPV6_MIN_MTU stop IPv6 on this interface.
3619                          */
3620                         if (dev->mtu < IPV6_MIN_MTU)
3621                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3622                 }
3623                 break;
3624
3625         case NETDEV_DOWN:
3626         case NETDEV_UNREGISTER:
3627                 /*
3628                  *      Remove all addresses from this interface.
3629                  */
3630                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3631                 break;
3632
3633         case NETDEV_CHANGENAME:
3634                 if (idev) {
3635                         snmp6_unregister_dev(idev);
3636                         addrconf_sysctl_unregister(idev);
3637                         err = addrconf_sysctl_register(idev);
3638                         if (err)
3639                                 return notifier_from_errno(err);
3640                         err = snmp6_register_dev(idev);
3641                         if (err) {
3642                                 addrconf_sysctl_unregister(idev);
3643                                 return notifier_from_errno(err);
3644                         }
3645                 }
3646                 break;
3647
3648         case NETDEV_PRE_TYPE_CHANGE:
3649         case NETDEV_POST_TYPE_CHANGE:
3650                 if (idev)
3651                         addrconf_type_change(dev, event);
3652                 break;
3653
3654         case NETDEV_CHANGEUPPER:
3655                 info = ptr;
3656
3657                 /* flush all routes if dev is linked to or unlinked from
3658                  * an L3 master device (e.g., VRF)
3659                  */
3660                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3661                         addrconf_ifdown(dev, 0);
3662         }
3663
3664         return NOTIFY_OK;
3665 }
3666
3667 /*
3668  *      addrconf module should be notified of a device going up
3669  */
3670 static struct notifier_block ipv6_dev_notf = {
3671         .notifier_call = addrconf_notify,
3672         .priority = ADDRCONF_NOTIFY_PRIORITY,
3673 };
3674
3675 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3676 {
3677         struct inet6_dev *idev;
3678         ASSERT_RTNL();
3679
3680         idev = __in6_dev_get(dev);
3681
3682         if (event == NETDEV_POST_TYPE_CHANGE)
3683                 ipv6_mc_remap(idev);
3684         else if (event == NETDEV_PRE_TYPE_CHANGE)
3685                 ipv6_mc_unmap(idev);
3686 }
3687
3688 static bool addr_is_local(const struct in6_addr *addr)
3689 {
3690         return ipv6_addr_type(addr) &
3691                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3692 }
3693
3694 static int addrconf_ifdown(struct net_device *dev, int how)
3695 {
3696         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3697         struct net *net = dev_net(dev);
3698         struct inet6_dev *idev;
3699         struct inet6_ifaddr *ifa, *tmp;
3700         bool keep_addr = false;
3701         int state, i;
3702
3703         ASSERT_RTNL();
3704
3705         rt6_disable_ip(dev, event);
3706
3707         idev = __in6_dev_get(dev);
3708         if (!idev)
3709                 return -ENODEV;
3710
3711         /*
3712          * Step 1: remove reference to ipv6 device from parent device.
3713          *         Do not dev_put!
3714          */
3715         if (how) {
3716                 idev->dead = 1;
3717
3718                 /* protected by rtnl_lock */
3719                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3720
3721                 /* Step 1.5: remove snmp6 entry */
3722                 snmp6_unregister_dev(idev);
3723
3724         }
3725
3726         /* combine the user config with event to determine if permanent
3727          * addresses are to be removed from address hash table
3728          */
3729         if (!how && !idev->cnf.disable_ipv6) {
3730                 /* aggregate the system setting and interface setting */
3731                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3732
3733                 if (!_keep_addr)
3734                         _keep_addr = idev->cnf.keep_addr_on_down;
3735
3736                 keep_addr = (_keep_addr > 0);
3737         }
3738
3739         /* Step 2: clear hash table */
3740         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3741                 struct hlist_head *h = &inet6_addr_lst[i];
3742
3743                 spin_lock_bh(&addrconf_hash_lock);
3744 restart:
3745                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3746                         if (ifa->idev == idev) {
3747                                 addrconf_del_dad_work(ifa);
3748                                 /* combined flag + permanent flag decide if
3749                                  * address is retained on a down event
3750                                  */
3751                                 if (!keep_addr ||
3752                                     !(ifa->flags & IFA_F_PERMANENT) ||
3753                                     addr_is_local(&ifa->addr)) {
3754                                         hlist_del_init_rcu(&ifa->addr_lst);
3755                                         goto restart;
3756                                 }
3757                         }
3758                 }
3759                 spin_unlock_bh(&addrconf_hash_lock);
3760         }
3761
3762         write_lock_bh(&idev->lock);
3763
3764         addrconf_del_rs_timer(idev);
3765
3766         /* Step 2: clear flags for stateless addrconf */
3767         if (!how)
3768                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3769
3770         /* Step 3: clear tempaddr list */
3771         while (!list_empty(&idev->tempaddr_list)) {
3772                 ifa = list_first_entry(&idev->tempaddr_list,
3773                                        struct inet6_ifaddr, tmp_list);
3774                 list_del(&ifa->tmp_list);
3775                 write_unlock_bh(&idev->lock);
3776                 spin_lock_bh(&ifa->lock);
3777
3778                 if (ifa->ifpub) {
3779                         in6_ifa_put(ifa->ifpub);
3780                         ifa->ifpub = NULL;
3781                 }
3782                 spin_unlock_bh(&ifa->lock);
3783                 in6_ifa_put(ifa);
3784                 write_lock_bh(&idev->lock);
3785         }
3786
3787         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3788                 struct fib6_info *rt = NULL;
3789                 bool keep;
3790
3791                 addrconf_del_dad_work(ifa);
3792
3793                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3794                         !addr_is_local(&ifa->addr);
3795
3796                 write_unlock_bh(&idev->lock);
3797                 spin_lock_bh(&ifa->lock);
3798
3799                 if (keep) {
3800                         /* set state to skip the notifier below */
3801                         state = INET6_IFADDR_STATE_DEAD;
3802                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3803                         if (!(ifa->flags & IFA_F_NODAD))
3804                                 ifa->flags |= IFA_F_TENTATIVE;
3805
3806                         rt = ifa->rt;
3807                         ifa->rt = NULL;
3808                 } else {
3809                         state = ifa->state;
3810                         ifa->state = INET6_IFADDR_STATE_DEAD;
3811                 }
3812
3813                 spin_unlock_bh(&ifa->lock);
3814
3815                 if (rt)
3816                         ip6_del_rt(net, rt);
3817
3818                 if (state != INET6_IFADDR_STATE_DEAD) {
3819                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3820                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3821                 } else {
3822                         if (idev->cnf.forwarding)
3823                                 addrconf_leave_anycast(ifa);
3824                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3825                 }
3826
3827                 write_lock_bh(&idev->lock);
3828                 if (!keep) {
3829                         list_del_rcu(&ifa->if_list);
3830                         in6_ifa_put(ifa);
3831                 }
3832         }
3833
3834         write_unlock_bh(&idev->lock);
3835
3836         /* Step 5: Discard anycast and multicast list */
3837         if (how) {
3838                 ipv6_ac_destroy_dev(idev);
3839                 ipv6_mc_destroy_dev(idev);
3840         } else {
3841                 ipv6_mc_down(idev);
3842         }
3843
3844         idev->tstamp = jiffies;
3845
3846         /* Last: Shot the device (if unregistered) */
3847         if (how) {
3848                 addrconf_sysctl_unregister(idev);
3849                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3850                 neigh_ifdown(&nd_tbl, dev);
3851                 in6_dev_put(idev);
3852         }
3853         return 0;
3854 }
3855
3856 static void addrconf_rs_timer(struct timer_list *t)
3857 {
3858         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3859         struct net_device *dev = idev->dev;
3860         struct in6_addr lladdr;
3861
3862         write_lock(&idev->lock);
3863         if (idev->dead || !(idev->if_flags & IF_READY))
3864                 goto out;
3865
3866         if (!ipv6_accept_ra(idev))
3867                 goto out;
3868
3869         /* Announcement received after solicitation was sent */
3870         if (idev->if_flags & IF_RA_RCVD)
3871                 goto out;
3872
3873         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3874                 write_unlock(&idev->lock);
3875                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3876                         ndisc_send_rs(dev, &lladdr,
3877                                       &in6addr_linklocal_allrouters);
3878                 else
3879                         goto put;
3880
3881                 write_lock(&idev->lock);
3882                 idev->rs_interval = rfc3315_s14_backoff_update(
3883                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3884                 /* The wait after the last probe can be shorter */
3885                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3886                                              idev->cnf.rtr_solicits) ?
3887                                       idev->cnf.rtr_solicit_delay :
3888                                       idev->rs_interval);
3889         } else {
3890                 /*
3891                  * Note: we do not support deprecated "all on-link"
3892                  * assumption any longer.
3893                  */
3894                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3895         }
3896
3897 out:
3898         write_unlock(&idev->lock);
3899 put:
3900         in6_dev_put(idev);
3901 }
3902
3903 /*
3904  *      Duplicate Address Detection
3905  */
3906 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3907 {
3908         unsigned long rand_num;
3909         struct inet6_dev *idev = ifp->idev;
3910         u64 nonce;
3911
3912         if (ifp->flags & IFA_F_OPTIMISTIC)
3913                 rand_num = 0;
3914         else
3915                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3916
3917         nonce = 0;
3918         if (idev->cnf.enhanced_dad ||
3919             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3920                 do
3921                         get_random_bytes(&nonce, 6);
3922                 while (nonce == 0);
3923         }
3924         ifp->dad_nonce = nonce;
3925         ifp->dad_probes = idev->cnf.dad_transmits;
3926         addrconf_mod_dad_work(ifp, rand_num);
3927 }
3928
3929 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3930 {
3931         struct inet6_dev *idev = ifp->idev;
3932         struct net_device *dev = idev->dev;
3933         bool bump_id, notify = false;
3934         struct net *net;
3935
3936         addrconf_join_solict(dev, &ifp->addr);
3937
3938         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3939
3940         read_lock_bh(&idev->lock);
3941         spin_lock(&ifp->lock);
3942         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3943                 goto out;
3944
3945         net = dev_net(dev);
3946         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3947             (net->ipv6.devconf_all->accept_dad < 1 &&
3948              idev->cnf.accept_dad < 1) ||
3949             !(ifp->flags&IFA_F_TENTATIVE) ||
3950             ifp->flags & IFA_F_NODAD) {
3951                 bool send_na = false;
3952
3953                 if (ifp->flags & IFA_F_TENTATIVE &&
3954                     !(ifp->flags & IFA_F_OPTIMISTIC))
3955                         send_na = true;
3956                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3957                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3958                 spin_unlock(&ifp->lock);
3959                 read_unlock_bh(&idev->lock);
3960
3961                 addrconf_dad_completed(ifp, bump_id, send_na);
3962                 return;
3963         }
3964
3965         if (!(idev->if_flags & IF_READY)) {
3966                 spin_unlock(&ifp->lock);
3967                 read_unlock_bh(&idev->lock);
3968                 /*
3969                  * If the device is not ready:
3970                  * - keep it tentative if it is a permanent address.
3971                  * - otherwise, kill it.
