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 idev;
 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                         atomic_read(&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         LIST_HEAD(tmp_addr_list);
 793
 794         if (!idev)
 795                 return;
 796         dev = idev->dev;
 797         if (idev->cnf.forwarding)
 798                 dev_disable_lro(dev);
 799         if (dev->flags & IFF_MULTICAST) {
 800                 if (idev->cnf.forwarding) {
 801                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 802                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 803                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 804                 } else {
 805                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 806                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 807                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 808                 }
 809         }
 810
 811         read_lock_bh(&idev->lock);
 812         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 813                 if (ifa->flags&IFA_F_TENTATIVE)
 814                         continue;
 815                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
 816         }
 817         read_unlock_bh(&idev->lock);
 818
 819         while (!list_empty(&tmp_addr_list)) {
 820                 ifa = list_first_entry(&tmp_addr_list,
 821                                        struct inet6_ifaddr, if_list_aux);
 822                 list_del(&ifa->if_list_aux);
 823                 if (idev->cnf.forwarding)
 824                         addrconf_join_anycast(ifa);
 825                 else
 826                         addrconf_leave_anycast(ifa);
 827         }
 828
 829         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 830                                      NETCONFA_FORWARDING,
 831                                      dev->ifindex, &idev->cnf);
 832 }
 833
 834
 835 static void addrconf_forward_change(struct net *net, __s32 newf)
 836 {
 837         struct net_device *dev;
 838         struct inet6_dev *idev;
 839
 840         for_each_netdev(net, dev) {
 841                 idev = __in6_dev_get(dev);
 842                 if (idev) {
 843                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 844                         idev->cnf.forwarding = newf;
 845                         if (changed)
 846                                 dev_forward_change(idev);
 847                 }
 848         }
 849 }
 850
 851 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 852 {
 853         struct net *net;
 854         int old;
 855
 856         if (!rtnl_trylock())
 857                 return restart_syscall();
 858
 859         net = (struct net *)table->extra2;
 860         old = *p;
 861         *p = newf;
 862
 863         if (p == &net->ipv6.devconf_dflt->forwarding) {
 864                 if ((!newf) ^ (!old))
 865                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 866                                                      NETCONFA_FORWARDING,
 867                                                      NETCONFA_IFINDEX_DEFAULT,
 868                                                      net->ipv6.devconf_dflt);
 869                 rtnl_unlock();
 870                 return 0;
 871         }
 872
 873         if (p == &net->ipv6.devconf_all->forwarding) {
 874                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 875
 876                 net->ipv6.devconf_dflt->forwarding = newf;
 877                 if ((!newf) ^ (!old_dflt))
 878                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 879                                                      NETCONFA_FORWARDING,
 880                                                      NETCONFA_IFINDEX_DEFAULT,
 881                                                      net->ipv6.devconf_dflt);
 882
 883                 addrconf_forward_change(net, newf);
 884                 if ((!newf) ^ (!old))
 885                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 886                                                      NETCONFA_FORWARDING,
 887                                                      NETCONFA_IFINDEX_ALL,
 888                                                      net->ipv6.devconf_all);
 889         } else if ((!newf) ^ (!old))
 890                 dev_forward_change((struct inet6_dev *)table->extra1);
 891         rtnl_unlock();
 892
 893         if (newf)
 894                 rt6_purge_dflt_routers(net);
 895         return 1;
 896 }
 897
 898 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 899 {
 900         struct net_device *dev;
 901         struct inet6_dev *idev;
 902
 903         for_each_netdev(net, dev) {
 904                 idev = __in6_dev_get(dev);
 905                 if (idev) {
 906                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 907
 908                         idev->cnf.ignore_routes_with_linkdown = newf;
 909                         if (changed)
 910                                 inet6_netconf_notify_devconf(dev_net(dev),
 911                                                              RTM_NEWNETCONF,
 912                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 913                                                              dev->ifindex,
 914                                                              &idev->cnf);
 915                 }
 916         }
 917 }
 918
 919 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 920 {
 921         struct net *net;
 922         int old;
 923
 924         if (!rtnl_trylock())
 925                 return restart_syscall();
 926
 927         net = (struct net *)table->extra2;
 928         old = *p;
 929         *p = newf;
 930
 931         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 932                 if ((!newf) ^ (!old))
 933                         inet6_netconf_notify_devconf(net,
 934                                                      RTM_NEWNETCONF,
 935                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 936                                                      NETCONFA_IFINDEX_DEFAULT,
 937                                                      net->ipv6.devconf_dflt);
 938                 rtnl_unlock();
 939                 return 0;
 940         }
 941
 942         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 943                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 944                 addrconf_linkdown_change(net, newf);
 945                 if ((!newf) ^ (!old))
 946                         inet6_netconf_notify_devconf(net,
 947                                                      RTM_NEWNETCONF,
 948                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 949                                                      NETCONFA_IFINDEX_ALL,
 950                                                      net->ipv6.devconf_all);
 951         }
 952         rtnl_unlock();
 953
 954         return 1;
 955 }
 956
 957 #endif
 958
 959 /* Nobody refers to this ifaddr, destroy it */
 960 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 961 {
 962         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 963
 964 #ifdef NET_REFCNT_DEBUG
 965         pr_debug("%s\n", __func__);
 966 #endif
 967
 968         in6_dev_put(ifp->idev);
 969
 970         if (cancel_delayed_work(&ifp->dad_work))
 971                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 972                           ifp);
 973
 974         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 975                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 976                 return;
 977         }
 978
 979         kfree_rcu(ifp, rcu);
 980 }
 981
 982 static void
 983 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 984 {
 985         struct list_head *p;
 986         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 987
 988         /*
 989          * Each device address list is sorted in order of scope -
 990          * global before linklocal.
 991          */
 992         list_for_each(p, &idev->addr_list) {
 993                 struct inet6_ifaddr *ifa
 994                         = list_entry(p, struct inet6_ifaddr, if_list);
 995                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 996                         break;
 997         }
 998
 999         list_add_tail_rcu(&ifp->if_list, p);
1000 }
1001
1002 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1003 {
1004         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1005
1006         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1007 }
1008
1009 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1010                                struct net_device *dev, unsigned int hash)
1011 {
1012         struct inet6_ifaddr *ifp;
1013
1014         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1015                 if (!net_eq(dev_net(ifp->idev->dev), net))
1016                         continue;
1017                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1018                         if (!dev || ifp->idev->dev == dev)
1019                                 return true;
1020                 }
1021         }
1022         return false;
1023 }
1024
1025 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1026 {
1027         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1028         int err = 0;
1029
1030         spin_lock(&addrconf_hash_lock);
1031
1032         /* Ignore adding duplicate addresses on an interface */
1033         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1034                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1035                 err = -EEXIST;
1036         } else {
1037                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1038         }
1039
1040         spin_unlock(&addrconf_hash_lock);
1041
1042         return err;
1043 }
1044
1045 /* On success it returns ifp with increased reference count */
1046
1047 static struct inet6_ifaddr *
1048 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1049               bool can_block, struct netlink_ext_ack *extack)
1050 {
1051         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1052         int addr_type = ipv6_addr_type(cfg->pfx);
1053         struct net *net = dev_net(idev->dev);
1054         struct inet6_ifaddr *ifa = NULL;
1055         struct fib6_info *f6i = NULL;
1056         int err = 0;
1057
1058         if (addr_type == IPV6_ADDR_ANY ||
1059             (addr_type & IPV6_ADDR_MULTICAST &&
1060              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1061             (!(idev->dev->flags & IFF_LOOPBACK) &&
1062              !netif_is_l3_master(idev->dev) &&
1063              addr_type & IPV6_ADDR_LOOPBACK))
1064                 return ERR_PTR(-EADDRNOTAVAIL);
1065
1066         if (idev->dead) {
1067                 err = -ENODEV;                  /*XXX*/
1068                 goto out;
1069         }
1070
1071         if (idev->cnf.disable_ipv6) {
1072                 err = -EACCES;
1073                 goto out;
1074         }
1075
1076         /* validator notifier needs to be blocking;
1077          * do not call in atomic context
1078          */
1079         if (can_block) {
1080                 struct in6_validator_info i6vi = {
1081                         .i6vi_addr = *cfg->pfx,
1082                         .i6vi_dev = idev,
1083                         .extack = extack,
1084                 };
1085
1086                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1087                 err = notifier_to_errno(err);
1088                 if (err < 0)
1089                         goto out;
1090         }
1091
1092         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1093         if (!ifa) {
1094                 err = -ENOBUFS;
1095                 goto out;
1096         }
1097
1098         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1099         if (IS_ERR(f6i)) {
1100                 err = PTR_ERR(f6i);
1101                 f6i = NULL;
1102                 goto out;
1103         }
1104
1105         neigh_parms_data_state_setall(idev->nd_parms);
1106
1107         ifa->addr = *cfg->pfx;
1108         if (cfg->peer_pfx)
1109                 ifa->peer_addr = *cfg->peer_pfx;
1110
1111         spin_lock_init(&ifa->lock);
1112         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1113         INIT_HLIST_NODE(&ifa->addr_lst);
1114         ifa->scope = cfg->scope;
1115         ifa->prefix_len = cfg->plen;
1116         ifa->rt_priority = cfg->rt_priority;
1117         ifa->flags = cfg->ifa_flags;
1118         /* No need to add the TENTATIVE flag for addresses with NODAD */
1119         if (!(cfg->ifa_flags & IFA_F_NODAD))
1120                 ifa->flags |= IFA_F_TENTATIVE;
1121         ifa->valid_lft = cfg->valid_lft;
1122         ifa->prefered_lft = cfg->preferred_lft;
1123         ifa->cstamp = ifa->tstamp = jiffies;
1124         ifa->tokenized = false;
1125
1126         ifa->rt = f6i;
1127
1128         ifa->idev = idev;
1129         in6_dev_hold(idev);
1130
1131         /* For caller */
1132         refcount_set(&ifa->refcnt, 1);
1133
1134         rcu_read_lock_bh();
1135
1136         err = ipv6_add_addr_hash(idev->dev, ifa);
1137         if (err < 0) {
1138                 rcu_read_unlock_bh();
1139                 goto out;
1140         }
1141
1142         write_lock(&idev->lock);
1143
1144         /* Add to inet6_dev unicast addr list. */
1145         ipv6_link_dev_addr(idev, ifa);
1146
1147         if (ifa->flags&IFA_F_TEMPORARY) {
1148                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1149                 in6_ifa_hold(ifa);
1150         }
1151
1152         in6_ifa_hold(ifa);
1153         write_unlock(&idev->lock);
1154
1155         rcu_read_unlock_bh();
1156
1157         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1158 out:
1159         if (unlikely(err < 0)) {
1160                 fib6_info_release(f6i);
1161
1162                 if (ifa) {
1163                         if (ifa->idev)
1164                                 in6_dev_put(ifa->idev);
1165                         kfree(ifa);
1166                 }
1167                 ifa = ERR_PTR(err);
1168         }
1169
1170         return ifa;
1171 }
1172
1173 enum cleanup_prefix_rt_t {
1174         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1175         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1176         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1177 };
1178
1179 /*
1180  * Check, whether the prefix for ifp would still need a prefix route
1181  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1182  * constants.
1183  *
1184  * 1) we don't purge prefix if address was not permanent.
1185  *    prefix is managed by its own lifetime.
1186  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1187  * 3) if there are no addresses, delete prefix.
1188  * 4) if there are still other permanent address(es),
1189  *    corresponding prefix is still permanent.
1190  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1191  *    don't purge the prefix, assume user space is managing it.
1192  * 6) otherwise, update prefix lifetime to the
1193  *    longest valid lifetime among the corresponding
1194  *    addresses on the device.
1195  *    Note: subsequent RA will update lifetime.
1196  **/
1197 static enum cleanup_prefix_rt_t
1198 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1199 {
1200         struct inet6_ifaddr *ifa;
1201         struct inet6_dev *idev = ifp->idev;
1202         unsigned long lifetime;
1203         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1204
1205         *expires = jiffies;
1206
1207         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1208                 if (ifa == ifp)
1209                         continue;
1210                 if (ifa->prefix_len != ifp->prefix_len ||
1211                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1212                                        ifp->prefix_len))
1213                         continue;
1214                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1215                         return CLEANUP_PREFIX_RT_NOP;
1216
1217                 action = CLEANUP_PREFIX_RT_EXPIRE;
1218
1219                 spin_lock(&ifa->lock);
1220
1221                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1222                 /*
1223                  * Note: Because this address is
1224                  * not permanent, lifetime <
1225                  * LONG_MAX / HZ here.
1226                  */
1227                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1228                         *expires = ifa->tstamp + lifetime * HZ;
1229                 spin_unlock(&ifa->lock);
1230         }
1231
1232         return action;
1233 }
1234
1235 static void
1236 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1237                      bool del_rt, bool del_peer)
1238 {
1239         struct fib6_info *f6i;
1240
1241         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1242                                         ifp->prefix_len,
1243                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1244         if (f6i) {
1245                 if (del_rt)
1246                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1247                 else {
1248                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1249                                 fib6_set_expires(f6i, expires);
1250                         fib6_info_release(f6i);
1251                 }
1252         }
1253 }
1254
1255
1256 /* This function wants to get referenced ifp and releases it before return */
1257
1258 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1259 {
1260         int state;
1261         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1262         unsigned long expires;
1263
1264         ASSERT_RTNL();
1265
1266         spin_lock_bh(&ifp->lock);
1267         state = ifp->state;
1268         ifp->state = INET6_IFADDR_STATE_DEAD;
1269         spin_unlock_bh(&ifp->lock);
1270
1271         if (state == INET6_IFADDR_STATE_DEAD)
1272                 goto out;
1273
1274         spin_lock_bh(&addrconf_hash_lock);
1275         hlist_del_init_rcu(&ifp->addr_lst);
1276         spin_unlock_bh(&addrconf_hash_lock);
1277
1278         write_lock_bh(&ifp->idev->lock);
1279
1280         if (ifp->flags&IFA_F_TEMPORARY) {
1281                 list_del(&ifp->tmp_list);
1282                 if (ifp->ifpub) {
1283                         in6_ifa_put(ifp->ifpub);
1284                         ifp->ifpub = NULL;
1285                 }
1286                 __in6_ifa_put(ifp);
1287         }
1288
1289         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1290                 action = check_cleanup_prefix_route(ifp, &expires);
1291
1292         list_del_rcu(&ifp->if_list);
1293         __in6_ifa_put(ifp);
1294
1295         write_unlock_bh(&ifp->idev->lock);
1296
1297         addrconf_del_dad_work(ifp);
1298
1299         ipv6_ifa_notify(RTM_DELADDR, ifp);
1300
1301         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1302
1303         if (action != CLEANUP_PREFIX_RT_NOP) {
1304                 cleanup_prefix_route(ifp, expires,
1305                         action == CLEANUP_PREFIX_RT_DEL, false);
1306         }
1307
1308         /* clean up prefsrc entries */
1309         rt6_remove_prefsrc(ifp);
1310 out:
1311         in6_ifa_put(ifp);
1312 }
1313
1314 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1315                                 struct inet6_ifaddr *ift,
1316                                 bool block)
1317 {
1318         struct inet6_dev *idev = ifp->idev;
1319         struct in6_addr addr, *tmpaddr;
1320         unsigned long tmp_tstamp, age;
1321         unsigned long regen_advance;
1322         struct ifa6_config cfg;
1323         int ret = 0;
1324         unsigned long now = jiffies;
1325         long max_desync_factor;
1326         s32 cnf_temp_preferred_lft;
1327
1328         write_lock_bh(&idev->lock);
1329         if (ift) {
1330                 spin_lock_bh(&ift->lock);
1331                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1332                 spin_unlock_bh(&ift->lock);
1333                 tmpaddr = &addr;
1334         } else {
1335                 tmpaddr = NULL;
1336         }
1337 retry:
1338         in6_dev_hold(idev);
1339         if (idev->cnf.use_tempaddr <= 0) {
1340                 write_unlock_bh(&idev->lock);
1341                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1342                 in6_dev_put(idev);
1343                 ret = -1;
1344                 goto out;
1345         }
1346         spin_lock_bh(&ifp->lock);
1347         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1348                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1349                 spin_unlock_bh(&ifp->lock);
1350                 write_unlock_bh(&idev->lock);
1351                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1352                         __func__);
1353                 in6_dev_put(idev);
1354                 ret = -1;
1355                 goto out;
1356         }
1357         in6_ifa_hold(ifp);
1358         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1359         ipv6_try_regen_rndid(idev, tmpaddr);
1360         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1361         age = (now - ifp->tstamp) / HZ;
1362
1363         regen_advance = idev->cnf.regen_max_retry *
1364                         idev->cnf.dad_transmits *
1365                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1366
1367         /* recalculate max_desync_factor each time and update
1368          * idev->desync_factor if it's larger
1369          */
1370         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1371         max_desync_factor = min_t(__u32,
1372                                   idev->cnf.max_desync_factor,
1373                                   cnf_temp_preferred_lft - regen_advance);
1374
1375         if (unlikely(idev->desync_factor > max_desync_factor)) {
1376                 if (max_desync_factor > 0) {
1377                         get_random_bytes(&idev->desync_factor,
1378                                          sizeof(idev->desync_factor));
1379                         idev->desync_factor %= max_desync_factor;
1380                 } else {
1381                         idev->desync_factor = 0;
1382                 }
1383         }
1384
1385         memset(&cfg, 0, sizeof(cfg));
1386         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1387                               idev->cnf.temp_valid_lft + age);
1388         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1389         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1390
1391         cfg.plen = ifp->prefix_len;
1392         tmp_tstamp = ifp->tstamp;
1393         spin_unlock_bh(&ifp->lock);
1394
1395         write_unlock_bh(&idev->lock);
1396
1397         /* A temporary address is created only if this calculated Preferred
1398          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1399          * an implementation must not create a temporary address with a zero
1400          * Preferred Lifetime.
1401          * Use age calculation as in addrconf_verify to avoid unnecessary
1402          * temporary addresses being generated.
1403          */
1404         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1405         if (cfg.preferred_lft <= regen_advance + age) {
1406                 in6_ifa_put(ifp);
1407                 in6_dev_put(idev);
1408                 ret = -1;
1409                 goto out;
1410         }
1411
1412         cfg.ifa_flags = IFA_F_TEMPORARY;
1413         /* set in addrconf_prefix_rcv() */
1414         if (ifp->flags & IFA_F_OPTIMISTIC)
1415                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1416
1417         cfg.pfx = &addr;
1418         cfg.scope = ipv6_addr_scope(cfg.pfx);
1419
1420         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1421         if (IS_ERR(ift)) {
1422                 in6_ifa_put(ifp);
1423                 in6_dev_put(idev);
1424                 pr_info("%s: retry temporary address regeneration\n", __func__);
1425                 tmpaddr = &addr;
1426                 write_lock_bh(&idev->lock);
1427                 goto retry;
1428         }
1429
1430         spin_lock_bh(&ift->lock);
1431         ift->ifpub = ifp;
1432         ift->cstamp = now;
1433         ift->tstamp = tmp_tstamp;
1434         spin_unlock_bh(&ift->lock);
1435
1436         addrconf_dad_start(ift);
1437         in6_ifa_put(ift);
1438         in6_dev_put(idev);
1439 out:
1440         return ret;
1441 }
1442
1443 /*
1444  *      Choose an appropriate source address (RFC3484)
1445  */
1446 enum {
1447         IPV6_SADDR_RULE_INIT = 0,
1448         IPV6_SADDR_RULE_LOCAL,
1449         IPV6_SADDR_RULE_SCOPE,
1450         IPV6_SADDR_RULE_PREFERRED,
1451 #ifdef CONFIG_IPV6_MIP6
1452         IPV6_SADDR_RULE_HOA,
1453 #endif
1454         IPV6_SADDR_RULE_OIF,
1455         IPV6_SADDR_RULE_LABEL,
1456         IPV6_SADDR_RULE_PRIVACY,
1457         IPV6_SADDR_RULE_ORCHID,
1458         IPV6_SADDR_RULE_PREFIX,
1459 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1460         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1461 #endif
1462         IPV6_SADDR_RULE_MAX
1463 };
1464
1465 struct ipv6_saddr_score {
1466         int                     rule;
1467         int                     addr_type;
1468         struct inet6_ifaddr     *ifa;
1469         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1470         int                     scopedist;
1471         int                     matchlen;
1472 };
1473
1474 struct ipv6_saddr_dst {
1475         const struct in6_addr *addr;
1476         int ifindex;
1477         int scope;
1478         int label;
1479         unsigned int prefs;
1480 };
1481
1482 static inline int ipv6_saddr_preferred(int type)
1483 {
1484         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1485                 return 1;
1486         return 0;
1487 }
1488
1489 static bool ipv6_use_optimistic_addr(struct net *net,
1490                                      struct inet6_dev *idev)
1491 {
1492 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1493         if (!idev)
1494                 return false;
1495         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1496                 return false;
1497         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1498                 return false;
1499
1500         return true;
1501 #else
1502         return false;
1503 #endif
1504 }
1505
1506 static bool ipv6_allow_optimistic_dad(struct net *net,
1507                                       struct inet6_dev *idev)
1508 {
1509 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1510         if (!idev)
1511                 return false;
1512         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1513                 return false;
1514
1515         return true;
1516 #else
1517         return false;
1518 #endif
1519 }
1520
1521 static int ipv6_get_saddr_eval(struct net *net,
1522                                struct ipv6_saddr_score *score,
1523                                struct ipv6_saddr_dst *dst,
1524                                int i)
1525 {
1526         int ret;
1527
1528         if (i <= score->rule) {
1529                 switch (i) {
1530                 case IPV6_SADDR_RULE_SCOPE:
1531                         ret = score->scopedist;
1532                         break;
1533                 case IPV6_SADDR_RULE_PREFIX:
1534                         ret = score->matchlen;
1535                         break;
1536                 default:
1537                         ret = !!test_bit(i, score->scorebits);
1538                 }
1539                 goto out;
1540         }
1541
1542         switch (i) {
1543         case IPV6_SADDR_RULE_INIT:
1544                 /* Rule 0: remember if hiscore is not ready yet */
1545                 ret = !!score->ifa;
1546                 break;
1547         case IPV6_SADDR_RULE_LOCAL:
1548                 /* Rule 1: Prefer same address */
1549                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1550                 break;
1551         case IPV6_SADDR_RULE_SCOPE:
1552                 /* Rule 2: Prefer appropriate scope
1553                  *
1554                  *      ret
1555                  *       ^
1556                  *    -1 |  d 15
1557                  *    ---+--+-+---> scope
1558                  *       |
1559                  *       |             d is scope of the destination.
1560                  *  B-d  |  \
1561                  *       |   \      <- smaller scope is better if
1562                  *  B-15 |    \        if scope is enough for destination.
1563                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1564                  * d-C-1 | /
1565                  *       |/         <- greater is better
1566                  *   -C  /             if scope is not enough for destination.
1567                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1568                  *
1569                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1570                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1571                  * Assume B = 0 and we get C > 29.
