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                 /* prefix routes only use builtin fib6_nh */
2421                 if (rt->nh)
2422                         continue;
2423
2424                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2425                         continue;
2426                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2427                         continue;
2428                 if ((rt->fib6_flags & flags) != flags)
2429                         continue;
2430                 if ((rt->fib6_flags & noflags) != 0)
2431                         continue;
2432                 if (!fib6_info_hold_safe(rt))
2433                         continue;
2434                 break;
2435         }
2436 out:
2437         rcu_read_unlock();
2438         return rt;
2439 }
2440
2441
2442 /* Create "default" multicast route to the interface */
2443
2444 static void addrconf_add_mroute(struct net_device *dev)
2445 {
2446         struct fib6_config cfg = {
2447                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2448                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2449                 .fc_ifindex = dev->ifindex,
2450                 .fc_dst_len = 8,
2451                 .fc_flags = RTF_UP,
2452                 .fc_type = RTN_UNICAST,
2453                 .fc_nlinfo.nl_net = dev_net(dev),
2454         };
2455
2456         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2457
2458         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2459 }
2460
2461 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2462 {
2463         struct inet6_dev *idev;
2464
2465         ASSERT_RTNL();
2466
2467         idev = ipv6_find_idev(dev);
2468         if (!idev)
2469                 return ERR_PTR(-ENOBUFS);
2470
2471         if (idev->cnf.disable_ipv6)
2472                 return ERR_PTR(-EACCES);
2473
2474         /* Add default multicast route */
2475         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2476                 addrconf_add_mroute(dev);
2477
2478         return idev;
2479 }
2480
2481 static void manage_tempaddrs(struct inet6_dev *idev,
2482                              struct inet6_ifaddr *ifp,
2483                              __u32 valid_lft, __u32 prefered_lft,
2484                              bool create, unsigned long now)
2485 {
2486         u32 flags;
2487         struct inet6_ifaddr *ift;
2488
2489         read_lock_bh(&idev->lock);
2490         /* update all temporary addresses in the list */
2491         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2492                 int age, max_valid, max_prefered;
2493
2494                 if (ifp != ift->ifpub)
2495                         continue;
2496
2497                 /* RFC 4941 section 3.3:
2498                  * If a received option will extend the lifetime of a public
2499                  * address, the lifetimes of temporary addresses should
2500                  * be extended, subject to the overall constraint that no
2501                  * temporary addresses should ever remain "valid" or "preferred"
2502                  * for a time longer than (TEMP_VALID_LIFETIME) or
2503                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2504                  */
2505                 age = (now - ift->cstamp) / HZ;
2506                 max_valid = idev->cnf.temp_valid_lft - age;
2507                 if (max_valid < 0)
2508                         max_valid = 0;
2509
2510                 max_prefered = idev->cnf.temp_prefered_lft -
2511                                idev->desync_factor - age;
2512                 if (max_prefered < 0)
2513                         max_prefered = 0;
2514
2515                 if (valid_lft > max_valid)
2516                         valid_lft = max_valid;
2517
2518                 if (prefered_lft > max_prefered)
2519                         prefered_lft = max_prefered;
2520
2521                 spin_lock(&ift->lock);
2522                 flags = ift->flags;
2523                 ift->valid_lft = valid_lft;
2524                 ift->prefered_lft = prefered_lft;
2525                 ift->tstamp = now;
2526                 if (prefered_lft > 0)
2527                         ift->flags &= ~IFA_F_DEPRECATED;
2528
2529                 spin_unlock(&ift->lock);
2530                 if (!(flags&IFA_F_TENTATIVE))
2531                         ipv6_ifa_notify(0, ift);
2532         }
2533
2534         if ((create || list_empty(&idev->tempaddr_list)) &&
2535             idev->cnf.use_tempaddr > 0) {
2536                 /* When a new public address is created as described
2537                  * in [ADDRCONF], also create a new temporary address.
2538                  * Also create a temporary address if it's enabled but
2539                  * no temporary address currently exists.
2540                  */
2541                 read_unlock_bh(&idev->lock);
2542                 ipv6_create_tempaddr(ifp, NULL, false);
2543         } else {
2544                 read_unlock_bh(&idev->lock);
2545         }
2546 }
2547
2548 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2549 {
2550         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2551                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2552 }
2553
2554 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2555                                  const struct prefix_info *pinfo,
2556                                  struct inet6_dev *in6_dev,
2557                                  const struct in6_addr *addr, int addr_type,
2558                                  u32 addr_flags, bool sllao, bool tokenized,
2559                                  __u32 valid_lft, u32 prefered_lft)
2560 {
2561         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2562         int create = 0, update_lft = 0;
2563
2564         if (!ifp && valid_lft) {
2565                 int max_addresses = in6_dev->cnf.max_addresses;
2566                 struct ifa6_config cfg = {
2567                         .pfx = addr,
2568                         .plen = pinfo->prefix_len,
2569                         .ifa_flags = addr_flags,
2570                         .valid_lft = valid_lft,
2571                         .preferred_lft = prefered_lft,
2572                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2573                 };
2574
2575 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2576                 if ((net->ipv6.devconf_all->optimistic_dad ||
2577                      in6_dev->cnf.optimistic_dad) &&
2578                     !net->ipv6.devconf_all->forwarding && sllao)
2579                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2580 #endif
2581
2582                 /* Do not allow to create too much of autoconfigured
2583                  * addresses; this would be too easy way to crash kernel.
2584                  */
2585                 if (!max_addresses ||
2586                     ipv6_count_addresses(in6_dev) < max_addresses)
2587                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2588
2589                 if (IS_ERR_OR_NULL(ifp))
2590                         return -1;
2591
2592                 create = 1;
2593                 spin_lock_bh(&ifp->lock);
2594                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2595                 ifp->cstamp = jiffies;
2596                 ifp->tokenized = tokenized;
2597                 spin_unlock_bh(&ifp->lock);
2598                 addrconf_dad_start(ifp);
2599         }
2600
2601         if (ifp) {
2602                 u32 flags;
2603                 unsigned long now;
2604                 u32 stored_lft;
2605
2606                 /* update lifetime (RFC2462 5.5.3 e) */
2607                 spin_lock_bh(&ifp->lock);
2608                 now = jiffies;
2609                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2610                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2611                 else
2612                         stored_lft = 0;
2613                 if (!create && stored_lft) {
2614                         const u32 minimum_lft = min_t(u32,
2615                                 stored_lft, MIN_VALID_LIFETIME);
2616                         valid_lft = max(valid_lft, minimum_lft);
2617
2618                         /* RFC4862 Section 5.5.3e:
2619                          * "Note that the preferred lifetime of the
2620                          *  corresponding address is always reset to
2621                          *  the Preferred Lifetime in the received
2622                          *  Prefix Information option, regardless of
2623                          *  whether the valid lifetime is also reset or
2624                          *  ignored."
2625                          *
2626                          * So we should always update prefered_lft here.
2627                          */
2628                         update_lft = 1;
2629                 }
2630
2631                 if (update_lft) {
2632                         ifp->valid_lft = valid_lft;
2633                         ifp->prefered_lft = prefered_lft;
2634                         ifp->tstamp = now;
2635                         flags = ifp->flags;
2636                         ifp->flags &= ~IFA_F_DEPRECATED;
2637                         spin_unlock_bh(&ifp->lock);
2638
2639                         if (!(flags&IFA_F_TENTATIVE))
2640                                 ipv6_ifa_notify(0, ifp);
2641                 } else
2642                         spin_unlock_bh(&ifp->lock);
2643
2644                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2645                                  create, now);
2646
2647                 in6_ifa_put(ifp);
2648                 addrconf_verify();
2649         }
2650
2651         return 0;
2652 }
2653 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2654
2655 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2656 {
2657         struct prefix_info *pinfo;
2658         __u32 valid_lft;
2659         __u32 prefered_lft;
2660         int addr_type, err;
2661         u32 addr_flags = 0;
2662         struct inet6_dev *in6_dev;
2663         struct net *net = dev_net(dev);
2664
2665         pinfo = (struct prefix_info *) opt;
2666
2667         if (len < sizeof(struct prefix_info)) {
2668                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2669                 return;
2670         }
2671
2672         /*
2673          *      Validation checks ([ADDRCONF], page 19)
2674          */
2675
2676         addr_type = ipv6_addr_type(&pinfo->prefix);
2677
2678         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2679                 return;
2680
2681         valid_lft = ntohl(pinfo->valid);
2682         prefered_lft = ntohl(pinfo->prefered);
2683
2684         if (prefered_lft > valid_lft) {
2685                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2686                 return;
2687         }
2688
2689         in6_dev = in6_dev_get(dev);
2690
2691         if (!in6_dev) {
2692                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2693                                     dev->name);
2694                 return;
2695         }
2696
2697         /*
2698          *      Two things going on here:
2699          *      1) Add routes for on-link prefixes
2700          *      2) Configure prefixes with the auto flag set
2701          */
2702
2703         if (pinfo->onlink) {
2704                 struct fib6_info *rt;
2705                 unsigned long rt_expires;
2706
2707                 /* Avoid arithmetic overflow. Really, we could
2708                  * save rt_expires in seconds, likely valid_lft,
2709                  * but it would require division in fib gc, that it
2710                  * not good.
2711                  */
2712                 if (HZ > USER_HZ)
2713                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2714                 else
2715                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2716
2717                 if (addrconf_finite_timeout(rt_expires))
2718                         rt_expires *= HZ;
2719
2720                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2721                                                pinfo->prefix_len,
2722                                                dev,
2723                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2724                                                RTF_DEFAULT, true);
2725
2726                 if (rt) {
2727                         /* Autoconf prefix route */
2728                         if (valid_lft == 0) {
2729                                 ip6_del_rt(net, rt);
2730                                 rt = NULL;
2731                         } else if (addrconf_finite_timeout(rt_expires)) {
2732                                 /* not infinity */
2733                                 fib6_set_expires(rt, jiffies + rt_expires);
2734                         } else {
2735                                 fib6_clean_expires(rt);
2736                         }
2737                 } else if (valid_lft) {
2738                         clock_t expires = 0;
2739                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2740                         if (addrconf_finite_timeout(rt_expires)) {
2741                                 /* not infinity */
2742                                 flags |= RTF_EXPIRES;
2743                                 expires = jiffies_to_clock_t(rt_expires);
2744                         }
2745                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2746                                               0, dev, expires, flags,
2747                                               GFP_ATOMIC);
2748                 }
2749                 fib6_info_release(rt);
2750         }
2751
2752         /* Try to figure out our local address for this prefix */
2753
2754         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2755                 struct in6_addr addr;
2756                 bool tokenized = false, dev_addr_generated = false;
2757
2758                 if (pinfo->prefix_len == 64) {
2759                         memcpy(&addr, &pinfo->prefix, 8);
2760
2761                         if (!ipv6_addr_any(&in6_dev->token)) {
2762                                 read_lock_bh(&in6_dev->lock);
2763                                 memcpy(addr.s6_addr + 8,
2764                                        in6_dev->token.s6_addr + 8, 8);
2765                                 read_unlock_bh(&in6_dev->lock);
2766                                 tokenized = true;
2767                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2768                                    !ipv6_generate_stable_address(&addr, 0,
2769                                                                  in6_dev)) {
2770                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2771                                 goto ok;
2772                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2773                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2774                                 goto put;
2775                         } else {
2776                                 dev_addr_generated = true;
2777                         }
2778                         goto ok;
2779                 }
2780                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2781                                     pinfo->prefix_len);
2782                 goto put;
2783
2784 ok:
2785                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2786                                                    &addr, addr_type,
2787                                                    addr_flags, sllao,
2788                                                    tokenized, valid_lft,
2789                                                    prefered_lft);
2790                 if (err)
2791                         goto put;
2792
2793                 /* Ignore error case here because previous prefix add addr was
2794                  * successful which will be notified.
2795                  */
2796                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2797                                               addr_type, addr_flags, sllao,
2798                                               tokenized, valid_lft,
2799                                               prefered_lft,
2800                                               dev_addr_generated);
2801         }
2802         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2803 put:
2804         in6_dev_put(in6_dev);
2805 }
2806
2807 /*
2808  *      Set destination address.
2809  *      Special case for SIT interfaces where we create a new "virtual"
2810  *      device.