3972                  */
3973                 in6_ifa_hold(ifp);
3974                 addrconf_dad_stop(ifp, 0);
3975                 return;
3976         }
3977
3978         /*
3979          * Optimistic nodes can start receiving
3980          * Frames right away
3981          */
3982         if (ifp->flags & IFA_F_OPTIMISTIC) {
3983                 ip6_ins_rt(net, ifp->rt);
3984                 if (ipv6_use_optimistic_addr(net, idev)) {
3985                         /* Because optimistic nodes can use this address,
3986                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3987                          */
3988                         notify = true;
3989                 }
3990         }
3991
3992         addrconf_dad_kick(ifp);
3993 out:
3994         spin_unlock(&ifp->lock);
3995         read_unlock_bh(&idev->lock);
3996         if (notify)
3997                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3998 }
3999
4000 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4001 {
4002         bool begin_dad = false;
4003
4004         spin_lock_bh(&ifp->lock);
4005         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4006                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4007                 begin_dad = true;
4008         }
4009         spin_unlock_bh(&ifp->lock);
4010
4011         if (begin_dad)
4012                 addrconf_mod_dad_work(ifp, 0);
4013 }
4014
4015 static void addrconf_dad_work(struct work_struct *w)
4016 {
4017         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4018                                                 struct inet6_ifaddr,
4019                                                 dad_work);
4020         struct inet6_dev *idev = ifp->idev;
4021         bool bump_id, disable_ipv6 = false;
4022         struct in6_addr mcaddr;
4023
4024         enum {
4025                 DAD_PROCESS,
4026                 DAD_BEGIN,
4027                 DAD_ABORT,
4028         } action = DAD_PROCESS;
4029
4030         rtnl_lock();
4031
4032         spin_lock_bh(&ifp->lock);
4033         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4034                 action = DAD_BEGIN;
4035                 ifp->state = INET6_IFADDR_STATE_DAD;
4036         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4037                 action = DAD_ABORT;
4038                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4039
4040                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4041                      idev->cnf.accept_dad > 1) &&
4042                     !idev->cnf.disable_ipv6 &&
4043                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4044                         struct in6_addr addr;
4045
4046                         addr.s6_addr32[0] = htonl(0xfe800000);
4047                         addr.s6_addr32[1] = 0;
4048
4049                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4050                             ipv6_addr_equal(&ifp->addr, &addr)) {
4051                                 /* DAD failed for link-local based on MAC */
4052                                 idev->cnf.disable_ipv6 = 1;
4053
4054                                 pr_info("%s: IPv6 being disabled!\n",
4055                                         ifp->idev->dev->name);
4056                                 disable_ipv6 = true;
4057                         }
4058                 }
4059         }
4060         spin_unlock_bh(&ifp->lock);
4061
4062         if (action == DAD_BEGIN) {
4063                 addrconf_dad_begin(ifp);
4064                 goto out;
4065         } else if (action == DAD_ABORT) {
4066                 in6_ifa_hold(ifp);
4067                 addrconf_dad_stop(ifp, 1);
4068                 if (disable_ipv6)
4069                         addrconf_ifdown(idev->dev, 0);
4070                 goto out;
4071         }
4072
4073         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4074                 goto out;
4075
4076         write_lock_bh(&idev->lock);
4077         if (idev->dead || !(idev->if_flags & IF_READY)) {
4078                 write_unlock_bh(&idev->lock);
4079                 goto out;
4080         }
4081
4082         spin_lock(&ifp->lock);
4083         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4084                 spin_unlock(&ifp->lock);
4085                 write_unlock_bh(&idev->lock);
4086                 goto out;
4087         }
4088
4089         if (ifp->dad_probes == 0) {
4090                 bool send_na = false;
4091
4092                 /*
4093                  * DAD was successful
4094                  */
4095
4096                 if (ifp->flags & IFA_F_TENTATIVE &&
4097                     !(ifp->flags & IFA_F_OPTIMISTIC))
4098                         send_na = true;
4099                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4100                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4101                 spin_unlock(&ifp->lock);
4102                 write_unlock_bh(&idev->lock);
4103
4104                 addrconf_dad_completed(ifp, bump_id, send_na);
4105
4106                 goto out;
4107         }
4108
4109         ifp->dad_probes--;
4110         addrconf_mod_dad_work(ifp,
4111                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4112         spin_unlock(&ifp->lock);
4113         write_unlock_bh(&idev->lock);
4114
4115         /* send a neighbour solicitation for our addr */
4116         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4117         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4118                       ifp->dad_nonce);
4119 out:
4120         in6_ifa_put(ifp);
4121         rtnl_unlock();
4122 }
4123
4124 /* ifp->idev must be at least read locked */
4125 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4126 {
4127         struct inet6_ifaddr *ifpiter;
4128         struct inet6_dev *idev = ifp->idev;
4129
4130         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4131                 if (ifpiter->scope > IFA_LINK)
4132                         break;
4133                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4134                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4135                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4136                     IFA_F_PERMANENT)
4137                         return false;
4138         }
4139         return true;
4140 }
4141
4142 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4143                                    bool send_na)
4144 {
4145         struct net_device *dev = ifp->idev->dev;
4146         struct in6_addr lladdr;
4147         bool send_rs, send_mld;
4148
4149         addrconf_del_dad_work(ifp);
4150
4151         /*
4152          *      Configure the address for reception. Now it is valid.
4153          */
4154
4155         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4156
4157         /* If added prefix is link local and we are prepared to process
4158            router advertisements, start sending router solicitations.
4159          */
4160
4161         read_lock_bh(&ifp->idev->lock);
4162         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4163         send_rs = send_mld &&
4164                   ipv6_accept_ra(ifp->idev) &&
4165                   ifp->idev->cnf.rtr_solicits != 0 &&
4166                   (dev->flags&IFF_LOOPBACK) == 0;
4167         read_unlock_bh(&ifp->idev->lock);
4168
4169         /* While dad is in progress mld report's source address is in6_addrany.
4170          * Resend with proper ll now.
4171          */
4172         if (send_mld)
4173                 ipv6_mc_dad_complete(ifp->idev);
4174
4175         /* send unsolicited NA if enabled */
4176         if (send_na &&
4177             (ifp->idev->cnf.ndisc_notify ||
4178              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4179                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4180                               /*router=*/ !!ifp->idev->cnf.forwarding,
4181                               /*solicited=*/ false, /*override=*/ true,
4182                               /*inc_opt=*/ true);
4183         }
4184
4185         if (send_rs) {
4186                 /*
4187                  *      If a host as already performed a random delay
4188                  *      [...] as part of DAD [...] there is no need
4189                  *      to delay again before sending the first RS
4190                  */
4191                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4192                         return;
4193                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4194
4195                 write_lock_bh(&ifp->idev->lock);
4196                 spin_lock(&ifp->lock);
4197                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4198                         ifp->idev->cnf.rtr_solicit_interval);
4199                 ifp->idev->rs_probes = 1;
4200                 ifp->idev->if_flags |= IF_RS_SENT;
4201                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4202                 spin_unlock(&ifp->lock);
4203                 write_unlock_bh(&ifp->idev->lock);
4204         }
4205
4206         if (bump_id)
4207                 rt_genid_bump_ipv6(dev_net(dev));
4208
4209         /* Make sure that a new temporary address will be created
4210          * before this temporary address becomes deprecated.
4211          */
4212         if (ifp->flags & IFA_F_TEMPORARY)
4213                 addrconf_verify_rtnl();
4214 }
4215
4216 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4217 {
4218         struct inet6_ifaddr *ifp;
4219
4220         read_lock_bh(&idev->lock);
4221         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4222                 spin_lock(&ifp->lock);
4223                 if ((ifp->flags & IFA_F_TENTATIVE &&
4224                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4225                         if (restart)
4226                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4227                         addrconf_dad_kick(ifp);
4228                 }
4229                 spin_unlock(&ifp->lock);
4230         }
4231         read_unlock_bh(&idev->lock);
4232 }
4233
4234 #ifdef CONFIG_PROC_FS
4235 struct if6_iter_state {
4236         struct seq_net_private p;
4237         int bucket;
4238         int offset;
4239 };
4240
4241 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4242 {
4243         struct if6_iter_state *state = seq->private;
4244         struct net *net = seq_file_net(seq);
4245         struct inet6_ifaddr *ifa = NULL;
4246         int p = 0;
4247
4248         /* initial bucket if pos is 0 */
4249         if (pos == 0) {
4250                 state->bucket = 0;
4251                 state->offset = 0;
4252         }
4253
4254         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4255                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4256                                          addr_lst) {
4257                         if (!net_eq(dev_net(ifa->idev->dev), net))
4258                                 continue;
4259                         /* sync with offset */
4260                         if (p < state->offset) {
4261                                 p++;
4262                                 continue;
4263                         }
4264                         return ifa;
4265                 }
4266
4267                 /* prepare for next bucket */
4268                 state->offset = 0;
4269                 p = 0;
4270         }
4271         return NULL;
4272 }
4273
4274 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4275                                          struct inet6_ifaddr *ifa)
4276 {
4277         struct if6_iter_state *state = seq->private;
4278         struct net *net = seq_file_net(seq);
4279
4280         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4281                 if (!net_eq(dev_net(ifa->idev->dev), net))
4282                         continue;
4283                 state->offset++;
4284                 return ifa;
4285         }
4286
4287         state->offset = 0;
4288         while (++state->bucket < IN6_ADDR_HSIZE) {
4289                 hlist_for_each_entry_rcu(ifa,
4290                                      &inet6_addr_lst[state->bucket], addr_lst) {
4291                         if (!net_eq(dev_net(ifa->idev->dev), net))
4292                                 continue;
4293                         return ifa;
4294                 }
4295         }
4296
4297         return NULL;
4298 }
4299
4300 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4301         __acquires(rcu)
4302 {
4303         rcu_read_lock();
4304         return if6_get_first(seq, *pos);
4305 }
4306
4307 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4308 {
4309         struct inet6_ifaddr *ifa;
4310
4311         ifa = if6_get_next(seq, v);
4312         ++*pos;
4313         return ifa;
4314 }
4315
4316 static void if6_seq_stop(struct seq_file *seq, void *v)
4317         __releases(rcu)
4318 {
4319         rcu_read_unlock();
4320 }
4321
4322 static int if6_seq_show(struct seq_file *seq, void *v)
4323 {
4324         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4325         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4326                    &ifp->addr,
4327                    ifp->idev->dev->ifindex,
4328                    ifp->prefix_len,
4329                    ifp->scope,
4330                    (u8) ifp->flags,
4331                    ifp->idev->dev->name);
4332         return 0;
4333 }
4334
4335 static const struct seq_operations if6_seq_ops = {
4336         .start  = if6_seq_start,
4337         .next   = if6_seq_next,
4338         .show   = if6_seq_show,
4339         .stop   = if6_seq_stop,
4340 };
4341
4342 static int __net_init if6_proc_net_init(struct net *net)
4343 {
4344         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4345                         sizeof(struct if6_iter_state)))
4346                 return -ENOMEM;
4347         return 0;
4348 }
4349
4350 static void __net_exit if6_proc_net_exit(struct net *net)
4351 {
4352         remove_proc_entry("if_inet6", net->proc_net);
4353 }
4354
4355 static struct pernet_operations if6_proc_net_ops = {
4356         .init = if6_proc_net_init,
4357         .exit = if6_proc_net_exit,
4358 };
4359
4360 int __init if6_proc_init(void)
4361 {
4362         return register_pernet_subsys(&if6_proc_net_ops);
4363 }
4364
4365 void if6_proc_exit(void)
4366 {
4367         unregister_pernet_subsys(&if6_proc_net_ops);
4368 }
4369 #endif  /* CONFIG_PROC_FS */
4370
4371 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4372 /* Check if address is a home address configured on any interface. */
4373 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4374 {
4375         unsigned int hash = inet6_addr_hash(net, addr);
4376         struct inet6_ifaddr *ifp = NULL;
4377         int ret = 0;
4378
4379         rcu_read_lock();
4380         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4381                 if (!net_eq(dev_net(ifp->idev->dev), net))
4382                         continue;
4383                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4384                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4385                         ret = 1;
4386                         break;
4387                 }
4388         }
4389         rcu_read_unlock();
4390         return ret;
4391 }
4392 #endif
4393
4394 /*
4395  *      Periodic address status verification
4396  */
4397
4398 static void addrconf_verify_rtnl(void)
4399 {
4400         unsigned long now, next, next_sec, next_sched;
4401         struct inet6_ifaddr *ifp;
4402         int i;
4403
4404         ASSERT_RTNL();
4405
4406         rcu_read_lock_bh();
4407         now = jiffies;
4408         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4409
4410         cancel_delayed_work(&addr_chk_work);
4411
4412         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4413 restart:
4414                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4415                         unsigned long age;
4416
4417                         /* When setting preferred_lft to a value not zero or
4418                          * infinity, while valid_lft is infinity
4419                          * IFA_F_PERMANENT has a non-infinity life time.