1572                  */
1573                 ret = __ipv6_addr_src_scope(score->addr_type);
1574                 if (ret >= dst->scope)
1575                         ret = -ret;
1576                 else
1577                         ret -= 128;     /* 30 is enough */
1578                 score->scopedist = ret;
1579                 break;
1580         case IPV6_SADDR_RULE_PREFERRED:
1581             {
1582                 /* Rule 3: Avoid deprecated and optimistic addresses */
1583                 u8 avoid = IFA_F_DEPRECATED;
1584
1585                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1586                         avoid |= IFA_F_OPTIMISTIC;
1587                 ret = ipv6_saddr_preferred(score->addr_type) ||
1588                       !(score->ifa->flags & avoid);
1589                 break;
1590             }
1591 #ifdef CONFIG_IPV6_MIP6
1592         case IPV6_SADDR_RULE_HOA:
1593             {
1594                 /* Rule 4: Prefer home address */
1595                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1596                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1597                 break;
1598             }
1599 #endif
1600         case IPV6_SADDR_RULE_OIF:
1601                 /* Rule 5: Prefer outgoing interface */
1602                 ret = (!dst->ifindex ||
1603                        dst->ifindex == score->ifa->idev->dev->ifindex);
1604                 break;
1605         case IPV6_SADDR_RULE_LABEL:
1606                 /* Rule 6: Prefer matching label */
1607                 ret = ipv6_addr_label(net,
1608                                       &score->ifa->addr, score->addr_type,
1609                                       score->ifa->idev->dev->ifindex) == dst->label;
1610                 break;
1611         case IPV6_SADDR_RULE_PRIVACY:
1612             {
1613                 /* Rule 7: Prefer public address
1614                  * Note: prefer temporary address if use_tempaddr >= 2
1615                  */
1616                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1617                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1618                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1619                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1620                 break;
1621             }
1622         case IPV6_SADDR_RULE_ORCHID:
1623                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1624                  *          non-ORCHID vs non-ORCHID
1625                  */
1626                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1627                         ipv6_addr_orchid(dst->addr));
1628                 break;
1629         case IPV6_SADDR_RULE_PREFIX:
1630                 /* Rule 8: Use longest matching prefix */
1631                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1632                 if (ret > score->ifa->prefix_len)
1633                         ret = score->ifa->prefix_len;
1634                 score->matchlen = ret;
1635                 break;
1636 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1637         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1638                 /* Optimistic addresses still have lower precedence than other
1639                  * preferred addresses.
1640                  */
1641                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1642                 break;
1643 #endif
1644         default:
1645                 ret = 0;
1646         }
1647
1648         if (ret)
1649                 __set_bit(i, score->scorebits);
1650         score->rule = i;
1651 out:
1652         return ret;
1653 }
1654
1655 static int __ipv6_dev_get_saddr(struct net *net,
1656                                 struct ipv6_saddr_dst *dst,
1657                                 struct inet6_dev *idev,
1658                                 struct ipv6_saddr_score *scores,
1659                                 int hiscore_idx)
1660 {
1661         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1662
1663         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1664                 int i;
1665
1666                 /*
1667                  * - Tentative Address (RFC2462 section 5.4)
1668                  *  - A tentative address is not considered
1669                  *    "assigned to an interface" in the traditional
1670                  *    sense, unless it is also flagged as optimistic.
1671                  * - Candidate Source Address (section 4)
1672                  *  - In any case, anycast addresses, multicast
1673                  *    addresses, and the unspecified address MUST
1674                  *    NOT be included in a candidate set.
1675                  */
1676                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1677                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1678                         continue;
1679
1680                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1681
1682                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1683                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1684                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1685                                             idev->dev->name);
1686                         continue;
1687                 }
1688
1689                 score->rule = -1;
1690                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1691
1692                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1693                         int minihiscore, miniscore;
1694
1695                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1696                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1697
1698                         if (minihiscore > miniscore) {
1699                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1700                                     score->scopedist > 0) {
1701                                         /*
1702                                          * special case:
1703                                          * each remaining entry
1704                                          * has too small (not enough)
1705                                          * scope, because ifa entries
1706                                          * are sorted by their scope
1707                                          * values.
1708                                          */
1709                                         goto out;
1710                                 }
1711                                 break;
1712                         } else if (minihiscore < miniscore) {
1713                                 swap(hiscore, score);
1714                                 hiscore_idx = 1 - hiscore_idx;
1715
1716                                 /* restore our iterator */
1717                                 score->ifa = hiscore->ifa;
1718
1719                                 break;
1720                         }
1721                 }
1722         }
1723 out:
1724         return hiscore_idx;
1725 }
1726
1727 static int ipv6_get_saddr_master(struct net *net,
1728                                  const struct net_device *dst_dev,
1729                                  const struct net_device *master,
1730                                  struct ipv6_saddr_dst *dst,
1731                                  struct ipv6_saddr_score *scores,
1732                                  int hiscore_idx)
1733 {
1734         struct inet6_dev *idev;
1735
1736         idev = __in6_dev_get(dst_dev);
1737         if (idev)
1738                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1739                                                    scores, hiscore_idx);
1740
1741         idev = __in6_dev_get(master);
1742         if (idev)
1743                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1744                                                    scores, hiscore_idx);
1745
1746         return hiscore_idx;
1747 }
1748
1749 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1750                        const struct in6_addr *daddr, unsigned int prefs,
1751                        struct in6_addr *saddr)
1752 {
1753         struct ipv6_saddr_score scores[2], *hiscore;
1754         struct ipv6_saddr_dst dst;
1755         struct inet6_dev *idev;
1756         struct net_device *dev;
1757         int dst_type;
1758         bool use_oif_addr = false;
1759         int hiscore_idx = 0;
1760         int ret = 0;
1761
1762         dst_type = __ipv6_addr_type(daddr);
1763         dst.addr = daddr;
1764         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1765         dst.scope = __ipv6_addr_src_scope(dst_type);
1766         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1767         dst.prefs = prefs;
1768
1769         scores[hiscore_idx].rule = -1;
1770         scores[hiscore_idx].ifa = NULL;
1771
1772         rcu_read_lock();
1773
1774         /* Candidate Source Address (section 4)
1775          *  - multicast and link-local destination address,
1776          *    the set of candidate source address MUST only
1777          *    include addresses assigned to interfaces
1778          *    belonging to the same link as the outgoing
1779          *    interface.
1780          * (- For site-local destination addresses, the
1781          *    set of candidate source addresses MUST only
1782          *    include addresses assigned to interfaces
1783          *    belonging to the same site as the outgoing
1784          *    interface.)
1785          *  - "It is RECOMMENDED that the candidate source addresses
1786          *    be the set of unicast addresses assigned to the
1787          *    interface that will be used to send to the destination
1788          *    (the 'outgoing' interface)." (RFC 6724)
1789          */
1790         if (dst_dev) {
1791                 idev = __in6_dev_get(dst_dev);
1792                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1793                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1794                     (idev && idev->cnf.use_oif_addrs_only)) {
1795                         use_oif_addr = true;
1796                 }
1797         }
1798
1799         if (use_oif_addr) {
1800                 if (idev)
1801                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1802         } else {
1803                 const struct net_device *master;
1804                 int master_idx = 0;
1805
1806                 /* if dst_dev exists and is enslaved to an L3 device, then
1807                  * prefer addresses from dst_dev and then the master over
1808                  * any other enslaved devices in the L3 domain.
1809                  */
1810                 master = l3mdev_master_dev_rcu(dst_dev);
1811                 if (master) {
1812                         master_idx = master->ifindex;
1813
1814                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1815                                                             master, &dst,
1816                                                             scores, hiscore_idx);
1817
1818                         if (scores[hiscore_idx].ifa)
1819                                 goto out;
1820                 }
1821
1822                 for_each_netdev_rcu(net, dev) {
1823                         /* only consider addresses on devices in the
1824                          * same L3 domain
1825                          */
1826                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1827                                 continue;
1828                         idev = __in6_dev_get(dev);
1829                         if (!idev)
1830                                 continue;
1831                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1832                 }
1833         }
1834
1835 out:
1836         hiscore = &scores[hiscore_idx];
1837         if (!hiscore->ifa)
1838                 ret = -EADDRNOTAVAIL;
1839         else
1840                 *saddr = hiscore->ifa->addr;
1841
1842         rcu_read_unlock();
1843         return ret;
1844 }
1845 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1846
1847 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1848                       u32 banned_flags)
1849 {
1850         struct inet6_ifaddr *ifp;
1851         int err = -EADDRNOTAVAIL;
1852
1853         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1854                 if (ifp->scope > IFA_LINK)
1855                         break;
1856                 if (ifp->scope == IFA_LINK &&
1857                     !(ifp->flags & banned_flags)) {
1858                         *addr = ifp->addr;
1859                         err = 0;
1860                         break;
1861                 }
1862         }
1863         return err;
1864 }
1865
1866 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1867                     u32 banned_flags)
1868 {
1869         struct inet6_dev *idev;
1870         int err = -EADDRNOTAVAIL;
1871
1872         rcu_read_lock();
1873         idev = __in6_dev_get(dev);
1874         if (idev) {
1875                 read_lock_bh(&idev->lock);
1876                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1877                 read_unlock_bh(&idev->lock);
1878         }
1879         rcu_read_unlock();
1880         return err;
1881 }
1882
1883 static int ipv6_count_addresses(const struct inet6_dev *idev)
1884 {
1885         const struct inet6_ifaddr *ifp;
1886         int cnt = 0;
1887
1888         rcu_read_lock();
1889         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1890                 cnt++;
1891         rcu_read_unlock();
1892         return cnt;
1893 }
1894
1895 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1896                   const struct net_device *dev, int strict)
1897 {
1898         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1899                                        strict, IFA_F_TENTATIVE);
1900 }
1901 EXPORT_SYMBOL(ipv6_chk_addr);
1902
1903 /* device argument is used to find the L3 domain of interest. If
1904  * skip_dev_check is set, then the ifp device is not checked against
1905  * the passed in dev argument. So the 2 cases for addresses checks are:
1906  *   1. does the address exist in the L3 domain that dev is part of
1907  *      (skip_dev_check = true), or
1908  *
1909  *   2. does the address exist on the specific device
1910  *      (skip_dev_check = false)
1911  */
1912 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1913                             const struct net_device *dev, bool skip_dev_check,
1914                             int strict, u32 banned_flags)
1915 {
1916         unsigned int hash = inet6_addr_hash(net, addr);
1917         const struct net_device *l3mdev;
1918         struct inet6_ifaddr *ifp;
1919         u32 ifp_flags;
1920
1921         rcu_read_lock();
1922
1923         l3mdev = l3mdev_master_dev_rcu(dev);
1924         if (skip_dev_check)
1925                 dev = NULL;
1926
1927         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1928                 if (!net_eq(dev_net(ifp->idev->dev), net))
1929                         continue;
1930
1931                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1932                         continue;
1933
1934                 /* Decouple optimistic from tentative for evaluation here.
1935                  * Ban optimistic addresses explicitly, when required.
1936                  */
1937                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1938                             ? (ifp->flags&~IFA_F_TENTATIVE)
1939                             : ifp->flags;
1940                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1941                     !(ifp_flags&banned_flags) &&
1942                     (!dev || ifp->idev->dev == dev ||
1943                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1944                         rcu_read_unlock();
1945                         return 1;
1946                 }
1947         }
1948
1949         rcu_read_unlock();
1950         return 0;
1951 }
1952 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1953
1954
1955 /* Compares an address/prefix_len with addresses on device @dev.
1956  * If one is found it returns true.
1957  */
1958 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1959         const unsigned int prefix_len, struct net_device *dev)
1960 {
1961         const struct inet6_ifaddr *ifa;
1962         const struct inet6_dev *idev;
1963         bool ret = false;
1964
1965         rcu_read_lock();
1966         idev = __in6_dev_get(dev);
1967         if (idev) {
1968                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1969                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1970                         if (ret)
1971                                 break;
1972                 }
1973         }
1974         rcu_read_unlock();
1975
1976         return ret;
1977 }
1978 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1979
1980 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1981 {
1982         const struct inet6_ifaddr *ifa;
1983         const struct inet6_dev *idev;
1984         int     onlink;
1985
1986         onlink = 0;
1987         rcu_read_lock();
1988         idev = __in6_dev_get(dev);
1989         if (idev) {
1990                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1991                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1992                                                    ifa->prefix_len);
1993                         if (onlink)
1994                                 break;
1995                 }
1996         }
1997         rcu_read_unlock();
1998         return onlink;
1999 }
2000 EXPORT_SYMBOL(ipv6_chk_prefix);
2001
2002 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2003                                      struct net_device *dev, int strict)
2004 {
2005         unsigned int hash = inet6_addr_hash(net, addr);
2006         struct inet6_ifaddr *ifp, *result = NULL;
2007
2008         rcu_read_lock();
2009         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2010                 if (!net_eq(dev_net(ifp->idev->dev), net))
2011                         continue;
2012                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2013                         if (!dev || ifp->idev->dev == dev ||
2014                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2015                                 result = ifp;
2016                                 in6_ifa_hold(ifp);
2017                                 break;
2018                         }
2019                 }
2020         }
2021         rcu_read_unlock();
2022
2023         return result;
2024 }
2025
2026 /* Gets referenced address, destroys ifaddr */
2027
2028 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2029 {
2030         if (dad_failed)
2031                 ifp->flags |= IFA_F_DADFAILED;
2032
2033         if (ifp->flags&IFA_F_TEMPORARY) {
2034                 struct inet6_ifaddr *ifpub;
2035                 spin_lock_bh(&ifp->lock);
2036                 ifpub = ifp->ifpub;
2037                 if (ifpub) {
2038                         in6_ifa_hold(ifpub);
2039                         spin_unlock_bh(&ifp->lock);
2040                         ipv6_create_tempaddr(ifpub, ifp, true);
2041                         in6_ifa_put(ifpub);
2042                 } else {
2043                         spin_unlock_bh(&ifp->lock);
2044                 }
2045                 ipv6_del_addr(ifp);
2046         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2047                 spin_lock_bh(&ifp->lock);
2048                 addrconf_del_dad_work(ifp);
2049                 ifp->flags |= IFA_F_TENTATIVE;
2050                 if (dad_failed)
2051                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2052                 spin_unlock_bh(&ifp->lock);
2053                 if (dad_failed)
2054                         ipv6_ifa_notify(0, ifp);
2055                 in6_ifa_put(ifp);
2056         } else {
2057                 ipv6_del_addr(ifp);
2058         }
2059 }
2060
2061 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2062 {
2063         int err = -ENOENT;
2064
2065         spin_lock_bh(&ifp->lock);
2066         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2067                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2068                 err = 0;
2069         }
2070         spin_unlock_bh(&ifp->lock);
2071
2072         return err;
2073 }
2074
2075 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2076 {
2077         struct inet6_dev *idev = ifp->idev;
2078         struct net *net = dev_net(ifp->idev->dev);
2079
2080         if (addrconf_dad_end(ifp)) {
2081                 in6_ifa_put(ifp);
2082                 return;
2083         }
2084
2085         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2086                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2087
2088         spin_lock_bh(&ifp->lock);
2089
2090         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2091                 struct in6_addr new_addr;
2092                 struct inet6_ifaddr *ifp2;
2093                 int retries = ifp->stable_privacy_retry + 1;
2094                 struct ifa6_config cfg = {
2095                         .pfx = &new_addr,
2096                         .plen = ifp->prefix_len,
2097                         .ifa_flags = ifp->flags,
2098                         .valid_lft = ifp->valid_lft,
2099                         .preferred_lft = ifp->prefered_lft,
2100                         .scope = ifp->scope,
2101                 };
2102
2103                 if (retries > net->ipv6.sysctl.idgen_retries) {
2104                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2105                                              ifp->idev->dev->name);
2106                         goto errdad;
2107                 }
2108
2109                 new_addr = ifp->addr;
2110                 if (ipv6_generate_stable_address(&new_addr, retries,
2111                                                  idev))
2112                         goto errdad;
2113
2114                 spin_unlock_bh(&ifp->lock);
2115
2116                 if (idev->cnf.max_addresses &&
2117                     ipv6_count_addresses(idev) >=
2118                     idev->cnf.max_addresses)
2119                         goto lock_errdad;
2120
2121                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2122                                      ifp->idev->dev->name);
2123
2124                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2125                 if (IS_ERR(ifp2))
2126                         goto lock_errdad;
2127
2128                 spin_lock_bh(&ifp2->lock);
2129                 ifp2->stable_privacy_retry = retries;
2130                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2131                 spin_unlock_bh(&ifp2->lock);
2132
2133                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2134                 in6_ifa_put(ifp2);
2135 lock_errdad:
2136                 spin_lock_bh(&ifp->lock);
2137         }
2138
2139 errdad:
2140         /* transition from _POSTDAD to _ERRDAD */
2141         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2142         spin_unlock_bh(&ifp->lock);
2143
2144         addrconf_mod_dad_work(ifp, 0);
2145         in6_ifa_put(ifp);
2146 }
2147
2148 /* Join to solicited addr multicast group.
2149  * caller must hold RTNL */
2150 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2151 {
2152         struct in6_addr maddr;
2153
2154         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2155                 return;
2156
2157         addrconf_addr_solict_mult(addr, &maddr);
2158         ipv6_dev_mc_inc(dev, &maddr);
2159 }
2160
2161 /* caller must hold RTNL */
2162 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2163 {
2164         struct in6_addr maddr;
2165
2166         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2167                 return;
2168
2169         addrconf_addr_solict_mult(addr, &maddr);
2170         __ipv6_dev_mc_dec(idev, &maddr);
2171 }
2172
2173 /* caller must hold RTNL */
2174 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2175 {
2176         struct in6_addr addr;
2177
2178         if (ifp->prefix_len >= 127) /* RFC 6164 */
2179                 return;
2180         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2181         if (ipv6_addr_any(&addr))
2182                 return;
2183         __ipv6_dev_ac_inc(ifp->idev, &addr);
2184 }
2185
2186 /* caller must hold RTNL */
2187 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2188 {
2189         struct in6_addr addr;
2190
2191         if (ifp->prefix_len >= 127) /* RFC 6164 */
2192                 return;
2193         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2194         if (ipv6_addr_any(&addr))
2195                 return;
2196         __ipv6_dev_ac_dec(ifp->idev, &addr);
2197 }
2198
2199 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2200 {
2201         switch (dev->addr_len) {
2202         case ETH_ALEN:
2203                 memcpy(eui, dev->dev_addr, 3);
2204                 eui[3] = 0xFF;
2205                 eui[4] = 0xFE;
2206                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2207                 break;
2208         case EUI64_ADDR_LEN:
2209                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2210                 eui[0] ^= 2;
2211                 break;
2212         default:
2213                 return -1;
2214         }
2215
2216         return 0;
2217 }
2218
2219 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2220 {
2221         union fwnet_hwaddr *ha;
2222
2223         if (dev->addr_len != FWNET_ALEN)
2224                 return -1;
2225
2226         ha = (union fwnet_hwaddr *)dev->dev_addr;
2227
2228         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2229         eui[0] ^= 2;
2230         return 0;
2231 }
2232
2233 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2234 {
2235         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2236         if (dev->addr_len != ARCNET_ALEN)
2237                 return -1;
2238         memset(eui, 0, 7);
2239         eui[7] = *(u8 *)dev->dev_addr;
2240         return 0;
2241 }
2242
2243 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2244 {
2245         if (dev->addr_len != INFINIBAND_ALEN)
2246                 return -1;
2247         memcpy(eui, dev->dev_addr + 12, 8);
2248         eui[0] |= 2;
2249         return 0;
2250 }
2251
2252 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2253 {
2254         if (addr == 0)
2255                 return -1;
2256         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2257                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2258                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2259                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2260                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2261                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2262         eui[1] = 0;
2263         eui[2] = 0x5E;
2264         eui[3] = 0xFE;
2265         memcpy(eui + 4, &addr, 4);
2266         return 0;
2267 }
2268
2269 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2270 {
2271         if (dev->priv_flags & IFF_ISATAP)
2272                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2273         return -1;
2274 }
2275
2276 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2277 {
2278         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2279 }
2280
2281 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2282 {
2283         memcpy(eui, dev->perm_addr, 3);
2284         memcpy(eui + 5, dev->perm_addr + 3, 3);
2285         eui[3] = 0xFF;
2286         eui[4] = 0xFE;
2287         eui[0] ^= 2;
2288         return 0;
2289 }
2290
2291 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2292 {
2293         switch (dev->type) {
2294         case ARPHRD_ETHER:
2295         case ARPHRD_FDDI:
2296                 return addrconf_ifid_eui48(eui, dev);
2297         case ARPHRD_ARCNET:
2298                 return addrconf_ifid_arcnet(eui, dev);
2299         case ARPHRD_INFINIBAND:
2300                 return addrconf_ifid_infiniband(eui, dev);
2301         case ARPHRD_SIT:
2302                 return addrconf_ifid_sit(eui, dev);
2303         case ARPHRD_IPGRE:
2304         case ARPHRD_TUNNEL:
2305                 return addrconf_ifid_gre(eui, dev);
2306         case ARPHRD_6LOWPAN:
2307                 return addrconf_ifid_6lowpan(eui, dev);
2308         case ARPHRD_IEEE1394:
2309                 return addrconf_ifid_ieee1394(eui, dev);
2310         case ARPHRD_TUNNEL6:
2311         case ARPHRD_IP6GRE:
2312         case ARPHRD_RAWIP:
2313                 return addrconf_ifid_ip6tnl(eui, dev);
2314         }
2315         return -1;
2316 }
2317
2318 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2319 {
2320         int err = -1;
2321         struct inet6_ifaddr *ifp;
2322
2323         read_lock_bh(&idev->lock);
2324         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2325                 if (ifp->scope > IFA_LINK)
2326                         break;
2327                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2328                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2329                         err = 0;
2330                         break;
2331                 }
2332         }
2333         read_unlock_bh(&idev->lock);
2334         return err;
2335 }
2336
2337 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2338 static void ipv6_regen_rndid(struct inet6_dev *idev)
2339 {
2340 regen:
2341         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2342         idev->rndid[0] &= ~0x02;
2343
2344         /*
2345          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2346          * check if generated address is not inappropriate
2347          *
2348          *  - Reserved subnet anycast (RFC 2526)
2349          *      11111101 11....11 1xxxxxxx
2350          *  - ISATAP (RFC4214) 6.1
2351          *      00-00-5E-FE-xx-xx-xx-xx
2352          *  - value 0
2353          *  - XXX: already assigned to an address on the device
2354          */
2355         if (idev->rndid[0] == 0xfd &&
2356             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2357             (idev->rndid[7]&0x80))
2358                 goto regen;
2359         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2360                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2361                         goto regen;
2362                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2363                         goto regen;
2364         }
2365 }
2366
2367 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2368 {
2369         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2370                 ipv6_regen_rndid(idev);
2371 }
2372
2373 /*
2374  *      Add prefix route.