2811  */
2812 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2813 {
2814         struct in6_ifreq ireq;
2815         struct net_device *dev;
2816         int err = -EINVAL;
2817
2818         rtnl_lock();
2819
2820         err = -EFAULT;
2821         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2822                 goto err_exit;
2823
2824         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2825
2826         err = -ENODEV;
2827         if (!dev)
2828                 goto err_exit;
2829
2830 #if IS_ENABLED(CONFIG_IPV6_SIT)
2831         if (dev->type == ARPHRD_SIT) {
2832                 const struct net_device_ops *ops = dev->netdev_ops;
2833                 struct ifreq ifr;
2834                 struct ip_tunnel_parm p;
2835
2836                 err = -EADDRNOTAVAIL;
2837                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2838                         goto err_exit;
2839
2840                 memset(&p, 0, sizeof(p));
2841                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2842                 p.iph.saddr = 0;
2843                 p.iph.version = 4;
2844                 p.iph.ihl = 5;
2845                 p.iph.protocol = IPPROTO_IPV6;
2846                 p.iph.ttl = 64;
2847                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2848
2849                 if (ops->ndo_do_ioctl) {
2850                         mm_segment_t oldfs = get_fs();
2851
2852                         set_fs(KERNEL_DS);
2853                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2854                         set_fs(oldfs);
2855                 } else
2856                         err = -EOPNOTSUPP;
2857
2858                 if (err == 0) {
2859                         err = -ENOBUFS;
2860                         dev = __dev_get_by_name(net, p.name);
2861                         if (!dev)
2862                                 goto err_exit;
2863                         err = dev_open(dev, NULL);
2864                 }
2865         }
2866 #endif
2867
2868 err_exit:
2869         rtnl_unlock();
2870         return err;
2871 }
2872
2873 static int ipv6_mc_config(struct sock *sk, bool join,
2874                           const struct in6_addr *addr, int ifindex)
2875 {
2876         int ret;
2877
2878         ASSERT_RTNL();
2879
2880         lock_sock(sk);
2881         if (join)
2882                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2883         else
2884                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2885         release_sock(sk);
2886
2887         return ret;
2888 }
2889
2890 /*
2891  *      Manual configuration of address on an interface
2892  */
2893 static int inet6_addr_add(struct net *net, int ifindex,
2894                           struct ifa6_config *cfg,
2895                           struct netlink_ext_ack *extack)
2896 {
2897         struct inet6_ifaddr *ifp;
2898         struct inet6_dev *idev;
2899         struct net_device *dev;
2900         unsigned long timeout;
2901         clock_t expires;
2902         u32 flags;
2903
2904         ASSERT_RTNL();
2905
2906         if (cfg->plen > 128)
2907                 return -EINVAL;
2908
2909         /* check the lifetime */
2910         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2911                 return -EINVAL;
2912
2913         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2914                 return -EINVAL;
2915
2916         dev = __dev_get_by_index(net, ifindex);
2917         if (!dev)
2918                 return -ENODEV;
2919
2920         idev = addrconf_add_dev(dev);
2921         if (IS_ERR(idev))
2922                 return PTR_ERR(idev);
2923
2924         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2925                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2926                                          true, cfg->pfx, ifindex);
2927
2928                 if (ret < 0)
2929                         return ret;
2930         }
2931
2932         cfg->scope = ipv6_addr_scope(cfg->pfx);
2933
2934         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2935         if (addrconf_finite_timeout(timeout)) {
2936                 expires = jiffies_to_clock_t(timeout * HZ);
2937                 cfg->valid_lft = timeout;
2938                 flags = RTF_EXPIRES;
2939         } else {
2940                 expires = 0;
2941                 flags = 0;
2942                 cfg->ifa_flags |= IFA_F_PERMANENT;
2943         }
2944
2945         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2946         if (addrconf_finite_timeout(timeout)) {
2947                 if (timeout == 0)
2948                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2949                 cfg->preferred_lft = timeout;
2950         }
2951
2952         ifp = ipv6_add_addr(idev, cfg, true, extack);
2953         if (!IS_ERR(ifp)) {
2954                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2955                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2956                                               ifp->rt_priority, dev, expires,
2957                                               flags, GFP_KERNEL);
2958                 }
2959
2960                 /* Send a netlink notification if DAD is enabled and
2961                  * optimistic flag is not set
2962                  */
2963                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2964                         ipv6_ifa_notify(0, ifp);
2965                 /*
2966                  * Note that section 3.1 of RFC 4429 indicates
2967                  * that the Optimistic flag should not be set for
2968                  * manually configured addresses
2969                  */
2970                 addrconf_dad_start(ifp);
2971                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2972                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2973                                          cfg->preferred_lft, true, jiffies);
2974                 in6_ifa_put(ifp);
2975                 addrconf_verify_rtnl();
2976                 return 0;
2977         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2978                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2979                                cfg->pfx, ifindex);
2980         }
2981
2982         return PTR_ERR(ifp);
2983 }
2984
2985 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2986                           const struct in6_addr *pfx, unsigned int plen)
2987 {
2988         struct inet6_ifaddr *ifp;
2989         struct inet6_dev *idev;
2990         struct net_device *dev;
2991
2992         if (plen > 128)
2993                 return -EINVAL;
2994
2995         dev = __dev_get_by_index(net, ifindex);
2996         if (!dev)
2997                 return -ENODEV;
2998
2999         idev = __in6_dev_get(dev);
3000         if (!idev)
3001                 return -ENXIO;
3002
3003         read_lock_bh(&idev->lock);
3004         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3005                 if (ifp->prefix_len == plen &&
3006                     ipv6_addr_equal(pfx, &ifp->addr)) {
3007                         in6_ifa_hold(ifp);
3008                         read_unlock_bh(&idev->lock);
3009
3010                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3011                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3012                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3013                                                  jiffies);
3014                         ipv6_del_addr(ifp);
3015                         addrconf_verify_rtnl();
3016                         if (ipv6_addr_is_multicast(pfx)) {
3017                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3018                                                false, pfx, dev->ifindex);
3019                         }
3020                         return 0;
3021                 }
3022         }
3023         read_unlock_bh(&idev->lock);
3024         return -EADDRNOTAVAIL;
3025 }
3026
3027
3028 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3029 {
3030         struct ifa6_config cfg = {
3031                 .ifa_flags = IFA_F_PERMANENT,
3032                 .preferred_lft = INFINITY_LIFE_TIME,
3033                 .valid_lft = INFINITY_LIFE_TIME,
3034         };
3035         struct in6_ifreq ireq;
3036         int err;
3037
3038         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3039                 return -EPERM;
3040
3041         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3042                 return -EFAULT;
3043
3044         cfg.pfx = &ireq.ifr6_addr;
3045         cfg.plen = ireq.ifr6_prefixlen;
3046
3047         rtnl_lock();
3048         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3049         rtnl_unlock();
3050         return err;
3051 }
3052
3053 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3054 {
3055         struct in6_ifreq ireq;
3056         int err;
3057
3058         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3059                 return -EPERM;
3060
3061         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3062                 return -EFAULT;
3063
3064         rtnl_lock();
3065         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3066                              ireq.ifr6_prefixlen);
3067         rtnl_unlock();
3068         return err;
3069 }
3070
3071 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3072                      int plen, int scope)
3073 {
3074         struct inet6_ifaddr *ifp;
3075         struct ifa6_config cfg = {
3076                 .pfx = addr,
3077                 .plen = plen,
3078                 .ifa_flags = IFA_F_PERMANENT,
3079                 .valid_lft = INFINITY_LIFE_TIME,
3080                 .preferred_lft = INFINITY_LIFE_TIME,
3081                 .scope = scope
3082         };
3083
3084         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3085         if (!IS_ERR(ifp)) {
3086                 spin_lock_bh(&ifp->lock);
3087                 ifp->flags &= ~IFA_F_TENTATIVE;
3088                 spin_unlock_bh(&ifp->lock);
3089                 rt_genid_bump_ipv6(dev_net(idev->dev));
3090                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3091                 in6_ifa_put(ifp);
3092         }
3093 }
3094
3095 #if IS_ENABLED(CONFIG_IPV6_SIT)
3096 static void sit_add_v4_addrs(struct inet6_dev *idev)
3097 {
3098         struct in6_addr addr;
3099         struct net_device *dev;
3100         struct net *net = dev_net(idev->dev);
3101         int scope, plen;
3102         u32 pflags = 0;
3103
3104         ASSERT_RTNL();
3105
3106         memset(&addr, 0, sizeof(struct in6_addr));
3107         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3108
3109         if (idev->dev->flags&IFF_POINTOPOINT) {
3110                 addr.s6_addr32[0] = htonl(0xfe800000);
3111                 scope = IFA_LINK;
3112                 plen = 64;
3113         } else {
3114                 scope = IPV6_ADDR_COMPATv4;
3115                 plen = 96;
3116                 pflags |= RTF_NONEXTHOP;
3117         }
3118
3119         if (addr.s6_addr32[3]) {
3120                 add_addr(idev, &addr, plen, scope);
3121                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3122                                       GFP_KERNEL);
3123                 return;
3124         }
3125
3126         for_each_netdev(net, dev) {
3127                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3128                 if (in_dev && (dev->flags & IFF_UP)) {
3129                         struct in_ifaddr *ifa;
3130                         int flag = scope;
3131
3132                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3133                                 addr.s6_addr32[3] = ifa->ifa_local;
3134
3135                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3136                                         continue;
3137                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3138                                         if (idev->dev->flags&IFF_POINTOPOINT)
3139                                                 continue;
3140                                         flag |= IFA_HOST;
3141                                 }
3142
3143                                 add_addr(idev, &addr, plen, flag);
3144                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3145                                                       0, pflags, GFP_KERNEL);
3146                         }
3147                 }
3148         }
3149 }
3150 #endif
3151
3152 static void init_loopback(struct net_device *dev)
3153 {
3154         struct inet6_dev  *idev;
3155
3156         /* ::1 */
3157
3158         ASSERT_RTNL();
3159
3160         idev = ipv6_find_idev(dev);
3161         if (!idev) {
3162                 pr_debug("%s: add_dev failed\n", __func__);
3163                 return;
3164         }
3165
3166         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3167 }
3168
3169 void addrconf_add_linklocal(struct inet6_dev *idev,
3170                             const struct in6_addr *addr, u32 flags)
3171 {
3172         struct ifa6_config cfg = {
3173                 .pfx = addr,
3174                 .plen = 64,
3175                 .ifa_flags = flags | IFA_F_PERMANENT,
3176                 .valid_lft = INFINITY_LIFE_TIME,
3177                 .preferred_lft = INFINITY_LIFE_TIME,
3178                 .scope = IFA_LINK
3179         };
3180         struct inet6_ifaddr *ifp;
3181
3182 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3183         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3184              idev->cnf.optimistic_dad) &&
3185             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3186                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3187 #endif
3188
3189         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3190         if (!IS_ERR(ifp)) {
3191                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3192                                       0, 0, GFP_ATOMIC);
3193                 addrconf_dad_start(ifp);
3194                 in6_ifa_put(ifp);
3195         }
3196 }
3197 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3198
3199 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3200 {
3201         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3202                 return true;
3203
3204         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3205             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3206                 return true;
3207
3208         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3209             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3210                 return true;
3211
3212         return false;
3213 }
3214
3215 static int ipv6_generate_stable_address(struct in6_addr *address,
3216                                         u8 dad_count,
3217                                         const struct inet6_dev *idev)
3218 {
3219         static DEFINE_SPINLOCK(lock);
3220         static __u32 digest[SHA_DIGEST_WORDS];
3221         static __u32 workspace[SHA_WORKSPACE_WORDS];
3222
3223         static union {
3224                 char __data[SHA_MESSAGE_BYTES];
3225                 struct {
3226                         struct in6_addr secret;
3227                         __be32 prefix[2];
3228                         unsigned char hwaddr[MAX_ADDR_LEN];
3229                         u8 dad_count;
3230                 } __packed;
3231         } data;
3232
3233         struct in6_addr secret;
3234         struct in6_addr temp;
3235         struct net *net = dev_net(idev->dev);
3236
3237         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3238
3239         if (idev->cnf.stable_secret.initialized)
3240                 secret = idev->cnf.stable_secret.secret;
3241         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3242                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3243         else
3244                 return -1;
3245
3246 retry:
3247         spin_lock_bh(&lock);
3248
3249         sha_init(digest);
3250         memset(&data, 0, sizeof(data));
3251         memset(workspace, 0, sizeof(workspace));
3252         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3253         data.prefix[0] = address->s6_addr32[0];
3254         data.prefix[1] = address->s6_addr32[1];
3255         data.secret = secret;
3256         data.dad_count = dad_count;
3257
3258         sha_transform(digest, data.__data, workspace);
3259
3260         temp = *address;
3261         temp.s6_addr32[2] = (__force __be32)digest[0];
3262         temp.s6_addr32[3] = (__force __be32)digest[1];
3263
3264         spin_unlock_bh(&lock);
3265
3266         if (ipv6_reserved_interfaceid(temp)) {
3267                 dad_count++;
3268                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3269                         return -1;
3270                 goto retry;
3271         }
3272
3273         *address = temp;
3274         return 0;
3275 }
3276
3277 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3278 {
3279         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3280
3281         if (s->initialized)
3282                 return;
3283         s = &idev->cnf.stable_secret;
3284         get_random_bytes(&s->secret, sizeof(s->secret));
3285         s->initialized = true;
3286 }
3287
3288 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3289 {
3290         struct in6_addr addr;
3291
3292         /* no link local addresses on L3 master devices */
3293         if (netif_is_l3_master(idev->dev))
3294                 return;
3295
3296         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3297
3298         switch (idev->cnf.addr_gen_mode) {
3299         case IN6_ADDR_GEN_MODE_RANDOM:
3300                 ipv6_gen_mode_random_init(idev);
3301                 /* fallthrough */
3302         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3303                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3304                         addrconf_add_linklocal(idev, &addr,
3305                                                IFA_F_STABLE_PRIVACY);
3306                 else if (prefix_route)
3307                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3308                                               0, 0, GFP_KERNEL);
3309                 break;
3310         case IN6_ADDR_GEN_MODE_EUI64:
3311                 /* addrconf_add_linklocal also adds a prefix_route and we
3312                  * only need to care about prefix routes if ipv6_generate_eui64
3313                  * couldn't generate one.
3314                  */
3315                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3316                         addrconf_add_linklocal(idev, &addr, 0);
3317                 else if (prefix_route)
3318                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3319                                               0, 0, GFP_KERNEL);
3320                 break;
3321         case IN6_ADDR_GEN_MODE_NONE:
3322         default:
3323                 /* will not add any link local address */
3324                 break;
3325         }
3326 }
3327
3328 static void addrconf_dev_config(struct net_device *dev)
3329 {
3330         struct inet6_dev *idev;
3331
3332         ASSERT_RTNL();
3333
3334         if ((dev->type != ARPHRD_ETHER) &&
3335             (dev->type != ARPHRD_FDDI) &&
3336             (dev->type != ARPHRD_ARCNET) &&
3337             (dev->type != ARPHRD_INFINIBAND) &&
3338             (dev->type != ARPHRD_IEEE1394) &&
3339             (dev->type != ARPHRD_TUNNEL6) &&
3340             (dev->type != ARPHRD_6LOWPAN) &&
3341             (dev->type != ARPHRD_IP6GRE) &&
3342             (dev->type != ARPHRD_IPGRE) &&
3343             (dev->type != ARPHRD_TUNNEL) &&
3344             (dev->type != ARPHRD_NONE) &&
3345             (dev->type != ARPHRD_RAWIP)) {
3346                 /* Alas, we support only Ethernet autoconfiguration. */
3347                 return;
3348         }
3349
3350         idev = addrconf_add_dev(dev);
3351         if (IS_ERR(idev))
3352                 return;
3353
3354         /* this device type has no EUI support */
3355         if (dev->type == ARPHRD_NONE &&
3356             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3357                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3358
3359         addrconf_addr_gen(idev, false);
3360 }
3361
3362 #if IS_ENABLED(CONFIG_IPV6_SIT)
3363 static void addrconf_sit_config(struct net_device *dev)
3364 {
3365         struct inet6_dev *idev;
3366
3367         ASSERT_RTNL();
3368
3369         /*
3370          * Configure the tunnel with one of our IPv4
3371          * addresses... we should configure all of
3372          * our v4 addrs in the tunnel
3373          */
3374
3375         idev = ipv6_find_idev(dev);
3376         if (!idev) {
3377                 pr_debug("%s: add_dev failed\n", __func__);
3378                 return;
3379         }
3380
3381         if (dev->priv_flags & IFF_ISATAP) {
3382                 addrconf_addr_gen(idev, false);
3383                 return;
3384         }
3385
3386         sit_add_v4_addrs(idev);
3387
3388         if (dev->flags&IFF_POINTOPOINT)
3389                 addrconf_add_mroute(dev);
3390 }
3391 #endif
3392
3393 #if IS_ENABLED(CONFIG_NET_IPGRE)
3394 static void addrconf_gre_config(struct net_device *dev)
3395 {
3396         struct inet6_dev *idev;
3397
3398         ASSERT_RTNL();
3399
3400         idev = ipv6_find_idev(dev);
3401         if (!idev) {
3402                 pr_debug("%s: add_dev failed\n", __func__);
3403                 return;
3404         }
3405
3406         addrconf_addr_gen(idev, true);
3407         if (dev->flags & IFF_POINTOPOINT)
3408                 addrconf_add_mroute(dev);
3409 }
3410 #endif
3411
3412 static int fixup_permanent_addr(struct net *net,
3413                                 struct inet6_dev *idev,
3414                                 struct inet6_ifaddr *ifp)
3415 {
3416         /* !fib6_node means the host route was removed from the
3417          * FIB, for example, if 'lo' device is taken down. In that
3418          * case regenerate the host route.
3419          */
3420         if (!ifp->rt || !ifp->rt->fib6_node) {
3421                 struct fib6_info *f6i, *prev;
3422
3423                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3424                                          GFP_ATOMIC);
3425                 if (IS_ERR(f6i))
3426                         return PTR_ERR(f6i);
3427
3428                 /* ifp->rt can be accessed outside of rtnl */
3429                 spin_lock(&ifp->lock);
3430                 prev = ifp->rt;
3431                 ifp->rt = f6i;
3432                 spin_unlock(&ifp->lock);
3433
3434                 fib6_info_release(prev);
3435         }
3436
3437         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3438                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3439                                       ifp->rt_priority, idev->dev, 0, 0,
3440                                       GFP_ATOMIC);
3441         }
3442
3443         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3444                 addrconf_dad_start(ifp);
3445
3446         return 0;
3447 }
3448
3449 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3450 {
3451         struct inet6_ifaddr *ifp, *tmp;
3452         struct inet6_dev *idev;
3453
3454         idev = __in6_dev_get(dev);
3455         if (!idev)
3456                 return;
3457
3458         write_lock_bh(&idev->lock);
3459
3460         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3461                 if ((ifp->flags & IFA_F_PERMANENT) &&
3462                     fixup_permanent_addr(net, idev, ifp) < 0) {
3463                         write_unlock_bh(&idev->lock);
3464                         in6_ifa_hold(ifp);
3465                         ipv6_del_addr(ifp);
3466                         write_lock_bh(&idev->lock);
3467
3468                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3469                                              idev->dev->name, &ifp->addr);
3470                 }
3471         }
3472
3473         write_unlock_bh(&idev->lock);
3474 }
3475
3476 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3477                            void *ptr)
3478 {
3479         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3480         struct netdev_notifier_change_info *change_info;
3481         struct netdev_notifier_changeupper_info *info;
3482         struct inet6_dev *idev = __in6_dev_get(dev);
3483         struct net *net = dev_net(dev);
3484         int run_pending = 0;
3485         int err;
3486
3487         switch (event) {
3488         case NETDEV_REGISTER:
3489                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3490                         idev = ipv6_add_dev(dev);
3491                         if (IS_ERR(idev))
3492                                 return notifier_from_errno(PTR_ERR(idev));
3493                 }
3494                 break;
3495
3496         case NETDEV_CHANGEMTU:
3497                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3498                 if (dev->mtu < IPV6_MIN_MTU) {
3499                         addrconf_ifdown(dev, dev != net->loopback_dev);
3500                         break;
3501                 }
3502
3503                 if (idev) {
3504                         rt6_mtu_change(dev, dev->mtu);
3505                         idev->cnf.mtu6 = dev->mtu;
3506                         break;
3507                 }
3508
3509                 /* allocate new idev */
3510                 idev = ipv6_add_dev(dev);
3511                 if (IS_ERR(idev))
3512                         break;
3513
3514                 /* device is still not ready */
3515                 if (!(idev->if_flags & IF_READY))
3516                         break;
3517
3518                 run_pending = 1;
3519
3520                 /* fall through */
3521
3522         case NETDEV_UP:
3523         case NETDEV_CHANGE:
3524                 if (dev->flags & IFF_SLAVE)
3525                         break;
3526
3527                 if (idev && idev->cnf.disable_ipv6)
3528                         break;
3529
3530                 if (event == NETDEV_UP) {
3531                         /* restore routes for permanent addresses */
3532                         addrconf_permanent_addr(net, dev);
3533
3534                         if (!addrconf_link_ready(dev)) {
3535                                 /* device is not ready yet. */
3536                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3537                                          dev->name);
3538                                 break;
3539                         }
3540
3541                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3542                                 idev = ipv6_add_dev(dev);
3543
3544                         if (!IS_ERR_OR_NULL(idev)) {
3545                                 idev->if_flags |= IF_READY;
3546                                 run_pending = 1;
3547                         }
3548                 } else if (event == NETDEV_CHANGE) {
3549                         if (!addrconf_link_ready(dev)) {
3550                                 /* device is still not ready. */
3551                                 rt6_sync_down_dev(dev, event);
3552                                 break;
3553                         }
3554
3555                         if (!IS_ERR_OR_NULL(idev)) {
3556                                 if (idev->if_flags & IF_READY) {
3557                                         /* device is already configured -
3558                                          * but resend MLD reports, we might
3559                                          * have roamed and need to update
3560                                          * multicast snooping switches
3561                                          */
3562                                         ipv6_mc_up(idev);
3563                                         change_info = ptr;
3564                                         if (change_info->flags_changed & IFF_NOARP)
3565                                                 addrconf_dad_run(idev, true);
3566                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3567                                         break;
3568                                 }
3569                                 idev->if_flags |= IF_READY;
3570                         }
3571
3572                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3573                                 dev->name);
3574
3575                         run_pending = 1;
3576                 }
3577
3578                 switch (dev->type) {
3579 #if IS_ENABLED(CONFIG_IPV6_SIT)
3580                 case ARPHRD_SIT:
3581                         addrconf_sit_config(dev);
3582                         break;
3583 #endif
3584 #if IS_ENABLED(CONFIG_NET_IPGRE)
3585                 case ARPHRD_IPGRE:
3586                         addrconf_gre_config(dev);
3587                         break;
3588 #endif
3589                 case ARPHRD_LOOPBACK:
3590                         init_loopback(dev);
3591                         break;
3592
3593                 default:
3594                         addrconf_dev_config(dev);
3595                         break;
3596                 }
3597
3598                 if (!IS_ERR_OR_NULL(idev)) {
3599                         if (run_pending)
3600                                 addrconf_dad_run(idev, false);
3601
3602                         /* Device has an address by now */
3603                         rt6_sync_up(dev, RTNH_F_DEAD);
3604
3605                         /*
3606                          * If the MTU changed during the interface down,
3607                          * when the interface up, the changed MTU must be
3608                          * reflected in the idev as well as routers.