4420                          */
4421                         if ((ifp->flags & IFA_F_PERMANENT) &&
4422                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4423                                 continue;
4424
4425                         spin_lock(&ifp->lock);
4426                         /* We try to batch several events at once. */
4427                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4428
4429                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4430                             age >= ifp->valid_lft) {
4431                                 spin_unlock(&ifp->lock);
4432                                 in6_ifa_hold(ifp);
4433                                 ipv6_del_addr(ifp);
4434                                 goto restart;
4435                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4436                                 spin_unlock(&ifp->lock);
4437                                 continue;
4438                         } else if (age >= ifp->prefered_lft) {
4439                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4440                                 int deprecate = 0;
4441
4442                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4443                                         deprecate = 1;
4444                                         ifp->flags |= IFA_F_DEPRECATED;
4445                                 }
4446
4447                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4448                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4449                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4450
4451                                 spin_unlock(&ifp->lock);
4452
4453                                 if (deprecate) {
4454                                         in6_ifa_hold(ifp);
4455
4456                                         ipv6_ifa_notify(0, ifp);
4457                                         in6_ifa_put(ifp);
4458                                         goto restart;
4459                                 }
4460                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4461                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4462                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4463                                         ifp->idev->cnf.dad_transmits *
4464                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4465
4466                                 if (age >= ifp->prefered_lft - regen_advance) {
4467                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4468                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4469                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4470                                         if (!ifp->regen_count && ifpub) {
4471                                                 ifp->regen_count++;
4472                                                 in6_ifa_hold(ifp);
4473                                                 in6_ifa_hold(ifpub);
4474                                                 spin_unlock(&ifp->lock);
4475
4476                                                 spin_lock(&ifpub->lock);
4477                                                 ifpub->regen_count = 0;
4478                                                 spin_unlock(&ifpub->lock);
4479                                                 rcu_read_unlock_bh();
4480                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4481                                                 in6_ifa_put(ifpub);
4482                                                 in6_ifa_put(ifp);
4483                                                 rcu_read_lock_bh();
4484                                                 goto restart;
4485                                         }
4486                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4487                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4488                                 spin_unlock(&ifp->lock);
4489                         } else {
4490                                 /* ifp->prefered_lft <= ifp->valid_lft */
4491                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4492                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4493                                 spin_unlock(&ifp->lock);
4494                         }
4495                 }
4496         }
4497
4498         next_sec = round_jiffies_up(next);
4499         next_sched = next;
4500
4501         /* If rounded timeout is accurate enough, accept it. */
4502         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4503                 next_sched = next_sec;
4504
4505         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4506         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4507                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4508
4509         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4510                  now, next, next_sec, next_sched);
4511         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4512         rcu_read_unlock_bh();
4513 }
4514
4515 static void addrconf_verify_work(struct work_struct *w)
4516 {
4517         rtnl_lock();
4518         addrconf_verify_rtnl();
4519         rtnl_unlock();
4520 }
4521
4522 static void addrconf_verify(void)
4523 {
4524         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4525 }
4526
4527 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4528                                      struct in6_addr **peer_pfx)
4529 {
4530         struct in6_addr *pfx = NULL;
4531
4532         *peer_pfx = NULL;
4533
4534         if (addr)
4535                 pfx = nla_data(addr);
4536
4537         if (local) {
4538                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4539                         *peer_pfx = pfx;
4540                 pfx = nla_data(local);
4541         }
4542
4543         return pfx;
4544 }
4545
4546 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4547         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4548         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4549         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4550         [IFA_FLAGS]             = { .len = sizeof(u32) },
4551         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4552         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4553 };
4554
4555 static int
4556 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4557                   struct netlink_ext_ack *extack)
4558 {
4559         struct net *net = sock_net(skb->sk);
4560         struct ifaddrmsg *ifm;
4561         struct nlattr *tb[IFA_MAX+1];
4562         struct in6_addr *pfx, *peer_pfx;
4563         u32 ifa_flags;
4564         int err;
4565
4566         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4567                                      ifa_ipv6_policy, extack);
4568         if (err < 0)
4569                 return err;
4570
4571         ifm = nlmsg_data(nlh);
4572         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4573         if (!pfx)
4574                 return -EINVAL;
4575
4576         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4577
4578         /* We ignore other flags so far. */
4579         ifa_flags &= IFA_F_MANAGETEMPADDR;
4580
4581         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4582                               ifm->ifa_prefixlen);
4583 }
4584
4585 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4586                                unsigned long expires, u32 flags)
4587 {
4588         struct fib6_info *f6i;
4589         u32 prio;
4590
4591         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
4592                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4593         if (!f6i)
4594                 return -ENOENT;
4595
4596         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4597         if (f6i->fib6_metric != prio) {
4598                 /* delete old one */
4599                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4600
4601                 /* add new one */
4602                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4603                                       ifp->rt_priority, ifp->idev->dev,
4604                                       expires, flags, GFP_KERNEL);
4605         } else {
4606                 if (!expires)
4607                         fib6_clean_expires(f6i);
4608                 else
4609                         fib6_set_expires(f6i, expires);
4610
4611                 fib6_info_release(f6i);
4612         }
4613
4614         return 0;
4615 }
4616
4617 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4618 {
4619         u32 flags;
4620         clock_t expires;
4621         unsigned long timeout;
4622         bool was_managetempaddr;
4623         bool had_prefixroute;
4624
4625         ASSERT_RTNL();
4626
4627         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4628                 return -EINVAL;
4629
4630         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4631             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4632                 return -EINVAL;
4633
4634         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4635                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4636
4637         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4638         if (addrconf_finite_timeout(timeout)) {
4639                 expires = jiffies_to_clock_t(timeout * HZ);
4640                 cfg->valid_lft = timeout;
4641                 flags = RTF_EXPIRES;
4642         } else {
4643                 expires = 0;
4644                 flags = 0;
4645                 cfg->ifa_flags |= IFA_F_PERMANENT;
4646         }
4647
4648         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4649         if (addrconf_finite_timeout(timeout)) {
4650                 if (timeout == 0)
4651                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4652                 cfg->preferred_lft = timeout;
4653         }
4654
4655         spin_lock_bh(&ifp->lock);
4656         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4657         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4658                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4659         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4660                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4661                         IFA_F_NOPREFIXROUTE);
4662         ifp->flags |= cfg->ifa_flags;
4663         ifp->tstamp = jiffies;
4664         ifp->valid_lft = cfg->valid_lft;
4665         ifp->prefered_lft = cfg->preferred_lft;
4666
4667         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4668                 ifp->rt_priority = cfg->rt_priority;
4669
4670         spin_unlock_bh(&ifp->lock);
4671         if (!(ifp->flags&IFA_F_TENTATIVE))
4672                 ipv6_ifa_notify(0, ifp);
4673
4674         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4675                 int rc = -ENOENT;
4676
4677                 if (had_prefixroute)
4678                         rc = modify_prefix_route(ifp, expires, flags);
4679
4680                 /* prefix route could have been deleted; if so restore it */
4681                 if (rc == -ENOENT) {
4682                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4683                                               ifp->rt_priority, ifp->idev->dev,
4684                                               expires, flags, GFP_KERNEL);
4685                 }
4686         } else if (had_prefixroute) {
4687                 enum cleanup_prefix_rt_t action;
4688                 unsigned long rt_expires;
4689
4690                 write_lock_bh(&ifp->idev->lock);
4691                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4692                 write_unlock_bh(&ifp->idev->lock);
4693
4694                 if (action != CLEANUP_PREFIX_RT_NOP) {
4695                         cleanup_prefix_route(ifp, rt_expires,
4696                                 action == CLEANUP_PREFIX_RT_DEL);
4697                 }
4698         }
4699
4700         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4701                 if (was_managetempaddr &&
4702                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4703                         cfg->valid_lft = 0;
4704                         cfg->preferred_lft = 0;
4705                 }
4706                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4707                                  cfg->preferred_lft, !was_managetempaddr,
4708                                  jiffies);
4709         }
4710
4711         addrconf_verify_rtnl();
4712
4713         return 0;
4714 }
4715
4716 static int
4717 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4718                   struct netlink_ext_ack *extack)
4719 {
4720         struct net *net = sock_net(skb->sk);
4721         struct ifaddrmsg *ifm;
4722         struct nlattr *tb[IFA_MAX+1];
4723         struct in6_addr *peer_pfx;
4724         struct inet6_ifaddr *ifa;
4725         struct net_device *dev;
4726         struct inet6_dev *idev;
4727         struct ifa6_config cfg;
4728         int err;
4729
4730         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4731                                      ifa_ipv6_policy, extack);
4732         if (err < 0)
4733                 return err;
4734
4735         memset(&cfg, 0, sizeof(cfg));
4736
4737         ifm = nlmsg_data(nlh);
4738         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4739         if (!cfg.pfx)
4740                 return -EINVAL;
4741
4742         cfg.peer_pfx = peer_pfx;
4743         cfg.plen = ifm->ifa_prefixlen;
4744         if (tb[IFA_RT_PRIORITY])
4745                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4746
4747         cfg.valid_lft = INFINITY_LIFE_TIME;
4748         cfg.preferred_lft = INFINITY_LIFE_TIME;
4749
4750         if (tb[IFA_CACHEINFO]) {
4751                 struct ifa_cacheinfo *ci;
4752
4753                 ci = nla_data(tb[IFA_CACHEINFO]);
4754                 cfg.valid_lft = ci->ifa_valid;
4755                 cfg.preferred_lft = ci->ifa_prefered;
4756         }
4757
4758         dev =  __dev_get_by_index(net, ifm->ifa_index);
4759         if (!dev)
4760                 return -ENODEV;
4761
4762         if (tb[IFA_FLAGS])
4763                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4764         else
4765                 cfg.ifa_flags = ifm->ifa_flags;
4766
4767         /* We ignore other flags so far. */
4768         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4769                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4770                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4771
4772         idev = ipv6_find_idev(dev);
4773         if (!idev)
4774                 return -ENOBUFS;
4775
4776         if (!ipv6_allow_optimistic_dad(net, idev))
4777                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4778
4779         if (cfg.ifa_flags & IFA_F_NODAD &&
4780             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4781                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4782                 return -EINVAL;
4783         }
4784
4785         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4786         if (!ifa) {
4787                 /*
4788                  * It would be best to check for !NLM_F_CREATE here but
4789                  * userspace already relies on not having to provide this.