2375  */
2376
2377 static void
2378 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2379                       struct net_device *dev, unsigned long expires,
2380                       u32 flags, gfp_t gfp_flags)
2381 {
2382         struct fib6_config cfg = {
2383                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2384                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2385                 .fc_ifindex = dev->ifindex,
2386                 .fc_expires = expires,
2387                 .fc_dst_len = plen,
2388                 .fc_flags = RTF_UP | flags,
2389                 .fc_nlinfo.nl_net = dev_net(dev),
2390                 .fc_protocol = RTPROT_KERNEL,
2391                 .fc_type = RTN_UNICAST,
2392         };
2393
2394         cfg.fc_dst = *pfx;
2395
2396         /* Prevent useless cloning on PtP SIT.
2397            This thing is done here expecting that the whole
2398            class of non-broadcast devices need not cloning.
2399          */
2400 #if IS_ENABLED(CONFIG_IPV6_SIT)
2401         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2402                 cfg.fc_flags |= RTF_NONEXTHOP;
2403 #endif
2404
2405         ip6_route_add(&cfg, gfp_flags, NULL);
2406 }
2407
2408
2409 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2410                                                   int plen,
2411                                                   const struct net_device *dev,
2412                                                   u32 flags, u32 noflags,
2413                                                   bool no_gw)
2414 {
2415         struct fib6_node *fn;
2416         struct fib6_info *rt = NULL;
2417         struct fib6_table *table;
2418         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2419
2420         table = fib6_get_table(dev_net(dev), tb_id);
2421         if (!table)
2422                 return NULL;
2423
2424         rcu_read_lock();
2425         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2426         if (!fn)
2427                 goto out;
2428
2429         for_each_fib6_node_rt_rcu(fn) {
2430                 /* prefix routes only use builtin fib6_nh */
2431                 if (rt->nh)
2432                         continue;
2433
2434                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2435                         continue;
2436                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2437                         continue;
2438                 if ((rt->fib6_flags & flags) != flags)
2439                         continue;
2440                 if ((rt->fib6_flags & noflags) != 0)
2441                         continue;
2442                 if (!fib6_info_hold_safe(rt))
2443                         continue;
2444                 break;
2445         }
2446 out:
2447         rcu_read_unlock();
2448         return rt;
2449 }
2450
2451
2452 /* Create "default" multicast route to the interface */
2453
2454 static void addrconf_add_mroute(struct net_device *dev)
2455 {
2456         struct fib6_config cfg = {
2457                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2458                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2459                 .fc_ifindex = dev->ifindex,
2460                 .fc_dst_len = 8,
2461                 .fc_flags = RTF_UP,
2462                 .fc_type = RTN_MULTICAST,
2463                 .fc_nlinfo.nl_net = dev_net(dev),
2464                 .fc_protocol = RTPROT_KERNEL,
2465         };
2466
2467         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2468
2469         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2470 }
2471
2472 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2473 {
2474         struct inet6_dev *idev;
2475
2476         ASSERT_RTNL();
2477
2478         idev = ipv6_find_idev(dev);
2479         if (IS_ERR(idev))
2480                 return idev;
2481
2482         if (idev->cnf.disable_ipv6)
2483                 return ERR_PTR(-EACCES);
2484
2485         /* Add default multicast route */
2486         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2487                 addrconf_add_mroute(dev);
2488
2489         return idev;
2490 }
2491
2492 static void manage_tempaddrs(struct inet6_dev *idev,
2493                              struct inet6_ifaddr *ifp,
2494                              __u32 valid_lft, __u32 prefered_lft,
2495                              bool create, unsigned long now)
2496 {
2497         u32 flags;
2498         struct inet6_ifaddr *ift;
2499
2500         read_lock_bh(&idev->lock);
2501         /* update all temporary addresses in the list */
2502         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2503                 int age, max_valid, max_prefered;
2504
2505                 if (ifp != ift->ifpub)
2506                         continue;
2507
2508                 /* RFC 4941 section 3.3:
2509                  * If a received option will extend the lifetime of a public
2510                  * address, the lifetimes of temporary addresses should
2511                  * be extended, subject to the overall constraint that no
2512                  * temporary addresses should ever remain "valid" or "preferred"
2513                  * for a time longer than (TEMP_VALID_LIFETIME) or
2514                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2515                  */
2516                 age = (now - ift->cstamp) / HZ;
2517                 max_valid = idev->cnf.temp_valid_lft - age;
2518                 if (max_valid < 0)
2519                         max_valid = 0;
2520
2521                 max_prefered = idev->cnf.temp_prefered_lft -
2522                                idev->desync_factor - age;
2523                 if (max_prefered < 0)
2524                         max_prefered = 0;
2525
2526                 if (valid_lft > max_valid)
2527                         valid_lft = max_valid;
2528
2529                 if (prefered_lft > max_prefered)
2530                         prefered_lft = max_prefered;
2531
2532                 spin_lock(&ift->lock);
2533                 flags = ift->flags;
2534                 ift->valid_lft = valid_lft;
2535                 ift->prefered_lft = prefered_lft;
2536                 ift->tstamp = now;
2537                 if (prefered_lft > 0)
2538                         ift->flags &= ~IFA_F_DEPRECATED;
2539
2540                 spin_unlock(&ift->lock);
2541                 if (!(flags&IFA_F_TENTATIVE))
2542                         ipv6_ifa_notify(0, ift);
2543         }
2544
2545         if ((create || list_empty(&idev->tempaddr_list)) &&
2546             idev->cnf.use_tempaddr > 0) {
2547                 /* When a new public address is created as described
2548                  * in [ADDRCONF], also create a new temporary address.
2549                  * Also create a temporary address if it's enabled but
2550                  * no temporary address currently exists.
2551                  */
2552                 read_unlock_bh(&idev->lock);
2553                 ipv6_create_tempaddr(ifp, NULL, false);
2554         } else {
2555                 read_unlock_bh(&idev->lock);
2556         }
2557 }
2558
2559 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2560 {
2561         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2562                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2563 }
2564
2565 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2566                                  const struct prefix_info *pinfo,
2567                                  struct inet6_dev *in6_dev,
2568                                  const struct in6_addr *addr, int addr_type,
2569                                  u32 addr_flags, bool sllao, bool tokenized,
2570                                  __u32 valid_lft, u32 prefered_lft)
2571 {
2572         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2573         int create = 0, update_lft = 0;
2574
2575         if (!ifp && valid_lft) {
2576                 int max_addresses = in6_dev->cnf.max_addresses;
2577                 struct ifa6_config cfg = {
2578                         .pfx = addr,
2579                         .plen = pinfo->prefix_len,
2580                         .ifa_flags = addr_flags,
2581                         .valid_lft = valid_lft,
2582                         .preferred_lft = prefered_lft,
2583                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2584                 };
2585
2586 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2587                 if ((net->ipv6.devconf_all->optimistic_dad ||
2588                      in6_dev->cnf.optimistic_dad) &&
2589                     !net->ipv6.devconf_all->forwarding && sllao)
2590                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2591 #endif
2592
2593                 /* Do not allow to create too much of autoconfigured
2594                  * addresses; this would be too easy way to crash kernel.
2595                  */
2596                 if (!max_addresses ||
2597                     ipv6_count_addresses(in6_dev) < max_addresses)
2598                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2599
2600                 if (IS_ERR_OR_NULL(ifp))
2601                         return -1;
2602
2603                 create = 1;
2604                 spin_lock_bh(&ifp->lock);
2605                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2606                 ifp->cstamp = jiffies;
2607                 ifp->tokenized = tokenized;
2608                 spin_unlock_bh(&ifp->lock);
2609                 addrconf_dad_start(ifp);
2610         }
2611
2612         if (ifp) {
2613                 u32 flags;
2614                 unsigned long now;
2615                 u32 stored_lft;
2616
2617                 /* update lifetime (RFC2462 5.5.3 e) */
2618                 spin_lock_bh(&ifp->lock);
2619                 now = jiffies;
2620                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2621                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2622                 else
2623                         stored_lft = 0;
2624                 if (!create && stored_lft) {
2625                         const u32 minimum_lft = min_t(u32,
2626                                 stored_lft, MIN_VALID_LIFETIME);
2627                         valid_lft = max(valid_lft, minimum_lft);
2628
2629                         /* RFC4862 Section 5.5.3e:
2630                          * "Note that the preferred lifetime of the
2631                          *  corresponding address is always reset to
2632                          *  the Preferred Lifetime in the received
2633                          *  Prefix Information option, regardless of
2634                          *  whether the valid lifetime is also reset or
2635                          *  ignored."
2636                          *
2637                          * So we should always update prefered_lft here.
2638                          */
2639                         update_lft = 1;
2640                 }
2641
2642                 if (update_lft) {
2643                         ifp->valid_lft = valid_lft;
2644                         ifp->prefered_lft = prefered_lft;
2645                         ifp->tstamp = now;
2646                         flags = ifp->flags;
2647                         ifp->flags &= ~IFA_F_DEPRECATED;
2648                         spin_unlock_bh(&ifp->lock);
2649
2650                         if (!(flags&IFA_F_TENTATIVE))
2651                                 ipv6_ifa_notify(0, ifp);
2652                 } else
2653                         spin_unlock_bh(&ifp->lock);
2654
2655                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2656                                  create, now);
2657
2658                 in6_ifa_put(ifp);
2659                 addrconf_verify();
2660         }
2661
2662         return 0;
2663 }
2664 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2665
2666 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2667 {
2668         struct prefix_info *pinfo;
2669         __u32 valid_lft;
2670         __u32 prefered_lft;
2671         int addr_type, err;
2672         u32 addr_flags = 0;
2673         struct inet6_dev *in6_dev;
2674         struct net *net = dev_net(dev);
2675
2676         pinfo = (struct prefix_info *) opt;
2677
2678         if (len < sizeof(struct prefix_info)) {
2679                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2680                 return;
2681         }
2682
2683         /*
2684          *      Validation checks ([ADDRCONF], page 19)
2685          */
2686
2687         addr_type = ipv6_addr_type(&pinfo->prefix);
2688
2689         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2690                 return;
2691
2692         valid_lft = ntohl(pinfo->valid);
2693         prefered_lft = ntohl(pinfo->prefered);
2694
2695         if (prefered_lft > valid_lft) {
2696                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2697                 return;
2698         }
2699
2700         in6_dev = in6_dev_get(dev);
2701
2702         if (!in6_dev) {
2703                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2704                                     dev->name);
2705                 return;
2706         }
2707
2708         /*
2709          *      Two things going on here:
2710          *      1) Add routes for on-link prefixes
2711          *      2) Configure prefixes with the auto flag set
2712          */
2713
2714         if (pinfo->onlink) {
2715                 struct fib6_info *rt;
2716                 unsigned long rt_expires;
2717
2718                 /* Avoid arithmetic overflow. Really, we could
2719                  * save rt_expires in seconds, likely valid_lft,
2720                  * but it would require division in fib gc, that it
2721                  * not good.
2722                  */
2723                 if (HZ > USER_HZ)
2724                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2725                 else
2726                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2727
2728                 if (addrconf_finite_timeout(rt_expires))
2729                         rt_expires *= HZ;
2730
2731                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2732                                                pinfo->prefix_len,
2733                                                dev,
2734                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2735                                                RTF_DEFAULT, true);
2736
2737                 if (rt) {
2738                         /* Autoconf prefix route */
2739                         if (valid_lft == 0) {
2740                                 ip6_del_rt(net, rt);
2741                                 rt = NULL;
2742                         } else if (addrconf_finite_timeout(rt_expires)) {
2743                                 /* not infinity */
2744                                 fib6_set_expires(rt, jiffies + rt_expires);
2745                         } else {
2746                                 fib6_clean_expires(rt);
2747                         }
2748                 } else if (valid_lft) {
2749                         clock_t expires = 0;
2750                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2751                         if (addrconf_finite_timeout(rt_expires)) {
2752                                 /* not infinity */
2753                                 flags |= RTF_EXPIRES;
2754                                 expires = jiffies_to_clock_t(rt_expires);
2755                         }
2756                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2757                                               0, dev, expires, flags,
2758                                               GFP_ATOMIC);
2759                 }
2760                 fib6_info_release(rt);
2761         }
2762
2763         /* Try to figure out our local address for this prefix */
2764
2765         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2766                 struct in6_addr addr;
2767                 bool tokenized = false, dev_addr_generated = false;
2768
2769                 if (pinfo->prefix_len == 64) {
2770                         memcpy(&addr, &pinfo->prefix, 8);
2771
2772                         if (!ipv6_addr_any(&in6_dev->token)) {
2773                                 read_lock_bh(&in6_dev->lock);
2774                                 memcpy(addr.s6_addr + 8,
2775                                        in6_dev->token.s6_addr + 8, 8);
2776                                 read_unlock_bh(&in6_dev->lock);
2777                                 tokenized = true;
2778                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2779                                    !ipv6_generate_stable_address(&addr, 0,
2780                                                                  in6_dev)) {
2781                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2782                                 goto ok;
2783                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2784                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2785                                 goto put;
2786                         } else {
2787                                 dev_addr_generated = true;
2788                         }
2789                         goto ok;
2790                 }
2791                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2792                                     pinfo->prefix_len);
2793                 goto put;
2794
2795 ok:
2796                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2797                                                    &addr, addr_type,
2798                                                    addr_flags, sllao,
2799                                                    tokenized, valid_lft,
2800                                                    prefered_lft);
2801                 if (err)
2802                         goto put;
2803
2804                 /* Ignore error case here because previous prefix add addr was
2805                  * successful which will be notified.
2806                  */
2807                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2808                                               addr_type, addr_flags, sllao,
2809                                               tokenized, valid_lft,
2810                                               prefered_lft,
2811                                               dev_addr_generated);
2812         }
2813         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2814 put:
2815         in6_dev_put(in6_dev);
2816 }
2817
2818 /*
2819  *      Set destination address.
2820  *      Special case for SIT interfaces where we create a new "virtual"
2821  *      device.
2822  */
2823 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2824 {
2825         struct in6_ifreq ireq;
2826         struct net_device *dev;
2827         int err = -EINVAL;
2828
2829         rtnl_lock();
2830
2831         err = -EFAULT;
2832         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2833                 goto err_exit;
2834
2835         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2836
2837         err = -ENODEV;
2838         if (!dev)
2839                 goto err_exit;
2840
2841 #if IS_ENABLED(CONFIG_IPV6_SIT)
2842         if (dev->type == ARPHRD_SIT) {
2843                 const struct net_device_ops *ops = dev->netdev_ops;
2844                 struct ifreq ifr;
2845                 struct ip_tunnel_parm p;
2846
2847                 err = -EADDRNOTAVAIL;
2848                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2849                         goto err_exit;
2850
2851                 memset(&p, 0, sizeof(p));
2852                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2853                 p.iph.saddr = 0;
2854                 p.iph.version = 4;
2855                 p.iph.ihl = 5;
2856                 p.iph.protocol = IPPROTO_IPV6;
2857                 p.iph.ttl = 64;
2858                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2859
2860                 if (ops->ndo_do_ioctl) {
2861                         mm_segment_t oldfs = get_fs();
2862
2863                         set_fs(KERNEL_DS);
2864                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2865                         set_fs(oldfs);
2866                 } else
2867                         err = -EOPNOTSUPP;
2868
2869                 if (err == 0) {
2870                         err = -ENOBUFS;
2871                         dev = __dev_get_by_name(net, p.name);
2872                         if (!dev)
2873                                 goto err_exit;
2874                         err = dev_open(dev, NULL);
2875                 }
2876         }
2877 #endif
2878
2879 err_exit:
2880         rtnl_unlock();
2881         return err;
2882 }
2883
2884 static int ipv6_mc_config(struct sock *sk, bool join,
2885                           const struct in6_addr *addr, int ifindex)
2886 {
2887         int ret;
2888
2889         ASSERT_RTNL();
2890
2891         lock_sock(sk);
2892         if (join)
2893                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2894         else
2895                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2896         release_sock(sk);
2897
2898         return ret;
2899 }
2900
2901 /*
2902  *      Manual configuration of address on an interface
2903  */
2904 static int inet6_addr_add(struct net *net, int ifindex,
2905                           struct ifa6_config *cfg,
2906                           struct netlink_ext_ack *extack)
2907 {
2908         struct inet6_ifaddr *ifp;
2909         struct inet6_dev *idev;
2910         struct net_device *dev;
2911         unsigned long timeout;
2912         clock_t expires;
2913         u32 flags;
2914
2915         ASSERT_RTNL();
2916
2917         if (cfg->plen > 128)
2918                 return -EINVAL;
2919
2920         /* check the lifetime */
2921         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2922                 return -EINVAL;
2923
2924         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2925                 return -EINVAL;
2926
2927         dev = __dev_get_by_index(net, ifindex);
2928         if (!dev)
2929                 return -ENODEV;
2930
2931         idev = addrconf_add_dev(dev);
2932         if (IS_ERR(idev))
2933                 return PTR_ERR(idev);
2934
2935         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2936                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2937                                          true, cfg->pfx, ifindex);
2938
2939                 if (ret < 0)
2940                         return ret;
2941         }
2942
2943         cfg->scope = ipv6_addr_scope(cfg->pfx);
2944
2945         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2946         if (addrconf_finite_timeout(timeout)) {
2947                 expires = jiffies_to_clock_t(timeout * HZ);
2948                 cfg->valid_lft = timeout;
2949                 flags = RTF_EXPIRES;
2950         } else {
2951                 expires = 0;
2952                 flags = 0;
2953                 cfg->ifa_flags |= IFA_F_PERMANENT;
2954         }
2955
2956         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2957         if (addrconf_finite_timeout(timeout)) {
2958                 if (timeout == 0)
2959                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2960                 cfg->preferred_lft = timeout;
2961         }
2962
2963         ifp = ipv6_add_addr(idev, cfg, true, extack);
2964         if (!IS_ERR(ifp)) {
2965                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2966                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2967                                               ifp->rt_priority, dev, expires,
2968                                               flags, GFP_KERNEL);
2969                 }
2970
2971                 /* Send a netlink notification if DAD is enabled and
2972                  * optimistic flag is not set
2973                  */
2974                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2975                         ipv6_ifa_notify(0, ifp);
2976                 /*
2977                  * Note that section 3.1 of RFC 4429 indicates
2978                  * that the Optimistic flag should not be set for
2979                  * manually configured addresses
2980                  */
2981                 addrconf_dad_start(ifp);
2982                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2983                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2984                                          cfg->preferred_lft, true, jiffies);
2985                 in6_ifa_put(ifp);
2986                 addrconf_verify_rtnl();
2987                 return 0;
2988         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2989                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2990                                cfg->pfx, ifindex);
2991         }
2992
2993         return PTR_ERR(ifp);
2994 }
2995
2996 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2997                           const struct in6_addr *pfx, unsigned int plen)
2998 {
2999         struct inet6_ifaddr *ifp;
3000         struct inet6_dev *idev;
3001         struct net_device *dev;
3002
3003         if (plen > 128)
3004                 return -EINVAL;
3005
3006         dev = __dev_get_by_index(net, ifindex);
3007         if (!dev)
3008                 return -ENODEV;
3009
3010         idev = __in6_dev_get(dev);
3011         if (!idev)
3012                 return -ENXIO;
3013
3014         read_lock_bh(&idev->lock);
3015         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3016                 if (ifp->prefix_len == plen &&
3017                     ipv6_addr_equal(pfx, &ifp->addr)) {
3018                         in6_ifa_hold(ifp);
3019                         read_unlock_bh(&idev->lock);
3020
3021                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3022                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3023                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3024                                                  jiffies);
3025                         ipv6_del_addr(ifp);
3026                         addrconf_verify_rtnl();
3027                         if (ipv6_addr_is_multicast(pfx)) {
3028                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3029                                                false, pfx, dev->ifindex);
3030                         }
3031                         return 0;
3032                 }
3033         }
3034         read_unlock_bh(&idev->lock);
3035         return -EADDRNOTAVAIL;
3036 }
3037
3038
3039 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3040 {
3041         struct ifa6_config cfg = {
3042                 .ifa_flags = IFA_F_PERMANENT,
3043                 .preferred_lft = INFINITY_LIFE_TIME,
3044                 .valid_lft = INFINITY_LIFE_TIME,
3045         };
3046         struct in6_ifreq ireq;
3047         int err;
3048
3049         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3050                 return -EPERM;
3051
3052         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3053                 return -EFAULT;
3054
3055         cfg.pfx = &ireq.ifr6_addr;
3056         cfg.plen = ireq.ifr6_prefixlen;
3057
3058         rtnl_lock();
3059         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3060         rtnl_unlock();
3061         return err;
3062 }
3063
3064 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3065 {
3066         struct in6_ifreq ireq;
3067         int err;
3068
3069         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3070                 return -EPERM;
3071
3072         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3073                 return -EFAULT;
3074
3075         rtnl_lock();
3076         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3077                              ireq.ifr6_prefixlen);
3078         rtnl_unlock();
3079         return err;
3080 }
3081
3082 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3083                      int plen, int scope)
3084 {
3085         struct inet6_ifaddr *ifp;
3086         struct ifa6_config cfg = {
3087                 .pfx = addr,
3088                 .plen = plen,
3089                 .ifa_flags = IFA_F_PERMANENT,
3090                 .valid_lft = INFINITY_LIFE_TIME,
3091                 .preferred_lft = INFINITY_LIFE_TIME,
3092                 .scope = scope
3093         };
3094
3095         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3096         if (!IS_ERR(ifp)) {
3097                 spin_lock_bh(&ifp->lock);
3098                 ifp->flags &= ~IFA_F_TENTATIVE;
3099                 spin_unlock_bh(&ifp->lock);
3100                 rt_genid_bump_ipv6(dev_net(idev->dev));
3101                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3102                 in6_ifa_put(ifp);
3103         }
3104 }
3105
3106 #if IS_ENABLED(CONFIG_IPV6_SIT)
3107 static void sit_add_v4_addrs(struct inet6_dev *idev)
3108 {
3109         struct in6_addr addr;
3110         struct net_device *dev;
3111         struct net *net = dev_net(idev->dev);
3112         int scope, plen;
3113         u32 pflags = 0;
3114
3115         ASSERT_RTNL();
3116
3117         memset(&addr, 0, sizeof(struct in6_addr));
3118         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3119
3120         if (idev->dev->flags&IFF_POINTOPOINT) {
3121                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3122                         return;
3123
3124                 addr.s6_addr32[0] = htonl(0xfe800000);
3125                 scope = IFA_LINK;
3126                 plen = 64;
3127         } else {
3128                 scope = IPV6_ADDR_COMPATv4;
3129                 plen = 96;
3130                 pflags |= RTF_NONEXTHOP;
3131         }
3132
3133         if (addr.s6_addr32[3]) {
3134                 add_addr(idev, &addr, plen, scope);
3135                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3136                                       GFP_KERNEL);
3137                 return;
3138         }
3139
3140         for_each_netdev(net, dev) {
3141                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3142                 if (in_dev && (dev->flags & IFF_UP)) {
3143                         struct in_ifaddr *ifa;
3144                         int flag = scope;
3145
3146                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3147                                 addr.s6_addr32[3] = ifa->ifa_local;
3148
3149                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3150                                         continue;
3151                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3152                                         if (idev->dev->flags&IFF_POINTOPOINT)
3153                                                 continue;
3154                                         flag |= IFA_HOST;
3155                                 }
3156
3157                                 add_addr(idev, &addr, plen, flag);
3158                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3159                                                       0, pflags, GFP_KERNEL);
3160                         }
3161                 }
3162         }
3163 }
3164 #endif
3165
3166 static void init_loopback(struct net_device *dev)
3167 {
3168         struct inet6_dev  *idev;
3169
3170         /* ::1 */
3171
3172         ASSERT_RTNL();
3173
3174         idev = ipv6_find_idev(dev);
3175         if (IS_ERR(idev)) {
3176                 pr_debug("%s: add_dev failed\n", __func__);
3177                 return;
3178         }
3179
3180         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3181 }
3182
3183 void addrconf_add_linklocal(struct inet6_dev *idev,
3184                             const struct in6_addr *addr, u32 flags)
3185 {
3186         struct ifa6_config cfg = {
3187                 .pfx = addr,
3188                 .plen = 64,
3189                 .ifa_flags = flags | IFA_F_PERMANENT,
3190                 .valid_lft = INFINITY_LIFE_TIME,
3191                 .preferred_lft = INFINITY_LIFE_TIME,
3192                 .scope = IFA_LINK
3193         };
3194         struct inet6_ifaddr *ifp;
3195
3196 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3197         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3198              idev->cnf.optimistic_dad) &&
3199             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3200                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3201 #endif
3202
3203         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3204         if (!IS_ERR(ifp)) {
3205                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3206                                       0, 0, GFP_ATOMIC);
3207                 addrconf_dad_start(ifp);
3208                 in6_ifa_put(ifp);
3209         }
3210 }
3211 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3212
3213 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3214 {
3215         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3216                 return true;
3217
3218         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3219             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3220                 return true;
3221
3222         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3223             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3224                 return true;
3225
3226         return false;
3227 }
3228
3229 static int ipv6_generate_stable_address(struct in6_addr *address,
3230                                         u8 dad_count,
3231                                         const struct inet6_dev *idev)
3232 {
3233         static DEFINE_SPINLOCK(lock);
3234         static __u32 digest[SHA_DIGEST_WORDS];
3235         static __u32 workspace[SHA_WORKSPACE_WORDS];
3236
3237         static union {
3238                 char __data[SHA_MESSAGE_BYTES];
3239                 struct {
3240                         struct in6_addr secret;
3241                         __be32 prefix[2];
3242                         unsigned char hwaddr[MAX_ADDR_LEN];
3243                         u8 dad_count;
3244                 } __packed;
3245         } data;
3246
3247         struct in6_addr secret;
3248         struct in6_addr temp;
3249         struct net *net = dev_net(idev->dev);
3250
3251         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3252
3253         if (idev->cnf.stable_secret.initialized)
3254                 secret = idev->cnf.stable_secret.secret;
3255         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3256                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3257         else
3258                 return -1;
3259
3260 retry:
3261         spin_lock_bh(&lock);
3262
3263         sha_init(digest);
3264         memset(&data, 0, sizeof(data));
3265         memset(workspace, 0, sizeof(workspace));
3266         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3267         data.prefix[0] = address->s6_addr32[0];
3268         data.prefix[1] = address->s6_addr32[1];
3269         data.secret = secret;
3270         data.dad_count = dad_count;
3271
3272         sha_transform(digest, data.__data, workspace);
3273
3274         temp = *address;
3275         temp.s6_addr32[2] = (__force __be32)digest[0];
3276         temp.s6_addr32[3] = (__force __be32)digest[1];
3277
3278         spin_unlock_bh(&lock);
3279
3280         if (ipv6_reserved_interfaceid(temp)) {
3281                 dad_count++;
3282                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3283                         return -1;
3284                 goto retry;
3285         }
3286
3287         *address = temp;
3288         return 0;
3289 }
3290
3291 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3292 {
3293         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3294
3295         if (s->initialized)
3296                 return;
3297         s = &idev->cnf.stable_secret;
3298         get_random_bytes(&s->secret, sizeof(s->secret));
3299         s->initialized = true;
3300 }
3301
3302 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3303 {
3304         struct in6_addr addr;
3305
3306         /* no link local addresses on L3 master devices */
3307         if (netif_is_l3_master(idev->dev))
3308                 return;
3309
3310         /* no link local addresses on devices flagged as slaves */
3311         if (idev->dev->flags & IFF_SLAVE)
3312                 return;
3313
3314         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3315
3316         switch (idev->cnf.addr_gen_mode) {
3317         case IN6_ADDR_GEN_MODE_RANDOM:
3318                 ipv6_gen_mode_random_init(idev);
3319                 /* fallthrough */
3320         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3321                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3322                         addrconf_add_linklocal(idev, &addr,
3323                                                IFA_F_STABLE_PRIVACY);
3324                 else if (prefix_route)
3325                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3326                                               0, 0, GFP_KERNEL);
3327                 break;
3328         case IN6_ADDR_GEN_MODE_EUI64:
3329                 /* addrconf_add_linklocal also adds a prefix_route and we
3330                  * only need to care about prefix routes if ipv6_generate_eui64
3331                  * couldn't generate one.