3609                          */
3610                         if (idev->cnf.mtu6 != dev->mtu &&
3611                             dev->mtu >= IPV6_MIN_MTU) {
3612                                 rt6_mtu_change(dev, dev->mtu);
3613                                 idev->cnf.mtu6 = dev->mtu;
3614                         }
3615                         idev->tstamp = jiffies;
3616                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3617
3618                         /*
3619                          * If the changed mtu during down is lower than
3620                          * IPV6_MIN_MTU stop IPv6 on this interface.
3621                          */
3622                         if (dev->mtu < IPV6_MIN_MTU)
3623                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3624                 }
3625                 break;
3626
3627         case NETDEV_DOWN:
3628         case NETDEV_UNREGISTER:
3629                 /*
3630                  *      Remove all addresses from this interface.
3631                  */
3632                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3633                 break;
3634
3635         case NETDEV_CHANGENAME:
3636                 if (idev) {
3637                         snmp6_unregister_dev(idev);
3638                         addrconf_sysctl_unregister(idev);
3639                         err = addrconf_sysctl_register(idev);
3640                         if (err)
3641                                 return notifier_from_errno(err);
3642                         err = snmp6_register_dev(idev);
3643                         if (err) {
3644                                 addrconf_sysctl_unregister(idev);
3645                                 return notifier_from_errno(err);
3646                         }
3647                 }
3648                 break;
3649
3650         case NETDEV_PRE_TYPE_CHANGE:
3651         case NETDEV_POST_TYPE_CHANGE:
3652                 if (idev)
3653                         addrconf_type_change(dev, event);
3654                 break;
3655
3656         case NETDEV_CHANGEUPPER:
3657                 info = ptr;
3658
3659                 /* flush all routes if dev is linked to or unlinked from
3660                  * an L3 master device (e.g., VRF)
3661                  */
3662                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3663                         addrconf_ifdown(dev, 0);
3664         }
3665
3666         return NOTIFY_OK;
3667 }
3668
3669 /*
3670  *      addrconf module should be notified of a device going up
3671  */
3672 static struct notifier_block ipv6_dev_notf = {
3673         .notifier_call = addrconf_notify,
3674         .priority = ADDRCONF_NOTIFY_PRIORITY,
3675 };
3676
3677 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3678 {
3679         struct inet6_dev *idev;
3680         ASSERT_RTNL();
3681
3682         idev = __in6_dev_get(dev);
3683
3684         if (event == NETDEV_POST_TYPE_CHANGE)
3685                 ipv6_mc_remap(idev);
3686         else if (event == NETDEV_PRE_TYPE_CHANGE)
3687                 ipv6_mc_unmap(idev);
3688 }
3689
3690 static bool addr_is_local(const struct in6_addr *addr)
3691 {
3692         return ipv6_addr_type(addr) &
3693                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3694 }
3695
3696 static int addrconf_ifdown(struct net_device *dev, int how)
3697 {
3698         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3699         struct net *net = dev_net(dev);
3700         struct inet6_dev *idev;
3701         struct inet6_ifaddr *ifa, *tmp;
3702         bool keep_addr = false;
3703         int state, i;
3704
3705         ASSERT_RTNL();
3706
3707         rt6_disable_ip(dev, event);
3708
3709         idev = __in6_dev_get(dev);
3710         if (!idev)
3711                 return -ENODEV;
3712
3713         /*
3714          * Step 1: remove reference to ipv6 device from parent device.
3715          *         Do not dev_put!
3716          */
3717         if (how) {
3718                 idev->dead = 1;
3719
3720                 /* protected by rtnl_lock */
3721                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3722
3723                 /* Step 1.5: remove snmp6 entry */
3724                 snmp6_unregister_dev(idev);
3725
3726         }
3727
3728         /* combine the user config with event to determine if permanent
3729          * addresses are to be removed from address hash table
3730          */
3731         if (!how && !idev->cnf.disable_ipv6) {
3732                 /* aggregate the system setting and interface setting */
3733                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3734
3735                 if (!_keep_addr)
3736                         _keep_addr = idev->cnf.keep_addr_on_down;
3737
3738                 keep_addr = (_keep_addr > 0);
3739         }
3740
3741         /* Step 2: clear hash table */
3742         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3743                 struct hlist_head *h = &inet6_addr_lst[i];
3744
3745                 spin_lock_bh(&addrconf_hash_lock);
3746 restart:
3747                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3748                         if (ifa->idev == idev) {
3749                                 addrconf_del_dad_work(ifa);
3750                                 /* combined flag + permanent flag decide if
3751                                  * address is retained on a down event
3752                                  */
3753                                 if (!keep_addr ||
3754                                     !(ifa->flags & IFA_F_PERMANENT) ||
3755                                     addr_is_local(&ifa->addr)) {
3756                                         hlist_del_init_rcu(&ifa->addr_lst);
3757                                         goto restart;
3758                                 }
3759                         }
3760                 }
3761                 spin_unlock_bh(&addrconf_hash_lock);
3762         }
3763
3764         write_lock_bh(&idev->lock);
3765
3766         addrconf_del_rs_timer(idev);
3767
3768         /* Step 2: clear flags for stateless addrconf */
3769         if (!how)
3770                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3771
3772         /* Step 3: clear tempaddr list */
3773         while (!list_empty(&idev->tempaddr_list)) {
3774                 ifa = list_first_entry(&idev->tempaddr_list,
3775                                        struct inet6_ifaddr, tmp_list);
3776                 list_del(&ifa->tmp_list);
3777                 write_unlock_bh(&idev->lock);
3778                 spin_lock_bh(&ifa->lock);
3779
3780                 if (ifa->ifpub) {
3781                         in6_ifa_put(ifa->ifpub);
3782                         ifa->ifpub = NULL;
3783                 }
3784                 spin_unlock_bh(&ifa->lock);
3785                 in6_ifa_put(ifa);
3786                 write_lock_bh(&idev->lock);
3787         }
3788
3789         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3790                 struct fib6_info *rt = NULL;
3791                 bool keep;
3792
3793                 addrconf_del_dad_work(ifa);
3794
3795                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3796                         !addr_is_local(&ifa->addr);
3797
3798                 write_unlock_bh(&idev->lock);
3799                 spin_lock_bh(&ifa->lock);
3800
3801                 if (keep) {
3802                         /* set state to skip the notifier below */
3803                         state = INET6_IFADDR_STATE_DEAD;
3804                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3805                         if (!(ifa->flags & IFA_F_NODAD))
3806                                 ifa->flags |= IFA_F_TENTATIVE;
3807
3808                         rt = ifa->rt;
3809                         ifa->rt = NULL;
3810                 } else {
3811                         state = ifa->state;
3812                         ifa->state = INET6_IFADDR_STATE_DEAD;
3813                 }
3814
3815                 spin_unlock_bh(&ifa->lock);
3816
3817                 if (rt)
3818                         ip6_del_rt(net, rt);
3819
3820                 if (state != INET6_IFADDR_STATE_DEAD) {
3821                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3822                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3823                 } else {
3824                         if (idev->cnf.forwarding)
3825                                 addrconf_leave_anycast(ifa);
3826                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3827                 }
3828
3829                 write_lock_bh(&idev->lock);
3830                 if (!keep) {
3831                         list_del_rcu(&ifa->if_list);
3832                         in6_ifa_put(ifa);
3833                 }
3834         }
3835
3836         write_unlock_bh(&idev->lock);
3837
3838         /* Step 5: Discard anycast and multicast list */
3839         if (how) {
3840                 ipv6_ac_destroy_dev(idev);
3841                 ipv6_mc_destroy_dev(idev);
3842         } else {
3843                 ipv6_mc_down(idev);
3844         }
3845
3846         idev->tstamp = jiffies;
3847
3848         /* Last: Shot the device (if unregistered) */
3849         if (how) {
3850                 addrconf_sysctl_unregister(idev);
3851                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3852                 neigh_ifdown(&nd_tbl, dev);
3853                 in6_dev_put(idev);
3854         }
3855         return 0;
3856 }
3857
3858 static void addrconf_rs_timer(struct timer_list *t)
3859 {
3860         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3861         struct net_device *dev = idev->dev;
3862         struct in6_addr lladdr;
3863
3864         write_lock(&idev->lock);
3865         if (idev->dead || !(idev->if_flags & IF_READY))
3866                 goto out;
3867
3868         if (!ipv6_accept_ra(idev))
3869                 goto out;
3870
3871         /* Announcement received after solicitation was sent */
3872         if (idev->if_flags & IF_RA_RCVD)
3873                 goto out;
3874
3875         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3876                 write_unlock(&idev->lock);
3877                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3878                         ndisc_send_rs(dev, &lladdr,
3879                                       &in6addr_linklocal_allrouters);
3880                 else
3881                         goto put;
3882
3883                 write_lock(&idev->lock);
3884                 idev->rs_interval = rfc3315_s14_backoff_update(
3885                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3886                 /* The wait after the last probe can be shorter */
3887                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3888                                              idev->cnf.rtr_solicits) ?
3889                                       idev->cnf.rtr_solicit_delay :
3890                                       idev->rs_interval);
3891         } else {
3892                 /*
3893                  * Note: we do not support deprecated "all on-link"
3894                  * assumption any longer.
3895                  */
3896                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3897         }
3898
3899 out:
3900         write_unlock(&idev->lock);
3901 put:
3902         in6_dev_put(idev);
3903 }
3904
3905 /*
3906  *      Duplicate Address Detection
3907  */
3908 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3909 {
3910         unsigned long rand_num;
3911         struct inet6_dev *idev = ifp->idev;
3912         u64 nonce;
3913
3914         if (ifp->flags & IFA_F_OPTIMISTIC)
3915                 rand_num = 0;
3916         else
3917                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3918
3919         nonce = 0;
3920         if (idev->cnf.enhanced_dad ||
3921             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3922                 do
3923                         get_random_bytes(&nonce, 6);
3924                 while (nonce == 0);
3925         }
3926         ifp->dad_nonce = nonce;
3927         ifp->dad_probes = idev->cnf.dad_transmits;
3928         addrconf_mod_dad_work(ifp, rand_num);
3929 }
3930
3931 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3932 {
3933         struct inet6_dev *idev = ifp->idev;
3934         struct net_device *dev = idev->dev;
3935         bool bump_id, notify = false;
3936         struct net *net;
3937
3938         addrconf_join_solict(dev, &ifp->addr);
3939
3940         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3941
3942         read_lock_bh(&idev->lock);
3943         spin_lock(&ifp->lock);
3944         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3945                 goto out;
3946
3947         net = dev_net(dev);
3948         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3949             (net->ipv6.devconf_all->accept_dad < 1 &&
3950              idev->cnf.accept_dad < 1) ||
3951             !(ifp->flags&IFA_F_TENTATIVE) ||
3952             ifp->flags & IFA_F_NODAD) {
3953                 bool send_na = false;
3954
3955                 if (ifp->flags & IFA_F_TENTATIVE &&
3956                     !(ifp->flags & IFA_F_OPTIMISTIC))
3957                         send_na = true;
3958                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3959                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3960                 spin_unlock(&ifp->lock);
3961                 read_unlock_bh(&idev->lock);
3962
3963                 addrconf_dad_completed(ifp, bump_id, send_na);
3964                 return;
3965         }
3966
3967         if (!(idev->if_flags & IF_READY)) {
3968                 spin_unlock(&ifp->lock);
3969                 read_unlock_bh(&idev->lock);
3970                 /*
3971                  * If the device is not ready:
3972                  * - keep it tentative if it is a permanent address.
3973                  * - otherwise, kill it.
3974                  */
3975                 in6_ifa_hold(ifp);
3976                 addrconf_dad_stop(ifp, 0);
3977                 return;
3978         }
3979
3980         /*
3981          * Optimistic nodes can start receiving
3982          * Frames right away
3983          */
3984         if (ifp->flags & IFA_F_OPTIMISTIC) {
3985                 ip6_ins_rt(net, ifp->rt);
3986                 if (ipv6_use_optimistic_addr(net, idev)) {
3987                         /* Because optimistic nodes can use this address,
3988                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3989                          */
3990                         notify = true;
3991                 }
3992         }
3993
3994         addrconf_dad_kick(ifp);
3995 out:
3996         spin_unlock(&ifp->lock);
3997         read_unlock_bh(&idev->lock);
3998         if (notify)
3999                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4000 }
4001
4002 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4003 {
4004         bool begin_dad = false;
4005
4006         spin_lock_bh(&ifp->lock);
4007         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4008                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4009                 begin_dad = true;
4010         }
4011         spin_unlock_bh(&ifp->lock);
4012
4013         if (begin_dad)
4014                 addrconf_mod_dad_work(ifp, 0);
4015 }
4016
4017 static void addrconf_dad_work(struct work_struct *w)
4018 {
4019         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4020                                                 struct inet6_ifaddr,
4021                                                 dad_work);
4022         struct inet6_dev *idev = ifp->idev;
4023         bool bump_id, disable_ipv6 = false;
4024         struct in6_addr mcaddr;
4025
4026         enum {
4027                 DAD_PROCESS,
4028                 DAD_BEGIN,
4029                 DAD_ABORT,
4030         } action = DAD_PROCESS;
4031
4032         rtnl_lock();
4033
4034         spin_lock_bh(&ifp->lock);
4035         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4036                 action = DAD_BEGIN;
4037                 ifp->state = INET6_IFADDR_STATE_DAD;
4038         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4039                 action = DAD_ABORT;
4040                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4041
4042                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4043                      idev->cnf.accept_dad > 1) &&
4044                     !idev->cnf.disable_ipv6 &&
4045                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4046                         struct in6_addr addr;
4047
4048                         addr.s6_addr32[0] = htonl(0xfe800000);
4049                         addr.s6_addr32[1] = 0;
4050
4051                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4052                             ipv6_addr_equal(&ifp->addr, &addr)) {
4053                                 /* DAD failed for link-local based on MAC */
4054                                 idev->cnf.disable_ipv6 = 1;
4055
4056                                 pr_info("%s: IPv6 being disabled!\n",
4057                                         ifp->idev->dev->name);
4058                                 disable_ipv6 = true;
4059                         }
4060                 }
4061         }
4062         spin_unlock_bh(&ifp->lock);
4063
4064         if (action == DAD_BEGIN) {
4065                 addrconf_dad_begin(ifp);
4066                 goto out;
4067         } else if (action == DAD_ABORT) {
4068                 in6_ifa_hold(ifp);
4069                 addrconf_dad_stop(ifp, 1);
4070                 if (disable_ipv6)
4071                         addrconf_ifdown(idev->dev, 0);
4072                 goto out;
4073         }
4074
4075         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4076                 goto out;
4077
4078         write_lock_bh(&idev->lock);
4079         if (idev->dead || !(idev->if_flags & IF_READY)) {
4080                 write_unlock_bh(&idev->lock);
4081                 goto out;
4082         }
4083
4084         spin_lock(&ifp->lock);
4085         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4086                 spin_unlock(&ifp->lock);
4087                 write_unlock_bh(&idev->lock);
4088                 goto out;
4089         }
4090
4091         if (ifp->dad_probes == 0) {
4092                 bool send_na = false;
4093
4094                 /*
4095                  * DAD was successful
4096                  */
4097
4098                 if (ifp->flags & IFA_F_TENTATIVE &&
4099                     !(ifp->flags & IFA_F_OPTIMISTIC))
4100                         send_na = true;
4101                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4102                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4103                 spin_unlock(&ifp->lock);
4104                 write_unlock_bh(&idev->lock);
4105
4106                 addrconf_dad_completed(ifp, bump_id, send_na);
4107
4108                 goto out;
4109         }
4110
4111         ifp->dad_probes--;
4112         addrconf_mod_dad_work(ifp,
4113                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4114         spin_unlock(&ifp->lock);
4115         write_unlock_bh(&idev->lock);
4116
4117         /* send a neighbour solicitation for our addr */
4118         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4119         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4120                       ifp->dad_nonce);
4121 out:
4122         in6_ifa_put(ifp);
4123         rtnl_unlock();
4124 }
4125
4126 /* ifp->idev must be at least read locked */
4127 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4128 {
4129         struct inet6_ifaddr *ifpiter;
4130         struct inet6_dev *idev = ifp->idev;
4131
4132         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4133                 if (ifpiter->scope > IFA_LINK)
4134                         break;
4135                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4136                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4137                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4138                     IFA_F_PERMANENT)
4139                         return false;
4140         }
4141         return true;
4142 }
4143
4144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4145                                    bool send_na)
4146 {
4147         struct net_device *dev = ifp->idev->dev;
4148         struct in6_addr lladdr;
4149         bool send_rs, send_mld;
4150
4151         addrconf_del_dad_work(ifp);
4152
4153         /*
4154          *      Configure the address for reception. Now it is valid.