4790                  */
4791                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4792         }
4793
4794         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4795             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4796                 err = -EEXIST;
4797         else
4798                 err = inet6_addr_modify(ifa, &cfg);
4799
4800         in6_ifa_put(ifa);
4801
4802         return err;
4803 }
4804
4805 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4806                           u8 scope, int ifindex)
4807 {
4808         struct ifaddrmsg *ifm;
4809
4810         ifm = nlmsg_data(nlh);
4811         ifm->ifa_family = AF_INET6;
4812         ifm->ifa_prefixlen = prefixlen;
4813         ifm->ifa_flags = flags;
4814         ifm->ifa_scope = scope;
4815         ifm->ifa_index = ifindex;
4816 }
4817
4818 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4819                          unsigned long tstamp, u32 preferred, u32 valid)
4820 {
4821         struct ifa_cacheinfo ci;
4822
4823         ci.cstamp = cstamp_delta(cstamp);
4824         ci.tstamp = cstamp_delta(tstamp);
4825         ci.ifa_prefered = preferred;
4826         ci.ifa_valid = valid;
4827
4828         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4829 }
4830
4831 static inline int rt_scope(int ifa_scope)
4832 {
4833         if (ifa_scope & IFA_HOST)
4834                 return RT_SCOPE_HOST;
4835         else if (ifa_scope & IFA_LINK)
4836                 return RT_SCOPE_LINK;
4837         else if (ifa_scope & IFA_SITE)
4838                 return RT_SCOPE_SITE;
4839         else
4840                 return RT_SCOPE_UNIVERSE;
4841 }
4842
4843 static inline int inet6_ifaddr_msgsize(void)
4844 {
4845         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4846                + nla_total_size(16) /* IFA_LOCAL */
4847                + nla_total_size(16) /* IFA_ADDRESS */
4848                + nla_total_size(sizeof(struct ifa_cacheinfo))
4849                + nla_total_size(4)  /* IFA_FLAGS */
4850                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4851 }
4852
4853 enum addr_type_t {
4854         UNICAST_ADDR,
4855         MULTICAST_ADDR,
4856         ANYCAST_ADDR,
4857 };
4858
4859 struct inet6_fill_args {
4860         u32 portid;
4861         u32 seq;
4862         int event;
4863         unsigned int flags;
4864         int netnsid;
4865         int ifindex;
4866         enum addr_type_t type;
4867 };
4868
4869 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4870                              struct inet6_fill_args *args)
4871 {
4872         struct nlmsghdr  *nlh;
4873         u32 preferred, valid;
4874
4875         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4876                         sizeof(struct ifaddrmsg), args->flags);
4877         if (!nlh)
4878                 return -EMSGSIZE;
4879
4880         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4881                       ifa->idev->dev->ifindex);
4882
4883         if (args->netnsid >= 0 &&
4884             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4885                 goto error;
4886
4887         if (!((ifa->flags&IFA_F_PERMANENT) &&
4888               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4889                 preferred = ifa->prefered_lft;
4890                 valid = ifa->valid_lft;
4891                 if (preferred != INFINITY_LIFE_TIME) {
4892                         long tval = (jiffies - ifa->tstamp)/HZ;
4893                         if (preferred > tval)
4894                                 preferred -= tval;
4895                         else
4896                                 preferred = 0;
4897                         if (valid != INFINITY_LIFE_TIME) {
4898                                 if (valid > tval)
4899                                         valid -= tval;
4900                                 else
4901                                         valid = 0;
4902                         }
4903                 }
4904         } else {
4905                 preferred = INFINITY_LIFE_TIME;
4906                 valid = INFINITY_LIFE_TIME;
4907         }
4908
4909         if (!ipv6_addr_any(&ifa->peer_addr)) {
4910                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4911                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4912                         goto error;
4913         } else
4914                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4915                         goto error;
4916
4917         if (ifa->rt_priority &&
4918             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4919                 goto error;
4920
4921         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4922                 goto error;
4923
4924         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4925                 goto error;
4926
4927         nlmsg_end(skb, nlh);
4928         return 0;
4929
4930 error:
4931         nlmsg_cancel(skb, nlh);
4932         return -EMSGSIZE;
4933 }
4934
4935 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4936                                struct inet6_fill_args *args)
4937 {
4938         struct nlmsghdr  *nlh;
4939         u8 scope = RT_SCOPE_UNIVERSE;
4940         int ifindex = ifmca->idev->dev->ifindex;
4941
4942         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4943                 scope = RT_SCOPE_SITE;
4944
4945         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4946                         sizeof(struct ifaddrmsg), args->flags);
4947         if (!nlh)
4948                 return -EMSGSIZE;
4949
4950         if (args->netnsid >= 0 &&
4951             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4952                 return -EMSGSIZE;
4953
4954         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4955         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4956             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4957                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4958                 nlmsg_cancel(skb, nlh);
4959                 return -EMSGSIZE;
4960         }
4961
4962         nlmsg_end(skb, nlh);
4963         return 0;
4964 }
4965
4966 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4967                                struct inet6_fill_args *args)
4968 {
4969         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4970         int ifindex = dev ? dev->ifindex : 1;
4971         struct nlmsghdr  *nlh;
4972         u8 scope = RT_SCOPE_UNIVERSE;
4973
4974         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4975                 scope = RT_SCOPE_SITE;
4976
4977         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4978                         sizeof(struct ifaddrmsg), args->flags);
4979         if (!nlh)
4980                 return -EMSGSIZE;
4981
4982         if (args->netnsid >= 0 &&
4983             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4984                 return -EMSGSIZE;
4985
4986         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4987         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4988             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4989                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4990                 nlmsg_cancel(skb, nlh);
4991                 return -EMSGSIZE;
4992         }
4993
4994         nlmsg_end(skb, nlh);
4995         return 0;
4996 }
4997
4998 /* called with rcu_read_lock() */
4999 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5000                           struct netlink_callback *cb, int s_ip_idx,
5001                           struct inet6_fill_args *fillargs)
5002 {
5003         struct ifmcaddr6 *ifmca;
5004         struct ifacaddr6 *ifaca;
5005         int ip_idx = 0;
5006         int err = 1;
5007
5008         read_lock_bh(&idev->lock);
5009         switch (fillargs->type) {
5010         case UNICAST_ADDR: {
5011                 struct inet6_ifaddr *ifa;
5012                 fillargs->event = RTM_NEWADDR;
5013
5014                 /* unicast address incl. temp addr */
5015                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5016                         if (ip_idx < s_ip_idx)
5017                                 goto next;
5018                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5019                         if (err < 0)
5020                                 break;
5021                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5022 next:
5023                         ip_idx++;
5024                 }
5025                 break;
5026         }
5027         case MULTICAST_ADDR:
5028                 fillargs->event = RTM_GETMULTICAST;
5029
5030                 /* multicast address */
5031                 for (ifmca = idev->mc_list; ifmca;
5032                      ifmca = ifmca->next, ip_idx++) {
5033                         if (ip_idx < s_ip_idx)
5034                                 continue;
5035                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5036                         if (err < 0)
5037                                 break;
5038                 }
5039                 break;
5040         case ANYCAST_ADDR:
5041                 fillargs->event = RTM_GETANYCAST;
5042                 /* anycast address */
5043                 for (ifaca = idev->ac_list; ifaca;
5044                      ifaca = ifaca->aca_next, ip_idx++) {
5045                         if (ip_idx < s_ip_idx)
5046                                 continue;
5047                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5048                         if (err < 0)
5049                                 break;
5050                 }
5051                 break;
5052         default:
5053                 break;
5054         }
5055         read_unlock_bh(&idev->lock);
5056         cb->args[2] = ip_idx;
5057         return err;
5058 }
5059
5060 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5061                                        struct inet6_fill_args *fillargs,
5062                                        struct net **tgt_net, struct sock *sk,
5063                                        struct netlink_callback *cb)
5064 {
5065         struct netlink_ext_ack *extack = cb->extack;
5066         struct nlattr *tb[IFA_MAX+1];
5067         struct ifaddrmsg *ifm;
5068         int err, i;
5069
5070         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5071                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5072                 return -EINVAL;
5073         }
5074
5075         ifm = nlmsg_data(nlh);
5076         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5077                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5078                 return -EINVAL;
5079         }
5080
5081         fillargs->ifindex = ifm->ifa_index;
5082         if (fillargs->ifindex) {
5083                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5084                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5085         }
5086
5087         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5088                                             ifa_ipv6_policy, extack);
5089         if (err < 0)
5090                 return err;
5091
5092         for (i = 0; i <= IFA_MAX; ++i) {
5093                 if (!tb[i])
5094                         continue;
5095
5096                 if (i == IFA_TARGET_NETNSID) {
5097                         struct net *net;
5098
5099                         fillargs->netnsid = nla_get_s32(tb[i]);
5100                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5101                         if (IS_ERR(net)) {
5102                                 fillargs->netnsid = -1;
5103                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5104                                 return PTR_ERR(net);
5105                         }
5106                         *tgt_net = net;
5107                 } else {
5108                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5109                         return -EINVAL;
5110                 }
5111         }
5112
5113         return 0;
5114 }
5115
5116 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5117                            enum addr_type_t type)
5118 {
5119         const struct nlmsghdr *nlh = cb->nlh;
5120         struct inet6_fill_args fillargs = {
5121                 .portid = NETLINK_CB(cb->skb).portid,
5122                 .seq = cb->nlh->nlmsg_seq,
5123                 .flags = NLM_F_MULTI,
5124                 .netnsid = -1,
5125                 .type = type,
5126         };
5127         struct net *net = sock_net(skb->sk);
5128         struct net *tgt_net = net;
5129         int idx, s_idx, s_ip_idx;
5130         int h, s_h;
5131         struct net_device *dev;
5132         struct inet6_dev *idev;
5133         struct hlist_head *head;
5134         int err = 0;
5135
5136         s_h = cb->args[0];
5137         s_idx = idx = cb->args[1];
5138         s_ip_idx = cb->args[2];
5139
5140         if (cb->strict_check) {
5141                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5142                                                   skb->sk, cb);
5143                 if (err < 0)
5144                         goto put_tgt_net;
5145
5146                 err = 0;
5147                 if (fillargs.ifindex) {
5148                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5149                         if (!dev) {
5150                                 err = -ENODEV;
5151                                 goto put_tgt_net;
5152                         }
5153                         idev = __in6_dev_get(dev);
5154                         if (idev) {
5155                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5156                                                      &fillargs);
5157                                 if (err > 0)
5158                                         err = 0;
5159                         }
5160                         goto put_tgt_net;
5161                 }
5162         }
5163
5164         rcu_read_lock();
5165         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5166         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5167                 idx = 0;
5168                 head = &tgt_net->dev_index_head[h];
5169                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5170                         if (idx < s_idx)
5171                                 goto cont;
5172                         if (h > s_h || idx > s_idx)
5173                                 s_ip_idx = 0;
5174                         idev = __in6_dev_get(dev);
5175                         if (!idev)
5176                                 goto cont;
5177
5178                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5179                                            &fillargs) < 0)
5180                                 goto done;
5181 cont:
5182                         idx++;
5183                 }
5184         }
5185 done:
5186         rcu_read_unlock();
5187         cb->args[0] = h;
5188         cb->args[1] = idx;
5189 put_tgt_net:
5190         if (fillargs.netnsid >= 0)
5191                 put_net(tgt_net);
5192
5193         return skb->len ? : err;
5194 }
5195
5196 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5197 {
5198         enum addr_type_t type = UNICAST_ADDR;
5199
5200         return inet6_dump_addr(skb, cb, type);
5201 }
5202
5203 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5204 {
5205         enum addr_type_t type = MULTICAST_ADDR;
5206
5207         return inet6_dump_addr(skb, cb, type);
5208 }
5209
5210
5211 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5212 {
5213         enum addr_type_t type = ANYCAST_ADDR;
5214
5215         return inet6_dump_addr(skb, cb, type);
5216 }
5217
5218 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5219                                        const struct nlmsghdr *nlh,
5220                                        struct nlattr **tb,
5221                                        struct netlink_ext_ack *extack)
5222 {
5223         struct ifaddrmsg *ifm;
5224         int i, err;
5225
5226         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5227                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5228                 return -EINVAL;
5229         }
5230
5231         ifm = nlmsg_data(nlh);
5232         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5233                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5234                 return -EINVAL;
5235         }
5236
5237         if (!netlink_strict_get_check(skb))
5238                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5239                                               ifa_ipv6_policy, extack);
5240
5241         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5242                                             ifa_ipv6_policy, extack);
5243         if (err)
5244                 return err;
5245
5246         for (i = 0; i <= IFA_MAX; i++) {
5247                 if (!tb[i])
5248                         continue;
5249
5250                 switch (i) {
5251                 case IFA_TARGET_NETNSID:
5252                 case IFA_ADDRESS:
5253                 case IFA_LOCAL:
5254                         break;
5255                 default:
5256                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5257                         return -EINVAL;
5258                 }
5259         }
5260
5261         return 0;
5262 }
5263
5264 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5265                              struct netlink_ext_ack *extack)
5266 {
5267         struct net *net = sock_net(in_skb->sk);
5268         struct inet6_fill_args fillargs = {
5269                 .portid = NETLINK_CB(in_skb).portid,
5270                 .seq = nlh->nlmsg_seq,
5271                 .event = RTM_NEWADDR,
5272                 .flags = 0,
5273                 .netnsid = -1,
5274         };
5275         struct net *tgt_net = net;
5276         struct ifaddrmsg *ifm;
5277         struct nlattr *tb[IFA_MAX+1];
5278         struct in6_addr *addr = NULL, *peer;
5279         struct net_device *dev = NULL;
5280         struct inet6_ifaddr *ifa;
5281         struct sk_buff *skb;
5282         int err;
5283
5284         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5285         if (err < 0)
5286                 return err;
5287
5288         if (tb[IFA_TARGET_NETNSID]) {
5289                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5290
5291                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5292                                                   fillargs.netnsid);
5293                 if (IS_ERR(tgt_net))
5294                         return PTR_ERR(tgt_net);
5295         }
5296
5297         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5298         if (!addr)
5299                 return -EINVAL;
5300
5301         ifm = nlmsg_data(nlh);
5302         if (ifm->ifa_index)
5303                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5304
5305         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5306         if (!ifa) {
5307                 err = -EADDRNOTAVAIL;
5308                 goto errout;
5309         }
5310
5311         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5312         if (!skb) {
5313                 err = -ENOBUFS;
5314                 goto errout_ifa;
5315         }
5316
5317         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5318         if (err < 0) {
5319                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5320                 WARN_ON(err == -EMSGSIZE);
5321                 kfree_skb(skb);
5322                 goto errout_ifa;
5323         }
5324         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5325 errout_ifa:
5326         in6_ifa_put(ifa);
5327 errout:
5328         if (dev)
5329                 dev_put(dev);
5330         if (fillargs.netnsid >= 0)
5331                 put_net(tgt_net);
5332
5333         return err;
5334 }
5335
5336 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5337 {
5338         struct sk_buff *skb;
5339         struct net *net = dev_net(ifa->idev->dev);
5340         struct inet6_fill_args fillargs = {
5341                 .portid = 0,
5342                 .seq = 0,
5343                 .event = event,
5344                 .flags = 0,
5345                 .netnsid = -1,
5346         };
5347         int err = -ENOBUFS;
5348
5349         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5350         if (!skb)
5351                 goto errout;
5352
5353         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5354         if (err < 0) {
5355                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5356                 WARN_ON(err == -EMSGSIZE);
5357                 kfree_skb(skb);
5358                 goto errout;
5359         }
5360         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5361         return;
5362 errout:
5363         if (err < 0)
5364                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5365 }
5366
5367 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5368                                 __s32 *array, int bytes)
5369 {
5370         BUG_ON(bytes < (DEVCONF_MAX * 4));
5371
5372         memset(array, 0, bytes);
5373         array[DEVCONF_FORWARDING] = cnf->forwarding;
5374         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5375         array[DEVCONF_MTU6] = cnf->mtu6;
5376         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5377         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5378         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5379         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5380         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5381         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5382                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5383         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5384                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5385         array[DEVCONF_RTR_SOLICIT_DELAY] =
5386                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5387         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5388         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5389                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5390         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5391                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5392         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5393         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5394         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5395         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5396         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5397         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5398         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5399         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5400         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5401 #ifdef CONFIG_IPV6_ROUTER_PREF
5402         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5403         array[DEVCONF_RTR_PROBE_INTERVAL] =
5404                 jiffies_to_msecs(cnf->rtr_probe_interval);