3332                  */
3333                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3334                         addrconf_add_linklocal(idev, &addr, 0);
3335                 else if (prefix_route)
3336                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3337                                               0, 0, GFP_KERNEL);
3338                 break;
3339         case IN6_ADDR_GEN_MODE_NONE:
3340         default:
3341                 /* will not add any link local address */
3342                 break;
3343         }
3344 }
3345
3346 static void addrconf_dev_config(struct net_device *dev)
3347 {
3348         struct inet6_dev *idev;
3349
3350         ASSERT_RTNL();
3351
3352         if ((dev->type != ARPHRD_ETHER) &&
3353             (dev->type != ARPHRD_FDDI) &&
3354             (dev->type != ARPHRD_ARCNET) &&
3355             (dev->type != ARPHRD_INFINIBAND) &&
3356             (dev->type != ARPHRD_IEEE1394) &&
3357             (dev->type != ARPHRD_TUNNEL6) &&
3358             (dev->type != ARPHRD_6LOWPAN) &&
3359             (dev->type != ARPHRD_IP6GRE) &&
3360             (dev->type != ARPHRD_IPGRE) &&
3361             (dev->type != ARPHRD_TUNNEL) &&
3362             (dev->type != ARPHRD_NONE) &&
3363             (dev->type != ARPHRD_RAWIP)) {
3364                 /* Alas, we support only Ethernet autoconfiguration. */
3365                 idev = __in6_dev_get(dev);
3366                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3367                     dev->flags & IFF_MULTICAST)
3368                         ipv6_mc_up(idev);
3369                 return;
3370         }
3371
3372         idev = addrconf_add_dev(dev);
3373         if (IS_ERR(idev))
3374                 return;
3375
3376         /* this device type has no EUI support */
3377         if (dev->type == ARPHRD_NONE &&
3378             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3379                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3380
3381         addrconf_addr_gen(idev, false);
3382 }
3383
3384 #if IS_ENABLED(CONFIG_IPV6_SIT)
3385 static void addrconf_sit_config(struct net_device *dev)
3386 {
3387         struct inet6_dev *idev;
3388
3389         ASSERT_RTNL();
3390
3391         /*
3392          * Configure the tunnel with one of our IPv4
3393          * addresses... we should configure all of
3394          * our v4 addrs in the tunnel
3395          */
3396
3397         idev = ipv6_find_idev(dev);
3398         if (IS_ERR(idev)) {
3399                 pr_debug("%s: add_dev failed\n", __func__);
3400                 return;
3401         }
3402
3403         if (dev->priv_flags & IFF_ISATAP) {
3404                 addrconf_addr_gen(idev, false);
3405                 return;
3406         }
3407
3408         sit_add_v4_addrs(idev);
3409
3410         if (dev->flags&IFF_POINTOPOINT)
3411                 addrconf_add_mroute(dev);
3412 }
3413 #endif
3414
3415 #if IS_ENABLED(CONFIG_NET_IPGRE)
3416 static void addrconf_gre_config(struct net_device *dev)
3417 {
3418         struct inet6_dev *idev;
3419
3420         ASSERT_RTNL();
3421
3422         idev = ipv6_find_idev(dev);
3423         if (IS_ERR(idev)) {
3424                 pr_debug("%s: add_dev failed\n", __func__);
3425                 return;
3426         }
3427
3428         addrconf_addr_gen(idev, true);
3429         if (dev->flags & IFF_POINTOPOINT)
3430                 addrconf_add_mroute(dev);
3431 }
3432 #endif
3433
3434 static int fixup_permanent_addr(struct net *net,
3435                                 struct inet6_dev *idev,
3436                                 struct inet6_ifaddr *ifp)
3437 {
3438         /* !fib6_node means the host route was removed from the
3439          * FIB, for example, if 'lo' device is taken down. In that
3440          * case regenerate the host route.
3441          */
3442         if (!ifp->rt || !ifp->rt->fib6_node) {
3443                 struct fib6_info *f6i, *prev;
3444
3445                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3446                                          GFP_ATOMIC);
3447                 if (IS_ERR(f6i))
3448                         return PTR_ERR(f6i);
3449
3450                 /* ifp->rt can be accessed outside of rtnl */
3451                 spin_lock(&ifp->lock);
3452                 prev = ifp->rt;
3453                 ifp->rt = f6i;
3454                 spin_unlock(&ifp->lock);
3455
3456                 fib6_info_release(prev);
3457         }
3458
3459         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3460                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3461                                       ifp->rt_priority, idev->dev, 0, 0,
3462                                       GFP_ATOMIC);
3463         }
3464
3465         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3466                 addrconf_dad_start(ifp);
3467
3468         return 0;
3469 }
3470
3471 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3472 {
3473         struct inet6_ifaddr *ifp, *tmp;
3474         struct inet6_dev *idev;
3475
3476         idev = __in6_dev_get(dev);
3477         if (!idev)
3478                 return;
3479
3480         write_lock_bh(&idev->lock);
3481
3482         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3483                 if ((ifp->flags & IFA_F_PERMANENT) &&
3484                     fixup_permanent_addr(net, idev, ifp) < 0) {
3485                         write_unlock_bh(&idev->lock);
3486                         in6_ifa_hold(ifp);
3487                         ipv6_del_addr(ifp);
3488                         write_lock_bh(&idev->lock);
3489
3490                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3491                                              idev->dev->name, &ifp->addr);
3492                 }
3493         }
3494
3495         write_unlock_bh(&idev->lock);
3496 }
3497
3498 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3499                            void *ptr)
3500 {
3501         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3502         struct netdev_notifier_change_info *change_info;
3503         struct netdev_notifier_changeupper_info *info;
3504         struct inet6_dev *idev = __in6_dev_get(dev);
3505         struct net *net = dev_net(dev);
3506         int run_pending = 0;
3507         int err;
3508
3509         switch (event) {
3510         case NETDEV_REGISTER:
3511                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3512                         idev = ipv6_add_dev(dev);
3513                         if (IS_ERR(idev))
3514                                 return notifier_from_errno(PTR_ERR(idev));
3515                 }
3516                 break;
3517
3518         case NETDEV_CHANGEMTU:
3519                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3520                 if (dev->mtu < IPV6_MIN_MTU) {
3521                         addrconf_ifdown(dev, dev != net->loopback_dev);
3522                         break;
3523                 }
3524
3525                 if (idev) {
3526                         rt6_mtu_change(dev, dev->mtu);
3527                         idev->cnf.mtu6 = dev->mtu;
3528                         break;
3529                 }
3530
3531                 /* allocate new idev */
3532                 idev = ipv6_add_dev(dev);
3533                 if (IS_ERR(idev))
3534                         break;
3535
3536                 /* device is still not ready */
3537                 if (!(idev->if_flags & IF_READY))
3538                         break;
3539
3540                 run_pending = 1;
3541
3542                 /* fall through */
3543
3544         case NETDEV_UP:
3545         case NETDEV_CHANGE:
3546                 if (dev->flags & IFF_SLAVE)
3547                         break;
3548
3549                 if (idev && idev->cnf.disable_ipv6)
3550                         break;
3551
3552                 if (event == NETDEV_UP) {
3553                         /* restore routes for permanent addresses */
3554                         addrconf_permanent_addr(net, dev);
3555
3556                         if (!addrconf_link_ready(dev)) {
3557                                 /* device is not ready yet. */
3558                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3559                                          dev->name);
3560                                 break;
3561                         }
3562
3563                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3564                                 idev = ipv6_add_dev(dev);
3565
3566                         if (!IS_ERR_OR_NULL(idev)) {
3567                                 idev->if_flags |= IF_READY;
3568                                 run_pending = 1;
3569                         }
3570                 } else if (event == NETDEV_CHANGE) {
3571                         if (!addrconf_link_ready(dev)) {
3572                                 /* device is still not ready. */
3573                                 rt6_sync_down_dev(dev, event);
3574                                 break;
3575                         }
3576
3577                         if (!IS_ERR_OR_NULL(idev)) {
3578                                 if (idev->if_flags & IF_READY) {
3579                                         /* device is already configured -
3580                                          * but resend MLD reports, we might
3581                                          * have roamed and need to update
3582                                          * multicast snooping switches
3583                                          */
3584                                         ipv6_mc_up(idev);
3585                                         change_info = ptr;
3586                                         if (change_info->flags_changed & IFF_NOARP)
3587                                                 addrconf_dad_run(idev, true);
3588                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3589                                         break;
3590                                 }
3591                                 idev->if_flags |= IF_READY;
3592                         }
3593
3594                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3595                                 dev->name);
3596
3597                         run_pending = 1;
3598                 }
3599
3600                 switch (dev->type) {
3601 #if IS_ENABLED(CONFIG_IPV6_SIT)
3602                 case ARPHRD_SIT:
3603                         addrconf_sit_config(dev);
3604                         break;
3605 #endif
3606 #if IS_ENABLED(CONFIG_NET_IPGRE)
3607                 case ARPHRD_IPGRE:
3608                         addrconf_gre_config(dev);
3609                         break;
3610 #endif
3611                 case ARPHRD_LOOPBACK:
3612                         init_loopback(dev);
3613                         break;
3614
3615                 default:
3616                         addrconf_dev_config(dev);
3617                         break;
3618                 }
3619
3620                 if (!IS_ERR_OR_NULL(idev)) {
3621                         if (run_pending)
3622                                 addrconf_dad_run(idev, false);
3623
3624                         /* Device has an address by now */
3625                         rt6_sync_up(dev, RTNH_F_DEAD);
3626
3627                         /*
3628                          * If the MTU changed during the interface down,
3629                          * when the interface up, the changed MTU must be
3630                          * reflected in the idev as well as routers.
3631                          */
3632                         if (idev->cnf.mtu6 != dev->mtu &&
3633                             dev->mtu >= IPV6_MIN_MTU) {
3634                                 rt6_mtu_change(dev, dev->mtu);
3635                                 idev->cnf.mtu6 = dev->mtu;
3636                         }
3637                         idev->tstamp = jiffies;
3638                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3639
3640                         /*
3641                          * If the changed mtu during down is lower than
3642                          * IPV6_MIN_MTU stop IPv6 on this interface.
3643                          */
3644                         if (dev->mtu < IPV6_MIN_MTU)
3645                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3646                 }
3647                 break;
3648
3649         case NETDEV_DOWN:
3650         case NETDEV_UNREGISTER:
3651                 /*
3652                  *      Remove all addresses from this interface.
3653                  */
3654                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3655                 break;
3656
3657         case NETDEV_CHANGENAME:
3658                 if (idev) {
3659                         snmp6_unregister_dev(idev);
3660                         addrconf_sysctl_unregister(idev);
3661                         err = addrconf_sysctl_register(idev);
3662                         if (err)
3663                                 return notifier_from_errno(err);
3664                         err = snmp6_register_dev(idev);
3665                         if (err) {
3666                                 addrconf_sysctl_unregister(idev);
3667                                 return notifier_from_errno(err);
3668                         }
3669                 }
3670                 break;
3671
3672         case NETDEV_PRE_TYPE_CHANGE:
3673         case NETDEV_POST_TYPE_CHANGE:
3674                 if (idev)
3675                         addrconf_type_change(dev, event);
3676                 break;
3677
3678         case NETDEV_CHANGEUPPER:
3679                 info = ptr;
3680
3681                 /* flush all routes if dev is linked to or unlinked from
3682                  * an L3 master device (e.g., VRF)
3683                  */
3684                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3685                         addrconf_ifdown(dev, 0);
3686         }
3687
3688         return NOTIFY_OK;
3689 }
3690
3691 /*
3692  *      addrconf module should be notified of a device going up
3693  */
3694 static struct notifier_block ipv6_dev_notf = {
3695         .notifier_call = addrconf_notify,
3696         .priority = ADDRCONF_NOTIFY_PRIORITY,
3697 };
3698
3699 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3700 {
3701         struct inet6_dev *idev;
3702         ASSERT_RTNL();
3703
3704         idev = __in6_dev_get(dev);
3705
3706         if (event == NETDEV_POST_TYPE_CHANGE)
3707                 ipv6_mc_remap(idev);
3708         else if (event == NETDEV_PRE_TYPE_CHANGE)
3709                 ipv6_mc_unmap(idev);
3710 }
3711
3712 static bool addr_is_local(const struct in6_addr *addr)
3713 {
3714         return ipv6_addr_type(addr) &
3715                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3716 }
3717
3718 static int addrconf_ifdown(struct net_device *dev, int how)
3719 {
3720         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3721         struct net *net = dev_net(dev);
3722         struct inet6_dev *idev;
3723         struct inet6_ifaddr *ifa;
3724         LIST_HEAD(tmp_addr_list);
3725         bool keep_addr = false;
3726         bool was_ready;
3727         int state, i;
3728
3729         ASSERT_RTNL();
3730
3731         rt6_disable_ip(dev, event);
3732
3733         idev = __in6_dev_get(dev);
3734         if (!idev)
3735                 return -ENODEV;
3736
3737         /*
3738          * Step 1: remove reference to ipv6 device from parent device.
3739          *         Do not dev_put!
3740          */
3741         if (how) {
3742                 idev->dead = 1;
3743
3744                 /* protected by rtnl_lock */
3745                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3746
3747                 /* Step 1.5: remove snmp6 entry */
3748                 snmp6_unregister_dev(idev);
3749
3750         }
3751
3752         /* combine the user config with event to determine if permanent
3753          * addresses are to be removed from address hash table
3754          */
3755         if (!how && !idev->cnf.disable_ipv6) {
3756                 /* aggregate the system setting and interface setting */
3757                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3758
3759                 if (!_keep_addr)
3760                         _keep_addr = idev->cnf.keep_addr_on_down;
3761
3762                 keep_addr = (_keep_addr > 0);
3763         }
3764
3765         /* Step 2: clear hash table */
3766         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3767                 struct hlist_head *h = &inet6_addr_lst[i];
3768
3769                 spin_lock_bh(&addrconf_hash_lock);
3770 restart:
3771                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3772                         if (ifa->idev == idev) {
3773                                 addrconf_del_dad_work(ifa);
3774                                 /* combined flag + permanent flag decide if
3775                                  * address is retained on a down event
3776                                  */
3777                                 if (!keep_addr ||
3778                                     !(ifa->flags & IFA_F_PERMANENT) ||
3779                                     addr_is_local(&ifa->addr)) {
3780                                         hlist_del_init_rcu(&ifa->addr_lst);
3781                                         goto restart;
3782                                 }
3783                         }
3784                 }
3785                 spin_unlock_bh(&addrconf_hash_lock);
3786         }
3787
3788         write_lock_bh(&idev->lock);
3789
3790         addrconf_del_rs_timer(idev);
3791
3792         /* Step 2: clear flags for stateless addrconf, repeated down
3793          *         detection
3794          */
3795         was_ready = idev->if_flags & IF_READY;
3796         if (!how)
3797                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3798
3799         /* Step 3: clear tempaddr list */
3800         while (!list_empty(&idev->tempaddr_list)) {
3801                 ifa = list_first_entry(&idev->tempaddr_list,
3802                                        struct inet6_ifaddr, tmp_list);
3803                 list_del(&ifa->tmp_list);
3804                 write_unlock_bh(&idev->lock);
3805                 spin_lock_bh(&ifa->lock);
3806
3807                 if (ifa->ifpub) {
3808                         in6_ifa_put(ifa->ifpub);
3809                         ifa->ifpub = NULL;
3810                 }
3811                 spin_unlock_bh(&ifa->lock);
3812                 in6_ifa_put(ifa);
3813                 write_lock_bh(&idev->lock);
3814         }
3815
3816         list_for_each_entry(ifa, &idev->addr_list, if_list)
3817                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3818         write_unlock_bh(&idev->lock);
3819
3820         while (!list_empty(&tmp_addr_list)) {
3821                 struct fib6_info *rt = NULL;
3822                 bool keep;
3823
3824                 ifa = list_first_entry(&tmp_addr_list,
3825                                        struct inet6_ifaddr, if_list_aux);
3826                 list_del(&ifa->if_list_aux);
3827
3828                 addrconf_del_dad_work(ifa);
3829
3830                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3831                         !addr_is_local(&ifa->addr);
3832
3833                 spin_lock_bh(&ifa->lock);
3834
3835                 if (keep) {
3836                         /* set state to skip the notifier below */
3837                         state = INET6_IFADDR_STATE_DEAD;
3838                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3839                         if (!(ifa->flags & IFA_F_NODAD))
3840                                 ifa->flags |= IFA_F_TENTATIVE;
3841
3842                         rt = ifa->rt;
3843                         ifa->rt = NULL;
3844                 } else {
3845                         state = ifa->state;
3846                         ifa->state = INET6_IFADDR_STATE_DEAD;
3847                 }
3848
3849                 spin_unlock_bh(&ifa->lock);
3850
3851                 if (rt)
3852                         ip6_del_rt(net, rt);
3853
3854                 if (state != INET6_IFADDR_STATE_DEAD) {
3855                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3856                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3857                 } else {
3858                         if (idev->cnf.forwarding)
3859                                 addrconf_leave_anycast(ifa);
3860                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3861                 }
3862
3863                 if (!keep) {
3864                         write_lock_bh(&idev->lock);
3865                         list_del_rcu(&ifa->if_list);
3866                         write_unlock_bh(&idev->lock);
3867                         in6_ifa_put(ifa);
3868                 }
3869         }
3870
3871         /* Step 5: Discard anycast and multicast list */
3872         if (how) {
3873                 ipv6_ac_destroy_dev(idev);
3874                 ipv6_mc_destroy_dev(idev);
3875         } else if (was_ready) {
3876                 ipv6_mc_down(idev);
3877         }
3878
3879         idev->tstamp = jiffies;
3880
3881         /* Last: Shot the device (if unregistered) */
3882         if (how) {
3883                 addrconf_sysctl_unregister(idev);
3884                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3885                 neigh_ifdown(&nd_tbl, dev);
3886                 in6_dev_put(idev);
3887         }
3888         return 0;
3889 }
3890
3891 static void addrconf_rs_timer(struct timer_list *t)
3892 {
3893         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3894         struct net_device *dev = idev->dev;
3895         struct in6_addr lladdr;
3896
3897         write_lock(&idev->lock);
3898         if (idev->dead || !(idev->if_flags & IF_READY))
3899                 goto out;
3900
3901         if (!ipv6_accept_ra(idev))
3902                 goto out;
3903
3904         /* Announcement received after solicitation was sent */
3905         if (idev->if_flags & IF_RA_RCVD)
3906                 goto out;
3907
3908         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3909                 write_unlock(&idev->lock);
3910                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3911                         ndisc_send_rs(dev, &lladdr,
3912                                       &in6addr_linklocal_allrouters);
3913                 else
3914                         goto put;
3915
3916                 write_lock(&idev->lock);
3917                 idev->rs_interval = rfc3315_s14_backoff_update(
3918                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3919                 /* The wait after the last probe can be shorter */
3920                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3921                                              idev->cnf.rtr_solicits) ?