4155          */
4156
4157         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4158
4159         /* If added prefix is link local and we are prepared to process
4160            router advertisements, start sending router solicitations.
4161          */
4162
4163         read_lock_bh(&ifp->idev->lock);
4164         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4165         send_rs = send_mld &&
4166                   ipv6_accept_ra(ifp->idev) &&
4167                   ifp->idev->cnf.rtr_solicits != 0 &&
4168                   (dev->flags&IFF_LOOPBACK) == 0;
4169         read_unlock_bh(&ifp->idev->lock);
4170
4171         /* While dad is in progress mld report's source address is in6_addrany.
4172          * Resend with proper ll now.
4173          */
4174         if (send_mld)
4175                 ipv6_mc_dad_complete(ifp->idev);
4176
4177         /* send unsolicited NA if enabled */
4178         if (send_na &&
4179             (ifp->idev->cnf.ndisc_notify ||
4180              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4181                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4182                               /*router=*/ !!ifp->idev->cnf.forwarding,
4183                               /*solicited=*/ false, /*override=*/ true,
4184                               /*inc_opt=*/ true);
4185         }
4186
4187         if (send_rs) {
4188                 /*
4189                  *      If a host as already performed a random delay
4190                  *      [...] as part of DAD [...] there is no need
4191                  *      to delay again before sending the first RS
4192                  */
4193                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4194                         return;
4195                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4196
4197                 write_lock_bh(&ifp->idev->lock);
4198                 spin_lock(&ifp->lock);
4199                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4200                         ifp->idev->cnf.rtr_solicit_interval);
4201                 ifp->idev->rs_probes = 1;
4202                 ifp->idev->if_flags |= IF_RS_SENT;
4203                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4204                 spin_unlock(&ifp->lock);
4205                 write_unlock_bh(&ifp->idev->lock);
4206         }
4207
4208         if (bump_id)
4209                 rt_genid_bump_ipv6(dev_net(dev));
4210
4211         /* Make sure that a new temporary address will be created
4212          * before this temporary address becomes deprecated.
4213          */
4214         if (ifp->flags & IFA_F_TEMPORARY)
4215                 addrconf_verify_rtnl();
4216 }
4217
4218 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4219 {
4220         struct inet6_ifaddr *ifp;
4221
4222         read_lock_bh(&idev->lock);
4223         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4224                 spin_lock(&ifp->lock);
4225                 if ((ifp->flags & IFA_F_TENTATIVE &&
4226                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4227                         if (restart)
4228                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4229                         addrconf_dad_kick(ifp);
4230                 }
4231                 spin_unlock(&ifp->lock);
4232         }
4233         read_unlock_bh(&idev->lock);
4234 }
4235
4236 #ifdef CONFIG_PROC_FS
4237 struct if6_iter_state {
4238         struct seq_net_private p;
4239         int bucket;
4240         int offset;
4241 };
4242
4243 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4244 {
4245         struct if6_iter_state *state = seq->private;
4246         struct net *net = seq_file_net(seq);
4247         struct inet6_ifaddr *ifa = NULL;
4248         int p = 0;
4249
4250         /* initial bucket if pos is 0 */
4251         if (pos == 0) {
4252                 state->bucket = 0;
4253                 state->offset = 0;
4254         }
4255
4256         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4257                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4258                                          addr_lst) {
4259                         if (!net_eq(dev_net(ifa->idev->dev), net))
4260                                 continue;
4261                         /* sync with offset */
4262                         if (p < state->offset) {
4263                                 p++;
4264                                 continue;
4265                         }
4266                         return ifa;
4267                 }
4268
4269                 /* prepare for next bucket */
4270                 state->offset = 0;
4271                 p = 0;
4272         }
4273         return NULL;
4274 }
4275
4276 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4277                                          struct inet6_ifaddr *ifa)
4278 {
4279         struct if6_iter_state *state = seq->private;
4280         struct net *net = seq_file_net(seq);
4281
4282         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4283                 if (!net_eq(dev_net(ifa->idev->dev), net))
4284                         continue;
4285                 state->offset++;
4286                 return ifa;
4287         }
4288
4289         state->offset = 0;
4290         while (++state->bucket < IN6_ADDR_HSIZE) {
4291                 hlist_for_each_entry_rcu(ifa,
4292                                      &inet6_addr_lst[state->bucket], addr_lst) {
4293                         if (!net_eq(dev_net(ifa->idev->dev), net))
4294                                 continue;
4295                         return ifa;
4296                 }
4297         }
4298
4299         return NULL;
4300 }
4301
4302 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4303         __acquires(rcu)
4304 {
4305         rcu_read_lock();
4306         return if6_get_first(seq, *pos);
4307 }
4308
4309 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4310 {
4311         struct inet6_ifaddr *ifa;
4312
4313         ifa = if6_get_next(seq, v);
4314         ++*pos;
4315         return ifa;
4316 }
4317
4318 static void if6_seq_stop(struct seq_file *seq, void *v)
4319         __releases(rcu)
4320 {
4321         rcu_read_unlock();
4322 }
4323
4324 static int if6_seq_show(struct seq_file *seq, void *v)
4325 {
4326         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4327         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4328                    &ifp->addr,
4329                    ifp->idev->dev->ifindex,
4330                    ifp->prefix_len,
4331                    ifp->scope,
4332                    (u8) ifp->flags,
4333                    ifp->idev->dev->name);
4334         return 0;
4335 }
4336
4337 static const struct seq_operations if6_seq_ops = {
4338         .start  = if6_seq_start,
4339         .next   = if6_seq_next,
4340         .show   = if6_seq_show,
4341         .stop   = if6_seq_stop,
4342 };
4343
4344 static int __net_init if6_proc_net_init(struct net *net)
4345 {
4346         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4347                         sizeof(struct if6_iter_state)))
4348                 return -ENOMEM;
4349         return 0;
4350 }
4351
4352 static void __net_exit if6_proc_net_exit(struct net *net)
4353 {
4354         remove_proc_entry("if_inet6", net->proc_net);
4355 }
4356
4357 static struct pernet_operations if6_proc_net_ops = {
4358         .init = if6_proc_net_init,
4359         .exit = if6_proc_net_exit,
4360 };
4361
4362 int __init if6_proc_init(void)
4363 {
4364         return register_pernet_subsys(&if6_proc_net_ops);
4365 }
4366
4367 void if6_proc_exit(void)
4368 {
4369         unregister_pernet_subsys(&if6_proc_net_ops);
4370 }
4371 #endif  /* CONFIG_PROC_FS */
4372
4373 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4374 /* Check if address is a home address configured on any interface. */
4375 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4376 {
4377         unsigned int hash = inet6_addr_hash(net, addr);
4378         struct inet6_ifaddr *ifp = NULL;
4379         int ret = 0;
4380
4381         rcu_read_lock();
4382         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4383                 if (!net_eq(dev_net(ifp->idev->dev), net))
4384                         continue;
4385                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4386                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4387                         ret = 1;
4388                         break;
4389                 }
4390         }
4391         rcu_read_unlock();
4392         return ret;
4393 }
4394 #endif
4395
4396 /*
4397  *      Periodic address status verification
4398  */
4399
4400 static void addrconf_verify_rtnl(void)
4401 {
4402         unsigned long now, next, next_sec, next_sched;
4403         struct inet6_ifaddr *ifp;
4404         int i;
4405
4406         ASSERT_RTNL();
4407
4408         rcu_read_lock_bh();
4409         now = jiffies;
4410         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4411
4412         cancel_delayed_work(&addr_chk_work);
4413
4414         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4415 restart:
4416                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4417                         unsigned long age;
4418
4419                         /* When setting preferred_lft to a value not zero or
4420                          * infinity, while valid_lft is infinity
4421                          * IFA_F_PERMANENT has a non-infinity life time.
4422                          */
4423                         if ((ifp->flags & IFA_F_PERMANENT) &&
4424                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4425                                 continue;
4426
4427                         spin_lock(&ifp->lock);
4428                         /* We try to batch several events at once. */
4429                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4430
4431                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4432                             age >= ifp->valid_lft) {
4433                                 spin_unlock(&ifp->lock);
4434                                 in6_ifa_hold(ifp);
4435                                 ipv6_del_addr(ifp);
4436                                 goto restart;
4437                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4438                                 spin_unlock(&ifp->lock);
4439                                 continue;
4440                         } else if (age >= ifp->prefered_lft) {
4441                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4442                                 int deprecate = 0;
4443
4444                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4445                                         deprecate = 1;
4446                                         ifp->flags |= IFA_F_DEPRECATED;
4447                                 }
4448
4449                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4450                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4451                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4452
4453                                 spin_unlock(&ifp->lock);
4454
4455                                 if (deprecate) {
4456                                         in6_ifa_hold(ifp);
4457
4458                                         ipv6_ifa_notify(0, ifp);
4459                                         in6_ifa_put(ifp);
4460                                         goto restart;
4461                                 }
4462                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4463                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4464                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4465                                         ifp->idev->cnf.dad_transmits *
4466                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4467
4468                                 if (age >= ifp->prefered_lft - regen_advance) {
4469                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4470                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4471                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4472                                         if (!ifp->regen_count && ifpub) {
4473                                                 ifp->regen_count++;
4474                                                 in6_ifa_hold(ifp);
4475                                                 in6_ifa_hold(ifpub);
4476                                                 spin_unlock(&ifp->lock);
4477
4478                                                 spin_lock(&ifpub->lock);
4479                                                 ifpub->regen_count = 0;
4480                                                 spin_unlock(&ifpub->lock);
4481                                                 rcu_read_unlock_bh();
4482                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4483                                                 in6_ifa_put(ifpub);
4484                                                 in6_ifa_put(ifp);
4485                                                 rcu_read_lock_bh();
4486                                                 goto restart;
4487                                         }
4488                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4489                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4490                                 spin_unlock(&ifp->lock);
4491                         } else {
4492                                 /* ifp->prefered_lft <= ifp->valid_lft */
4493                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4494                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4495                                 spin_unlock(&ifp->lock);
4496                         }
4497                 }
4498         }
4499
4500         next_sec = round_jiffies_up(next);
4501         next_sched = next;
4502
4503         /* If rounded timeout is accurate enough, accept it. */
4504         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4505                 next_sched = next_sec;
4506
4507         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4508         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4509                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4510
4511         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4512                  now, next, next_sec, next_sched);
4513         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4514         rcu_read_unlock_bh();
4515 }
4516
4517 static void addrconf_verify_work(struct work_struct *w)
4518 {
4519         rtnl_lock();
4520         addrconf_verify_rtnl();
4521         rtnl_unlock();
4522 }
4523
4524 static void addrconf_verify(void)
4525 {
4526         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4527 }
4528
4529 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4530                                      struct in6_addr **peer_pfx)
4531 {
4532         struct in6_addr *pfx = NULL;
4533
4534         *peer_pfx = NULL;
4535
4536         if (addr)
4537                 pfx = nla_data(addr);
4538
4539         if (local) {
4540                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4541                         *peer_pfx = pfx;
4542                 pfx = nla_data(local);
4543         }
4544
4545         return pfx;
4546 }
4547
4548 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4549         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4550         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4551         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4552         [IFA_FLAGS]             = { .len = sizeof(u32) },
4553         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4554         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4555 };
4556
4557 static int
4558 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4559                   struct netlink_ext_ack *extack)
4560 {
4561         struct net *net = sock_net(skb->sk);
4562         struct ifaddrmsg *ifm;
4563         struct nlattr *tb[IFA_MAX+1];
4564         struct in6_addr *pfx, *peer_pfx;
4565         u32 ifa_flags;
4566         int err;
4567
4568         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4569                                      ifa_ipv6_policy, extack);
4570         if (err < 0)
4571                 return err;
4572
4573         ifm = nlmsg_data(nlh);
4574         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4575         if (!pfx)
4576                 return -EINVAL;
4577
4578         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4579
4580         /* We ignore other flags so far. */
4581         ifa_flags &= IFA_F_MANAGETEMPADDR;
4582
4583         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4584                               ifm->ifa_prefixlen);
4585 }
4586
4587 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4588                                unsigned long expires, u32 flags)
4589 {
4590         struct fib6_info *f6i;
4591         u32 prio;
4592
4593         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
4594                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4595         if (!f6i)
4596                 return -ENOENT;
4597
4598         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4599         if (f6i->fib6_metric != prio) {
4600                 /* delete old one */
4601                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4602
4603                 /* add new one */
4604                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4605                                       ifp->rt_priority, ifp->idev->dev,
4606                                       expires, flags, GFP_KERNEL);
4607         } else {
4608                 if (!expires)
4609                         fib6_clean_expires(f6i);
4610                 else
4611                         fib6_set_expires(f6i, expires);
4612
4613                 fib6_info_release(f6i);
4614         }
4615
4616         return 0;
4617 }
4618
4619 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4620 {
4621         u32 flags;
4622         clock_t expires;
4623         unsigned long timeout;
4624         bool was_managetempaddr;
4625         bool had_prefixroute;
4626
4627         ASSERT_RTNL();
4628
4629         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4630                 return -EINVAL;
4631
4632         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4633             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4634                 return -EINVAL;
4635
4636         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4637                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4638
4639         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4640         if (addrconf_finite_timeout(timeout)) {
4641                 expires = jiffies_to_clock_t(timeout * HZ);
4642                 cfg->valid_lft = timeout;
4643                 flags = RTF_EXPIRES;
4644         } else {
4645                 expires = 0;
4646                 flags = 0;
4647                 cfg->ifa_flags |= IFA_F_PERMANENT;
4648         }
4649
4650         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4651         if (addrconf_finite_timeout(timeout)) {
4652                 if (timeout == 0)
4653                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4654                 cfg->preferred_lft = timeout;
4655         }
4656
4657         spin_lock_bh(&ifp->lock);
4658         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4659         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4660                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4661         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4662                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4663                         IFA_F_NOPREFIXROUTE);
4664         ifp->flags |= cfg->ifa_flags;
4665         ifp->tstamp = jiffies;
4666         ifp->valid_lft = cfg->valid_lft;
4667         ifp->prefered_lft = cfg->preferred_lft;
4668
4669         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4670                 ifp->rt_priority = cfg->rt_priority;
4671
4672         spin_unlock_bh(&ifp->lock);
4673         if (!(ifp->flags&IFA_F_TENTATIVE))
4674                 ipv6_ifa_notify(0, ifp);
4675
4676         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4677                 int rc = -ENOENT;
4678
4679                 if (had_prefixroute)
4680                         rc = modify_prefix_route(ifp, expires, flags);
4681
4682                 /* prefix route could have been deleted; if so restore it */
4683                 if (rc == -ENOENT) {
4684                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4685                                               ifp->rt_priority, ifp->idev->dev,
4686                                               expires, flags, GFP_KERNEL);
4687                 }
4688         } else if (had_prefixroute) {
4689                 enum cleanup_prefix_rt_t action;
4690                 unsigned long rt_expires;
4691
4692                 write_lock_bh(&ifp->idev->lock);
4693                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4694                 write_unlock_bh(&ifp->idev->lock);
4695
4696                 if (action != CLEANUP_PREFIX_RT_NOP) {
4697                         cleanup_prefix_route(ifp, rt_expires,
4698                                 action == CLEANUP_PREFIX_RT_DEL);
4699                 }
4700         }
4701
4702         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4703                 if (was_managetempaddr &&
4704                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4705                         cfg->valid_lft = 0;
4706                         cfg->preferred_lft = 0;
4707                 }
4708                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4709                                  cfg->preferred_lft, !was_managetempaddr,
4710                                  jiffies);
4711         }
4712
4713         addrconf_verify_rtnl();
4714
4715         return 0;
4716 }
4717
4718 static int
4719 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4720                   struct netlink_ext_ack *extack)
4721 {
4722         struct net *net = sock_net(skb->sk);
4723         struct ifaddrmsg *ifm;
4724         struct nlattr *tb[IFA_MAX+1];
4725         struct in6_addr *peer_pfx;
4726         struct inet6_ifaddr *ifa;
4727         struct net_device *dev;
4728         struct inet6_dev *idev;
4729         struct ifa6_config cfg;
4730         int err;
4731
4732         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4733                                      ifa_ipv6_policy, extack);
4734         if (err < 0)
4735                 return err;
4736
4737         memset(&cfg, 0, sizeof(cfg));
4738
4739         ifm = nlmsg_data(nlh);
4740         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4741         if (!cfg.pfx)
4742                 return -EINVAL;
4743
4744         cfg.peer_pfx = peer_pfx;
4745         cfg.plen = ifm->ifa_prefixlen;
4746         if (tb[IFA_RT_PRIORITY])
4747                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4748
4749         cfg.valid_lft = INFINITY_LIFE_TIME;
4750         cfg.preferred_lft = INFINITY_LIFE_TIME;
4751
4752         if (tb[IFA_CACHEINFO]) {
4753                 struct ifa_cacheinfo *ci;
4754
4755                 ci = nla_data(tb[IFA_CACHEINFO]);
4756                 cfg.valid_lft = ci->ifa_valid;
4757                 cfg.preferred_lft = ci->ifa_prefered;
4758         }
4759
4760         dev =  __dev_get_by_index(net, ifm->ifa_index);
4761         if (!dev)
4762                 return -ENODEV;
4763
4764         if (tb[IFA_FLAGS])
4765                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4766         else
4767                 cfg.ifa_flags = ifm->ifa_flags;
4768
4769         /* We ignore other flags so far. */
4770         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4771                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4772                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4773
4774         idev = ipv6_find_idev(dev);
4775         if (!idev)
4776                 return -ENOBUFS;
4777
4778         if (!ipv6_allow_optimistic_dad(net, idev))
4779                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4780
4781         if (cfg.ifa_flags & IFA_F_NODAD &&
4782             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4783                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4784                 return -EINVAL;
4785         }
4786
4787         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4788         if (!ifa) {
4789                 /*
4790                  * It would be best to check for !NLM_F_CREATE here but
4791                  * userspace already relies on not having to provide this.