5405 #ifdef CONFIG_IPV6_ROUTE_INFO
5406         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5407         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5408 #endif
5409 #endif
5410         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5411         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5412 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5413         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5414         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5415 #endif
5416 #ifdef CONFIG_IPV6_MROUTE
5417         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5418 #endif
5419         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5420         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5421         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5422         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5423         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5424         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5425         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5426         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5427         /* we omit DEVCONF_STABLE_SECRET for now */
5428         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5429         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5430         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5431         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5432         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5433 #ifdef CONFIG_IPV6_SEG6_HMAC
5434         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5435 #endif
5436         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5437         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5438         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5439         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5440 }
5441
5442 static inline size_t inet6_ifla6_size(void)
5443 {
5444         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5445              + nla_total_size(sizeof(struct ifla_cacheinfo))
5446              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5447              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5448              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5449              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5450              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5451              + 0;
5452 }
5453
5454 static inline size_t inet6_if_nlmsg_size(void)
5455 {
5456         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5457                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5458                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5459                + nla_total_size(4) /* IFLA_MTU */
5460                + nla_total_size(4) /* IFLA_LINK */
5461                + nla_total_size(1) /* IFLA_OPERSTATE */
5462                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5463 }
5464
5465 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5466                                         int bytes)
5467 {
5468         int i;
5469         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5470         BUG_ON(pad < 0);
5471
5472         /* Use put_unaligned() because stats may not be aligned for u64. */
5473         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5474         for (i = 1; i < ICMP6_MIB_MAX; i++)
5475                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5476
5477         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5478 }
5479
5480 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5481                                         int bytes, size_t syncpoff)
5482 {
5483         int i, c;
5484         u64 buff[IPSTATS_MIB_MAX];
5485         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5486
5487         BUG_ON(pad < 0);
5488
5489         memset(buff, 0, sizeof(buff));
5490         buff[0] = IPSTATS_MIB_MAX;
5491
5492         for_each_possible_cpu(c) {
5493                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5494                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5495         }
5496
5497         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5498         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5499 }
5500
5501 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5502                              int bytes)
5503 {
5504         switch (attrtype) {
5505         case IFLA_INET6_STATS:
5506                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5507                                      offsetof(struct ipstats_mib, syncp));
5508                 break;
5509         case IFLA_INET6_ICMP6STATS:
5510                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5511                 break;
5512         }
5513 }
5514
5515 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5516                                   u32 ext_filter_mask)
5517 {
5518         struct nlattr *nla;
5519         struct ifla_cacheinfo ci;
5520
5521         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5522                 goto nla_put_failure;
5523         ci.max_reasm_len = IPV6_MAXPLEN;
5524         ci.tstamp = cstamp_delta(idev->tstamp);
5525         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5526         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5527         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5528                 goto nla_put_failure;
5529         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5530         if (!nla)
5531                 goto nla_put_failure;
5532         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5533
5534         /* XXX - MC not implemented */
5535
5536         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5537                 return 0;
5538
5539         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5540         if (!nla)
5541                 goto nla_put_failure;
5542         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5543
5544         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5545         if (!nla)
5546                 goto nla_put_failure;
5547         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5548
5549         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5550         if (!nla)
5551                 goto nla_put_failure;
5552
5553         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5554                 goto nla_put_failure;
5555
5556         read_lock_bh(&idev->lock);
5557         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5558         read_unlock_bh(&idev->lock);
5559
5560         return 0;
5561
5562 nla_put_failure:
5563         return -EMSGSIZE;
5564 }
5565
5566 static size_t inet6_get_link_af_size(const struct net_device *dev,
5567                                      u32 ext_filter_mask)
5568 {
5569         if (!__in6_dev_get(dev))
5570                 return 0;
5571
5572         return inet6_ifla6_size();
5573 }
5574
5575 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5576                               u32 ext_filter_mask)
5577 {
5578         struct inet6_dev *idev = __in6_dev_get(dev);
5579
5580         if (!idev)
5581                 return -ENODATA;
5582
5583         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5584                 return -EMSGSIZE;
5585
5586         return 0;
5587 }
5588
5589 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5590 {
5591         struct inet6_ifaddr *ifp;
5592         struct net_device *dev = idev->dev;
5593         bool clear_token, update_rs = false;
5594         struct in6_addr ll_addr;
5595
5596         ASSERT_RTNL();
5597
5598         if (!token)
5599                 return -EINVAL;
5600         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5601                 return -EINVAL;
5602         if (!ipv6_accept_ra(idev))
5603                 return -EINVAL;
5604         if (idev->cnf.rtr_solicits == 0)
5605                 return -EINVAL;
5606
5607         write_lock_bh(&idev->lock);
5608
5609         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5610         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5611
5612         write_unlock_bh(&idev->lock);
5613
5614         clear_token = ipv6_addr_any(token);
5615         if (clear_token)
5616                 goto update_lft;
5617
5618         if (!idev->dead && (idev->if_flags & IF_READY) &&
5619             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5620                              IFA_F_OPTIMISTIC)) {
5621                 /* If we're not ready, then normal ifup will take care
5622                  * of this. Otherwise, we need to request our rs here.
5623                  */
5624                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5625                 update_rs = true;
5626         }
5627
5628 update_lft:
5629         write_lock_bh(&idev->lock);
5630
5631         if (update_rs) {
5632                 idev->if_flags |= IF_RS_SENT;
5633                 idev->rs_interval = rfc3315_s14_backoff_init(
5634                         idev->cnf.rtr_solicit_interval);
5635                 idev->rs_probes = 1;
5636                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5637         }
5638
5639         /* Well, that's kinda nasty ... */
5640         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5641                 spin_lock(&ifp->lock);
5642                 if (ifp->tokenized) {
5643                         ifp->valid_lft = 0;
5644                         ifp->prefered_lft = 0;
5645                 }
5646                 spin_unlock(&ifp->lock);
5647         }
5648
5649         write_unlock_bh(&idev->lock);
5650         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5651         addrconf_verify_rtnl();
5652         return 0;
5653 }
5654
5655 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5656         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5657         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5658 };
5659
5660 static int check_addr_gen_mode(int mode)
5661 {
5662         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5663             mode != IN6_ADDR_GEN_MODE_NONE &&
5664             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5665             mode != IN6_ADDR_GEN_MODE_RANDOM)
5666                 return -EINVAL;
5667         return 1;
5668 }
5669
5670 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5671                                 int mode)
5672 {
5673         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5674             !idev->cnf.stable_secret.initialized &&
5675             !net->ipv6.devconf_dflt->stable_secret.initialized)
5676                 return -EINVAL;
5677         return 1;
5678 }
5679
5680 static int inet6_validate_link_af(const struct net_device *dev,
5681                                   const struct nlattr *nla)
5682 {
5683         struct nlattr *tb[IFLA_INET6_MAX + 1];
5684         struct inet6_dev *idev = NULL;
5685         int err;
5686
5687         if (dev) {
5688                 idev = __in6_dev_get(dev);
5689                 if (!idev)
5690                         return -EAFNOSUPPORT;
5691         }
5692
5693         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5694                                           inet6_af_policy, NULL);
5695         if (err)
5696                 return err;
5697
5698         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5699                 return -EINVAL;
5700
5701         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5702                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5703
5704                 if (check_addr_gen_mode(mode) < 0)
5705                         return -EINVAL;
5706                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5707                         return -EINVAL;
5708         }
5709
5710         return 0;
5711 }
5712
5713 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5714 {
5715         struct inet6_dev *idev = __in6_dev_get(dev);
5716         struct nlattr *tb[IFLA_INET6_MAX + 1];
5717         int err;
5718
5719         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5720                 BUG();
5721
5722         if (tb[IFLA_INET6_TOKEN]) {
5723                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5724                 if (err)
5725                         return err;
5726         }
5727
5728         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5729                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5730
5731                 idev->cnf.addr_gen_mode = mode;
5732         }
5733
5734         return 0;
5735 }
5736
5737 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5738                              u32 portid, u32 seq, int event, unsigned int flags)
5739 {
5740         struct net_device *dev = idev->dev;
5741         struct ifinfomsg *hdr;
5742         struct nlmsghdr *nlh;
5743         void *protoinfo;
5744
5745         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5746         if (!nlh)
5747                 return -EMSGSIZE;
5748
5749         hdr = nlmsg_data(nlh);
5750         hdr->ifi_family = AF_INET6;
5751         hdr->__ifi_pad = 0;
5752         hdr->ifi_type = dev->type;
5753         hdr->ifi_index = dev->ifindex;
5754         hdr->ifi_flags = dev_get_flags(dev);
5755         hdr->ifi_change = 0;
5756
5757         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5758             (dev->addr_len &&
5759              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5760             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5761             (dev->ifindex != dev_get_iflink(dev) &&
5762              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5763             nla_put_u8(skb, IFLA_OPERSTATE,
5764                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5765                 goto nla_put_failure;
5766         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5767         if (!protoinfo)
5768                 goto nla_put_failure;
5769
5770         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5771                 goto nla_put_failure;
5772
5773         nla_nest_end(skb, protoinfo);
5774         nlmsg_end(skb, nlh);
5775         return 0;
5776
5777 nla_put_failure:
5778         nlmsg_cancel(skb, nlh);
5779         return -EMSGSIZE;
5780 }
5781
5782 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5783                                    struct netlink_ext_ack *extack)
5784 {
5785         struct ifinfomsg *ifm;
5786
5787         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5788                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5789                 return -EINVAL;
5790         }
5791
5792         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5793                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5794                 return -EINVAL;
5795         }
5796
5797         ifm = nlmsg_data(nlh);
5798         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5799             ifm->ifi_change || ifm->ifi_index) {
5800                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5801                 return -EINVAL;
5802         }
5803
5804         return 0;
5805 }
5806
5807 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5808 {
5809         struct net *net = sock_net(skb->sk);
5810         int h, s_h;
5811         int idx = 0, s_idx;
5812         struct net_device *dev;
5813         struct inet6_dev *idev;
5814         struct hlist_head *head;
5815
5816         /* only requests using strict checking can pass data to
5817          * influence the dump
5818          */
5819         if (cb->strict_check) {
5820                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5821
5822                 if (err < 0)
5823                         return err;
5824         }
5825
5826         s_h = cb->args[0];
5827         s_idx = cb->args[1];
5828
5829         rcu_read_lock();
5830         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5831                 idx = 0;
5832                 head = &net->dev_index_head[h];
5833                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5834                         if (idx < s_idx)
5835                                 goto cont;
5836                         idev = __in6_dev_get(dev);
5837                         if (!idev)
5838                                 goto cont;
5839                         if (inet6_fill_ifinfo(skb, idev,
5840                                               NETLINK_CB(cb->skb).portid,
5841                                               cb->nlh->nlmsg_seq,
5842                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5843                                 goto out;
5844 cont:
5845                         idx++;
5846                 }
5847         }
5848 out:
5849         rcu_read_unlock();
5850         cb->args[1] = idx;
5851         cb->args[0] = h;
5852
5853         return skb->len;
5854 }
5855
5856 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5857 {
5858         struct sk_buff *skb;
5859         struct net *net = dev_net(idev->dev);
5860         int err = -ENOBUFS;
5861
5862         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5863         if (!skb)
5864                 goto errout;
5865
5866         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5867         if (err < 0) {
5868                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5869                 WARN_ON(err == -EMSGSIZE);
5870                 kfree_skb(skb);
5871                 goto errout;
5872         }
5873         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5874         return;
5875 errout:
5876         if (err < 0)
5877                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5878 }
5879
5880 static inline size_t inet6_prefix_nlmsg_size(void)
5881 {
5882         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5883                + nla_total_size(sizeof(struct in6_addr))
5884                + nla_total_size(sizeof(struct prefix_cacheinfo));
5885 }
5886
5887 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5888                              struct prefix_info *pinfo, u32 portid, u32 seq,
5889                              int event, unsigned int flags)
5890 {
5891         struct prefixmsg *pmsg;
5892         struct nlmsghdr *nlh;
5893         struct prefix_cacheinfo ci;
5894
5895         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5896         if (!nlh)
5897                 return -EMSGSIZE;
5898
5899         pmsg = nlmsg_data(nlh);
5900         pmsg->prefix_family = AF_INET6;
5901         pmsg->prefix_pad1 = 0;
5902         pmsg->prefix_pad2 = 0;
5903         pmsg->prefix_ifindex = idev->dev->ifindex;
5904         pmsg->prefix_len = pinfo->prefix_len;
5905         pmsg->prefix_type = pinfo->type;
5906         pmsg->prefix_pad3 = 0;
5907         pmsg->prefix_flags = 0;
5908         if (pinfo->onlink)
5909                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5910         if (pinfo->autoconf)
5911                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5912
5913         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5914                 goto nla_put_failure;
5915         ci.preferred_time = ntohl(pinfo->prefered);
5916         ci.valid_time = ntohl(pinfo->valid);
5917         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5918                 goto nla_put_failure;
5919         nlmsg_end(skb, nlh);
5920         return 0;
5921
5922 nla_put_failure:
5923         nlmsg_cancel(skb, nlh);
5924         return -EMSGSIZE;
5925 }
5926
5927 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5928                          struct prefix_info *pinfo)
5929 {
5930         struct sk_buff *skb;
5931         struct net *net = dev_net(idev->dev);
5932         int err = -ENOBUFS;
5933
5934         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5935         if (!skb)
5936                 goto errout;
5937
5938         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5939         if (err < 0) {
5940                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5941                 WARN_ON(err == -EMSGSIZE);
5942                 kfree_skb(skb);
5943                 goto errout;
5944         }
5945         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5946         return;
5947 errout:
5948         if (err < 0)
5949                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5950 }
5951
5952 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5953 {
5954         struct net *net = dev_net(ifp->idev->dev);
5955
5956         if (event)
5957                 ASSERT_RTNL();
5958
5959         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5960
5961         switch (event) {
5962         case RTM_NEWADDR:
5963                 /*
5964                  * If the address was optimistic
5965                  * we inserted the route at the start of
5966                  * our DAD process, so we don't need
5967                  * to do it again
5968                  */
5969                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5970                         ip6_ins_rt(net, ifp->rt);
5971                 if (ifp->idev->cnf.forwarding)
5972                         addrconf_join_anycast(ifp);
5973                 if (!ipv6_addr_any(&ifp->peer_addr))
5974                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5975                                               ifp->idev->dev, 0, 0,
5976                                               GFP_ATOMIC);
5977                 break;
5978         case RTM_DELADDR:
5979                 if (ifp->idev->cnf.forwarding)
5980                         addrconf_leave_anycast(ifp);
5981                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5982                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5983                         struct fib6_info *rt;
5984
5985                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5986                                                        ifp->idev->dev, 0, 0,
5987                                                        false);
5988                         if (rt)
5989                                 ip6_del_rt(net, rt);
5990                 }
5991                 if (ifp->rt) {
5992                         ip6_del_rt(net, ifp->rt);
5993                         ifp->rt = NULL;
5994                 }
5995                 rt_genid_bump_ipv6(net);
5996                 break;
5997         }
5998         atomic_inc(&net->ipv6.dev_addr_genid);
5999 }
6000
6001 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6002 {
6003         rcu_read_lock_bh();
6004         if (likely(ifp->idev->dead == 0))
6005                 __ipv6_ifa_notify(event, ifp);
6006         rcu_read_unlock_bh();
6007 }
6008
6009 #ifdef CONFIG_SYSCTL
6010
6011 static
6012 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6013                            void __user *buffer, size_t *lenp, loff_t *ppos)
6014 {
6015         int *valp = ctl->data;
6016         int val = *valp;
6017         loff_t pos = *ppos;
6018         struct ctl_table lctl;
6019         int ret;
6020
6021         /*
6022          * ctl->data points to idev->cnf.forwarding, we should
6023          * not modify it until we get the rtnl lock.