3922                                       idev->cnf.rtr_solicit_delay :
3923                                       idev->rs_interval);
3924         } else {
3925                 /*
3926                  * Note: we do not support deprecated "all on-link"
3927                  * assumption any longer.
3928                  */
3929                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3930         }
3931
3932 out:
3933         write_unlock(&idev->lock);
3934 put:
3935         in6_dev_put(idev);
3936 }
3937
3938 /*
3939  *      Duplicate Address Detection
3940  */
3941 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3942 {
3943         unsigned long rand_num;
3944         struct inet6_dev *idev = ifp->idev;
3945         u64 nonce;
3946
3947         if (ifp->flags & IFA_F_OPTIMISTIC)
3948                 rand_num = 0;
3949         else
3950                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3951
3952         nonce = 0;
3953         if (idev->cnf.enhanced_dad ||
3954             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3955                 do
3956                         get_random_bytes(&nonce, 6);
3957                 while (nonce == 0);
3958         }
3959         ifp->dad_nonce = nonce;
3960         ifp->dad_probes = idev->cnf.dad_transmits;
3961         addrconf_mod_dad_work(ifp, rand_num);
3962 }
3963
3964 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3965 {
3966         struct inet6_dev *idev = ifp->idev;
3967         struct net_device *dev = idev->dev;
3968         bool bump_id, notify = false;
3969         struct net *net;
3970
3971         addrconf_join_solict(dev, &ifp->addr);
3972
3973         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3974
3975         read_lock_bh(&idev->lock);
3976         spin_lock(&ifp->lock);
3977         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3978                 goto out;
3979
3980         net = dev_net(dev);
3981         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3982             (net->ipv6.devconf_all->accept_dad < 1 &&
3983              idev->cnf.accept_dad < 1) ||
3984             !(ifp->flags&IFA_F_TENTATIVE) ||
3985             ifp->flags & IFA_F_NODAD) {
3986                 bool send_na = false;
3987
3988                 if (ifp->flags & IFA_F_TENTATIVE &&
3989                     !(ifp->flags & IFA_F_OPTIMISTIC))
3990                         send_na = true;
3991                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3992                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3993                 spin_unlock(&ifp->lock);
3994                 read_unlock_bh(&idev->lock);
3995
3996                 addrconf_dad_completed(ifp, bump_id, send_na);
3997                 return;
3998         }
3999
4000         if (!(idev->if_flags & IF_READY)) {
4001                 spin_unlock(&ifp->lock);
4002                 read_unlock_bh(&idev->lock);
4003                 /*
4004                  * If the device is not ready:
4005                  * - keep it tentative if it is a permanent address.
4006                  * - otherwise, kill it.
4007                  */
4008                 in6_ifa_hold(ifp);
4009                 addrconf_dad_stop(ifp, 0);
4010                 return;
4011         }
4012
4013         /*
4014          * Optimistic nodes can start receiving
4015          * Frames right away
4016          */
4017         if (ifp->flags & IFA_F_OPTIMISTIC) {
4018                 ip6_ins_rt(net, ifp->rt);
4019                 if (ipv6_use_optimistic_addr(net, idev)) {
4020                         /* Because optimistic nodes can use this address,
4021                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4022                          */
4023                         notify = true;
4024                 }
4025         }
4026
4027         addrconf_dad_kick(ifp);
4028 out:
4029         spin_unlock(&ifp->lock);
4030         read_unlock_bh(&idev->lock);
4031         if (notify)
4032                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4033 }
4034
4035 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4036 {
4037         bool begin_dad = false;
4038
4039         spin_lock_bh(&ifp->lock);
4040         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4041                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4042                 begin_dad = true;
4043         }
4044         spin_unlock_bh(&ifp->lock);
4045
4046         if (begin_dad)
4047                 addrconf_mod_dad_work(ifp, 0);
4048 }
4049
4050 static void addrconf_dad_work(struct work_struct *w)
4051 {
4052         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4053                                                 struct inet6_ifaddr,
4054                                                 dad_work);
4055         struct inet6_dev *idev = ifp->idev;
4056         bool bump_id, disable_ipv6 = false;
4057         struct in6_addr mcaddr;
4058
4059         enum {
4060                 DAD_PROCESS,
4061                 DAD_BEGIN,
4062                 DAD_ABORT,
4063         } action = DAD_PROCESS;
4064
4065         rtnl_lock();
4066
4067         spin_lock_bh(&ifp->lock);
4068         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4069                 action = DAD_BEGIN;
4070                 ifp->state = INET6_IFADDR_STATE_DAD;
4071         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4072                 action = DAD_ABORT;
4073                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4074
4075                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4076                      idev->cnf.accept_dad > 1) &&
4077                     !idev->cnf.disable_ipv6 &&
4078                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4079                         struct in6_addr addr;
4080
4081                         addr.s6_addr32[0] = htonl(0xfe800000);
4082                         addr.s6_addr32[1] = 0;
4083
4084                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4085                             ipv6_addr_equal(&ifp->addr, &addr)) {
4086                                 /* DAD failed for link-local based on MAC */
4087                                 idev->cnf.disable_ipv6 = 1;
4088
4089                                 pr_info("%s: IPv6 being disabled!\n",
4090                                         ifp->idev->dev->name);
4091                                 disable_ipv6 = true;
4092                         }
4093                 }
4094         }
4095         spin_unlock_bh(&ifp->lock);
4096
4097         if (action == DAD_BEGIN) {
4098                 addrconf_dad_begin(ifp);
4099                 goto out;
4100         } else if (action == DAD_ABORT) {
4101                 in6_ifa_hold(ifp);
4102                 addrconf_dad_stop(ifp, 1);
4103                 if (disable_ipv6)
4104                         addrconf_ifdown(idev->dev, 0);
4105                 goto out;
4106         }
4107
4108         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4109                 goto out;
4110
4111         write_lock_bh(&idev->lock);
4112         if (idev->dead || !(idev->if_flags & IF_READY)) {
4113                 write_unlock_bh(&idev->lock);
4114                 goto out;
4115         }
4116
4117         spin_lock(&ifp->lock);
4118         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4119                 spin_unlock(&ifp->lock);
4120                 write_unlock_bh(&idev->lock);
4121                 goto out;
4122         }
4123
4124         if (ifp->dad_probes == 0) {
4125                 bool send_na = false;
4126
4127                 /*
4128                  * DAD was successful
4129                  */
4130
4131                 if (ifp->flags & IFA_F_TENTATIVE &&
4132                     !(ifp->flags & IFA_F_OPTIMISTIC))
4133                         send_na = true;
4134                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4135                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4136                 spin_unlock(&ifp->lock);
4137                 write_unlock_bh(&idev->lock);
4138
4139                 addrconf_dad_completed(ifp, bump_id, send_na);
4140
4141                 goto out;
4142         }
4143
4144         ifp->dad_probes--;
4145         addrconf_mod_dad_work(ifp,
4146                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4147         spin_unlock(&ifp->lock);
4148         write_unlock_bh(&idev->lock);
4149
4150         /* send a neighbour solicitation for our addr */
4151         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4152         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4153                       ifp->dad_nonce);
4154 out:
4155         in6_ifa_put(ifp);
4156         rtnl_unlock();
4157 }
4158
4159 /* ifp->idev must be at least read locked */
4160 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4161 {
4162         struct inet6_ifaddr *ifpiter;
4163         struct inet6_dev *idev = ifp->idev;
4164
4165         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4166                 if (ifpiter->scope > IFA_LINK)
4167                         break;
4168                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4169                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4170                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4171                     IFA_F_PERMANENT)
4172                         return false;
4173         }
4174         return true;
4175 }
4176
4177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4178                                    bool send_na)
4179 {
4180         struct net_device *dev = ifp->idev->dev;
4181         struct in6_addr lladdr;
4182         bool send_rs, send_mld;
4183
4184         addrconf_del_dad_work(ifp);
4185
4186         /*
4187          *      Configure the address for reception. Now it is valid.
4188          */
4189
4190         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4191
4192         /* If added prefix is link local and we are prepared to process
4193            router advertisements, start sending router solicitations.
4194          */
4195
4196         read_lock_bh(&ifp->idev->lock);
4197         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4198         send_rs = send_mld &&
4199                   ipv6_accept_ra(ifp->idev) &&
4200                   ifp->idev->cnf.rtr_solicits != 0 &&
4201                   (dev->flags & IFF_LOOPBACK) == 0 &&
4202                   (dev->type != ARPHRD_TUNNEL);
4203         read_unlock_bh(&ifp->idev->lock);
4204
4205         /* While dad is in progress mld report's source address is in6_addrany.
4206          * Resend with proper ll now.
4207          */
4208         if (send_mld)
4209                 ipv6_mc_dad_complete(ifp->idev);
4210
4211         /* send unsolicited NA if enabled */
4212         if (send_na &&
4213             (ifp->idev->cnf.ndisc_notify ||
4214              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4215                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4216                               /*router=*/ !!ifp->idev->cnf.forwarding,
4217                               /*solicited=*/ false, /*override=*/ true,
4218                               /*inc_opt=*/ true);
4219         }
4220
4221         if (send_rs) {
4222                 /*
4223                  *      If a host as already performed a random delay
4224                  *      [...] as part of DAD [...] there is no need
4225                  *      to delay again before sending the first RS
4226                  */
4227                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4228                         return;
4229                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4230
4231                 write_lock_bh(&ifp->idev->lock);
4232                 spin_lock(&ifp->lock);
4233                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4234                         ifp->idev->cnf.rtr_solicit_interval);
4235                 ifp->idev->rs_probes = 1;
4236                 ifp->idev->if_flags |= IF_RS_SENT;
4237                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4238                 spin_unlock(&ifp->lock);
4239                 write_unlock_bh(&ifp->idev->lock);
4240         }
4241
4242         if (bump_id)
4243                 rt_genid_bump_ipv6(dev_net(dev));
4244
4245         /* Make sure that a new temporary address will be created
4246          * before this temporary address becomes deprecated.
4247          */
4248         if (ifp->flags & IFA_F_TEMPORARY)
4249                 addrconf_verify_rtnl();
4250 }
4251
4252 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4253 {
4254         struct inet6_ifaddr *ifp;
4255
4256         read_lock_bh(&idev->lock);
4257         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4258                 spin_lock(&ifp->lock);
4259                 if ((ifp->flags & IFA_F_TENTATIVE &&
4260                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4261                         if (restart)
4262                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4263                         addrconf_dad_kick(ifp);
4264                 }
4265                 spin_unlock(&ifp->lock);
4266         }
4267         read_unlock_bh(&idev->lock);
4268 }
4269
4270 #ifdef CONFIG_PROC_FS
4271 struct if6_iter_state {
4272         struct seq_net_private p;
4273         int bucket;
4274         int offset;
4275 };
4276
4277 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4278 {
4279         struct if6_iter_state *state = seq->private;
4280         struct net *net = seq_file_net(seq);
4281         struct inet6_ifaddr *ifa = NULL;
4282         int p = 0;
4283
4284         /* initial bucket if pos is 0 */
4285         if (pos == 0) {
4286                 state->bucket = 0;
4287                 state->offset = 0;
4288         }
4289
4290         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4291                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4292                                          addr_lst) {
4293                         if (!net_eq(dev_net(ifa->idev->dev), net))
4294                                 continue;
4295                         /* sync with offset */
4296                         if (p < state->offset) {
4297                                 p++;
4298                                 continue;
4299                         }
4300                         return ifa;
4301                 }
4302
4303                 /* prepare for next bucket */
4304                 state->offset = 0;
4305                 p = 0;
4306         }
4307         return NULL;
4308 }
4309
4310 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4311                                          struct inet6_ifaddr *ifa)
4312 {
4313         struct if6_iter_state *state = seq->private;
4314         struct net *net = seq_file_net(seq);
4315
4316         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4317                 if (!net_eq(dev_net(ifa->idev->dev), net))
4318                         continue;
4319                 state->offset++;
4320                 return ifa;
4321         }
4322
4323         state->offset = 0;
4324         while (++state->bucket < IN6_ADDR_HSIZE) {
4325                 hlist_for_each_entry_rcu(ifa,
4326                                      &inet6_addr_lst[state->bucket], addr_lst) {
4327                         if (!net_eq(dev_net(ifa->idev->dev), net))
4328                                 continue;
4329                         return ifa;
4330                 }
4331         }
4332
4333         return NULL;
4334 }
4335
4336 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4337         __acquires(rcu)
4338 {
4339         rcu_read_lock();
4340         return if6_get_first(seq, *pos);
4341 }
4342
4343 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4344 {
4345         struct inet6_ifaddr *ifa;
4346
4347         ifa = if6_get_next(seq, v);
4348         ++*pos;
4349         return ifa;
4350 }
4351
4352 static void if6_seq_stop(struct seq_file *seq, void *v)
4353         __releases(rcu)
4354 {
4355         rcu_read_unlock();
4356 }
4357
4358 static int if6_seq_show(struct seq_file *seq, void *v)
4359 {
4360         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4361         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4362                    &ifp->addr,
4363                    ifp->idev->dev->ifindex,
4364                    ifp->prefix_len,
4365                    ifp->scope,
4366                    (u8) ifp->flags,
4367                    ifp->idev->dev->name);
4368         return 0;
4369 }
4370
4371 static const struct seq_operations if6_seq_ops = {
4372         .start  = if6_seq_start,
4373         .next   = if6_seq_next,
4374         .show   = if6_seq_show,
4375         .stop   = if6_seq_stop,
4376 };
4377
4378 static int __net_init if6_proc_net_init(struct net *net)
4379 {
4380         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4381                         sizeof(struct if6_iter_state)))
4382                 return -ENOMEM;
4383         return 0;
4384 }
4385
4386 static void __net_exit if6_proc_net_exit(struct net *net)
4387 {
4388         remove_proc_entry("if_inet6", net->proc_net);
4389 }
4390
4391 static struct pernet_operations if6_proc_net_ops = {
4392         .init = if6_proc_net_init,
4393         .exit = if6_proc_net_exit,
4394 };
4395
4396 int __init if6_proc_init(void)
4397 {
4398         return register_pernet_subsys(&if6_proc_net_ops);
4399 }
4400
4401 void if6_proc_exit(void)
4402 {
4403         unregister_pernet_subsys(&if6_proc_net_ops);
4404 }
4405 #endif  /* CONFIG_PROC_FS */
4406
4407 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4408 /* Check if address is a home address configured on any interface. */
4409 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4410 {
4411         unsigned int hash = inet6_addr_hash(net, addr);
4412         struct inet6_ifaddr *ifp = NULL;
4413         int ret = 0;
4414
4415         rcu_read_lock();
4416         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4417                 if (!net_eq(dev_net(ifp->idev->dev), net))
4418                         continue;
4419                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4420                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4421                         ret = 1;
4422                         break;
4423                 }
4424         }
4425         rcu_read_unlock();
4426         return ret;
4427 }
4428 #endif
4429
4430 /*
4431  *      Periodic address status verification
4432  */
4433
4434 static void addrconf_verify_rtnl(void)
4435 {
4436         unsigned long now, next, next_sec, next_sched;
4437         struct inet6_ifaddr *ifp;
4438         int i;
4439
4440         ASSERT_RTNL();
4441
4442         rcu_read_lock_bh();
4443         now = jiffies;
4444         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4445
4446         cancel_delayed_work(&addr_chk_work);
4447
4448         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4449 restart:
4450                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4451                         unsigned long age;
4452
4453                         /* When setting preferred_lft to a value not zero or
4454                          * infinity, while valid_lft is infinity
4455                          * IFA_F_PERMANENT has a non-infinity life time.
4456                          */
4457                         if ((ifp->flags & IFA_F_PERMANENT) &&
4458                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4459                                 continue;
4460
4461                         spin_lock(&ifp->lock);
4462                         /* We try to batch several events at once. */
4463                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4464
4465                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4466                             age >= ifp->valid_lft) {
4467                                 spin_unlock(&ifp->lock);
4468                                 in6_ifa_hold(ifp);
4469                                 ipv6_del_addr(ifp);
4470                                 goto restart;
4471                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4472                                 spin_unlock(&ifp->lock);
4473                                 continue;
4474                         } else if (age >= ifp->prefered_lft) {
4475                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4476                                 int deprecate = 0;
4477
4478                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4479                                         deprecate = 1;
4480                                         ifp->flags |= IFA_F_DEPRECATED;
4481                                 }
4482
4483                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4484                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4485                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4486
4487                                 spin_unlock(&ifp->lock);
4488
4489                                 if (deprecate) {
4490                                         in6_ifa_hold(ifp);
4491
4492                                         ipv6_ifa_notify(0, ifp);
4493                                         in6_ifa_put(ifp);
4494                                         goto restart;
4495                                 }
4496                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4497                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4498                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4499                                         ifp->idev->cnf.dad_transmits *
4500                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4501
4502                                 if (age >= ifp->prefered_lft - regen_advance) {
4503                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4504                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4505                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4506                                         if (!ifp->regen_count && ifpub) {
4507                                                 ifp->regen_count++;
4508                                                 in6_ifa_hold(ifp);
4509                                                 in6_ifa_hold(ifpub);
4510                                                 spin_unlock(&ifp->lock);
4511
4512                                                 spin_lock(&ifpub->lock);
4513                                                 ifpub->regen_count = 0;
4514                                                 spin_unlock(&ifpub->lock);
4515                                                 rcu_read_unlock_bh();
4516                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4517                                                 in6_ifa_put(ifpub);
4518                                                 in6_ifa_put(ifp);
4519                                                 rcu_read_lock_bh();
4520                                                 goto restart;
4521                                         }
4522                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4523                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4524                                 spin_unlock(&ifp->lock);
4525                         } else {
4526                                 /* ifp->prefered_lft <= ifp->valid_lft */
4527                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4528                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4529                                 spin_unlock(&ifp->lock);
4530                         }
4531                 }
4532         }
4533
4534         next_sec = round_jiffies_up(next);
4535         next_sched = next;
4536
4537         /* If rounded timeout is accurate enough, accept it. */
4538         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4539                 next_sched = next_sec;
4540
4541         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4542         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4543                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4544
4545         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4546                  now, next, next_sec, next_sched);
4547         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4548         rcu_read_unlock_bh();
4549 }
4550
4551 static void addrconf_verify_work(struct work_struct *w)
4552 {
4553         rtnl_lock();
4554         addrconf_verify_rtnl();
4555         rtnl_unlock();
4556 }
4557
4558 static void addrconf_verify(void)
4559 {
4560         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4561 }
4562
4563 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4564                                      struct in6_addr **peer_pfx)
4565 {
4566         struct in6_addr *pfx = NULL;
4567
4568         *peer_pfx = NULL;
4569
4570         if (addr)
4571                 pfx = nla_data(addr);
4572
4573         if (local) {
4574                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4575                         *peer_pfx = pfx;
4576                 pfx = nla_data(local);
4577         }
4578
4579         return pfx;
4580 }
4581
4582 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4583         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4584         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4585         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4586         [IFA_FLAGS]             = { .len = sizeof(u32) },
4587         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4588         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4589 };
4590
4591 static int
4592 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4593                   struct netlink_ext_ack *extack)
4594 {
4595         struct net *net = sock_net(skb->sk);
4596         struct ifaddrmsg *ifm;
4597         struct nlattr *tb[IFA_MAX+1];
4598         struct in6_addr *pfx, *peer_pfx;
4599         u32 ifa_flags;
4600         int err;
4601
4602         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4603                                      ifa_ipv6_policy, extack);
4604         if (err < 0)
4605                 return err;
4606
4607         ifm = nlmsg_data(nlh);
4608         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4609         if (!pfx)
4610                 return -EINVAL;
4611
4612         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4613
4614         /* We ignore other flags so far. */
4615         ifa_flags &= IFA_F_MANAGETEMPADDR;
4616
4617         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4618                               ifm->ifa_prefixlen);
4619 }
4620
4621 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4622                                unsigned long expires, u32 flags,
4623                                bool modify_peer)
4624 {
4625         struct fib6_info *f6i;
4626         u32 prio;
4627
4628         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4629                                         ifp->prefix_len,
4630                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4631         if (!f6i)
4632                 return -ENOENT;
4633
4634         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4635         if (f6i->fib6_metric != prio) {
4636                 /* delete old one */
4637                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4638
4639                 /* add new one */
4640                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4641                                       ifp->prefix_len,
4642                                       ifp->rt_priority, ifp->idev->dev,
4643                                       expires, flags, GFP_KERNEL);
4644         } else {
4645                 if (!expires)
4646                         fib6_clean_expires(f6i);
4647                 else
4648                         fib6_set_expires(f6i, expires);
4649
4650                 fib6_info_release(f6i);
4651         }
4652
4653         return 0;
4654 }
4655
4656 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4657 {
4658         u32 flags;
4659         clock_t expires;
4660         unsigned long timeout;
4661         bool was_managetempaddr;
4662         bool had_prefixroute;
4663         bool new_peer = false;
4664
4665         ASSERT_RTNL();
4666
4667         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4668                 return -EINVAL;
4669
4670         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4671             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4672                 return -EINVAL;
4673
4674         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4675                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4676
4677         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4678         if (addrconf_finite_timeout(timeout)) {
4679                 expires = jiffies_to_clock_t(timeout * HZ);
4680                 cfg->valid_lft = timeout;
4681                 flags = RTF_EXPIRES;
4682         } else {
4683                 expires = 0;
4684                 flags = 0;
4685                 cfg->ifa_flags |= IFA_F_PERMANENT;
4686         }
4687
4688         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4689         if (addrconf_finite_timeout(timeout)) {
4690                 if (timeout == 0)
4691                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4692                 cfg->preferred_lft = timeout;
4693         }
4694
4695         if (cfg->peer_pfx &&
4696             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4697                 if (!ipv6_addr_any(&ifp->peer_addr))
4698                         cleanup_prefix_route(ifp, expires, true, true);
4699                 new_peer = true;
4700         }
4701
4702         spin_lock_bh(&ifp->lock);
4703         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4704         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4705                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4706         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4707                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4708                         IFA_F_NOPREFIXROUTE);
4709         ifp->flags |= cfg->ifa_flags;
4710         ifp->tstamp = jiffies;
4711         ifp->valid_lft = cfg->valid_lft;
4712         ifp->prefered_lft = cfg->preferred_lft;
4713
4714         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4715                 ifp->rt_priority = cfg->rt_priority;
4716
4717         if (new_peer)
4718                 ifp->peer_addr = *cfg->peer_pfx;
4719
4720         spin_unlock_bh(&ifp->lock);
4721         if (!(ifp->flags&IFA_F_TENTATIVE))
4722                 ipv6_ifa_notify(0, ifp);
4723
4724         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4725                 int rc = -ENOENT;
4726
4727                 if (had_prefixroute)
4728                         rc = modify_prefix_route(ifp, expires, flags, false);
4729
4730                 /* prefix route could have been deleted; if so restore it */
4731                 if (rc == -ENOENT) {
4732                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4733                                               ifp->rt_priority, ifp->idev->dev,
4734                                               expires, flags, GFP_KERNEL);
4735                 }
4736
4737                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4738                         rc = modify_prefix_route(ifp, expires, flags, true);
4739
4740                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4741                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4742                                               ifp->rt_priority, ifp->idev->dev,
4743                                               expires, flags, GFP_KERNEL);
4744                 }
4745         } else if (had_prefixroute) {
4746                 enum cleanup_prefix_rt_t action;
4747                 unsigned long rt_expires;
4748
4749                 write_lock_bh(&ifp->idev->lock);
4750                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4751                 write_unlock_bh(&ifp->idev->lock);
4752
4753                 if (action != CLEANUP_PREFIX_RT_NOP) {
4754                         cleanup_prefix_route(ifp, rt_expires,
4755                                 action == CLEANUP_PREFIX_RT_DEL, false);
4756                 }
4757         }
4758
4759         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4760                 if (was_managetempaddr &&
4761                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4762                         cfg->valid_lft = 0;
4763                         cfg->preferred_lft = 0;
4764                 }
4765                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4766                                  cfg->preferred_lft, !was_managetempaddr,
4767                                  jiffies);
4768         }
4769
4770         addrconf_verify_rtnl();
4771
4772         return 0;
4773 }
4774
4775 static int
4776 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4777                   struct netlink_ext_ack *extack)
4778 {
4779         struct net *net = sock_net(skb->sk);
4780         struct ifaddrmsg *ifm;
4781         struct nlattr *tb[IFA_MAX+1];
4782         struct in6_addr *peer_pfx;
4783         struct inet6_ifaddr *ifa;
4784         struct net_device *dev;
4785         struct inet6_dev *idev;
4786         struct ifa6_config cfg;
4787         int err;
4788
4789         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4790                                      ifa_ipv6_policy, extack);
4791         if (err < 0)
4792                 return err;
4793
4794         memset(&cfg, 0, sizeof(cfg));
4795
4796         ifm = nlmsg_data(nlh);
4797         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4798         if (!cfg.pfx)
4799                 return -EINVAL;
4800
4801         cfg.peer_pfx = peer_pfx;
4802         cfg.plen = ifm->ifa_prefixlen;
4803         if (tb[IFA_RT_PRIORITY])
4804                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4805
4806         cfg.valid_lft = INFINITY_LIFE_TIME;
4807         cfg.preferred_lft = INFINITY_LIFE_TIME;
4808
4809         if (tb[IFA_CACHEINFO]) {
4810                 struct ifa_cacheinfo *ci;
4811
4812                 ci = nla_data(tb[IFA_CACHEINFO]);
4813                 cfg.valid_lft = ci->ifa_valid;
4814                 cfg.preferred_lft = ci->ifa_prefered;
4815         }
4816
4817         dev =  __dev_get_by_index(net, ifm->ifa_index);
4818         if (!dev)
4819                 return -ENODEV;
4820
4821         if (tb[IFA_FLAGS])
4822                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4823         else
4824                 cfg.ifa_flags = ifm->ifa_flags;
4825
4826         /* We ignore other flags so far. */
4827         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4828                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4829                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4830
4831         idev = ipv6_find_idev(dev);
4832         if (IS_ERR(idev))
4833                 return PTR_ERR(idev);
4834
4835         if (!ipv6_allow_optimistic_dad(net, idev))
4836                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4837
4838         if (cfg.ifa_flags & IFA_F_NODAD &&
4839             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4840                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4841                 return -EINVAL;
4842         }
4843
4844         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4845         if (!ifa) {
4846                 /*
4847                  * It would be best to check for !NLM_F_CREATE here but
4848                  * userspace already relies on not having to provide this.