4792                  */
4793                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4794         }
4795
4796         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4797             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4798                 err = -EEXIST;
4799         else
4800                 err = inet6_addr_modify(ifa, &cfg);
4801
4802         in6_ifa_put(ifa);
4803
4804         return err;
4805 }
4806
4807 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4808                           u8 scope, int ifindex)
4809 {
4810         struct ifaddrmsg *ifm;
4811
4812         ifm = nlmsg_data(nlh);
4813         ifm->ifa_family = AF_INET6;
4814         ifm->ifa_prefixlen = prefixlen;
4815         ifm->ifa_flags = flags;
4816         ifm->ifa_scope = scope;
4817         ifm->ifa_index = ifindex;
4818 }
4819
4820 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4821                          unsigned long tstamp, u32 preferred, u32 valid)
4822 {
4823         struct ifa_cacheinfo ci;
4824
4825         ci.cstamp = cstamp_delta(cstamp);
4826         ci.tstamp = cstamp_delta(tstamp);
4827         ci.ifa_prefered = preferred;
4828         ci.ifa_valid = valid;
4829
4830         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4831 }
4832
4833 static inline int rt_scope(int ifa_scope)
4834 {
4835         if (ifa_scope & IFA_HOST)
4836                 return RT_SCOPE_HOST;
4837         else if (ifa_scope & IFA_LINK)
4838                 return RT_SCOPE_LINK;
4839         else if (ifa_scope & IFA_SITE)
4840                 return RT_SCOPE_SITE;
4841         else
4842                 return RT_SCOPE_UNIVERSE;
4843 }
4844
4845 static inline int inet6_ifaddr_msgsize(void)
4846 {
4847         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4848                + nla_total_size(16) /* IFA_LOCAL */
4849                + nla_total_size(16) /* IFA_ADDRESS */
4850                + nla_total_size(sizeof(struct ifa_cacheinfo))
4851                + nla_total_size(4)  /* IFA_FLAGS */
4852                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4853 }
4854
4855 enum addr_type_t {
4856         UNICAST_ADDR,
4857         MULTICAST_ADDR,
4858         ANYCAST_ADDR,
4859 };
4860
4861 struct inet6_fill_args {
4862         u32 portid;
4863         u32 seq;
4864         int event;
4865         unsigned int flags;
4866         int netnsid;
4867         int ifindex;
4868         enum addr_type_t type;
4869 };
4870
4871 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4872                              struct inet6_fill_args *args)
4873 {
4874         struct nlmsghdr  *nlh;
4875         u32 preferred, valid;
4876
4877         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4878                         sizeof(struct ifaddrmsg), args->flags);
4879         if (!nlh)
4880                 return -EMSGSIZE;
4881
4882         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4883                       ifa->idev->dev->ifindex);
4884
4885         if (args->netnsid >= 0 &&
4886             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4887                 goto error;
4888
4889         if (!((ifa->flags&IFA_F_PERMANENT) &&
4890               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4891                 preferred = ifa->prefered_lft;
4892                 valid = ifa->valid_lft;
4893                 if (preferred != INFINITY_LIFE_TIME) {
4894                         long tval = (jiffies - ifa->tstamp)/HZ;
4895                         if (preferred > tval)
4896                                 preferred -= tval;
4897                         else
4898                                 preferred = 0;
4899                         if (valid != INFINITY_LIFE_TIME) {
4900                                 if (valid > tval)
4901                                         valid -= tval;
4902                                 else
4903                                         valid = 0;
4904                         }
4905                 }
4906         } else {
4907                 preferred = INFINITY_LIFE_TIME;
4908                 valid = INFINITY_LIFE_TIME;
4909         }
4910
4911         if (!ipv6_addr_any(&ifa->peer_addr)) {
4912                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4913                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4914                         goto error;
4915         } else
4916                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4917                         goto error;
4918
4919         if (ifa->rt_priority &&
4920             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4921                 goto error;
4922
4923         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4924                 goto error;
4925
4926         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4927                 goto error;
4928
4929         nlmsg_end(skb, nlh);
4930         return 0;
4931
4932 error:
4933         nlmsg_cancel(skb, nlh);
4934         return -EMSGSIZE;
4935 }
4936
4937 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4938                                struct inet6_fill_args *args)
4939 {
4940         struct nlmsghdr  *nlh;
4941         u8 scope = RT_SCOPE_UNIVERSE;
4942         int ifindex = ifmca->idev->dev->ifindex;
4943
4944         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4945                 scope = RT_SCOPE_SITE;
4946
4947         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4948                         sizeof(struct ifaddrmsg), args->flags);
4949         if (!nlh)
4950                 return -EMSGSIZE;
4951
4952         if (args->netnsid >= 0 &&
4953             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4954                 return -EMSGSIZE;
4955
4956         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4957         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4958             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4959                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4960                 nlmsg_cancel(skb, nlh);
4961                 return -EMSGSIZE;
4962         }
4963
4964         nlmsg_end(skb, nlh);
4965         return 0;
4966 }
4967
4968 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4969                                struct inet6_fill_args *args)
4970 {
4971         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4972         int ifindex = dev ? dev->ifindex : 1;
4973         struct nlmsghdr  *nlh;
4974         u8 scope = RT_SCOPE_UNIVERSE;
4975
4976         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4977                 scope = RT_SCOPE_SITE;
4978
4979         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4980                         sizeof(struct ifaddrmsg), args->flags);
4981         if (!nlh)
4982                 return -EMSGSIZE;
4983
4984         if (args->netnsid >= 0 &&
4985             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4986                 return -EMSGSIZE;
4987
4988         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4989         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4990             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4991                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4992                 nlmsg_cancel(skb, nlh);
4993                 return -EMSGSIZE;
4994         }
4995
4996         nlmsg_end(skb, nlh);
4997         return 0;
4998 }
4999
5000 /* called with rcu_read_lock() */
5001 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5002                           struct netlink_callback *cb, int s_ip_idx,
5003                           struct inet6_fill_args *fillargs)
5004 {
5005         struct ifmcaddr6 *ifmca;
5006         struct ifacaddr6 *ifaca;
5007         int ip_idx = 0;
5008         int err = 1;
5009
5010         read_lock_bh(&idev->lock);
5011         switch (fillargs->type) {
5012         case UNICAST_ADDR: {
5013                 struct inet6_ifaddr *ifa;
5014                 fillargs->event = RTM_NEWADDR;
5015
5016                 /* unicast address incl. temp addr */
5017                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5018                         if (ip_idx < s_ip_idx)
5019                                 goto next;
5020                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5021                         if (err < 0)
5022                                 break;
5023                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5024 next:
5025                         ip_idx++;
5026                 }
5027                 break;
5028         }
5029         case MULTICAST_ADDR:
5030                 fillargs->event = RTM_GETMULTICAST;
5031
5032                 /* multicast address */
5033                 for (ifmca = idev->mc_list; ifmca;
5034                      ifmca = ifmca->next, ip_idx++) {
5035                         if (ip_idx < s_ip_idx)
5036                                 continue;
5037                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5038                         if (err < 0)
5039                                 break;
5040                 }
5041                 break;
5042         case ANYCAST_ADDR:
5043                 fillargs->event = RTM_GETANYCAST;
5044                 /* anycast address */
5045                 for (ifaca = idev->ac_list; ifaca;
5046                      ifaca = ifaca->aca_next, ip_idx++) {
5047                         if (ip_idx < s_ip_idx)
5048                                 continue;
5049                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5050                         if (err < 0)
5051                                 break;
5052                 }
5053                 break;
5054         default:
5055                 break;
5056         }
5057         read_unlock_bh(&idev->lock);
5058         cb->args[2] = ip_idx;
5059         return err;
5060 }
5061
5062 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5063                                        struct inet6_fill_args *fillargs,
5064                                        struct net **tgt_net, struct sock *sk,
5065                                        struct netlink_callback *cb)
5066 {
5067         struct netlink_ext_ack *extack = cb->extack;
5068         struct nlattr *tb[IFA_MAX+1];
5069         struct ifaddrmsg *ifm;
5070         int err, i;
5071
5072         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5073                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5074                 return -EINVAL;
5075         }
5076
5077         ifm = nlmsg_data(nlh);
5078         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5079                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5080                 return -EINVAL;
5081         }
5082
5083         fillargs->ifindex = ifm->ifa_index;
5084         if (fillargs->ifindex) {
5085                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5086                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5087         }
5088
5089         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5090                                             ifa_ipv6_policy, extack);
5091         if (err < 0)
5092                 return err;
5093
5094         for (i = 0; i <= IFA_MAX; ++i) {
5095                 if (!tb[i])
5096                         continue;
5097
5098                 if (i == IFA_TARGET_NETNSID) {
5099                         struct net *net;
5100
5101                         fillargs->netnsid = nla_get_s32(tb[i]);
5102                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5103                         if (IS_ERR(net)) {
5104                                 fillargs->netnsid = -1;
5105                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5106                                 return PTR_ERR(net);
5107                         }
5108                         *tgt_net = net;
5109                 } else {
5110                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5111                         return -EINVAL;
5112                 }
5113         }
5114
5115         return 0;
5116 }
5117
5118 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5119                            enum addr_type_t type)
5120 {
5121         const struct nlmsghdr *nlh = cb->nlh;
5122         struct inet6_fill_args fillargs = {
5123                 .portid = NETLINK_CB(cb->skb).portid,
5124                 .seq = cb->nlh->nlmsg_seq,
5125                 .flags = NLM_F_MULTI,
5126                 .netnsid = -1,
5127                 .type = type,
5128         };
5129         struct net *net = sock_net(skb->sk);
5130         struct net *tgt_net = net;
5131         int idx, s_idx, s_ip_idx;
5132         int h, s_h;
5133         struct net_device *dev;
5134         struct inet6_dev *idev;
5135         struct hlist_head *head;
5136         int err = 0;
5137
5138         s_h = cb->args[0];
5139         s_idx = idx = cb->args[1];
5140         s_ip_idx = cb->args[2];
5141
5142         if (cb->strict_check) {
5143                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5144                                                   skb->sk, cb);
5145                 if (err < 0)
5146                         goto put_tgt_net;
5147
5148                 err = 0;
5149                 if (fillargs.ifindex) {
5150                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5151                         if (!dev) {
5152                                 err = -ENODEV;
5153                                 goto put_tgt_net;
5154                         }
5155                         idev = __in6_dev_get(dev);
5156                         if (idev) {
5157                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5158                                                      &fillargs);
5159                                 if (err > 0)
5160                                         err = 0;
5161                         }
5162                         goto put_tgt_net;
5163                 }
5164         }
5165
5166         rcu_read_lock();
5167         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5168         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5169                 idx = 0;
5170                 head = &tgt_net->dev_index_head[h];
5171                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5172                         if (idx < s_idx)
5173                                 goto cont;
5174                         if (h > s_h || idx > s_idx)
5175                                 s_ip_idx = 0;
5176                         idev = __in6_dev_get(dev);
5177                         if (!idev)
5178                                 goto cont;
5179
5180                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5181                                            &fillargs) < 0)
5182                                 goto done;
5183 cont:
5184                         idx++;
5185                 }
5186         }
5187 done:
5188         rcu_read_unlock();
5189         cb->args[0] = h;
5190         cb->args[1] = idx;
5191 put_tgt_net:
5192         if (fillargs.netnsid >= 0)
5193                 put_net(tgt_net);
5194
5195         return skb->len ? : err;
5196 }
5197
5198 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5199 {
5200         enum addr_type_t type = UNICAST_ADDR;
5201
5202         return inet6_dump_addr(skb, cb, type);
5203 }
5204
5205 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5206 {
5207         enum addr_type_t type = MULTICAST_ADDR;
5208
5209         return inet6_dump_addr(skb, cb, type);
5210 }
5211
5212
5213 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5214 {
5215         enum addr_type_t type = ANYCAST_ADDR;
5216
5217         return inet6_dump_addr(skb, cb, type);
5218 }
5219
5220 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5221                                        const struct nlmsghdr *nlh,
5222                                        struct nlattr **tb,
5223                                        struct netlink_ext_ack *extack)
5224 {
5225         struct ifaddrmsg *ifm;
5226         int i, err;
5227
5228         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5229                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5230                 return -EINVAL;
5231         }
5232
5233         ifm = nlmsg_data(nlh);
5234         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5235                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5236                 return -EINVAL;
5237         }
5238
5239         if (!netlink_strict_get_check(skb))
5240                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5241                                               ifa_ipv6_policy, extack);
5242
5243         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5244                                             ifa_ipv6_policy, extack);
5245         if (err)
5246                 return err;
5247
5248         for (i = 0; i <= IFA_MAX; i++) {
5249                 if (!tb[i])
5250                         continue;
5251
5252                 switch (i) {
5253                 case IFA_TARGET_NETNSID:
5254                 case IFA_ADDRESS:
5255                 case IFA_LOCAL:
5256                         break;
5257                 default:
5258                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5259                         return -EINVAL;
5260                 }
5261         }
5262
5263         return 0;
5264 }
5265
5266 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5267                              struct netlink_ext_ack *extack)
5268 {
5269         struct net *net = sock_net(in_skb->sk);
5270         struct inet6_fill_args fillargs = {
5271                 .portid = NETLINK_CB(in_skb).portid,
5272                 .seq = nlh->nlmsg_seq,
5273                 .event = RTM_NEWADDR,
5274                 .flags = 0,
5275                 .netnsid = -1,
5276         };
5277         struct net *tgt_net = net;
5278         struct ifaddrmsg *ifm;
5279         struct nlattr *tb[IFA_MAX+1];
5280         struct in6_addr *addr = NULL, *peer;
5281         struct net_device *dev = NULL;
5282         struct inet6_ifaddr *ifa;
5283         struct sk_buff *skb;
5284         int err;
5285
5286         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5287         if (err < 0)
5288                 return err;
5289
5290         if (tb[IFA_TARGET_NETNSID]) {
5291                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5292
5293                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5294                                                   fillargs.netnsid);
5295                 if (IS_ERR(tgt_net))
5296                         return PTR_ERR(tgt_net);
5297         }
5298
5299         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5300         if (!addr)
5301                 return -EINVAL;
5302
5303         ifm = nlmsg_data(nlh);
5304         if (ifm->ifa_index)
5305                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5306
5307         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5308         if (!ifa) {
5309                 err = -EADDRNOTAVAIL;
5310                 goto errout;
5311         }
5312
5313         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5314         if (!skb) {
5315                 err = -ENOBUFS;
5316                 goto errout_ifa;
5317         }
5318
5319         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5320         if (err < 0) {
5321                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5322                 WARN_ON(err == -EMSGSIZE);
5323                 kfree_skb(skb);
5324                 goto errout_ifa;
5325         }
5326         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5327 errout_ifa:
5328         in6_ifa_put(ifa);
5329 errout:
5330         if (dev)
5331                 dev_put(dev);
5332         if (fillargs.netnsid >= 0)
5333                 put_net(tgt_net);
5334
5335         return err;
5336 }
5337
5338 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5339 {
5340         struct sk_buff *skb;
5341         struct net *net = dev_net(ifa->idev->dev);
5342         struct inet6_fill_args fillargs = {
5343                 .portid = 0,
5344                 .seq = 0,
5345                 .event = event,
5346                 .flags = 0,
5347                 .netnsid = -1,
5348         };
5349         int err = -ENOBUFS;
5350
5351         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5352         if (!skb)
5353                 goto errout;
5354
5355         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5356         if (err < 0) {
5357                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5358                 WARN_ON(err == -EMSGSIZE);
5359                 kfree_skb(skb);
5360                 goto errout;
5361         }
5362         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5363         return;
5364 errout:
5365         if (err < 0)
5366                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5367 }
5368
5369 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5370                                 __s32 *array, int bytes)
5371 {
5372         BUG_ON(bytes < (DEVCONF_MAX * 4));
5373
5374         memset(array, 0, bytes);
5375         array[DEVCONF_FORWARDING] = cnf->forwarding;
5376         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5377         array[DEVCONF_MTU6] = cnf->mtu6;
5378         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5379         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5380         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5381         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5382         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5383         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5384                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5385         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5386                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5387         array[DEVCONF_RTR_SOLICIT_DELAY] =
5388                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5389         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5390         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5391                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5392         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5393                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5394         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5395         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5396         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5397         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5398         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5399         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5400         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5401         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5402         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5403 #ifdef CONFIG_IPV6_ROUTER_PREF
5404         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5405         array[DEVCONF_RTR_PROBE_INTERVAL] =
5406                 jiffies_to_msecs(cnf->rtr_probe_interval);