6024          */
6025         lctl = *ctl;
6026         lctl.data = &val;
6027
6028         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6029
6030         if (write)
6031                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6032         if (ret)
6033                 *ppos = pos;
6034         return ret;
6035 }
6036
6037 static
6038 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6039                         void __user *buffer, size_t *lenp, loff_t *ppos)
6040 {
6041         struct inet6_dev *idev = ctl->extra1;
6042         int min_mtu = IPV6_MIN_MTU;
6043         struct ctl_table lctl;
6044
6045         lctl = *ctl;
6046         lctl.extra1 = &min_mtu;
6047         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6048
6049         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6050 }
6051
6052 static void dev_disable_change(struct inet6_dev *idev)
6053 {
6054         struct netdev_notifier_info info;
6055
6056         if (!idev || !idev->dev)
6057                 return;
6058
6059         netdev_notifier_info_init(&info, idev->dev);
6060         if (idev->cnf.disable_ipv6)
6061                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6062         else
6063                 addrconf_notify(NULL, NETDEV_UP, &info);
6064 }
6065
6066 static void addrconf_disable_change(struct net *net, __s32 newf)
6067 {
6068         struct net_device *dev;
6069         struct inet6_dev *idev;
6070
6071         for_each_netdev(net, dev) {
6072                 idev = __in6_dev_get(dev);
6073                 if (idev) {
6074                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6075                         idev->cnf.disable_ipv6 = newf;
6076                         if (changed)
6077                                 dev_disable_change(idev);
6078                 }
6079         }
6080 }
6081
6082 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6083 {
6084         struct net *net;
6085         int old;
6086
6087         if (!rtnl_trylock())
6088                 return restart_syscall();
6089
6090         net = (struct net *)table->extra2;
6091         old = *p;
6092         *p = newf;
6093
6094         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6095                 rtnl_unlock();
6096                 return 0;
6097         }
6098
6099         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6100                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6101                 addrconf_disable_change(net, newf);
6102         } else if ((!newf) ^ (!old))
6103                 dev_disable_change((struct inet6_dev *)table->extra1);
6104
6105         rtnl_unlock();
6106         return 0;
6107 }
6108
6109 static
6110 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6111                             void __user *buffer, size_t *lenp, loff_t *ppos)
6112 {
6113         int *valp = ctl->data;
6114         int val = *valp;
6115         loff_t pos = *ppos;
6116         struct ctl_table lctl;
6117         int ret;
6118
6119         /*
6120          * ctl->data points to idev->cnf.disable_ipv6, we should
6121          * not modify it until we get the rtnl lock.
6122          */
6123         lctl = *ctl;
6124         lctl.data = &val;
6125
6126         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6127
6128         if (write)
6129                 ret = addrconf_disable_ipv6(ctl, valp, val);
6130         if (ret)
6131                 *ppos = pos;
6132         return ret;
6133 }
6134
6135 static
6136 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6137                               void __user *buffer, size_t *lenp, loff_t *ppos)
6138 {
6139         int *valp = ctl->data;
6140         int ret;
6141         int old, new;
6142
6143         old = *valp;
6144         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6145         new = *valp;
6146
6147         if (write && old != new) {
6148                 struct net *net = ctl->extra2;
6149
6150                 if (!rtnl_trylock())
6151                         return restart_syscall();
6152
6153                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6154                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6155                                                      NETCONFA_PROXY_NEIGH,
6156                                                      NETCONFA_IFINDEX_DEFAULT,
6157                                                      net->ipv6.devconf_dflt);
6158                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6159                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6160                                                      NETCONFA_PROXY_NEIGH,
6161                                                      NETCONFA_IFINDEX_ALL,
6162                                                      net->ipv6.devconf_all);
6163                 else {
6164                         struct inet6_dev *idev = ctl->extra1;
6165
6166                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6167                                                      NETCONFA_PROXY_NEIGH,
6168                                                      idev->dev->ifindex,
6169                                                      &idev->cnf);
6170                 }
6171                 rtnl_unlock();
6172         }
6173
6174         return ret;
6175 }
6176
6177 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6178                                          void __user *buffer, size_t *lenp,
6179                                          loff_t *ppos)
6180 {
6181         int ret = 0;
6182         u32 new_val;
6183         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6184         struct net *net = (struct net *)ctl->extra2;
6185         struct ctl_table tmp = {
6186                 .data = &new_val,
6187                 .maxlen = sizeof(new_val),
6188                 .mode = ctl->mode,
6189         };
6190
6191         if (!rtnl_trylock())
6192                 return restart_syscall();
6193
6194         new_val = *((u32 *)ctl->data);
6195
6196         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6197         if (ret != 0)
6198                 goto out;
6199
6200         if (write) {
6201                 if (check_addr_gen_mode(new_val) < 0) {
6202                         ret = -EINVAL;
6203                         goto out;
6204                 }
6205
6206                 if (idev) {
6207                         if (check_stable_privacy(idev, net, new_val) < 0) {
6208                                 ret = -EINVAL;
6209                                 goto out;
6210                         }
6211
6212                         if (idev->cnf.addr_gen_mode != new_val) {
6213                                 idev->cnf.addr_gen_mode = new_val;
6214                                 addrconf_dev_config(idev->dev);
6215                         }
6216                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6217                         struct net_device *dev;
6218
6219                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6220                         for_each_netdev(net, dev) {
6221                                 idev = __in6_dev_get(dev);
6222                                 if (idev &&
6223                                     idev->cnf.addr_gen_mode != new_val) {
6224                                         idev->cnf.addr_gen_mode = new_val;
6225                                         addrconf_dev_config(idev->dev);
6226                                 }
6227                         }
6228                 }
6229
6230                 *((u32 *)ctl->data) = new_val;
6231         }
6232
6233 out:
6234         rtnl_unlock();
6235
6236         return ret;
6237 }
6238
6239 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6240                                          void __user *buffer, size_t *lenp,
6241                                          loff_t *ppos)
6242 {
6243         int err;
6244         struct in6_addr addr;
6245         char str[IPV6_MAX_STRLEN];
6246         struct ctl_table lctl = *ctl;
6247         struct net *net = ctl->extra2;
6248         struct ipv6_stable_secret *secret = ctl->data;
6249
6250         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6251                 return -EIO;
6252
6253         lctl.maxlen = IPV6_MAX_STRLEN;
6254         lctl.data = str;
6255
6256         if (!rtnl_trylock())
6257                 return restart_syscall();
6258
6259         if (!write && !secret->initialized) {
6260                 err = -EIO;
6261                 goto out;
6262         }
6263
6264         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6265         if (err >= sizeof(str)) {
6266                 err = -EIO;
6267                 goto out;
6268         }
6269
6270         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6271         if (err || !write)
6272                 goto out;
6273
6274         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6275                 err = -EIO;
6276                 goto out;
6277         }
6278
6279         secret->initialized = true;
6280         secret->secret = addr;
6281
6282         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6283                 struct net_device *dev;
6284
6285                 for_each_netdev(net, dev) {
6286                         struct inet6_dev *idev = __in6_dev_get(dev);
6287
6288                         if (idev) {
6289                                 idev->cnf.addr_gen_mode =
6290                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6291                         }
6292                 }
6293         } else {
6294                 struct inet6_dev *idev = ctl->extra1;
6295
6296                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6297         }
6298
6299 out:
6300         rtnl_unlock();
6301
6302         return err;
6303 }
6304
6305 static
6306 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6307                                                 int write,
6308                                                 void __user *buffer,
6309                                                 size_t *lenp,
6310                                                 loff_t *ppos)
6311 {
6312         int *valp = ctl->data;
6313         int val = *valp;
6314         loff_t pos = *ppos;
6315         struct ctl_table lctl;
6316         int ret;
6317
6318         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6319          * we should not modify it until we get the rtnl lock.