4849                  */
4850                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4851         }
4852
4853         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4854             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4855                 err = -EEXIST;
4856         else
4857                 err = inet6_addr_modify(ifa, &cfg);
4858
4859         in6_ifa_put(ifa);
4860
4861         return err;
4862 }
4863
4864 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4865                           u8 scope, int ifindex)
4866 {
4867         struct ifaddrmsg *ifm;
4868
4869         ifm = nlmsg_data(nlh);
4870         ifm->ifa_family = AF_INET6;
4871         ifm->ifa_prefixlen = prefixlen;
4872         ifm->ifa_flags = flags;
4873         ifm->ifa_scope = scope;
4874         ifm->ifa_index = ifindex;
4875 }
4876
4877 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4878                          unsigned long tstamp, u32 preferred, u32 valid)
4879 {
4880         struct ifa_cacheinfo ci;
4881
4882         ci.cstamp = cstamp_delta(cstamp);
4883         ci.tstamp = cstamp_delta(tstamp);
4884         ci.ifa_prefered = preferred;
4885         ci.ifa_valid = valid;
4886
4887         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4888 }
4889
4890 static inline int rt_scope(int ifa_scope)
4891 {
4892         if (ifa_scope & IFA_HOST)
4893                 return RT_SCOPE_HOST;
4894         else if (ifa_scope & IFA_LINK)
4895                 return RT_SCOPE_LINK;
4896         else if (ifa_scope & IFA_SITE)
4897                 return RT_SCOPE_SITE;
4898         else
4899                 return RT_SCOPE_UNIVERSE;
4900 }
4901
4902 static inline int inet6_ifaddr_msgsize(void)
4903 {
4904         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4905                + nla_total_size(16) /* IFA_LOCAL */
4906                + nla_total_size(16) /* IFA_ADDRESS */
4907                + nla_total_size(sizeof(struct ifa_cacheinfo))
4908                + nla_total_size(4)  /* IFA_FLAGS */
4909                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4910 }
4911
4912 enum addr_type_t {
4913         UNICAST_ADDR,
4914         MULTICAST_ADDR,
4915         ANYCAST_ADDR,
4916 };
4917
4918 struct inet6_fill_args {
4919         u32 portid;
4920         u32 seq;
4921         int event;
4922         unsigned int flags;
4923         int netnsid;
4924         int ifindex;
4925         enum addr_type_t type;
4926 };
4927
4928 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4929                              struct inet6_fill_args *args)
4930 {
4931         struct nlmsghdr  *nlh;
4932         u32 preferred, valid;
4933
4934         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4935                         sizeof(struct ifaddrmsg), args->flags);
4936         if (!nlh)
4937                 return -EMSGSIZE;
4938
4939         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4940                       ifa->idev->dev->ifindex);
4941
4942         if (args->netnsid >= 0 &&
4943             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4944                 goto error;
4945
4946         spin_lock_bh(&ifa->lock);
4947         if (!((ifa->flags&IFA_F_PERMANENT) &&
4948               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4949                 preferred = ifa->prefered_lft;
4950                 valid = ifa->valid_lft;
4951                 if (preferred != INFINITY_LIFE_TIME) {
4952                         long tval = (jiffies - ifa->tstamp)/HZ;
4953                         if (preferred > tval)
4954                                 preferred -= tval;
4955                         else
4956                                 preferred = 0;
4957                         if (valid != INFINITY_LIFE_TIME) {
4958                                 if (valid > tval)
4959                                         valid -= tval;
4960                                 else
4961                                         valid = 0;
4962                         }
4963                 }
4964         } else {
4965                 preferred = INFINITY_LIFE_TIME;
4966                 valid = INFINITY_LIFE_TIME;
4967         }
4968         spin_unlock_bh(&ifa->lock);
4969
4970         if (!ipv6_addr_any(&ifa->peer_addr)) {
4971                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4972                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4973                         goto error;
4974         } else
4975                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4976                         goto error;
4977
4978         if (ifa->rt_priority &&
4979             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4980                 goto error;
4981
4982         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4983                 goto error;
4984
4985         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4986                 goto error;
4987
4988         nlmsg_end(skb, nlh);
4989         return 0;
4990
4991 error:
4992         nlmsg_cancel(skb, nlh);
4993         return -EMSGSIZE;
4994 }
4995
4996 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4997                                struct inet6_fill_args *args)
4998 {
4999         struct nlmsghdr  *nlh;
5000         u8 scope = RT_SCOPE_UNIVERSE;
5001         int ifindex = ifmca->idev->dev->ifindex;
5002
5003         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5004                 scope = RT_SCOPE_SITE;
5005
5006         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5007                         sizeof(struct ifaddrmsg), args->flags);
5008         if (!nlh)
5009                 return -EMSGSIZE;
5010
5011         if (args->netnsid >= 0 &&
5012             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5013                 nlmsg_cancel(skb, nlh);
5014                 return -EMSGSIZE;
5015         }
5016
5017         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5018         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5019             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5020                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5021                 nlmsg_cancel(skb, nlh);
5022                 return -EMSGSIZE;
5023         }
5024
5025         nlmsg_end(skb, nlh);
5026         return 0;
5027 }
5028
5029 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5030                                struct inet6_fill_args *args)
5031 {
5032         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5033         int ifindex = dev ? dev->ifindex : 1;
5034         struct nlmsghdr  *nlh;
5035         u8 scope = RT_SCOPE_UNIVERSE;
5036
5037         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5038                 scope = RT_SCOPE_SITE;
5039
5040         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5041                         sizeof(struct ifaddrmsg), args->flags);
5042         if (!nlh)
5043                 return -EMSGSIZE;
5044
5045         if (args->netnsid >= 0 &&
5046             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5047                 nlmsg_cancel(skb, nlh);
5048                 return -EMSGSIZE;
5049         }
5050
5051         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5052         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5053             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5054                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5055                 nlmsg_cancel(skb, nlh);
5056                 return -EMSGSIZE;
5057         }
5058
5059         nlmsg_end(skb, nlh);
5060         return 0;
5061 }
5062
5063 /* called with rcu_read_lock() */
5064 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5065                           struct netlink_callback *cb, int s_ip_idx,
5066                           struct inet6_fill_args *fillargs)
5067 {
5068         struct ifmcaddr6 *ifmca;
5069         struct ifacaddr6 *ifaca;
5070         int ip_idx = 0;
5071         int err = 1;
5072
5073         read_lock_bh(&idev->lock);
5074         switch (fillargs->type) {
5075         case UNICAST_ADDR: {
5076                 struct inet6_ifaddr *ifa;
5077                 fillargs->event = RTM_NEWADDR;
5078
5079                 /* unicast address incl. temp addr */
5080                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5081                         if (ip_idx < s_ip_idx)
5082                                 goto next;
5083                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5084                         if (err < 0)
5085                                 break;
5086                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5087 next:
5088                         ip_idx++;
5089                 }
5090                 break;
5091         }
5092         case MULTICAST_ADDR:
5093                 fillargs->event = RTM_GETMULTICAST;
5094
5095                 /* multicast address */
5096                 for (ifmca = idev->mc_list; ifmca;
5097                      ifmca = ifmca->next, ip_idx++) {
5098                         if (ip_idx < s_ip_idx)
5099                                 continue;
5100                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5101                         if (err < 0)
5102                                 break;
5103                 }
5104                 break;
5105         case ANYCAST_ADDR:
5106                 fillargs->event = RTM_GETANYCAST;
5107                 /* anycast address */
5108                 for (ifaca = idev->ac_list; ifaca;
5109                      ifaca = ifaca->aca_next, ip_idx++) {
5110                         if (ip_idx < s_ip_idx)
5111                                 continue;
5112                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5113                         if (err < 0)
5114                                 break;
5115                 }
5116                 break;
5117         default:
5118                 break;
5119         }
5120         read_unlock_bh(&idev->lock);
5121         cb->args[2] = ip_idx;
5122         return err;
5123 }
5124
5125 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5126                                        struct inet6_fill_args *fillargs,
5127                                        struct net **tgt_net, struct sock *sk,
5128                                        struct netlink_callback *cb)
5129 {
5130         struct netlink_ext_ack *extack = cb->extack;
5131         struct nlattr *tb[IFA_MAX+1];
5132         struct ifaddrmsg *ifm;
5133         int err, i;
5134
5135         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5136                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5137                 return -EINVAL;
5138         }
5139
5140         ifm = nlmsg_data(nlh);
5141         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5142                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5143                 return -EINVAL;
5144         }
5145
5146         fillargs->ifindex = ifm->ifa_index;
5147         if (fillargs->ifindex) {
5148                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5149                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5150         }
5151
5152         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5153                                             ifa_ipv6_policy, extack);
5154         if (err < 0)
5155                 return err;
5156
5157         for (i = 0; i <= IFA_MAX; ++i) {
5158                 if (!tb[i])
5159                         continue;
5160
5161                 if (i == IFA_TARGET_NETNSID) {
5162                         struct net *net;
5163
5164                         fillargs->netnsid = nla_get_s32(tb[i]);
5165                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5166                         if (IS_ERR(net)) {
5167                                 fillargs->netnsid = -1;
5168                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5169                                 return PTR_ERR(net);
5170                         }
5171                         *tgt_net = net;
5172                 } else {
5173                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5174                         return -EINVAL;
5175                 }
5176         }
5177
5178         return 0;
5179 }
5180
5181 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5182                            enum addr_type_t type)
5183 {
5184         const struct nlmsghdr *nlh = cb->nlh;
5185         struct inet6_fill_args fillargs = {
5186                 .portid = NETLINK_CB(cb->skb).portid,
5187                 .seq = cb->nlh->nlmsg_seq,
5188                 .flags = NLM_F_MULTI,
5189                 .netnsid = -1,
5190                 .type = type,
5191         };
5192         struct net *net = sock_net(skb->sk);
5193         struct net *tgt_net = net;
5194         int idx, s_idx, s_ip_idx;
5195         int h, s_h;
5196         struct net_device *dev;
5197         struct inet6_dev *idev;
5198         struct hlist_head *head;
5199         int err = 0;
5200
5201         s_h = cb->args[0];
5202         s_idx = idx = cb->args[1];
5203         s_ip_idx = cb->args[2];
5204
5205         if (cb->strict_check) {
5206                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5207                                                   skb->sk, cb);
5208                 if (err < 0)
5209                         goto put_tgt_net;
5210
5211                 err = 0;
5212                 if (fillargs.ifindex) {
5213                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5214                         if (!dev) {
5215                                 err = -ENODEV;
5216                                 goto put_tgt_net;
5217                         }
5218                         idev = __in6_dev_get(dev);
5219                         if (idev) {
5220                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5221                                                      &fillargs);
5222                                 if (err > 0)
5223                                         err = 0;
5224                         }
5225                         goto put_tgt_net;
5226                 }
5227         }
5228
5229         rcu_read_lock();
5230         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5231         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5232                 idx = 0;
5233                 head = &tgt_net->dev_index_head[h];
5234                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5235                         if (idx < s_idx)
5236                                 goto cont;
5237                         if (h > s_h || idx > s_idx)
5238                                 s_ip_idx = 0;
5239                         idev = __in6_dev_get(dev);
5240                         if (!idev)
5241                                 goto cont;
5242
5243                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5244                                            &fillargs) < 0)
5245                                 goto done;
5246 cont:
5247                         idx++;
5248                 }
5249         }
5250 done:
5251         rcu_read_unlock();
5252         cb->args[0] = h;
5253         cb->args[1] = idx;
5254 put_tgt_net:
5255         if (fillargs.netnsid >= 0)
5256                 put_net(tgt_net);
5257
5258         return skb->len ? : err;
5259 }
5260
5261 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5262 {
5263         enum addr_type_t type = UNICAST_ADDR;
5264
5265         return inet6_dump_addr(skb, cb, type);
5266 }
5267
5268 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5269 {
5270         enum addr_type_t type = MULTICAST_ADDR;
5271
5272         return inet6_dump_addr(skb, cb, type);
5273 }
5274
5275
5276 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5277 {
5278         enum addr_type_t type = ANYCAST_ADDR;
5279
5280         return inet6_dump_addr(skb, cb, type);
5281 }
5282
5283 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5284                                        const struct nlmsghdr *nlh,
5285                                        struct nlattr **tb,
5286                                        struct netlink_ext_ack *extack)
5287 {
5288         struct ifaddrmsg *ifm;
5289         int i, err;
5290
5291         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5292                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5293                 return -EINVAL;
5294         }
5295
5296         if (!netlink_strict_get_check(skb))
5297                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5298                                               ifa_ipv6_policy, extack);
5299
5300         ifm = nlmsg_data(nlh);
5301         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5302                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5303                 return -EINVAL;
5304         }
5305
5306         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5307                                             ifa_ipv6_policy, extack);
5308         if (err)
5309                 return err;
5310
5311         for (i = 0; i <= IFA_MAX; i++) {
5312                 if (!tb[i])
5313                         continue;
5314
5315                 switch (i) {
5316                 case IFA_TARGET_NETNSID:
5317                 case IFA_ADDRESS:
5318                 case IFA_LOCAL:
5319                         break;
5320                 default:
5321                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5322                         return -EINVAL;
5323                 }
5324         }
5325
5326         return 0;
5327 }
5328
5329 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5330                              struct netlink_ext_ack *extack)
5331 {
5332         struct net *net = sock_net(in_skb->sk);
5333         struct inet6_fill_args fillargs = {
5334                 .portid = NETLINK_CB(in_skb).portid,
5335                 .seq = nlh->nlmsg_seq,
5336                 .event = RTM_NEWADDR,
5337                 .flags = 0,
5338                 .netnsid = -1,
5339         };
5340         struct net *tgt_net = net;
5341         struct ifaddrmsg *ifm;
5342         struct nlattr *tb[IFA_MAX+1];
5343         struct in6_addr *addr = NULL, *peer;
5344         struct net_device *dev = NULL;
5345         struct inet6_ifaddr *ifa;
5346         struct sk_buff *skb;
5347         int err;
5348
5349         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5350         if (err < 0)
5351                 return err;
5352
5353         if (tb[IFA_TARGET_NETNSID]) {
5354                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5355
5356                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5357                                                   fillargs.netnsid);
5358                 if (IS_ERR(tgt_net))
5359                         return PTR_ERR(tgt_net);
5360         }
5361
5362         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5363         if (!addr)
5364                 return -EINVAL;
5365
5366         ifm = nlmsg_data(nlh);
5367         if (ifm->ifa_index)
5368                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5369
5370         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5371         if (!ifa) {
5372                 err = -EADDRNOTAVAIL;
5373                 goto errout;
5374         }
5375
5376         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5377         if (!skb) {
5378                 err = -ENOBUFS;
5379                 goto errout_ifa;
5380         }
5381
5382         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5383         if (err < 0) {
5384                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5385                 WARN_ON(err == -EMSGSIZE);
5386                 kfree_skb(skb);
5387                 goto errout_ifa;
5388         }
5389         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5390 errout_ifa:
5391         in6_ifa_put(ifa);
5392 errout:
5393         if (dev)
5394                 dev_put(dev);
5395         if (fillargs.netnsid >= 0)
5396                 put_net(tgt_net);
5397
5398         return err;
5399 }
5400
5401 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5402 {
5403         struct sk_buff *skb;
5404         struct net *net = dev_net(ifa->idev->dev);
5405         struct inet6_fill_args fillargs = {
5406                 .portid = 0,
5407                 .seq = 0,
5408                 .event = event,
5409                 .flags = 0,
5410                 .netnsid = -1,
5411         };
5412         int err = -ENOBUFS;
5413
5414         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5415         if (!skb)
5416                 goto errout;
5417
5418         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5419         if (err < 0) {
5420                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5421                 WARN_ON(err == -EMSGSIZE);
5422                 kfree_skb(skb);
5423                 goto errout;
5424         }
5425         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5426         return;
5427 errout:
5428         if (err < 0)
5429                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5430 }
5431
5432 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5433                                 __s32 *array, int bytes)
5434 {
5435         BUG_ON(bytes < (DEVCONF_MAX * 4));
5436
5437         memset(array, 0, bytes);
5438         array[DEVCONF_FORWARDING] = cnf->forwarding;
5439         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5440         array[DEVCONF_MTU6] = cnf->mtu6;
5441         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5442         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5443         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5444         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5445         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5446         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5447                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5448         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5449                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5450         array[DEVCONF_RTR_SOLICIT_DELAY] =
5451                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5452         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5453         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5454                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5455         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5456                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5457         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5458         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5459         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5460         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5461         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5462         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5463         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5464         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5465         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5466 #ifdef CONFIG_IPV6_ROUTER_PREF
5467         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5468         array[DEVCONF_RTR_PROBE_INTERVAL] =
5469                 jiffies_to_msecs(cnf->rtr_probe_interval);
5470 #ifdef CONFIG_IPV6_ROUTE_INFO
5471         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5472         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5473 #endif
5474 #endif
5475         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5476         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5477 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5478         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5479         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5480 #endif
5481 #ifdef CONFIG_IPV6_MROUTE
5482         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5483 #endif
5484         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5485         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5486         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5487         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5488         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5489         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5490         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5491         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5492         /* we omit DEVCONF_STABLE_SECRET for now */
5493         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5494         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5495         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5496         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5497         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5498 #ifdef CONFIG_IPV6_SEG6_HMAC
5499         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5500 #endif
5501         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5502         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5503         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5504         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5505 }
5506
5507 static inline size_t inet6_ifla6_size(void)
5508 {
5509         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5510              + nla_total_size(sizeof(struct ifla_cacheinfo))
5511              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5512              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5513              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5514              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5515              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5516              + 0;
5517 }
5518
5519 static inline size_t inet6_if_nlmsg_size(void)
5520 {
5521         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5522                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5523                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5524                + nla_total_size(4) /* IFLA_MTU */
5525                + nla_total_size(4) /* IFLA_LINK */
5526                + nla_total_size(1) /* IFLA_OPERSTATE */
5527                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5528 }
5529
5530 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5531                                         int bytes)
5532 {
5533         int i;
5534         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5535         BUG_ON(pad < 0);
5536
5537         /* Use put_unaligned() because stats may not be aligned for u64. */
5538         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5539         for (i = 1; i < ICMP6_MIB_MAX; i++)
5540                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5541
5542         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5543 }
5544
5545 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5546                                         int bytes, size_t syncpoff)
5547 {
5548         int i, c;
5549         u64 buff[IPSTATS_MIB_MAX];
5550         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5551
5552         BUG_ON(pad < 0);
5553
5554         memset(buff, 0, sizeof(buff));
5555         buff[0] = IPSTATS_MIB_MAX;
5556
5557         for_each_possible_cpu(c) {
5558                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5559                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5560         }
5561
5562         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5563         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5564 }
5565
5566 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5567                              int bytes)
5568 {
5569         switch (attrtype) {
5570         case IFLA_INET6_STATS:
5571                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5572                                      offsetof(struct ipstats_mib, syncp));
5573                 break;
5574         case IFLA_INET6_ICMP6STATS:
5575                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5576                 break;
5577         }
5578 }
5579
5580 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5581                                   u32 ext_filter_mask)
5582 {
5583         struct nlattr *nla;
5584         struct ifla_cacheinfo ci;
5585
5586         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5587                 goto nla_put_failure;
5588         ci.max_reasm_len = IPV6_MAXPLEN;
5589         ci.tstamp = cstamp_delta(idev->tstamp);
5590         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5591         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5592         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5593                 goto nla_put_failure;
5594         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5595         if (!nla)
5596                 goto nla_put_failure;
5597         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5598
5599         /* XXX - MC not implemented */
5600
5601         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5602                 return 0;
5603
5604         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5605         if (!nla)
5606                 goto nla_put_failure;
5607         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5608
5609         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5610         if (!nla)
5611                 goto nla_put_failure;
5612         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5613
5614         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5615         if (!nla)
5616                 goto nla_put_failure;
5617
5618         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5619                 goto nla_put_failure;
5620
5621         read_lock_bh(&idev->lock);
5622         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5623         read_unlock_bh(&idev->lock);
5624
5625         return 0;
5626
5627 nla_put_failure:
5628         return -EMSGSIZE;
5629 }
5630
5631 static size_t inet6_get_link_af_size(const struct net_device *dev,
5632                                      u32 ext_filter_mask)
5633 {
5634         if (!__in6_dev_get(dev))
5635                 return 0;
5636
5637         return inet6_ifla6_size();
5638 }
5639
5640 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5641                               u32 ext_filter_mask)
5642 {
5643         struct inet6_dev *idev = __in6_dev_get(dev);
5644
5645         if (!idev)
5646                 return -ENODATA;
5647
5648         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5649                 return -EMSGSIZE;
5650
5651         return 0;
5652 }
5653
5654 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5655 {
5656         struct inet6_ifaddr *ifp;
5657         struct net_device *dev = idev->dev;
5658         bool clear_token, update_rs = false;
5659         struct in6_addr ll_addr;
5660
5661         ASSERT_RTNL();
5662
5663         if (!token)
5664                 return -EINVAL;
5665         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5666                 return -EINVAL;
5667         if (!ipv6_accept_ra(idev))
5668                 return -EINVAL;
5669         if (idev->cnf.rtr_solicits == 0)
5670                 return -EINVAL;
5671
5672         write_lock_bh(&idev->lock);
5673
5674         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5675         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5676
5677         write_unlock_bh(&idev->lock);
5678
5679         clear_token = ipv6_addr_any(token);
5680         if (clear_token)
5681                 goto update_lft;
5682
5683         if (!idev->dead && (idev->if_flags & IF_READY) &&
5684             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5685                              IFA_F_OPTIMISTIC)) {
5686                 /* If we're not ready, then normal ifup will take care
5687                  * of this. Otherwise, we need to request our rs here.