5407 #ifdef CONFIG_IPV6_ROUTE_INFO
5408         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5409         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5410 #endif
5411 #endif
5412         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5413         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5415         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5416         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5417 #endif
5418 #ifdef CONFIG_IPV6_MROUTE
5419         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5420 #endif
5421         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5422         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5423         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5424         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5425         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5426         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5427         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5428         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5429         /* we omit DEVCONF_STABLE_SECRET for now */
5430         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5431         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5432         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5433         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5434         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5435 #ifdef CONFIG_IPV6_SEG6_HMAC
5436         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5437 #endif
5438         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5439         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5440         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5441         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5442 }
5443
5444 static inline size_t inet6_ifla6_size(void)
5445 {
5446         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5447              + nla_total_size(sizeof(struct ifla_cacheinfo))
5448              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5449              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5450              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5451              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5452              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5453              + 0;
5454 }
5455
5456 static inline size_t inet6_if_nlmsg_size(void)
5457 {
5458         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5459                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5460                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5461                + nla_total_size(4) /* IFLA_MTU */
5462                + nla_total_size(4) /* IFLA_LINK */
5463                + nla_total_size(1) /* IFLA_OPERSTATE */
5464                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5465 }
5466
5467 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5468                                         int bytes)
5469 {
5470         int i;
5471         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5472         BUG_ON(pad < 0);
5473
5474         /* Use put_unaligned() because stats may not be aligned for u64. */
5475         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5476         for (i = 1; i < ICMP6_MIB_MAX; i++)
5477                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5478
5479         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5480 }
5481
5482 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5483                                         int bytes, size_t syncpoff)
5484 {
5485         int i, c;
5486         u64 buff[IPSTATS_MIB_MAX];
5487         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5488
5489         BUG_ON(pad < 0);
5490
5491         memset(buff, 0, sizeof(buff));
5492         buff[0] = IPSTATS_MIB_MAX;
5493
5494         for_each_possible_cpu(c) {
5495                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5496                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5497         }
5498
5499         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5500         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5501 }
5502
5503 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5504                              int bytes)
5505 {
5506         switch (attrtype) {
5507         case IFLA_INET6_STATS:
5508                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5509                                      offsetof(struct ipstats_mib, syncp));
5510                 break;
5511         case IFLA_INET6_ICMP6STATS:
5512                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5513                 break;
5514         }
5515 }
5516
5517 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5518                                   u32 ext_filter_mask)
5519 {
5520         struct nlattr *nla;
5521         struct ifla_cacheinfo ci;
5522
5523         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5524                 goto nla_put_failure;
5525         ci.max_reasm_len = IPV6_MAXPLEN;
5526         ci.tstamp = cstamp_delta(idev->tstamp);
5527         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5528         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5529         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5530                 goto nla_put_failure;
5531         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5532         if (!nla)
5533                 goto nla_put_failure;
5534         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5535
5536         /* XXX - MC not implemented */
5537
5538         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5539                 return 0;
5540
5541         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5542         if (!nla)
5543                 goto nla_put_failure;
5544         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5545
5546         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5547         if (!nla)
5548                 goto nla_put_failure;
5549         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5550
5551         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5552         if (!nla)
5553                 goto nla_put_failure;
5554
5555         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5556                 goto nla_put_failure;
5557
5558         read_lock_bh(&idev->lock);
5559         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5560         read_unlock_bh(&idev->lock);
5561
5562         return 0;
5563
5564 nla_put_failure:
5565         return -EMSGSIZE;
5566 }
5567
5568 static size_t inet6_get_link_af_size(const struct net_device *dev,
5569                                      u32 ext_filter_mask)
5570 {
5571         if (!__in6_dev_get(dev))
5572                 return 0;
5573
5574         return inet6_ifla6_size();
5575 }
5576
5577 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5578                               u32 ext_filter_mask)
5579 {
5580         struct inet6_dev *idev = __in6_dev_get(dev);
5581
5582         if (!idev)
5583                 return -ENODATA;
5584
5585         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5586                 return -EMSGSIZE;
5587
5588         return 0;
5589 }
5590
5591 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5592 {
5593         struct inet6_ifaddr *ifp;
5594         struct net_device *dev = idev->dev;
5595         bool clear_token, update_rs = false;
5596         struct in6_addr ll_addr;
5597
5598         ASSERT_RTNL();
5599
5600         if (!token)
5601                 return -EINVAL;
5602         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5603                 return -EINVAL;
5604         if (!ipv6_accept_ra(idev))
5605                 return -EINVAL;
5606         if (idev->cnf.rtr_solicits == 0)
5607                 return -EINVAL;
5608
5609         write_lock_bh(&idev->lock);
5610
5611         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5612         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5613
5614         write_unlock_bh(&idev->lock);
5615
5616         clear_token = ipv6_addr_any(token);
5617         if (clear_token)
5618                 goto update_lft;
5619
5620         if (!idev->dead && (idev->if_flags & IF_READY) &&
5621             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5622                              IFA_F_OPTIMISTIC)) {
5623                 /* If we're not ready, then normal ifup will take care
5624                  * of this. Otherwise, we need to request our rs here.
5625                  */
5626                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5627                 update_rs = true;
5628         }
5629
5630 update_lft:
5631         write_lock_bh(&idev->lock);
5632
5633         if (update_rs) {
5634                 idev->if_flags |= IF_RS_SENT;
5635                 idev->rs_interval = rfc3315_s14_backoff_init(
5636                         idev->cnf.rtr_solicit_interval);
5637                 idev->rs_probes = 1;
5638                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5639         }
5640
5641         /* Well, that's kinda nasty ... */
5642         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5643                 spin_lock(&ifp->lock);
5644                 if (ifp->tokenized) {
5645                         ifp->valid_lft = 0;
5646                         ifp->prefered_lft = 0;
5647                 }
5648                 spin_unlock(&ifp->lock);
5649         }
5650
5651         write_unlock_bh(&idev->lock);
5652         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5653         addrconf_verify_rtnl();
5654         return 0;
5655 }
5656
5657 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5658         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5659         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5660 };
5661
5662 static int check_addr_gen_mode(int mode)
5663 {
5664         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5665             mode != IN6_ADDR_GEN_MODE_NONE &&
5666             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5667             mode != IN6_ADDR_GEN_MODE_RANDOM)
5668                 return -EINVAL;
5669         return 1;
5670 }
5671
5672 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5673                                 int mode)
5674 {
5675         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5676             !idev->cnf.stable_secret.initialized &&
5677             !net->ipv6.devconf_dflt->stable_secret.initialized)
5678                 return -EINVAL;
5679         return 1;
5680 }
5681
5682 static int inet6_validate_link_af(const struct net_device *dev,
5683                                   const struct nlattr *nla)
5684 {
5685         struct nlattr *tb[IFLA_INET6_MAX + 1];
5686         struct inet6_dev *idev = NULL;
5687         int err;
5688
5689         if (dev) {
5690                 idev = __in6_dev_get(dev);
5691                 if (!idev)
5692                         return -EAFNOSUPPORT;
5693         }
5694
5695         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5696                                           inet6_af_policy, NULL);
5697         if (err)
5698                 return err;
5699
5700         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5701                 return -EINVAL;
5702
5703         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5704                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5705
5706                 if (check_addr_gen_mode(mode) < 0)
5707                         return -EINVAL;
5708                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5709                         return -EINVAL;
5710         }
5711
5712         return 0;
5713 }
5714
5715 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5716 {
5717         struct inet6_dev *idev = __in6_dev_get(dev);
5718         struct nlattr *tb[IFLA_INET6_MAX + 1];
5719         int err;
5720
5721         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5722                 BUG();
5723
5724         if (tb[IFLA_INET6_TOKEN]) {
5725                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5726                 if (err)
5727                         return err;
5728         }
5729
5730         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5731                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5732
5733                 idev->cnf.addr_gen_mode = mode;
5734         }
5735
5736         return 0;
5737 }
5738
5739 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5740                              u32 portid, u32 seq, int event, unsigned int flags)
5741 {
5742         struct net_device *dev = idev->dev;
5743         struct ifinfomsg *hdr;
5744         struct nlmsghdr *nlh;
5745         void *protoinfo;
5746
5747         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5748         if (!nlh)
5749                 return -EMSGSIZE;
5750
5751         hdr = nlmsg_data(nlh);
5752         hdr->ifi_family = AF_INET6;
5753         hdr->__ifi_pad = 0;
5754         hdr->ifi_type = dev->type;
5755         hdr->ifi_index = dev->ifindex;
5756         hdr->ifi_flags = dev_get_flags(dev);
5757         hdr->ifi_change = 0;
5758
5759         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5760             (dev->addr_len &&
5761              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5762             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5763             (dev->ifindex != dev_get_iflink(dev) &&
5764              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5765             nla_put_u8(skb, IFLA_OPERSTATE,
5766                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5767                 goto nla_put_failure;
5768         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5769         if (!protoinfo)
5770                 goto nla_put_failure;
5771
5772         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5773                 goto nla_put_failure;
5774
5775         nla_nest_end(skb, protoinfo);
5776         nlmsg_end(skb, nlh);
5777         return 0;
5778
5779 nla_put_failure:
5780         nlmsg_cancel(skb, nlh);
5781         return -EMSGSIZE;
5782 }
5783
5784 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5785                                    struct netlink_ext_ack *extack)
5786 {
5787         struct ifinfomsg *ifm;
5788
5789         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5790                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5791                 return -EINVAL;
5792         }
5793
5794         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5795                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5796                 return -EINVAL;
5797         }
5798
5799         ifm = nlmsg_data(nlh);
5800         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5801             ifm->ifi_change || ifm->ifi_index) {
5802                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5803                 return -EINVAL;
5804         }
5805
5806         return 0;
5807 }
5808
5809 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5810 {
5811         struct net *net = sock_net(skb->sk);
5812         int h, s_h;
5813         int idx = 0, s_idx;
5814         struct net_device *dev;
5815         struct inet6_dev *idev;
5816         struct hlist_head *head;
5817
5818         /* only requests using strict checking can pass data to
5819          * influence the dump
5820          */
5821         if (cb->strict_check) {
5822                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5823
5824                 if (err < 0)
5825                         return err;
5826         }
5827
5828         s_h = cb->args[0];
5829         s_idx = cb->args[1];
5830
5831         rcu_read_lock();
5832         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5833                 idx = 0;
5834                 head = &net->dev_index_head[h];
5835                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5836                         if (idx < s_idx)
5837                                 goto cont;
5838                         idev = __in6_dev_get(dev);
5839                         if (!idev)
5840                                 goto cont;
5841                         if (inet6_fill_ifinfo(skb, idev,
5842                                               NETLINK_CB(cb->skb).portid,
5843                                               cb->nlh->nlmsg_seq,
5844                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5845                                 goto out;
5846 cont:
5847                         idx++;
5848                 }
5849         }
5850 out:
5851         rcu_read_unlock();
5852         cb->args[1] = idx;
5853         cb->args[0] = h;
5854
5855         return skb->len;
5856 }
5857
5858 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5859 {
5860         struct sk_buff *skb;
5861         struct net *net = dev_net(idev->dev);
5862         int err = -ENOBUFS;
5863
5864         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5865         if (!skb)
5866                 goto errout;
5867
5868         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5869         if (err < 0) {
5870                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5871                 WARN_ON(err == -EMSGSIZE);
5872                 kfree_skb(skb);
5873                 goto errout;
5874         }
5875         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5876         return;
5877 errout:
5878         if (err < 0)
5879                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5880 }
5881
5882 static inline size_t inet6_prefix_nlmsg_size(void)
5883 {
5884         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5885                + nla_total_size(sizeof(struct in6_addr))
5886                + nla_total_size(sizeof(struct prefix_cacheinfo));
5887 }
5888
5889 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5890                              struct prefix_info *pinfo, u32 portid, u32 seq,
5891                              int event, unsigned int flags)
5892 {
5893         struct prefixmsg *pmsg;
5894         struct nlmsghdr *nlh;
5895         struct prefix_cacheinfo ci;
5896
5897         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5898         if (!nlh)
5899                 return -EMSGSIZE;
5900
5901         pmsg = nlmsg_data(nlh);
5902         pmsg->prefix_family = AF_INET6;
5903         pmsg->prefix_pad1 = 0;
5904         pmsg->prefix_pad2 = 0;
5905         pmsg->prefix_ifindex = idev->dev->ifindex;
5906         pmsg->prefix_len = pinfo->prefix_len;
5907         pmsg->prefix_type = pinfo->type;
5908         pmsg->prefix_pad3 = 0;
5909         pmsg->prefix_flags = 0;
5910         if (pinfo->onlink)
5911                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5912         if (pinfo->autoconf)
5913                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5914
5915         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5916                 goto nla_put_failure;
5917         ci.preferred_time = ntohl(pinfo->prefered);
5918         ci.valid_time = ntohl(pinfo->valid);
5919         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5920                 goto nla_put_failure;
5921         nlmsg_end(skb, nlh);
5922         return 0;
5923
5924 nla_put_failure:
5925         nlmsg_cancel(skb, nlh);
5926         return -EMSGSIZE;
5927 }
5928
5929 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5930                          struct prefix_info *pinfo)
5931 {
5932         struct sk_buff *skb;
5933         struct net *net = dev_net(idev->dev);
5934         int err = -ENOBUFS;
5935
5936         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5937         if (!skb)
5938                 goto errout;
5939
5940         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5941         if (err < 0) {
5942                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5943                 WARN_ON(err == -EMSGSIZE);
5944                 kfree_skb(skb);
5945                 goto errout;
5946         }
5947         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5948         return;
5949 errout:
5950         if (err < 0)
5951                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5952 }
5953
5954 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5955 {
5956         struct net *net = dev_net(ifp->idev->dev);
5957
5958         if (event)
5959                 ASSERT_RTNL();
5960
5961         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5962
5963         switch (event) {
5964         case RTM_NEWADDR:
5965                 /*
5966                  * If the address was optimistic
5967                  * we inserted the route at the start of
5968                  * our DAD process, so we don't need
5969                  * to do it again
5970                  */
5971                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5972                         ip6_ins_rt(net, ifp->rt);
5973                 if (ifp->idev->cnf.forwarding)
5974                         addrconf_join_anycast(ifp);
5975                 if (!ipv6_addr_any(&ifp->peer_addr))
5976                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5977                                               ifp->idev->dev, 0, 0,
5978                                               GFP_ATOMIC);
5979                 break;
5980         case RTM_DELADDR:
5981                 if (ifp->idev->cnf.forwarding)
5982                         addrconf_leave_anycast(ifp);
5983                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5984                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5985                         struct fib6_info *rt;
5986
5987                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5988                                                        ifp->idev->dev, 0, 0,
5989                                                        false);
5990                         if (rt)
5991                                 ip6_del_rt(net, rt);
5992                 }
5993                 if (ifp->rt) {
5994                         ip6_del_rt(net, ifp->rt);
5995                         ifp->rt = NULL;
5996                 }
5997                 rt_genid_bump_ipv6(net);
5998                 break;
5999         }
6000         atomic_inc(&net->ipv6.dev_addr_genid);
6001 }
6002
6003 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6004 {
6005         rcu_read_lock_bh();
6006         if (likely(ifp->idev->dead == 0))
6007                 __ipv6_ifa_notify(event, ifp);
6008         rcu_read_unlock_bh();
6009 }
6010
6011 #ifdef CONFIG_SYSCTL
6012
6013 static
6014 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6015                            void __user *buffer, size_t *lenp, loff_t *ppos)
6016 {
6017         int *valp = ctl->data;
6018         int val = *valp;
6019         loff_t pos = *ppos;
6020         struct ctl_table lctl;
6021         int ret;
6022
6023         /*
6024          * ctl->data points to idev->cnf.forwarding, we should
6025          * not modify it until we get the rtnl lock.