6320          */
6321         lctl = *ctl;
6322         lctl.data = &val;
6323
6324         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6325
6326         if (write)
6327                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6328         if (ret)
6329                 *ppos = pos;
6330         return ret;
6331 }
6332
6333 static
6334 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6335 {
6336         if (rt) {
6337                 if (action)
6338                         rt->dst.flags |= DST_NOPOLICY;
6339                 else
6340                         rt->dst.flags &= ~DST_NOPOLICY;
6341         }
6342 }
6343
6344 static
6345 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6346 {
6347         struct inet6_ifaddr *ifa;
6348
6349         read_lock_bh(&idev->lock);
6350         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6351                 spin_lock(&ifa->lock);
6352                 if (ifa->rt) {
6353                         struct fib6_info *rt = ifa->rt;
6354                         int cpu;
6355
6356                         rcu_read_lock();
6357                         ifa->rt->dst_nopolicy = val ? true : false;
6358                         if (rt->rt6i_pcpu) {
6359                                 for_each_possible_cpu(cpu) {
6360                                         struct rt6_info **rtp;
6361
6362                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6363                                         addrconf_set_nopolicy(*rtp, val);
6364                                 }
6365                         }
6366                         rcu_read_unlock();
6367                 }
6368                 spin_unlock(&ifa->lock);
6369         }
6370         read_unlock_bh(&idev->lock);
6371 }
6372
6373 static
6374 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6375 {
6376         struct inet6_dev *idev;
6377         struct net *net;
6378
6379         if (!rtnl_trylock())
6380                 return restart_syscall();
6381
6382         *valp = val;
6383
6384         net = (struct net *)ctl->extra2;
6385         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6386                 rtnl_unlock();
6387                 return 0;
6388         }
6389
6390         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6391                 struct net_device *dev;
6392
6393                 for_each_netdev(net, dev) {
6394                         idev = __in6_dev_get(dev);
6395                         if (idev)
6396                                 addrconf_disable_policy_idev(idev, val);
6397                 }
6398         } else {
6399                 idev = (struct inet6_dev *)ctl->extra1;
6400                 addrconf_disable_policy_idev(idev, val);
6401         }
6402
6403         rtnl_unlock();
6404         return 0;
6405 }
6406
6407 static
6408 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6409                                    void __user *buffer, size_t *lenp,
6410                                    loff_t *ppos)
6411 {
6412         int *valp = ctl->data;
6413         int val = *valp;
6414         loff_t pos = *ppos;
6415         struct ctl_table lctl;
6416         int ret;
6417
6418         lctl = *ctl;
6419         lctl.data = &val;
6420         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6421
6422         if (write && (*valp != val))
6423                 ret = addrconf_disable_policy(ctl, valp, val);
6424
6425         if (ret)
6426                 *ppos = pos;
6427
6428         return ret;
6429 }
6430
6431 static int minus_one = -1;
6432 static const int zero = 0;
6433 static const int one = 1;
6434 static const int two_five_five = 255;
6435
6436 static const struct ctl_table addrconf_sysctl[] = {
6437         {
6438                 .procname       = "forwarding",
6439                 .data           = &ipv6_devconf.forwarding,
6440                 .maxlen         = sizeof(int),
6441                 .mode           = 0644,
6442                 .proc_handler   = addrconf_sysctl_forward,
6443         },
6444         {
6445                 .procname       = "hop_limit",
6446                 .data           = &ipv6_devconf.hop_limit,
6447                 .maxlen         = sizeof(int),
6448                 .mode           = 0644,
6449                 .proc_handler   = proc_dointvec_minmax,
6450                 .extra1         = (void *)&one,
6451                 .extra2         = (void *)&two_five_five,
6452         },
6453         {
6454                 .procname       = "mtu",
6455                 .data           = &ipv6_devconf.mtu6,
6456                 .maxlen         = sizeof(int),
6457                 .mode           = 0644,
6458                 .proc_handler   = addrconf_sysctl_mtu,
6459         },
6460         {
6461                 .procname       = "accept_ra",
6462                 .data           = &ipv6_devconf.accept_ra,
6463                 .maxlen         = sizeof(int),
6464                 .mode           = 0644,
6465                 .proc_handler   = proc_dointvec,
6466         },
6467         {
6468                 .procname       = "accept_redirects",
6469                 .data           = &ipv6_devconf.accept_redirects,
6470                 .maxlen         = sizeof(int),
6471                 .mode           = 0644,
6472                 .proc_handler   = proc_dointvec,
6473         },
6474         {
6475                 .procname       = "autoconf",
6476                 .data           = &ipv6_devconf.autoconf,
6477                 .maxlen         = sizeof(int),
6478                 .mode           = 0644,
6479                 .proc_handler   = proc_dointvec,
6480         },
6481         {
6482                 .procname       = "dad_transmits",
6483                 .data           = &ipv6_devconf.dad_transmits,
6484                 .maxlen         = sizeof(int),
6485                 .mode           = 0644,
6486                 .proc_handler   = proc_dointvec,
6487         },
6488         {
6489                 .procname       = "router_solicitations",
6490                 .data           = &ipv6_devconf.rtr_solicits,
6491                 .maxlen         = sizeof(int),
6492                 .mode           = 0644,
6493                 .proc_handler   = proc_dointvec_minmax,
6494                 .extra1         = &minus_one,
6495         },
6496         {
6497                 .procname       = "router_solicitation_interval",
6498                 .data           = &ipv6_devconf.rtr_solicit_interval,
6499                 .maxlen         = sizeof(int),
6500                 .mode           = 0644,
6501                 .proc_handler   = proc_dointvec_jiffies,
6502         },
6503         {
6504                 .procname       = "router_solicitation_max_interval",
6505                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6506                 .maxlen         = sizeof(int),
6507                 .mode           = 0644,
6508                 .proc_handler   = proc_dointvec_jiffies,
6509         },
6510         {
6511                 .procname       = "router_solicitation_delay",
6512                 .data           = &ipv6_devconf.rtr_solicit_delay,
6513                 .maxlen         = sizeof(int),
6514                 .mode           = 0644,
6515                 .proc_handler   = proc_dointvec_jiffies,
6516         },
6517         {
6518                 .procname       = "force_mld_version",
6519                 .data           = &ipv6_devconf.force_mld_version,
6520                 .maxlen         = sizeof(int),
6521                 .mode           = 0644,
6522                 .proc_handler   = proc_dointvec,
6523         },
6524         {
6525                 .procname       = "mldv1_unsolicited_report_interval",
6526                 .data           =
6527                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6528                 .maxlen         = sizeof(int),
6529                 .mode           = 0644,
6530                 .proc_handler   = proc_dointvec_ms_jiffies,
6531         },
6532         {
6533                 .procname       = "mldv2_unsolicited_report_interval",
6534                 .data           =
6535                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6536                 .maxlen         = sizeof(int),
6537                 .mode           = 0644,
6538                 .proc_handler   = proc_dointvec_ms_jiffies,
6539         },
6540         {
6541                 .procname       = "use_tempaddr",
6542                 .data           = &ipv6_devconf.use_tempaddr,
6543                 .maxlen         = sizeof(int),
6544                 .mode           = 0644,
6545                 .proc_handler   = proc_dointvec,
6546         },
6547         {
6548                 .procname       = "temp_valid_lft",
6549                 .data           = &ipv6_devconf.temp_valid_lft,
6550                 .maxlen         = sizeof(int),
6551                 .mode           = 0644,
6552                 .proc_handler   = proc_dointvec,
6553         },
6554         {
6555                 .procname       = "temp_prefered_lft",
6556                 .data           = &ipv6_devconf.temp_prefered_lft,
6557                 .maxlen         = sizeof(int),
6558                 .mode           = 0644,
6559                 .proc_handler   = proc_dointvec,
6560         },
6561         {
6562                 .procname       = "regen_max_retry",
6563                 .data           = &ipv6_devconf.regen_max_retry,
6564                 .maxlen         = sizeof(int),
6565                 .mode           = 0644,
6566                 .proc_handler   = proc_dointvec,
6567         },
6568         {
6569                 .procname       = "max_desync_factor",
6570                 .data           = &ipv6_devconf.max_desync_factor,
6571                 .maxlen         = sizeof(int),
6572                 .mode           = 0644,
6573                 .proc_handler   = proc_dointvec,
6574         },
6575         {
6576                 .procname       = "max_addresses",
6577                 .data           = &ipv6_devconf.max_addresses,
6578                 .maxlen         = sizeof(int),
6579                 .mode           = 0644,
6580                 .proc_handler   = proc_dointvec,
6581         },
6582         {
6583                 .procname       = "accept_ra_defrtr",
6584                 .data           = &ipv6_devconf.accept_ra_defrtr,
6585                 .maxlen         = sizeof(int),
6586                 .mode           = 0644,
6587                 .proc_handler   = proc_dointvec,
6588         },
6589         {
6590                 .procname       = "accept_ra_min_hop_limit",
6591                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6592                 .maxlen         = sizeof(int),
6593                 .mode           = 0644,
6594                 .proc_handler   = proc_dointvec,
6595         },
6596         {
6597                 .procname       = "accept_ra_pinfo",
6598                 .data           = &ipv6_devconf.accept_ra_pinfo,
6599                 .maxlen         = sizeof(int),
6600                 .mode           = 0644,
6601                 .proc_handler   = proc_dointvec,
6602         },
6603 #ifdef CONFIG_IPV6_ROUTER_PREF
6604         {
6605                 .procname       = "accept_ra_rtr_pref",
6606                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6607                 .maxlen         = sizeof(int),
6608                 .mode           = 0644,
6609                 .proc_handler   = proc_dointvec,
6610         },
6611         {
6612                 .procname       = "router_probe_interval",
6613                 .data           = &ipv6_devconf.rtr_probe_interval,
6614                 .maxlen         = sizeof(int),
6615                 .mode           = 0644,
6616                 .proc_handler   = proc_dointvec_jiffies,
6617         },
6618 #ifdef CONFIG_IPV6_ROUTE_INFO
6619         {
6620                 .procname       = "accept_ra_rt_info_min_plen",
6621                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6622                 .maxlen         = sizeof(int),
6623                 .mode           = 0644,
6624                 .proc_handler   = proc_dointvec,
6625         },
6626         {
6627                 .procname       = "accept_ra_rt_info_max_plen",
6628                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6629                 .maxlen         = sizeof(int),
6630                 .mode           = 0644,
6631                 .proc_handler   = proc_dointvec,
6632         },
6633 #endif
6634 #endif
6635         {
6636                 .procname       = "proxy_ndp",
6637                 .data           = &ipv6_devconf.proxy_ndp,
6638                 .maxlen         = sizeof(int),
6639                 .mode           = 0644,
6640                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6641         },
6642         {
6643                 .procname       = "accept_source_route",
6644                 .data           = &ipv6_devconf.accept_source_route,
6645                 .maxlen         = sizeof(int),
6646                 .mode           = 0644,
6647                 .proc_handler   = proc_dointvec,
6648         },
6649 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6650         {
6651                 .procname       = "optimistic_dad",
6652                 .data           = &ipv6_devconf.optimistic_dad,
6653                 .maxlen         = sizeof(int),
6654                 .mode           = 0644,
6655                 .proc_handler   = proc_dointvec,
6656         },
6657         {
6658                 .procname       = "use_optimistic",
6659                 .data           = &ipv6_devconf.use_optimistic,
6660                 .maxlen         = sizeof(int),
6661                 .mode           = 0644,
6662                 .proc_handler   = proc_dointvec,
6663         },
6664 #endif
6665 #ifdef CONFIG_IPV6_MROUTE
6666         {
6667                 .procname       = "mc_forwarding",
6668                 .data           = &ipv6_devconf.mc_forwarding,
6669                 .maxlen         = sizeof(int),
6670                 .mode           = 0444,
6671                 .proc_handler   = proc_dointvec,
6672         },
6673 #endif
6674         {
6675                 .procname       = "disable_ipv6",
6676                 .data           = &ipv6_devconf.disable_ipv6,
6677                 .maxlen         = sizeof(int),
6678                 .mode           = 0644,
6679                 .proc_handler   = addrconf_sysctl_disable,
6680         },
6681         {
6682                 .procname       = "accept_dad",
6683                 .data           = &ipv6_devconf.accept_dad,
6684                 .maxlen         = sizeof(int),
6685                 .mode           = 0644,
6686                 .proc_handler   = proc_dointvec,
6687         },
6688         {
6689                 .procname       = "force_tllao",
6690                 .data           = &ipv6_devconf.force_tllao,
6691                 .maxlen         = sizeof(int),
6692                 .mode           = 0644,
6693                 .proc_handler   = proc_dointvec
6694         },
6695         {
6696                 .procname       = "ndisc_notify",
6697                 .data           = &ipv6_devconf.ndisc_notify,
6698                 .maxlen         = sizeof(int),
6699                 .mode           = 0644,
6700                 .proc_handler   = proc_dointvec
6701         },
6702         {
6703                 .procname       = "suppress_frag_ndisc",
6704                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6705                 .maxlen         = sizeof(int),
6706                 .mode           = 0644,
6707                 .proc_handler   = proc_dointvec
6708         },
6709         {
6710                 .procname       = "accept_ra_from_local",
6711                 .data           = &ipv6_devconf.accept_ra_from_local,
6712                 .maxlen         = sizeof(int),
6713                 .mode           = 0644,
6714                 .proc_handler   = proc_dointvec,
6715         },
6716         {
6717                 .procname       = "accept_ra_mtu",
6718                 .data           = &ipv6_devconf.accept_ra_mtu,
6719                 .maxlen         = sizeof(int),
6720                 .mode           = 0644,
6721                 .proc_handler   = proc_dointvec,
6722         },
6723         {
6724                 .procname       = "stable_secret",
6725                 .data           = &ipv6_devconf.stable_secret,
6726                 .maxlen         = IPV6_MAX_STRLEN,
6727                 .mode           = 0600,
6728                 .proc_handler   = addrconf_sysctl_stable_secret,
6729         },
6730         {
6731                 .procname       = "use_oif_addrs_only",
6732                 .data           = &ipv6_devconf.use_oif_addrs_only,
6733                 .maxlen         = sizeof(int),
6734                 .mode           = 0644,
6735                 .proc_handler   = proc_dointvec,
6736         },
6737         {
6738                 .procname       = "ignore_routes_with_linkdown",
6739                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6740                 .maxlen         = sizeof(int),
6741                 .mode           = 0644,
6742                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6743         },
6744         {
6745                 .procname       = "drop_unicast_in_l2_multicast",
6746                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6747                 .maxlen         = sizeof(int),
6748                 .mode           = 0644,
6749                 .proc_handler   = proc_dointvec,
6750         },
6751         {
6752                 .procname       = "drop_unsolicited_na",
6753                 .data           = &ipv6_devconf.drop_unsolicited_na,
6754                 .maxlen         = sizeof(int),
6755                 .mode           = 0644,
6756                 .proc_handler   = proc_dointvec,
6757         },
6758         {
6759                 .procname       = "keep_addr_on_down",
6760                 .data           = &ipv6_devconf.keep_addr_on_down,
6761                 .maxlen         = sizeof(int),
6762                 .mode           = 0644,
6763                 .proc_handler   = proc_dointvec,
6764
6765         },
6766         {
6767                 .procname       = "seg6_enabled",
6768                 .data           = &ipv6_devconf.seg6_enabled,
6769                 .maxlen         = sizeof(int),
6770                 .mode           = 0644,
6771                 .proc_handler   = proc_dointvec,
6772         },
6773 #ifdef CONFIG_IPV6_SEG6_HMAC
6774         {
6775                 .procname       = "seg6_require_hmac",
6776                 .data           = &ipv6_devconf.seg6_require_hmac,
6777                 .maxlen         = sizeof(int),
6778                 .mode           = 0644,
6779                 .proc_handler   = proc_dointvec,
6780         },
6781 #endif
6782         {
6783                 .procname       = "enhanced_dad",
6784                 .data           = &ipv6_devconf.enhanced_dad,
6785                 .maxlen         = sizeof(int),
6786                 .mode           = 0644,
6787                 .proc_handler   = proc_dointvec,
6788         },
6789         {
6790                 .procname               = "addr_gen_mode",
6791                 .data                   = &ipv6_devconf.addr_gen_mode,
6792                 .maxlen                 = sizeof(int),
6793                 .mode                   = 0644,
6794                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6795         },
6796         {
6797                 .procname       = "disable_policy",
6798                 .data           = &ipv6_devconf.disable_policy,
6799                 .maxlen         = sizeof(int),
6800                 .mode           = 0644,
6801                 .proc_handler   = addrconf_sysctl_disable_policy,
6802         },
6803         {
6804                 .procname       = "ndisc_tclass",
6805                 .data           = &ipv6_devconf.ndisc_tclass,
6806                 .maxlen         = sizeof(int),
6807                 .mode           = 0644,
6808                 .proc_handler   = proc_dointvec_minmax,
6809                 .extra1         = (void *)&zero,
6810                 .extra2         = (void *)&two_five_five,
6811         },
6812         {
6813                 /* sentinel */
6814         }
6815 };
6816
6817 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6818                 struct inet6_dev *idev, struct ipv6_devconf *p)
6819 {
6820         int i, ifindex;
6821         struct ctl_table *table;
6822         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6823
6824         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6825         if (!table)
6826                 goto out;
6827
6828         for (i = 0; table[i].data; i++) {
6829                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6830                 /* If one of these is already set, then it is not safe to
6831                  * overwrite either of them: this makes proc_dointvec_minmax
6832                  * usable.