5688                  */
5689                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5690                 update_rs = true;
5691         }
5692
5693 update_lft:
5694         write_lock_bh(&idev->lock);
5695
5696         if (update_rs) {
5697                 idev->if_flags |= IF_RS_SENT;
5698                 idev->rs_interval = rfc3315_s14_backoff_init(
5699                         idev->cnf.rtr_solicit_interval);
5700                 idev->rs_probes = 1;
5701                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5702         }
5703
5704         /* Well, that's kinda nasty ... */
5705         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5706                 spin_lock(&ifp->lock);
5707                 if (ifp->tokenized) {
5708                         ifp->valid_lft = 0;
5709                         ifp->prefered_lft = 0;
5710                 }
5711                 spin_unlock(&ifp->lock);
5712         }
5713
5714         write_unlock_bh(&idev->lock);
5715         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5716         addrconf_verify_rtnl();
5717         return 0;
5718 }
5719
5720 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5721         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5722         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5723 };
5724
5725 static int check_addr_gen_mode(int mode)
5726 {
5727         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5728             mode != IN6_ADDR_GEN_MODE_NONE &&
5729             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5730             mode != IN6_ADDR_GEN_MODE_RANDOM)
5731                 return -EINVAL;
5732         return 1;
5733 }
5734
5735 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5736                                 int mode)
5737 {
5738         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5739             !idev->cnf.stable_secret.initialized &&
5740             !net->ipv6.devconf_dflt->stable_secret.initialized)
5741                 return -EINVAL;
5742         return 1;
5743 }
5744
5745 static int inet6_validate_link_af(const struct net_device *dev,
5746                                   const struct nlattr *nla)
5747 {
5748         struct nlattr *tb[IFLA_INET6_MAX + 1];
5749         struct inet6_dev *idev = NULL;
5750         int err;
5751
5752         if (dev) {
5753                 idev = __in6_dev_get(dev);
5754                 if (!idev)
5755                         return -EAFNOSUPPORT;
5756         }
5757
5758         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5759                                           inet6_af_policy, NULL);
5760         if (err)
5761                 return err;
5762
5763         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5764                 return -EINVAL;
5765
5766         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5767                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5768
5769                 if (check_addr_gen_mode(mode) < 0)
5770                         return -EINVAL;
5771                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5772                         return -EINVAL;
5773         }
5774
5775         return 0;
5776 }
5777
5778 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5779 {
5780         struct inet6_dev *idev = __in6_dev_get(dev);
5781         struct nlattr *tb[IFLA_INET6_MAX + 1];
5782         int err;
5783
5784         if (!idev)
5785                 return -EAFNOSUPPORT;
5786
5787         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5788                 return -EINVAL;
5789
5790         if (tb[IFLA_INET6_TOKEN]) {
5791                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5792                 if (err)
5793                         return err;
5794         }
5795
5796         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5797                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5798
5799                 idev->cnf.addr_gen_mode = mode;
5800         }
5801
5802         return 0;
5803 }
5804
5805 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5806                              u32 portid, u32 seq, int event, unsigned int flags)
5807 {
5808         struct net_device *dev = idev->dev;
5809         struct ifinfomsg *hdr;
5810         struct nlmsghdr *nlh;
5811         void *protoinfo;
5812
5813         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5814         if (!nlh)
5815                 return -EMSGSIZE;
5816
5817         hdr = nlmsg_data(nlh);
5818         hdr->ifi_family = AF_INET6;
5819         hdr->__ifi_pad = 0;
5820         hdr->ifi_type = dev->type;
5821         hdr->ifi_index = dev->ifindex;
5822         hdr->ifi_flags = dev_get_flags(dev);
5823         hdr->ifi_change = 0;
5824
5825         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5826             (dev->addr_len &&
5827              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5828             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5829             (dev->ifindex != dev_get_iflink(dev) &&
5830              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5831             nla_put_u8(skb, IFLA_OPERSTATE,
5832                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5833                 goto nla_put_failure;
5834         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5835         if (!protoinfo)
5836                 goto nla_put_failure;
5837
5838         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5839                 goto nla_put_failure;
5840
5841         nla_nest_end(skb, protoinfo);
5842         nlmsg_end(skb, nlh);
5843         return 0;
5844
5845 nla_put_failure:
5846         nlmsg_cancel(skb, nlh);
5847         return -EMSGSIZE;
5848 }
5849
5850 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5851                                    struct netlink_ext_ack *extack)
5852 {
5853         struct ifinfomsg *ifm;
5854
5855         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5856                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5857                 return -EINVAL;
5858         }
5859
5860         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5861                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5862                 return -EINVAL;
5863         }
5864
5865         ifm = nlmsg_data(nlh);
5866         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5867             ifm->ifi_change || ifm->ifi_index) {
5868                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5869                 return -EINVAL;
5870         }
5871
5872         return 0;
5873 }
5874
5875 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5876 {
5877         struct net *net = sock_net(skb->sk);
5878         int h, s_h;
5879         int idx = 0, s_idx;
5880         struct net_device *dev;
5881         struct inet6_dev *idev;
5882         struct hlist_head *head;
5883
5884         /* only requests using strict checking can pass data to
5885          * influence the dump
5886          */
5887         if (cb->strict_check) {
5888                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5889
5890                 if (err < 0)
5891                         return err;
5892         }
5893
5894         s_h = cb->args[0];
5895         s_idx = cb->args[1];
5896
5897         rcu_read_lock();
5898         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5899                 idx = 0;
5900                 head = &net->dev_index_head[h];
5901                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5902                         if (idx < s_idx)
5903                                 goto cont;
5904                         idev = __in6_dev_get(dev);
5905                         if (!idev)
5906                                 goto cont;
5907                         if (inet6_fill_ifinfo(skb, idev,
5908                                               NETLINK_CB(cb->skb).portid,
5909                                               cb->nlh->nlmsg_seq,
5910                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5911                                 goto out;
5912 cont:
5913                         idx++;
5914                 }
5915         }
5916 out:
5917         rcu_read_unlock();
5918         cb->args[1] = idx;
5919         cb->args[0] = h;
5920
5921         return skb->len;
5922 }
5923
5924 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5925 {
5926         struct sk_buff *skb;
5927         struct net *net = dev_net(idev->dev);
5928         int err = -ENOBUFS;
5929
5930         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5931         if (!skb)
5932                 goto errout;
5933
5934         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5935         if (err < 0) {
5936                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5937                 WARN_ON(err == -EMSGSIZE);
5938                 kfree_skb(skb);
5939                 goto errout;
5940         }
5941         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5942         return;
5943 errout:
5944         if (err < 0)
5945                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5946 }
5947
5948 static inline size_t inet6_prefix_nlmsg_size(void)
5949 {
5950         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5951                + nla_total_size(sizeof(struct in6_addr))
5952                + nla_total_size(sizeof(struct prefix_cacheinfo));
5953 }
5954
5955 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5956                              struct prefix_info *pinfo, u32 portid, u32 seq,
5957                              int event, unsigned int flags)
5958 {
5959         struct prefixmsg *pmsg;
5960         struct nlmsghdr *nlh;
5961         struct prefix_cacheinfo ci;
5962
5963         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5964         if (!nlh)
5965                 return -EMSGSIZE;
5966
5967         pmsg = nlmsg_data(nlh);
5968         pmsg->prefix_family = AF_INET6;
5969         pmsg->prefix_pad1 = 0;
5970         pmsg->prefix_pad2 = 0;
5971         pmsg->prefix_ifindex = idev->dev->ifindex;
5972         pmsg->prefix_len = pinfo->prefix_len;
5973         pmsg->prefix_type = pinfo->type;
5974         pmsg->prefix_pad3 = 0;
5975         pmsg->prefix_flags = 0;
5976         if (pinfo->onlink)
5977                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5978         if (pinfo->autoconf)
5979                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5980
5981         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5982                 goto nla_put_failure;
5983         ci.preferred_time = ntohl(pinfo->prefered);
5984         ci.valid_time = ntohl(pinfo->valid);
5985         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5986                 goto nla_put_failure;
5987         nlmsg_end(skb, nlh);
5988         return 0;
5989
5990 nla_put_failure:
5991         nlmsg_cancel(skb, nlh);
5992         return -EMSGSIZE;
5993 }
5994
5995 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5996                          struct prefix_info *pinfo)
5997 {
5998         struct sk_buff *skb;
5999         struct net *net = dev_net(idev->dev);
6000         int err = -ENOBUFS;
6001
6002         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6003         if (!skb)
6004                 goto errout;
6005
6006         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6007         if (err < 0) {
6008                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6009                 WARN_ON(err == -EMSGSIZE);
6010                 kfree_skb(skb);
6011                 goto errout;
6012         }
6013         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6014         return;
6015 errout:
6016         if (err < 0)
6017                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6018 }
6019
6020 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6021 {
6022         struct net *net = dev_net(ifp->idev->dev);
6023
6024         if (event)
6025                 ASSERT_RTNL();
6026
6027         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6028
6029         switch (event) {
6030         case RTM_NEWADDR:
6031                 /*
6032                  * If the address was optimistic we inserted the route at the
6033                  * start of our DAD process, so we don't need to do it again.
6034                  * If the device was taken down in the middle of the DAD
6035                  * cycle there is a race where we could get here without a
6036                  * host route, so nothing to insert. That will be fixed when
6037                  * the device is brought up.
6038                  */
6039                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6040                         ip6_ins_rt(net, ifp->rt);
6041                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6042                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6043                                 &ifp->addr, ifp->idev->dev->name);
6044                 }
6045
6046                 if (ifp->idev->cnf.forwarding)
6047                         addrconf_join_anycast(ifp);
6048                 if (!ipv6_addr_any(&ifp->peer_addr))
6049                         addrconf_prefix_route(&ifp->peer_addr, 128,
6050                                               ifp->rt_priority, ifp->idev->dev,
6051                                               0, 0, GFP_ATOMIC);
6052                 break;
6053         case RTM_DELADDR:
6054                 if (ifp->idev->cnf.forwarding)
6055                         addrconf_leave_anycast(ifp);
6056                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6057                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6058                         struct fib6_info *rt;
6059
6060                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6061                                                        ifp->idev->dev, 0, 0,
6062                                                        false);
6063                         if (rt)
6064                                 ip6_del_rt(net, rt);
6065                 }
6066                 if (ifp->rt) {
6067                         ip6_del_rt(net, ifp->rt);
6068                         ifp->rt = NULL;
6069                 }
6070                 rt_genid_bump_ipv6(net);
6071                 break;
6072         }
6073         atomic_inc(&net->ipv6.dev_addr_genid);
6074 }
6075
6076 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6077 {
6078         rcu_read_lock_bh();
6079         if (likely(ifp->idev->dead == 0))
6080                 __ipv6_ifa_notify(event, ifp);
6081         rcu_read_unlock_bh();
6082 }
6083
6084 #ifdef CONFIG_SYSCTL
6085
6086 static
6087 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6088                            void __user *buffer, size_t *lenp, loff_t *ppos)
6089 {
6090         int *valp = ctl->data;
6091         int val = *valp;
6092         loff_t pos = *ppos;
6093         struct ctl_table lctl;
6094         int ret;
6095
6096         /*
6097          * ctl->data points to idev->cnf.forwarding, we should
6098          * not modify it until we get the rtnl lock.
6099          */
6100         lctl = *ctl;
6101         lctl.data = &val;
6102
6103         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6104
6105         if (write)
6106                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6107         if (ret)
6108                 *ppos = pos;
6109         return ret;
6110 }
6111
6112 static
6113 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6114                         void __user *buffer, size_t *lenp, loff_t *ppos)
6115 {
6116         struct inet6_dev *idev = ctl->extra1;
6117         int min_mtu = IPV6_MIN_MTU;
6118         struct ctl_table lctl;
6119
6120         lctl = *ctl;
6121         lctl.extra1 = &min_mtu;
6122         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6123
6124         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6125 }
6126
6127 static void dev_disable_change(struct inet6_dev *idev)
6128 {
6129         struct netdev_notifier_info info;
6130
6131         if (!idev || !idev->dev)
6132                 return;
6133
6134         netdev_notifier_info_init(&info, idev->dev);
6135         if (idev->cnf.disable_ipv6)
6136                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6137         else
6138                 addrconf_notify(NULL, NETDEV_UP, &info);
6139 }
6140
6141 static void addrconf_disable_change(struct net *net, __s32 newf)
6142 {
6143         struct net_device *dev;
6144         struct inet6_dev *idev;
6145
6146         for_each_netdev(net, dev) {
6147                 idev = __in6_dev_get(dev);
6148                 if (idev) {
6149                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6150                         idev->cnf.disable_ipv6 = newf;
6151                         if (changed)
6152                                 dev_disable_change(idev);
6153                 }
6154         }
6155 }
6156
6157 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6158 {
6159         struct net *net;
6160         int old;
6161
6162         if (!rtnl_trylock())
6163                 return restart_syscall();
6164
6165         net = (struct net *)table->extra2;
6166         old = *p;
6167         *p = newf;
6168
6169         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6170                 rtnl_unlock();
6171                 return 0;
6172         }
6173
6174         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6175                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6176                 addrconf_disable_change(net, newf);
6177         } else if ((!newf) ^ (!old))
6178                 dev_disable_change((struct inet6_dev *)table->extra1);
6179
6180         rtnl_unlock();
6181         return 0;
6182 }
6183
6184 static
6185 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6186                             void __user *buffer, size_t *lenp, loff_t *ppos)
6187 {
6188         int *valp = ctl->data;
6189         int val = *valp;
6190         loff_t pos = *ppos;
6191         struct ctl_table lctl;
6192         int ret;
6193
6194         /*
6195          * ctl->data points to idev->cnf.disable_ipv6, we should
6196          * not modify it until we get the rtnl lock.
6197          */
6198         lctl = *ctl;
6199         lctl.data = &val;
6200
6201         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6202
6203         if (write)
6204                 ret = addrconf_disable_ipv6(ctl, valp, val);
6205         if (ret)
6206                 *ppos = pos;
6207         return ret;
6208 }
6209
6210 static
6211 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6212                               void __user *buffer, size_t *lenp, loff_t *ppos)
6213 {
6214         int *valp = ctl->data;
6215         int ret;
6216         int old, new;
6217
6218         old = *valp;
6219         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6220         new = *valp;
6221
6222         if (write && old != new) {
6223                 struct net *net = ctl->extra2;
6224
6225                 if (!rtnl_trylock())
6226                         return restart_syscall();
6227
6228                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6229                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6230                                                      NETCONFA_PROXY_NEIGH,
6231                                                      NETCONFA_IFINDEX_DEFAULT,
6232                                                      net->ipv6.devconf_dflt);
6233                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6234                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6235                                                      NETCONFA_PROXY_NEIGH,
6236                                                      NETCONFA_IFINDEX_ALL,
6237                                                      net->ipv6.devconf_all);
6238                 else {
6239                         struct inet6_dev *idev = ctl->extra1;
6240
6241                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6242                                                      NETCONFA_PROXY_NEIGH,
6243                                                      idev->dev->ifindex,
6244                                                      &idev->cnf);
6245                 }
6246                 rtnl_unlock();
6247         }
6248
6249         return ret;
6250 }
6251
6252 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6253                                          void __user *buffer, size_t *lenp,
6254                                          loff_t *ppos)
6255 {
6256         int ret = 0;
6257         u32 new_val;
6258         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6259         struct net *net = (struct net *)ctl->extra2;
6260         struct ctl_table tmp = {
6261                 .data = &new_val,
6262                 .maxlen = sizeof(new_val),
6263                 .mode = ctl->mode,
6264         };
6265
6266         if (!rtnl_trylock())
6267                 return restart_syscall();
6268
6269         new_val = *((u32 *)ctl->data);
6270
6271         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6272         if (ret != 0)
6273                 goto out;
6274
6275         if (write) {
6276                 if (check_addr_gen_mode(new_val) < 0) {
6277                         ret = -EINVAL;
6278                         goto out;
6279                 }
6280
6281                 if (idev) {
6282                         if (check_stable_privacy(idev, net, new_val) < 0) {
6283                                 ret = -EINVAL;
6284                                 goto out;
6285                         }
6286
6287                         if (idev->cnf.addr_gen_mode != new_val) {
6288                                 idev->cnf.addr_gen_mode = new_val;
6289                                 addrconf_dev_config(idev->dev);
6290                         }
6291                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6292                         struct net_device *dev;
6293
6294                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6295                         for_each_netdev(net, dev) {
6296                                 idev = __in6_dev_get(dev);
6297                                 if (idev &&
6298                                     idev->cnf.addr_gen_mode != new_val) {
6299                                         idev->cnf.addr_gen_mode = new_val;
6300                                         addrconf_dev_config(idev->dev);
6301                                 }
6302                         }
6303                 }
6304
6305                 *((u32 *)ctl->data) = new_val;
6306         }
6307
6308 out:
6309         rtnl_unlock();
6310
6311         return ret;
6312 }
6313
6314 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6315                                          void __user *buffer, size_t *lenp,
6316                                          loff_t *ppos)
6317 {
6318         int err;
6319         struct in6_addr addr;
6320         char str[IPV6_MAX_STRLEN];
6321         struct ctl_table lctl = *ctl;
6322         struct net *net = ctl->extra2;
6323         struct ipv6_stable_secret *secret = ctl->data;
6324
6325         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6326                 return -EIO;
6327
6328         lctl.maxlen = IPV6_MAX_STRLEN;
6329         lctl.data = str;
6330
6331         if (!rtnl_trylock())
6332                 return restart_syscall();
6333
6334         if (!write && !secret->initialized) {
6335                 err = -EIO;
6336                 goto out;
6337         }
6338
6339         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6340         if (err >= sizeof(str)) {
6341                 err = -EIO;
6342                 goto out;
6343         }
6344
6345         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6346         if (err || !write)
6347                 goto out;
6348
6349         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6350                 err = -EIO;
6351                 goto out;
6352         }
6353
6354         secret->initialized = true;
6355         secret->secret = addr;
6356
6357         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6358                 struct net_device *dev;
6359
6360                 for_each_netdev(net, dev) {
6361                         struct inet6_dev *idev = __in6_dev_get(dev);
6362
6363                         if (idev) {
6364                                 idev->cnf.addr_gen_mode =
6365                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6366                         }
6367                 }
6368         } else {
6369                 struct inet6_dev *idev = ctl->extra1;
6370
6371                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6372         }
6373
6374 out:
6375         rtnl_unlock();
6376
6377         return err;
6378 }
6379
6380 static
6381 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6382                                                 int write,
6383                                                 void __user *buffer,
6384                                                 size_t *lenp,
6385                                                 loff_t *ppos)
6386 {
6387         int *valp = ctl->data;
6388         int val = *valp;
6389         loff_t pos = *ppos;
6390         struct ctl_table lctl;
6391         int ret;
6392
6393         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6394          * we should not modify it until we get the rtnl lock.