6026          */
6027         lctl = *ctl;
6028         lctl.data = &val;
6029
6030         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6031
6032         if (write)
6033                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6034         if (ret)
6035                 *ppos = pos;
6036         return ret;
6037 }
6038
6039 static
6040 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6041                         void __user *buffer, size_t *lenp, loff_t *ppos)
6042 {
6043         struct inet6_dev *idev = ctl->extra1;
6044         int min_mtu = IPV6_MIN_MTU;
6045         struct ctl_table lctl;
6046
6047         lctl = *ctl;
6048         lctl.extra1 = &min_mtu;
6049         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6050
6051         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6052 }
6053
6054 static void dev_disable_change(struct inet6_dev *idev)
6055 {
6056         struct netdev_notifier_info info;
6057
6058         if (!idev || !idev->dev)
6059                 return;
6060
6061         netdev_notifier_info_init(&info, idev->dev);
6062         if (idev->cnf.disable_ipv6)
6063                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6064         else
6065                 addrconf_notify(NULL, NETDEV_UP, &info);
6066 }
6067
6068 static void addrconf_disable_change(struct net *net, __s32 newf)
6069 {
6070         struct net_device *dev;
6071         struct inet6_dev *idev;
6072
6073         for_each_netdev(net, dev) {
6074                 idev = __in6_dev_get(dev);
6075                 if (idev) {
6076                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6077                         idev->cnf.disable_ipv6 = newf;
6078                         if (changed)
6079                                 dev_disable_change(idev);
6080                 }
6081         }
6082 }
6083
6084 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6085 {
6086         struct net *net;
6087         int old;
6088
6089         if (!rtnl_trylock())
6090                 return restart_syscall();
6091
6092         net = (struct net *)table->extra2;
6093         old = *p;
6094         *p = newf;
6095
6096         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6097                 rtnl_unlock();
6098                 return 0;
6099         }
6100
6101         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6102                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6103                 addrconf_disable_change(net, newf);
6104         } else if ((!newf) ^ (!old))
6105                 dev_disable_change((struct inet6_dev *)table->extra1);
6106
6107         rtnl_unlock();
6108         return 0;
6109 }
6110
6111 static
6112 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6113                             void __user *buffer, size_t *lenp, loff_t *ppos)
6114 {
6115         int *valp = ctl->data;
6116         int val = *valp;
6117         loff_t pos = *ppos;
6118         struct ctl_table lctl;
6119         int ret;
6120
6121         /*
6122          * ctl->data points to idev->cnf.disable_ipv6, we should
6123          * not modify it until we get the rtnl lock.
6124          */
6125         lctl = *ctl;
6126         lctl.data = &val;
6127
6128         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6129
6130         if (write)
6131                 ret = addrconf_disable_ipv6(ctl, valp, val);
6132         if (ret)
6133                 *ppos = pos;
6134         return ret;
6135 }
6136
6137 static
6138 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6139                               void __user *buffer, size_t *lenp, loff_t *ppos)
6140 {
6141         int *valp = ctl->data;
6142         int ret;
6143         int old, new;
6144
6145         old = *valp;
6146         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6147         new = *valp;
6148
6149         if (write && old != new) {
6150                 struct net *net = ctl->extra2;
6151
6152                 if (!rtnl_trylock())
6153                         return restart_syscall();
6154
6155                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6156                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6157                                                      NETCONFA_PROXY_NEIGH,
6158                                                      NETCONFA_IFINDEX_DEFAULT,
6159                                                      net->ipv6.devconf_dflt);
6160                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6161                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6162                                                      NETCONFA_PROXY_NEIGH,
6163                                                      NETCONFA_IFINDEX_ALL,
6164                                                      net->ipv6.devconf_all);
6165                 else {
6166                         struct inet6_dev *idev = ctl->extra1;
6167
6168                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6169                                                      NETCONFA_PROXY_NEIGH,
6170                                                      idev->dev->ifindex,
6171                                                      &idev->cnf);
6172                 }
6173                 rtnl_unlock();
6174         }
6175
6176         return ret;
6177 }
6178
6179 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6180                                          void __user *buffer, size_t *lenp,
6181                                          loff_t *ppos)
6182 {
6183         int ret = 0;
6184         u32 new_val;
6185         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6186         struct net *net = (struct net *)ctl->extra2;
6187         struct ctl_table tmp = {
6188                 .data = &new_val,
6189                 .maxlen = sizeof(new_val),
6190                 .mode = ctl->mode,
6191         };
6192
6193         if (!rtnl_trylock())
6194                 return restart_syscall();
6195
6196         new_val = *((u32 *)ctl->data);
6197
6198         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6199         if (ret != 0)
6200                 goto out;
6201
6202         if (write) {
6203                 if (check_addr_gen_mode(new_val) < 0) {
6204                         ret = -EINVAL;
6205                         goto out;
6206                 }
6207
6208                 if (idev) {
6209                         if (check_stable_privacy(idev, net, new_val) < 0) {
6210                                 ret = -EINVAL;
6211                                 goto out;
6212                         }
6213
6214                         if (idev->cnf.addr_gen_mode != new_val) {
6215                                 idev->cnf.addr_gen_mode = new_val;
6216                                 addrconf_dev_config(idev->dev);
6217                         }
6218                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6219                         struct net_device *dev;
6220
6221                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6222                         for_each_netdev(net, dev) {
6223                                 idev = __in6_dev_get(dev);
6224                                 if (idev &&
6225                                     idev->cnf.addr_gen_mode != new_val) {
6226                                         idev->cnf.addr_gen_mode = new_val;
6227                                         addrconf_dev_config(idev->dev);
6228                                 }
6229                         }
6230                 }
6231
6232                 *((u32 *)ctl->data) = new_val;
6233         }
6234
6235 out:
6236         rtnl_unlock();
6237
6238         return ret;
6239 }
6240
6241 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6242                                          void __user *buffer, size_t *lenp,
6243                                          loff_t *ppos)
6244 {
6245         int err;
6246         struct in6_addr addr;
6247         char str[IPV6_MAX_STRLEN];
6248         struct ctl_table lctl = *ctl;
6249         struct net *net = ctl->extra2;
6250         struct ipv6_stable_secret *secret = ctl->data;
6251
6252         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6253                 return -EIO;
6254
6255         lctl.maxlen = IPV6_MAX_STRLEN;
6256         lctl.data = str;
6257
6258         if (!rtnl_trylock())
6259                 return restart_syscall();
6260
6261         if (!write && !secret->initialized) {
6262                 err = -EIO;
6263                 goto out;
6264         }
6265
6266         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6267         if (err >= sizeof(str)) {
6268                 err = -EIO;
6269                 goto out;
6270         }
6271
6272         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6273         if (err || !write)
6274                 goto out;
6275
6276         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6277                 err = -EIO;
6278                 goto out;
6279         }
6280
6281         secret->initialized = true;
6282         secret->secret = addr;
6283
6284         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6285                 struct net_device *dev;
6286
6287                 for_each_netdev(net, dev) {
6288                         struct inet6_dev *idev = __in6_dev_get(dev);
6289
6290                         if (idev) {
6291                                 idev->cnf.addr_gen_mode =
6292                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6293                         }
6294                 }
6295         } else {
6296                 struct inet6_dev *idev = ctl->extra1;
6297
6298                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6299         }
6300
6301 out:
6302         rtnl_unlock();
6303
6304         return err;
6305 }
6306
6307 static
6308 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6309                                                 int write,
6310                                                 void __user *buffer,
6311                                                 size_t *lenp,
6312                                                 loff_t *ppos)
6313 {
6314         int *valp = ctl->data;
6315         int val = *valp;
6316         loff_t pos = *ppos;
6317         struct ctl_table lctl;
6318         int ret;
6319
6320         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6321          * we should not modify it until we get the rtnl lock.
6322          */
6323         lctl = *ctl;
6324         lctl.data = &val;
6325
6326         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6327
6328         if (write)
6329                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6330         if (ret)
6331                 *ppos = pos;
6332         return ret;
6333 }
6334
6335 static
6336 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6337 {
6338         if (rt) {
6339                 if (action)
6340                         rt->dst.flags |= DST_NOPOLICY;
6341                 else
6342                         rt->dst.flags &= ~DST_NOPOLICY;
6343         }
6344 }
6345
6346 static
6347 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6348 {
6349         struct inet6_ifaddr *ifa;
6350
6351         read_lock_bh(&idev->lock);
6352         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6353                 spin_lock(&ifa->lock);
6354                 if (ifa->rt) {
6355                         /* host routes only use builtin fib6_nh */
6356                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6357                         int cpu;
6358
6359                         rcu_read_lock();
6360                         ifa->rt->dst_nopolicy = val ? true : false;
6361                         if (nh->rt6i_pcpu) {
6362                                 for_each_possible_cpu(cpu) {
6363                                         struct rt6_info **rtp;
6364
6365                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6366                                         addrconf_set_nopolicy(*rtp, val);
6367                                 }
6368                         }
6369                         rcu_read_unlock();
6370                 }
6371                 spin_unlock(&ifa->lock);
6372         }
6373         read_unlock_bh(&idev->lock);
6374 }
6375
6376 static
6377 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6378 {
6379         struct inet6_dev *idev;
6380         struct net *net;
6381
6382         if (!rtnl_trylock())
6383                 return restart_syscall();
6384
6385         *valp = val;
6386
6387         net = (struct net *)ctl->extra2;
6388         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6389                 rtnl_unlock();
6390                 return 0;
6391         }
6392
6393         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6394                 struct net_device *dev;
6395
6396                 for_each_netdev(net, dev) {
6397                         idev = __in6_dev_get(dev);
6398                         if (idev)
6399                                 addrconf_disable_policy_idev(idev, val);
6400                 }
6401         } else {
6402                 idev = (struct inet6_dev *)ctl->extra1;
6403                 addrconf_disable_policy_idev(idev, val);
6404         }
6405
6406         rtnl_unlock();
6407         return 0;
6408 }
6409
6410 static
6411 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6412                                    void __user *buffer, size_t *lenp,
6413                                    loff_t *ppos)
6414 {
6415         int *valp = ctl->data;
6416         int val = *valp;
6417         loff_t pos = *ppos;
6418         struct ctl_table lctl;
6419         int ret;
6420
6421         lctl = *ctl;
6422         lctl.data = &val;
6423         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6424
6425         if (write && (*valp != val))
6426                 ret = addrconf_disable_policy(ctl, valp, val);
6427
6428         if (ret)
6429                 *ppos = pos;
6430
6431         return ret;
6432 }
6433
6434 static int minus_one = -1;
6435 static const int two_five_five = 255;
6436
6437 static const struct ctl_table addrconf_sysctl[] = {
6438         {
6439                 .procname       = "forwarding",
6440                 .data           = &ipv6_devconf.forwarding,
6441                 .maxlen         = sizeof(int),
6442                 .mode           = 0644,
6443                 .proc_handler   = addrconf_sysctl_forward,
6444         },
6445         {
6446                 .procname       = "hop_limit",
6447                 .data           = &ipv6_devconf.hop_limit,
6448                 .maxlen         = sizeof(int),
6449                 .mode           = 0644,
6450                 .proc_handler   = proc_dointvec_minmax,
6451                 .extra1         = (void *)SYSCTL_ONE,
6452                 .extra2         = (void *)&two_five_five,
6453         },
6454         {
6455                 .procname       = "mtu",
6456                 .data           = &ipv6_devconf.mtu6,
6457                 .maxlen         = sizeof(int),
6458                 .mode           = 0644,
6459                 .proc_handler   = addrconf_sysctl_mtu,
6460         },
6461         {
6462                 .procname       = "accept_ra",
6463                 .data           = &ipv6_devconf.accept_ra,
6464                 .maxlen         = sizeof(int),
6465                 .mode           = 0644,
6466                 .proc_handler   = proc_dointvec,
6467         },
6468         {
6469                 .procname       = "accept_redirects",
6470                 .data           = &ipv6_devconf.accept_redirects,
6471                 .maxlen         = sizeof(int),
6472                 .mode           = 0644,
6473                 .proc_handler   = proc_dointvec,
6474         },
6475         {
6476                 .procname       = "autoconf",
6477                 .data           = &ipv6_devconf.autoconf,
6478                 .maxlen         = sizeof(int),
6479                 .mode           = 0644,
6480                 .proc_handler   = proc_dointvec,
6481         },
6482         {
6483                 .procname       = "dad_transmits",
6484                 .data           = &ipv6_devconf.dad_transmits,
6485                 .maxlen         = sizeof(int),
6486                 .mode           = 0644,
6487                 .proc_handler   = proc_dointvec,
6488         },
6489         {
6490                 .procname       = "router_solicitations",
6491                 .data           = &ipv6_devconf.rtr_solicits,
6492                 .maxlen         = sizeof(int),
6493                 .mode           = 0644,
6494                 .proc_handler   = proc_dointvec_minmax,
6495                 .extra1         = &minus_one,
6496         },
6497         {
6498                 .procname       = "router_solicitation_interval",
6499                 .data           = &ipv6_devconf.rtr_solicit_interval,
6500                 .maxlen         = sizeof(int),
6501                 .mode           = 0644,
6502                 .proc_handler   = proc_dointvec_jiffies,
6503         },
6504         {
6505                 .procname       = "router_solicitation_max_interval",
6506                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6507                 .maxlen         = sizeof(int),
6508                 .mode           = 0644,
6509                 .proc_handler   = proc_dointvec_jiffies,
6510         },
6511         {
6512                 .procname       = "router_solicitation_delay",
6513                 .data           = &ipv6_devconf.rtr_solicit_delay,
6514                 .maxlen         = sizeof(int),
6515                 .mode           = 0644,
6516                 .proc_handler   = proc_dointvec_jiffies,
6517         },
6518         {
6519                 .procname       = "force_mld_version",
6520                 .data           = &ipv6_devconf.force_mld_version,
6521                 .maxlen         = sizeof(int),
6522                 .mode           = 0644,
6523                 .proc_handler   = proc_dointvec,
6524         },
6525         {
6526                 .procname       = "mldv1_unsolicited_report_interval",
6527                 .data           =
6528                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6529                 .maxlen         = sizeof(int),
6530                 .mode           = 0644,
6531                 .proc_handler   = proc_dointvec_ms_jiffies,
6532         },
6533         {
6534                 .procname       = "mldv2_unsolicited_report_interval",
6535                 .data           =
6536                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6537                 .maxlen         = sizeof(int),
6538                 .mode           = 0644,
6539                 .proc_handler   = proc_dointvec_ms_jiffies,
6540         },
6541         {
6542                 .procname       = "use_tempaddr",
6543                 .data           = &ipv6_devconf.use_tempaddr,
6544                 .maxlen         = sizeof(int),
6545                 .mode           = 0644,
6546                 .proc_handler   = proc_dointvec,
6547         },
6548         {
6549                 .procname       = "temp_valid_lft",
6550                 .data           = &ipv6_devconf.temp_valid_lft,
6551                 .maxlen         = sizeof(int),
6552                 .mode           = 0644,
6553                 .proc_handler   = proc_dointvec,
6554         },
6555         {
6556                 .procname       = "temp_prefered_lft",
6557                 .data           = &ipv6_devconf.temp_prefered_lft,
6558                 .maxlen         = sizeof(int),
6559                 .mode           = 0644,
6560                 .proc_handler   = proc_dointvec,
6561         },
6562         {
6563                 .procname       = "regen_max_retry",
6564                 .data           = &ipv6_devconf.regen_max_retry,
6565                 .maxlen         = sizeof(int),
6566                 .mode           = 0644,
6567                 .proc_handler   = proc_dointvec,
6568         },
6569         {
6570                 .procname       = "max_desync_factor",
6571                 .data           = &ipv6_devconf.max_desync_factor,
6572                 .maxlen         = sizeof(int),
6573                 .mode           = 0644,
6574                 .proc_handler   = proc_dointvec,
6575         },
6576         {
6577                 .procname       = "max_addresses",
6578                 .data           = &ipv6_devconf.max_addresses,
6579                 .maxlen         = sizeof(int),
6580                 .mode           = 0644,
6581                 .proc_handler   = proc_dointvec,
6582         },
6583         {
6584                 .procname       = "accept_ra_defrtr",
6585                 .data           = &ipv6_devconf.accept_ra_defrtr,
6586                 .maxlen         = sizeof(int),
6587                 .mode           = 0644,
6588                 .proc_handler   = proc_dointvec,
6589         },
6590         {
6591                 .procname       = "accept_ra_min_hop_limit",