6833                  */
6834                 if (!table[i].extra1 && !table[i].extra2) {
6835                         table[i].extra1 = idev; /* embedded; no ref */
6836                         table[i].extra2 = net;
6837                 }
6838         }
6839
6840         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6841
6842         p->sysctl_header = register_net_sysctl(net, path, table);
6843         if (!p->sysctl_header)
6844                 goto free;
6845
6846         if (!strcmp(dev_name, "all"))
6847                 ifindex = NETCONFA_IFINDEX_ALL;
6848         else if (!strcmp(dev_name, "default"))
6849                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6850         else
6851                 ifindex = idev->dev->ifindex;
6852         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6853                                      ifindex, p);
6854         return 0;
6855
6856 free:
6857         kfree(table);
6858 out:
6859         return -ENOBUFS;
6860 }
6861
6862 static void __addrconf_sysctl_unregister(struct net *net,
6863                                          struct ipv6_devconf *p, int ifindex)
6864 {
6865         struct ctl_table *table;
6866
6867         if (!p->sysctl_header)
6868                 return;
6869
6870         table = p->sysctl_header->ctl_table_arg;
6871         unregister_net_sysctl_table(p->sysctl_header);
6872         p->sysctl_header = NULL;
6873         kfree(table);
6874
6875         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6876 }
6877
6878 static int addrconf_sysctl_register(struct inet6_dev *idev)
6879 {
6880         int err;
6881
6882         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6883                 return -EINVAL;
6884
6885         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6886                                     &ndisc_ifinfo_sysctl_change);
6887         if (err)
6888                 return err;
6889         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6890                                          idev, &idev->cnf);
6891         if (err)
6892                 neigh_sysctl_unregister(idev->nd_parms);
6893
6894         return err;
6895 }
6896
6897 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6898 {
6899         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6900                                      idev->dev->ifindex);
6901         neigh_sysctl_unregister(idev->nd_parms);
6902 }
6903
6904
6905 #endif
6906
6907 static int __net_init addrconf_init_net(struct net *net)
6908 {
6909         int err = -ENOMEM;
6910         struct ipv6_devconf *all, *dflt;
6911
6912         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6913         if (!all)
6914                 goto err_alloc_all;
6915
6916         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6917         if (!dflt)
6918                 goto err_alloc_dflt;
6919
6920         if (IS_ENABLED(CONFIG_SYSCTL) &&
6921             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6922                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6923                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6924         }
6925
6926         /* these will be inherited by all namespaces */
6927         dflt->autoconf = ipv6_defaults.autoconf;
6928         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6929
6930         dflt->stable_secret.initialized = false;
6931         all->stable_secret.initialized = false;
6932
6933         net->ipv6.devconf_all = all;
6934         net->ipv6.devconf_dflt = dflt;
6935
6936 #ifdef CONFIG_SYSCTL
6937         err = __addrconf_sysctl_register(net, "all", NULL, all);
6938         if (err < 0)
6939                 goto err_reg_all;
6940
6941         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6942         if (err < 0)
6943                 goto err_reg_dflt;
6944 #endif
6945         return 0;
6946
6947 #ifdef CONFIG_SYSCTL
6948 err_reg_dflt:
6949         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6950 err_reg_all:
6951         kfree(dflt);
6952 #endif
6953 err_alloc_dflt:
6954         kfree(all);
6955 err_alloc_all:
6956         return err;
6957 }
6958
6959 static void __net_exit addrconf_exit_net(struct net *net)
6960 {
6961 #ifdef CONFIG_SYSCTL
6962         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6963                                      NETCONFA_IFINDEX_DEFAULT);
6964         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6965                                      NETCONFA_IFINDEX_ALL);
6966 #endif
6967         kfree(net->ipv6.devconf_dflt);
6968         kfree(net->ipv6.devconf_all);
6969 }
6970
6971 static struct pernet_operations addrconf_ops = {
6972         .init = addrconf_init_net,
6973         .exit = addrconf_exit_net,
6974 };
6975
6976 static struct rtnl_af_ops inet6_ops __read_mostly = {
6977         .family           = AF_INET6,
6978         .fill_link_af     = inet6_fill_link_af,
6979         .get_link_af_size = inet6_get_link_af_size,
6980         .validate_link_af = inet6_validate_link_af,
6981         .set_link_af      = inet6_set_link_af,
6982 };
6983
6984 /*
6985  *      Init / cleanup code
6986  */
6987
6988 int __init addrconf_init(void)
6989 {
6990         struct inet6_dev *idev;
6991         int i, err;
6992
6993         err = ipv6_addr_label_init();
6994         if (err < 0) {
6995                 pr_crit("%s: cannot initialize default policy table: %d\n",
6996                         __func__, err);
6997                 goto out;
6998         }
6999
7000         err = register_pernet_subsys(&addrconf_ops);
7001         if (err < 0)
7002                 goto out_addrlabel;
7003
7004         addrconf_wq = create_workqueue("ipv6_addrconf");
7005         if (!addrconf_wq) {
7006                 err = -ENOMEM;
7007                 goto out_nowq;
7008         }
7009
7010         /* The addrconf netdev notifier requires that loopback_dev
7011          * has it's ipv6 private information allocated and setup
7012          * before it can bring up and give link-local addresses
7013          * to other devices which are up.
7014          *
7015          * Unfortunately, loopback_dev is not necessarily the first
7016          * entry in the global dev_base list of net devices.  In fact,
7017          * it is likely to be the very last entry on that list.
7018          * So this causes the notifier registry below to try and
7019          * give link-local addresses to all devices besides loopback_dev
7020          * first, then loopback_dev, which cases all the non-loopback_dev
7021          * devices to fail to get a link-local address.
7022          *
7023          * So, as a temporary fix, allocate the ipv6 structure for
7024          * loopback_dev first by hand.
7025          * Longer term, all of the dependencies ipv6 has upon the loopback
7026          * device and it being up should be removed.
7027          */
7028         rtnl_lock();
7029         idev = ipv6_add_dev(init_net.loopback_dev);
7030         rtnl_unlock();
7031         if (IS_ERR(idev)) {
7032                 err = PTR_ERR(idev);
7033                 goto errlo;
7034         }
7035
7036         ip6_route_init_special_entries();
7037
7038         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7039                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7040
7041         register_netdevice_notifier(&ipv6_dev_notf);
7042
7043         addrconf_verify();
7044
7045         rtnl_af_register(&inet6_ops);
7046
7047         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7048                                    NULL, inet6_dump_ifinfo, 0);
7049         if (err < 0)
7050                 goto errout;
7051
7052         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7053                                    inet6_rtm_newaddr, NULL, 0);
7054         if (err < 0)
7055                 goto errout;
7056         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7057                                    inet6_rtm_deladdr, NULL, 0);
7058         if (err < 0)
7059                 goto errout;
7060         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7061                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7062                                    RTNL_FLAG_DOIT_UNLOCKED);
7063         if (err < 0)
7064                 goto errout;
7065         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7066                                    NULL, inet6_dump_ifmcaddr, 0);
7067         if (err < 0)
7068                 goto errout;
7069         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7070                                    NULL, inet6_dump_ifacaddr, 0);
7071         if (err < 0)
7072                 goto errout;
7073         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7074                                    inet6_netconf_get_devconf,
7075                                    inet6_netconf_dump_devconf,
7076                                    RTNL_FLAG_DOIT_UNLOCKED);
7077         if (err < 0)
7078                 goto errout;
7079         err = ipv6_addr_label_rtnl_register();
7080         if (err < 0)
7081                 goto errout;
7082
7083         return 0;
7084 errout:
7085         rtnl_unregister_all(PF_INET6);
7086         rtnl_af_unregister(&inet6_ops);
7087         unregister_netdevice_notifier(&ipv6_dev_notf);
7088 errlo:
7089         destroy_workqueue(addrconf_wq);
7090 out_nowq:
7091         unregister_pernet_subsys(&addrconf_ops);
7092 out_addrlabel:
7093         ipv6_addr_label_cleanup();
7094 out:
7095         return err;
7096 }
7097
7098 void addrconf_cleanup(void)
7099 {
7100         struct net_device *dev;
7101         int i;
7102
7103         unregister_netdevice_notifier(&ipv6_dev_notf);
7104         unregister_pernet_subsys(&addrconf_ops);
7105         ipv6_addr_label_cleanup();
7106
7107         rtnl_af_unregister(&inet6_ops);
7108
7109         rtnl_lock();
7110
7111         /* clean dev list */
7112         for_each_netdev(&init_net, dev) {
7113                 if (__in6_dev_get(dev) == NULL)
7114                         continue;
7115                 addrconf_ifdown(dev, 1);
7116         }
7117         addrconf_ifdown(init_net.loopback_dev, 2);
7118
7119         /*
7120          *      Check hash table.
7121          */
7122         spin_lock_bh(&addrconf_hash_lock);
7123         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7124                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7125         spin_unlock_bh(&addrconf_hash_lock);
7126         cancel_delayed_work(&addr_chk_work);
7127         rtnl_unlock();
7128
7129         destroy_workqueue(addrconf_wq);
7130 }