6592                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6593                 .maxlen         = sizeof(int),
6594                 .mode           = 0644,
6595                 .proc_handler   = proc_dointvec,
6596         },
6597         {
6598                 .procname       = "accept_ra_pinfo",
6599                 .data           = &ipv6_devconf.accept_ra_pinfo,
6600                 .maxlen         = sizeof(int),
6601                 .mode           = 0644,
6602                 .proc_handler   = proc_dointvec,
6603         },
6604 #ifdef CONFIG_IPV6_ROUTER_PREF
6605         {
6606                 .procname       = "accept_ra_rtr_pref",
6607                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6608                 .maxlen         = sizeof(int),
6609                 .mode           = 0644,
6610                 .proc_handler   = proc_dointvec,
6611         },
6612         {
6613                 .procname       = "router_probe_interval",
6614                 .data           = &ipv6_devconf.rtr_probe_interval,
6615                 .maxlen         = sizeof(int),
6616                 .mode           = 0644,
6617                 .proc_handler   = proc_dointvec_jiffies,
6618         },
6619 #ifdef CONFIG_IPV6_ROUTE_INFO
6620         {
6621                 .procname       = "accept_ra_rt_info_min_plen",
6622                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6623                 .maxlen         = sizeof(int),
6624                 .mode           = 0644,
6625                 .proc_handler   = proc_dointvec,
6626         },
6627         {
6628                 .procname       = "accept_ra_rt_info_max_plen",
6629                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6630                 .maxlen         = sizeof(int),
6631                 .mode           = 0644,
6632                 .proc_handler   = proc_dointvec,
6633         },
6634 #endif
6635 #endif
6636         {
6637                 .procname       = "proxy_ndp",
6638                 .data           = &ipv6_devconf.proxy_ndp,
6639                 .maxlen         = sizeof(int),
6640                 .mode           = 0644,
6641                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6642         },
6643         {
6644                 .procname       = "accept_source_route",
6645                 .data           = &ipv6_devconf.accept_source_route,
6646                 .maxlen         = sizeof(int),
6647                 .mode           = 0644,
6648                 .proc_handler   = proc_dointvec,
6649         },
6650 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6651         {
6652                 .procname       = "optimistic_dad",
6653                 .data           = &ipv6_devconf.optimistic_dad,
6654                 .maxlen         = sizeof(int),
6655                 .mode           = 0644,
6656                 .proc_handler   = proc_dointvec,
6657         },
6658         {
6659                 .procname       = "use_optimistic",
6660                 .data           = &ipv6_devconf.use_optimistic,
6661                 .maxlen         = sizeof(int),
6662                 .mode           = 0644,
6663                 .proc_handler   = proc_dointvec,
6664         },
6665 #endif
6666 #ifdef CONFIG_IPV6_MROUTE
6667         {
6668                 .procname       = "mc_forwarding",
6669                 .data           = &ipv6_devconf.mc_forwarding,
6670                 .maxlen         = sizeof(int),
6671                 .mode           = 0444,
6672                 .proc_handler   = proc_dointvec,
6673         },
6674 #endif
6675         {
6676                 .procname       = "disable_ipv6",
6677                 .data           = &ipv6_devconf.disable_ipv6,
6678                 .maxlen         = sizeof(int),
6679                 .mode           = 0644,
6680                 .proc_handler   = addrconf_sysctl_disable,
6681         },
6682         {
6683                 .procname       = "accept_dad",
6684                 .data           = &ipv6_devconf.accept_dad,
6685                 .maxlen         = sizeof(int),
6686                 .mode           = 0644,
6687                 .proc_handler   = proc_dointvec,
6688         },
6689         {
6690                 .procname       = "force_tllao",
6691                 .data           = &ipv6_devconf.force_tllao,
6692                 .maxlen         = sizeof(int),
6693                 .mode           = 0644,
6694                 .proc_handler   = proc_dointvec
6695         },
6696         {
6697                 .procname       = "ndisc_notify",
6698                 .data           = &ipv6_devconf.ndisc_notify,
6699                 .maxlen         = sizeof(int),
6700                 .mode           = 0644,
6701                 .proc_handler   = proc_dointvec
6702         },
6703         {
6704                 .procname       = "suppress_frag_ndisc",
6705                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6706                 .maxlen         = sizeof(int),
6707                 .mode           = 0644,
6708                 .proc_handler   = proc_dointvec
6709         },
6710         {
6711                 .procname       = "accept_ra_from_local",
6712                 .data           = &ipv6_devconf.accept_ra_from_local,
6713                 .maxlen         = sizeof(int),
6714                 .mode           = 0644,
6715                 .proc_handler   = proc_dointvec,
6716         },
6717         {
6718                 .procname       = "accept_ra_mtu",
6719                 .data           = &ipv6_devconf.accept_ra_mtu,
6720                 .maxlen         = sizeof(int),
6721                 .mode           = 0644,
6722                 .proc_handler   = proc_dointvec,
6723         },
6724         {
6725                 .procname       = "stable_secret",
6726                 .data           = &ipv6_devconf.stable_secret,
6727                 .maxlen         = IPV6_MAX_STRLEN,
6728                 .mode           = 0600,
6729                 .proc_handler   = addrconf_sysctl_stable_secret,
6730         },
6731         {
6732                 .procname       = "use_oif_addrs_only",
6733                 .data           = &ipv6_devconf.use_oif_addrs_only,
6734                 .maxlen         = sizeof(int),
6735                 .mode           = 0644,
6736                 .proc_handler   = proc_dointvec,
6737         },
6738         {
6739                 .procname       = "ignore_routes_with_linkdown",
6740                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6741                 .maxlen         = sizeof(int),
6742                 .mode           = 0644,
6743                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6744         },
6745         {
6746                 .procname       = "drop_unicast_in_l2_multicast",
6747                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6748                 .maxlen         = sizeof(int),
6749                 .mode           = 0644,
6750                 .proc_handler   = proc_dointvec,
6751         },
6752         {
6753                 .procname       = "drop_unsolicited_na",
6754                 .data           = &ipv6_devconf.drop_unsolicited_na,
6755                 .maxlen         = sizeof(int),
6756                 .mode           = 0644,
6757                 .proc_handler   = proc_dointvec,
6758         },
6759         {
6760                 .procname       = "keep_addr_on_down",
6761                 .data           = &ipv6_devconf.keep_addr_on_down,
6762                 .maxlen         = sizeof(int),
6763                 .mode           = 0644,
6764                 .proc_handler   = proc_dointvec,
6765
6766         },
6767         {
6768                 .procname       = "seg6_enabled",
6769                 .data           = &ipv6_devconf.seg6_enabled,
6770                 .maxlen         = sizeof(int),
6771                 .mode           = 0644,
6772                 .proc_handler   = proc_dointvec,
6773         },
6774 #ifdef CONFIG_IPV6_SEG6_HMAC
6775         {
6776                 .procname       = "seg6_require_hmac",
6777                 .data           = &ipv6_devconf.seg6_require_hmac,
6778                 .maxlen         = sizeof(int),
6779                 .mode           = 0644,
6780                 .proc_handler   = proc_dointvec,
6781         },
6782 #endif
6783         {
6784                 .procname       = "enhanced_dad",
6785                 .data           = &ipv6_devconf.enhanced_dad,
6786                 .maxlen         = sizeof(int),
6787                 .mode           = 0644,
6788                 .proc_handler   = proc_dointvec,
6789         },
6790         {
6791                 .procname               = "addr_gen_mode",
6792                 .data                   = &ipv6_devconf.addr_gen_mode,
6793                 .maxlen                 = sizeof(int),
6794                 .mode                   = 0644,
6795                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6796         },
6797         {
6798                 .procname       = "disable_policy",
6799                 .data           = &ipv6_devconf.disable_policy,
6800                 .maxlen         = sizeof(int),
6801                 .mode           = 0644,
6802                 .proc_handler   = addrconf_sysctl_disable_policy,
6803         },
6804         {
6805                 .procname       = "ndisc_tclass",
6806                 .data           = &ipv6_devconf.ndisc_tclass,
6807                 .maxlen         = sizeof(int),
6808                 .mode           = 0644,
6809                 .proc_handler   = proc_dointvec_minmax,
6810                 .extra1         = (void *)SYSCTL_ZERO,
6811                 .extra2         = (void *)&two_five_five,
6812         },
6813         {
6814                 /* sentinel */
6815         }
6816 };
6817
6818 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6819                 struct inet6_dev *idev, struct ipv6_devconf *p)
6820 {
6821         int i, ifindex;
6822         struct ctl_table *table;
6823         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6824
6825         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6826         if (!table)
6827                 goto out;
6828
6829         for (i = 0; table[i].data; i++) {
6830                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6831                 /* If one of these is already set, then it is not safe to
6832                  * overwrite either of them: this makes proc_dointvec_minmax
6833                  * usable.
6834                  */
6835                 if (!table[i].extra1 && !table[i].extra2) {
6836                         table[i].extra1 = idev; /* embedded; no ref */
6837                         table[i].extra2 = net;
6838                 }
6839         }
6840
6841         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6842
6843         p->sysctl_header = register_net_sysctl(net, path, table);
6844         if (!p->sysctl_header)
6845                 goto free;
6846
6847         if (!strcmp(dev_name, "all"))
6848                 ifindex = NETCONFA_IFINDEX_ALL;
6849         else if (!strcmp(dev_name, "default"))
6850                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6851         else
6852                 ifindex = idev->dev->ifindex;
6853         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6854                                      ifindex, p);
6855         return 0;
6856
6857 free:
6858         kfree(table);
6859 out:
6860         return -ENOBUFS;
6861 }
6862
6863 static void __addrconf_sysctl_unregister(struct net *net,
6864                                          struct ipv6_devconf *p, int ifindex)
6865 {
6866         struct ctl_table *table;
6867
6868         if (!p->sysctl_header)
6869                 return;
6870
6871         table = p->sysctl_header->ctl_table_arg;
6872         unregister_net_sysctl_table(p->sysctl_header);
6873         p->sysctl_header = NULL;
6874         kfree(table);
6875
6876         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6877 }
6878
6879 static int addrconf_sysctl_register(struct inet6_dev *idev)
6880 {
6881         int err;
6882
6883         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6884                 return -EINVAL;
6885
6886         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6887                                     &ndisc_ifinfo_sysctl_change);
6888         if (err)
6889                 return err;
6890         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6891                                          idev, &idev->cnf);
6892         if (err)
6893                 neigh_sysctl_unregister(idev->nd_parms);
6894
6895         return err;
6896 }
6897
6898 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6899 {
6900         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6901                                      idev->dev->ifindex);
6902         neigh_sysctl_unregister(idev->nd_parms);
6903 }
6904
6905
6906 #endif
6907
6908 static int __net_init addrconf_init_net(struct net *net)
6909 {
6910         int err = -ENOMEM;
6911         struct ipv6_devconf *all, *dflt;
6912
6913         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6914         if (!all)
6915                 goto err_alloc_all;
6916
6917         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6918         if (!dflt)
6919                 goto err_alloc_dflt;
6920
6921         if (IS_ENABLED(CONFIG_SYSCTL) &&
6922             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6923                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6924                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6925         }
6926
6927         /* these will be inherited by all namespaces */
6928         dflt->autoconf = ipv6_defaults.autoconf;
6929         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6930
6931         dflt->stable_secret.initialized = false;
6932         all->stable_secret.initialized = false;
6933
6934         net->ipv6.devconf_all = all;
6935         net->ipv6.devconf_dflt = dflt;
6936
6937 #ifdef CONFIG_SYSCTL
6938         err = __addrconf_sysctl_register(net, "all", NULL, all);
6939         if (err < 0)
6940                 goto err_reg_all;
6941
6942         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6943         if (err < 0)
6944                 goto err_reg_dflt;
6945 #endif
6946         return 0;
6947
6948 #ifdef CONFIG_SYSCTL
6949 err_reg_dflt:
6950         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6951 err_reg_all:
6952         kfree(dflt);
6953 #endif
6954 err_alloc_dflt:
6955         kfree(all);
6956 err_alloc_all:
6957         return err;
6958 }
6959
6960 static void __net_exit addrconf_exit_net(struct net *net)
6961 {
6962 #ifdef CONFIG_SYSCTL
6963         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6964                                      NETCONFA_IFINDEX_DEFAULT);
6965         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6966                                      NETCONFA_IFINDEX_ALL);
6967 #endif
6968         kfree(net->ipv6.devconf_dflt);
6969         kfree(net->ipv6.devconf_all);
6970 }
6971
6972 static struct pernet_operations addrconf_ops = {
6973         .init = addrconf_init_net,
6974         .exit = addrconf_exit_net,
6975 };
6976
6977 static struct rtnl_af_ops inet6_ops __read_mostly = {
6978         .family           = AF_INET6,
6979         .fill_link_af     = inet6_fill_link_af,
6980         .get_link_af_size = inet6_get_link_af_size,
6981         .validate_link_af = inet6_validate_link_af,
6982         .set_link_af      = inet6_set_link_af,
6983 };
6984
6985 /*
6986  *      Init / cleanup code
6987  */
6988
6989 int __init addrconf_init(void)
6990 {
6991         struct inet6_dev *idev;
6992         int i, err;
6993
6994         err = ipv6_addr_label_init();
6995         if (err < 0) {
6996                 pr_crit("%s: cannot initialize default policy table: %d\n",
6997                         __func__, err);
6998                 goto out;
6999         }
7000
7001         err = register_pernet_subsys(&addrconf_ops);
7002         if (err < 0)
7003                 goto out_addrlabel;
7004
7005         addrconf_wq = create_workqueue("ipv6_addrconf");
7006         if (!addrconf_wq) {
7007                 err = -ENOMEM;
7008                 goto out_nowq;
7009         }
7010
7011         /* The addrconf netdev notifier requires that loopback_dev
7012          * has it's ipv6 private information allocated and setup
7013          * before it can bring up and give link-local addresses
7014          * to other devices which are up.
7015          *
7016          * Unfortunately, loopback_dev is not necessarily the first
7017          * entry in the global dev_base list of net devices.  In fact,
7018          * it is likely to be the very last entry on that list.
7019          * So this causes the notifier registry below to try and
7020          * give link-local addresses to all devices besides loopback_dev
7021          * first, then loopback_dev, which cases all the non-loopback_dev
7022          * devices to fail to get a link-local address.
7023          *
7024          * So, as a temporary fix, allocate the ipv6 structure for
7025          * loopback_dev first by hand.
7026          * Longer term, all of the dependencies ipv6 has upon the loopback
7027          * device and it being up should be removed.
7028          */
7029         rtnl_lock();
7030         idev = ipv6_add_dev(init_net.loopback_dev);
7031         rtnl_unlock();
7032         if (IS_ERR(idev)) {
7033                 err = PTR_ERR(idev);
7034                 goto errlo;
7035         }
7036
7037         ip6_route_init_special_entries();
7038
7039         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7040                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7041
7042         register_netdevice_notifier(&ipv6_dev_notf);
7043
7044         addrconf_verify();
7045
7046         rtnl_af_register(&inet6_ops);
7047
7048         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7049                                    NULL, inet6_dump_ifinfo, 0);
7050         if (err < 0)
7051                 goto errout;
7052
7053         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7054                                    inet6_rtm_newaddr, NULL, 0);
7055         if (err < 0)
7056                 goto errout;
7057         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7058                                    inet6_rtm_deladdr, NULL, 0);
7059         if (err < 0)
7060                 goto errout;
7061         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7062                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7063                                    RTNL_FLAG_DOIT_UNLOCKED);
7064         if (err < 0)
7065                 goto errout;
7066         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7067                                    NULL, inet6_dump_ifmcaddr, 0);
7068         if (err < 0)
7069                 goto errout;
7070         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7071                                    NULL, inet6_dump_ifacaddr, 0);
7072         if (err < 0)
7073                 goto errout;
7074         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7075                                    inet6_netconf_get_devconf,
7076                                    inet6_netconf_dump_devconf,
7077                                    RTNL_FLAG_DOIT_UNLOCKED);
7078         if (err < 0)
7079                 goto errout;
7080         err = ipv6_addr_label_rtnl_register();
7081         if (err < 0)
7082                 goto errout;
7083
7084         return 0;
7085 errout:
7086         rtnl_unregister_all(PF_INET6);
7087         rtnl_af_unregister(&inet6_ops);
7088         unregister_netdevice_notifier(&ipv6_dev_notf);
7089 errlo:
7090         destroy_workqueue(addrconf_wq);
7091 out_nowq:
7092         unregister_pernet_subsys(&addrconf_ops);
7093 out_addrlabel:
7094         ipv6_addr_label_cleanup();
7095 out:
7096         return err;
7097 }
7098
7099 void addrconf_cleanup(void)
7100 {
7101         struct net_device *dev;
7102         int i;
7103
7104         unregister_netdevice_notifier(&ipv6_dev_notf);
7105         unregister_pernet_subsys(&addrconf_ops);
7106         ipv6_addr_label_cleanup();
7107
7108         rtnl_af_unregister(&inet6_ops);
7109
7110         rtnl_lock();
7111
7112         /* clean dev list */
7113         for_each_netdev(&init_net, dev) {
7114                 if (__in6_dev_get(dev) == NULL)
7115                         continue;
7116                 addrconf_ifdown(dev, 1);
7117         }
7118         addrconf_ifdown(init_net.loopback_dev, 2);
7119
7120         /*
7121          *      Check hash table.
7122          */
7123         spin_lock_bh(&addrconf_hash_lock);
7124         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7125                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7126         spin_unlock_bh(&addrconf_hash_lock);
7127         cancel_delayed_work(&addr_chk_work);
7128         rtnl_unlock();
7129
7130         destroy_workqueue(addrconf_wq);
7131 }