6395          */
6396         lctl = *ctl;
6397         lctl.data = &val;
6398
6399         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6400
6401         if (write)
6402                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6403         if (ret)
6404                 *ppos = pos;
6405         return ret;
6406 }
6407
6408 static
6409 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6410 {
6411         if (rt) {
6412                 if (action)
6413                         rt->dst.flags |= DST_NOPOLICY;
6414                 else
6415                         rt->dst.flags &= ~DST_NOPOLICY;
6416         }
6417 }
6418
6419 static
6420 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6421 {
6422         struct inet6_ifaddr *ifa;
6423
6424         read_lock_bh(&idev->lock);
6425         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6426                 spin_lock(&ifa->lock);
6427                 if (ifa->rt) {
6428                         /* host routes only use builtin fib6_nh */
6429                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6430                         int cpu;
6431
6432                         rcu_read_lock();
6433                         ifa->rt->dst_nopolicy = val ? true : false;
6434                         if (nh->rt6i_pcpu) {
6435                                 for_each_possible_cpu(cpu) {
6436                                         struct rt6_info **rtp;
6437
6438                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6439                                         addrconf_set_nopolicy(*rtp, val);
6440                                 }
6441                         }
6442                         rcu_read_unlock();
6443                 }
6444                 spin_unlock(&ifa->lock);
6445         }
6446         read_unlock_bh(&idev->lock);
6447 }
6448
6449 static
6450 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6451 {
6452         struct inet6_dev *idev;
6453         struct net *net;
6454
6455         if (!rtnl_trylock())
6456                 return restart_syscall();
6457
6458         *valp = val;
6459
6460         net = (struct net *)ctl->extra2;
6461         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6462                 rtnl_unlock();
6463                 return 0;
6464         }
6465
6466         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6467                 struct net_device *dev;
6468
6469                 for_each_netdev(net, dev) {
6470                         idev = __in6_dev_get(dev);
6471                         if (idev)
6472                                 addrconf_disable_policy_idev(idev, val);
6473                 }
6474         } else {
6475                 idev = (struct inet6_dev *)ctl->extra1;
6476                 addrconf_disable_policy_idev(idev, val);
6477         }
6478
6479         rtnl_unlock();
6480         return 0;
6481 }
6482
6483 static
6484 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6485                                    void __user *buffer, size_t *lenp,
6486                                    loff_t *ppos)
6487 {
6488         int *valp = ctl->data;
6489         int val = *valp;
6490         loff_t pos = *ppos;
6491         struct ctl_table lctl;
6492         int ret;
6493
6494         lctl = *ctl;
6495         lctl.data = &val;
6496         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6497
6498         if (write && (*valp != val))
6499                 ret = addrconf_disable_policy(ctl, valp, val);
6500
6501         if (ret)
6502                 *ppos = pos;
6503
6504         return ret;
6505 }
6506
6507 static int minus_one = -1;
6508 static const int two_five_five = 255;
6509
6510 static const struct ctl_table addrconf_sysctl[] = {
6511         {
6512                 .procname       = "forwarding",
6513                 .data           = &ipv6_devconf.forwarding,
6514                 .maxlen         = sizeof(int),
6515                 .mode           = 0644,
6516                 .proc_handler   = addrconf_sysctl_forward,
6517         },
6518         {
6519                 .procname       = "hop_limit",
6520                 .data           = &ipv6_devconf.hop_limit,
6521                 .maxlen         = sizeof(int),
6522                 .mode           = 0644,
6523                 .proc_handler   = proc_dointvec_minmax,
6524                 .extra1         = (void *)SYSCTL_ONE,
6525                 .extra2         = (void *)&two_five_five,
6526         },
6527         {
6528                 .procname       = "mtu",
6529                 .data           = &ipv6_devconf.mtu6,
6530                 .maxlen         = sizeof(int),
6531                 .mode           = 0644,
6532                 .proc_handler   = addrconf_sysctl_mtu,
6533         },
6534         {
6535                 .procname       = "accept_ra",
6536                 .data           = &ipv6_devconf.accept_ra,
6537                 .maxlen         = sizeof(int),
6538                 .mode           = 0644,
6539                 .proc_handler   = proc_dointvec,
6540         },
6541         {
6542                 .procname       = "accept_redirects",
6543                 .data           = &ipv6_devconf.accept_redirects,
6544                 .maxlen         = sizeof(int),
6545                 .mode           = 0644,
6546                 .proc_handler   = proc_dointvec,
6547         },
6548         {
6549                 .procname       = "autoconf",
6550                 .data           = &ipv6_devconf.autoconf,
6551                 .maxlen         = sizeof(int),
6552                 .mode           = 0644,
6553                 .proc_handler   = proc_dointvec,
6554         },
6555         {
6556                 .procname       = "dad_transmits",
6557                 .data           = &ipv6_devconf.dad_transmits,
6558                 .maxlen         = sizeof(int),
6559                 .mode           = 0644,
6560                 .proc_handler   = proc_dointvec,
6561         },
6562         {
6563                 .procname       = "router_solicitations",
6564                 .data           = &ipv6_devconf.rtr_solicits,
6565                 .maxlen         = sizeof(int),
6566                 .mode           = 0644,
6567                 .proc_handler   = proc_dointvec_minmax,
6568                 .extra1         = &minus_one,
6569         },
6570         {
6571                 .procname       = "router_solicitation_interval",
6572                 .data           = &ipv6_devconf.rtr_solicit_interval,
6573                 .maxlen         = sizeof(int),
6574                 .mode           = 0644,
6575                 .proc_handler   = proc_dointvec_jiffies,
6576         },
6577         {
6578                 .procname       = "router_solicitation_max_interval",
6579                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6580                 .maxlen         = sizeof(int),
6581                 .mode           = 0644,
6582                 .proc_handler   = proc_dointvec_jiffies,
6583         },
6584         {
6585                 .procname       = "router_solicitation_delay",
6586                 .data           = &ipv6_devconf.rtr_solicit_delay,
6587                 .maxlen         = sizeof(int),
6588                 .mode           = 0644,
6589                 .proc_handler   = proc_dointvec_jiffies,
6590         },
6591         {
6592                 .procname       = "force_mld_version",
6593                 .data           = &ipv6_devconf.force_mld_version,
6594                 .maxlen         = sizeof(int),
6595                 .mode           = 0644,
6596                 .proc_handler   = proc_dointvec,
6597         },
6598         {
6599                 .procname       = "mldv1_unsolicited_report_interval",
6600                 .data           =
6601                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6602                 .maxlen         = sizeof(int),
6603                 .mode           = 0644,
6604                 .proc_handler   = proc_dointvec_ms_jiffies,
6605         },
6606         {
6607                 .procname       = "mldv2_unsolicited_report_interval",
6608                 .data           =
6609                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6610                 .maxlen         = sizeof(int),
6611                 .mode           = 0644,
6612                 .proc_handler   = proc_dointvec_ms_jiffies,
6613         },
6614         {
6615                 .procname       = "use_tempaddr",
6616                 .data           = &ipv6_devconf.use_tempaddr,
6617                 .maxlen         = sizeof(int),
6618                 .mode           = 0644,
6619                 .proc_handler   = proc_dointvec,
6620         },
6621         {
6622                 .procname       = "temp_valid_lft",
6623                 .data           = &ipv6_devconf.temp_valid_lft,
6624                 .maxlen         = sizeof(int),
6625                 .mode           = 0644,
6626                 .proc_handler   = proc_dointvec,
6627         },
6628         {
6629                 .procname       = "temp_prefered_lft",
6630                 .data           = &ipv6_devconf.temp_prefered_lft,
6631                 .maxlen         = sizeof(int),
6632                 .mode           = 0644,
6633                 .proc_handler   = proc_dointvec,
6634         },
6635         {
6636                 .procname       = "regen_max_retry",
6637                 .data           = &ipv6_devconf.regen_max_retry,
6638                 .maxlen         = sizeof(int),
6639                 .mode           = 0644,
6640                 .proc_handler   = proc_dointvec,
6641         },
6642         {
6643                 .procname       = "max_desync_factor",
6644                 .data           = &ipv6_devconf.max_desync_factor,
6645                 .maxlen         = sizeof(int),
6646                 .mode           = 0644,
6647                 .proc_handler   = proc_dointvec,
6648         },
6649         {
6650                 .procname       = "max_addresses",
6651                 .data           = &ipv6_devconf.max_addresses,
6652                 .maxlen         = sizeof(int),
6653                 .mode           = 0644,
6654                 .proc_handler   = proc_dointvec,
6655         },
6656         {
6657                 .procname       = "accept_ra_defrtr",
6658                 .data           = &ipv6_devconf.accept_ra_defrtr,
6659                 .maxlen         = sizeof(int),
6660                 .mode           = 0644,
6661                 .proc_handler   = proc_dointvec,
6662         },
6663         {
6664                 .procname       = "accept_ra_min_hop_limit",
6665                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6666                 .maxlen         = sizeof(int),
6667                 .mode           = 0644,
6668                 .proc_handler   = proc_dointvec,
6669         },
6670         {
6671                 .procname       = "accept_ra_pinfo",
6672                 .data           = &ipv6_devconf.accept_ra_pinfo,
6673                 .maxlen         = sizeof(int),
6674                 .mode           = 0644,
6675                 .proc_handler   = proc_dointvec,
6676         },
6677 #ifdef CONFIG_IPV6_ROUTER_PREF
6678         {
6679                 .procname       = "accept_ra_rtr_pref",
6680                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6681                 .maxlen         = sizeof(int),
6682                 .mode           = 0644,
6683                 .proc_handler   = proc_dointvec,
6684         },
6685         {
6686                 .procname       = "router_probe_interval",
6687                 .data           = &ipv6_devconf.rtr_probe_interval,
6688                 .maxlen         = sizeof(int),
6689                 .mode           = 0644,
6690                 .proc_handler   = proc_dointvec_jiffies,
6691         },
6692 #ifdef CONFIG_IPV6_ROUTE_INFO
6693         {
6694                 .procname       = "accept_ra_rt_info_min_plen",
6695                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6696                 .maxlen         = sizeof(int),
6697                 .mode           = 0644,
6698                 .proc_handler   = proc_dointvec,
6699         },
6700         {
6701                 .procname       = "accept_ra_rt_info_max_plen",
6702                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6703                 .maxlen         = sizeof(int),
6704                 .mode           = 0644,
6705                 .proc_handler   = proc_dointvec,
6706         },
6707 #endif
6708 #endif
6709         {
6710                 .procname       = "proxy_ndp",
6711                 .data           = &ipv6_devconf.proxy_ndp,
6712                 .maxlen         = sizeof(int),
6713                 .mode           = 0644,
6714                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6715         },
6716         {
6717                 .procname       = "accept_source_route",
6718                 .data           = &ipv6_devconf.accept_source_route,
6719                 .maxlen         = sizeof(int),
6720                 .mode           = 0644,
6721                 .proc_handler   = proc_dointvec,
6722         },
6723 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6724         {
6725                 .procname       = "optimistic_dad",
6726                 .data           = &ipv6_devconf.optimistic_dad,
6727                 .maxlen         = sizeof(int),
6728                 .mode           = 0644,
6729                 .proc_handler   = proc_dointvec,
6730         },
6731         {
6732                 .procname       = "use_optimistic",
6733                 .data           = &ipv6_devconf.use_optimistic,
6734                 .maxlen         = sizeof(int),
6735                 .mode           = 0644,
6736                 .proc_handler   = proc_dointvec,
6737         },
6738 #endif
6739 #ifdef CONFIG_IPV6_MROUTE
6740         {
6741                 .procname       = "mc_forwarding",
6742                 .data           = &ipv6_devconf.mc_forwarding,
6743                 .maxlen         = sizeof(int),
6744                 .mode           = 0444,
6745                 .proc_handler   = proc_dointvec,
6746         },
6747 #endif
6748         {
6749                 .procname       = "disable_ipv6",
6750                 .data           = &ipv6_devconf.disable_ipv6,
6751                 .maxlen         = sizeof(int),
6752                 .mode           = 0644,
6753                 .proc_handler   = addrconf_sysctl_disable,
6754         },
6755         {
6756                 .procname       = "accept_dad",
6757                 .data           = &ipv6_devconf.accept_dad,
6758                 .maxlen         = sizeof(int),
6759                 .mode           = 0644,
6760                 .proc_handler   = proc_dointvec,
6761         },
6762         {
6763                 .procname       = "force_tllao",
6764                 .data           = &ipv6_devconf.force_tllao,
6765                 .maxlen         = sizeof(int),
6766                 .mode           = 0644,
6767                 .proc_handler   = proc_dointvec
6768         },
6769         {
6770                 .procname       = "ndisc_notify",
6771                 .data           = &ipv6_devconf.ndisc_notify,
6772                 .maxlen         = sizeof(int),
6773                 .mode           = 0644,
6774                 .proc_handler   = proc_dointvec
6775         },
6776         {
6777                 .procname       = "suppress_frag_ndisc",
6778                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6779                 .maxlen         = sizeof(int),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_dointvec
6782         },
6783         {
6784                 .procname       = "accept_ra_from_local",
6785                 .data           = &ipv6_devconf.accept_ra_from_local,
6786                 .maxlen         = sizeof(int),
6787                 .mode           = 0644,
6788                 .proc_handler   = proc_dointvec,
6789         },
6790         {
6791                 .procname       = "accept_ra_mtu",
6792                 .data           = &ipv6_devconf.accept_ra_mtu,
6793                 .maxlen         = sizeof(int),
6794                 .mode           = 0644,
6795                 .proc_handler   = proc_dointvec,
6796         },
6797         {
6798                 .procname       = "stable_secret",
6799                 .data           = &ipv6_devconf.stable_secret,
6800                 .maxlen         = IPV6_MAX_STRLEN,
6801                 .mode           = 0600,
6802                 .proc_handler   = addrconf_sysctl_stable_secret,
6803         },
6804         {
6805                 .procname       = "use_oif_addrs_only",
6806                 .data           = &ipv6_devconf.use_oif_addrs_only,
6807                 .maxlen         = sizeof(int),
6808                 .mode           = 0644,
6809                 .proc_handler   = proc_dointvec,
6810         },
6811         {
6812                 .procname       = "ignore_routes_with_linkdown",
6813                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6814                 .maxlen         = sizeof(int),
6815                 .mode           = 0644,
6816                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6817         },
6818         {
6819                 .procname       = "drop_unicast_in_l2_multicast",
6820                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6821                 .maxlen         = sizeof(int),
6822                 .mode           = 0644,
6823                 .proc_handler   = proc_dointvec,
6824         },
6825         {
6826                 .procname       = "drop_unsolicited_na",
6827                 .data           = &ipv6_devconf.drop_unsolicited_na,
6828                 .maxlen         = sizeof(int),
6829                 .mode           = 0644,
6830                 .proc_handler   = proc_dointvec,
6831         },
6832         {
6833                 .procname       = "keep_addr_on_down",
6834                 .data           = &ipv6_devconf.keep_addr_on_down,
6835                 .maxlen         = sizeof(int),
6836                 .mode           = 0644,
6837                 .proc_handler   = proc_dointvec,
6838
6839         },
6840         {
6841                 .procname       = "seg6_enabled",
6842                 .data           = &ipv6_devconf.seg6_enabled,
6843                 .maxlen         = sizeof(int),
6844                 .mode           = 0644,
6845                 .proc_handler   = proc_dointvec,
6846         },
6847 #ifdef CONFIG_IPV6_SEG6_HMAC
6848         {
6849                 .procname       = "seg6_require_hmac",
6850                 .data           = &ipv6_devconf.seg6_require_hmac,
6851                 .maxlen         = sizeof(int),
6852                 .mode           = 0644,
6853                 .proc_handler   = proc_dointvec,
6854         },
6855 #endif
6856         {
6857                 .procname       = "enhanced_dad",
6858                 .data           = &ipv6_devconf.enhanced_dad,
6859                 .maxlen         = sizeof(int),
6860                 .mode           = 0644,
6861                 .proc_handler   = proc_dointvec,
6862         },
6863         {
6864                 .procname               = "addr_gen_mode",
6865                 .data                   = &ipv6_devconf.addr_gen_mode,
6866                 .maxlen                 = sizeof(int),
6867                 .mode                   = 0644,
6868                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6869         },
6870         {
6871                 .procname       = "disable_policy",
6872                 .data           = &ipv6_devconf.disable_policy,
6873                 .maxlen         = sizeof(int),
6874                 .mode           = 0644,
6875                 .proc_handler   = addrconf_sysctl_disable_policy,
6876         },
6877         {
6878                 .procname       = "ndisc_tclass",
6879                 .data           = &ipv6_devconf.ndisc_tclass,
6880                 .maxlen         = sizeof(int),
6881                 .mode           = 0644,
6882                 .proc_handler   = proc_dointvec_minmax,
6883                 .extra1         = (void *)SYSCTL_ZERO,
6884                 .extra2         = (void *)&two_five_five,
6885         },
6886         {
6887                 /* sentinel */
6888         }
6889 };
6890
6891 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6892                 struct inet6_dev *idev, struct ipv6_devconf *p)
6893 {
6894         int i, ifindex;
6895         struct ctl_table *table;
6896         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6897
6898         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6899         if (!table)
6900                 goto out;
6901
6902         for (i = 0; table[i].data; i++) {
6903                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6904                 /* If one of these is already set, then it is not safe to
6905                  * overwrite either of them: this makes proc_dointvec_minmax
6906                  * usable.
6907                  */
6908                 if (!table[i].extra1 && !table[i].extra2) {
6909                         table[i].extra1 = idev; /* embedded; no ref */
6910                         table[i].extra2 = net;
6911                 }
6912         }
6913
6914         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6915
6916         p->sysctl_header = register_net_sysctl(net, path, table);
6917         if (!p->sysctl_header)
6918                 goto free;
6919
6920         if (!strcmp(dev_name, "all"))
6921                 ifindex = NETCONFA_IFINDEX_ALL;
6922         else if (!strcmp(dev_name, "default"))
6923                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6924         else
6925                 ifindex = idev->dev->ifindex;
6926         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6927                                      ifindex, p);
6928         return 0;
6929
6930 free:
6931         kfree(table);
6932 out:
6933         return -ENOBUFS;
6934 }
6935
6936 static void __addrconf_sysctl_unregister(struct net *net,
6937                                          struct ipv6_devconf *p, int ifindex)
6938 {
6939         struct ctl_table *table;
6940
6941         if (!p->sysctl_header)
6942                 return;
6943
6944         table = p->sysctl_header->ctl_table_arg;
6945         unregister_net_sysctl_table(p->sysctl_header);
6946         p->sysctl_header = NULL;
6947         kfree(table);
6948
6949         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6950 }
6951
6952 static int addrconf_sysctl_register(struct inet6_dev *idev)
6953 {
6954         int err;
6955
6956         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6957                 return -EINVAL;
6958
6959         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6960                                     &ndisc_ifinfo_sysctl_change);
6961         if (err)
6962                 return err;
6963         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6964                                          idev, &idev->cnf);
6965         if (err)
6966                 neigh_sysctl_unregister(idev->nd_parms);
6967
6968         return err;
6969 }
6970
6971 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6972 {
6973         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6974                                      idev->dev->ifindex);
6975         neigh_sysctl_unregister(idev->nd_parms);
6976 }
6977
6978
6979 #endif
6980
6981 static int __net_init addrconf_init_net(struct net *net)
6982 {
6983         int err = -ENOMEM;
6984         struct ipv6_devconf *all, *dflt;
6985
6986         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6987         if (!all)
6988                 goto err_alloc_all;
6989
6990         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6991         if (!dflt)
6992                 goto err_alloc_dflt;
6993
6994         if (IS_ENABLED(CONFIG_SYSCTL) &&
6995             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6996                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6997                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6998         }
6999
7000         /* these will be inherited by all namespaces */
7001         dflt->autoconf = ipv6_defaults.autoconf;
7002         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7003
7004         dflt->stable_secret.initialized = false;
7005         all->stable_secret.initialized = false;
7006
7007         net->ipv6.devconf_all = all;
7008         net->ipv6.devconf_dflt = dflt;
7009
7010 #ifdef CONFIG_SYSCTL
7011         err = __addrconf_sysctl_register(net, "all", NULL, all);
7012         if (err < 0)
7013                 goto err_reg_all;
7014
7015         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7016         if (err < 0)
7017                 goto err_reg_dflt;
7018 #endif
7019         return 0;
7020
7021 #ifdef CONFIG_SYSCTL
7022 err_reg_dflt:
7023         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7024 err_reg_all:
7025         kfree(dflt);
7026 #endif
7027 err_alloc_dflt:
7028         kfree(all);
7029 err_alloc_all:
7030         return err;
7031 }
7032
7033 static void __net_exit addrconf_exit_net(struct net *net)
7034 {
7035 #ifdef CONFIG_SYSCTL
7036         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7037                                      NETCONFA_IFINDEX_DEFAULT);
7038         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7039                                      NETCONFA_IFINDEX_ALL);
7040 #endif
7041         kfree(net->ipv6.devconf_dflt);
7042         kfree(net->ipv6.devconf_all);
7043 }
7044
7045 static struct pernet_operations addrconf_ops = {
7046         .init = addrconf_init_net,
7047         .exit = addrconf_exit_net,
7048 };
7049
7050 static struct rtnl_af_ops inet6_ops __read_mostly = {
7051         .family           = AF_INET6,
7052         .fill_link_af     = inet6_fill_link_af,
7053         .get_link_af_size = inet6_get_link_af_size,
7054         .validate_link_af = inet6_validate_link_af,
7055         .set_link_af      = inet6_set_link_af,
7056 };
7057
7058 /*
7059  *      Init / cleanup code
7060  */
7061
7062 int __init addrconf_init(void)
7063 {
7064         struct inet6_dev *idev;
7065         int i, err;
7066
7067         err = ipv6_addr_label_init();
7068         if (err < 0) {
7069                 pr_crit("%s: cannot initialize default policy table: %d\n",
7070                         __func__, err);
7071                 goto out;
7072         }
7073
7074         err = register_pernet_subsys(&addrconf_ops);
7075         if (err < 0)
7076                 goto out_addrlabel;
7077
7078         addrconf_wq = create_workqueue("ipv6_addrconf");
7079         if (!addrconf_wq) {
7080                 err = -ENOMEM;
7081                 goto out_nowq;
7082         }
7083
7084         /* The addrconf netdev notifier requires that loopback_dev
7085          * has it's ipv6 private information allocated and setup
7086          * before it can bring up and give link-local addresses
7087          * to other devices which are up.
7088          *
7089          * Unfortunately, loopback_dev is not necessarily the first
7090          * entry in the global dev_base list of net devices.  In fact,
7091          * it is likely to be the very last entry on that list.
7092          * So this causes the notifier registry below to try and
7093          * give link-local addresses to all devices besides loopback_dev
7094          * first, then loopback_dev, which cases all the non-loopback_dev
7095          * devices to fail to get a link-local address.
7096          *
7097          * So, as a temporary fix, allocate the ipv6 structure for
7098          * loopback_dev first by hand.
7099          * Longer term, all of the dependencies ipv6 has upon the loopback
7100          * device and it being up should be removed.
7101          */
7102         rtnl_lock();
7103         idev = ipv6_add_dev(init_net.loopback_dev);
7104         rtnl_unlock();
7105         if (IS_ERR(idev)) {
7106                 err = PTR_ERR(idev);
7107                 goto errlo;
7108         }
7109
7110         ip6_route_init_special_entries();
7111
7112         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7113                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7114
7115         register_netdevice_notifier(&ipv6_dev_notf);
7116
7117         addrconf_verify();
7118
7119         rtnl_af_register(&inet6_ops);
7120
7121         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7122                                    NULL, inet6_dump_ifinfo, 0);
7123         if (err < 0)
7124                 goto errout;
7125
7126         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7127                                    inet6_rtm_newaddr, NULL, 0);
7128         if (err < 0)
7129                 goto errout;
7130         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7131                                    inet6_rtm_deladdr, NULL, 0);
7132         if (err < 0)
7133                 goto errout;
7134         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7135                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7136                                    RTNL_FLAG_DOIT_UNLOCKED);
7137         if (err < 0)
7138                 goto errout;
7139         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7140                                    NULL, inet6_dump_ifmcaddr, 0);
7141         if (err < 0)
7142                 goto errout;
7143         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7144                                    NULL, inet6_dump_ifacaddr, 0);
7145         if (err < 0)
7146                 goto errout;
7147         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7148                                    inet6_netconf_get_devconf,
7149                                    inet6_netconf_dump_devconf,
7150                                    RTNL_FLAG_DOIT_UNLOCKED);
7151         if (err < 0)
7152                 goto errout;
7153         err = ipv6_addr_label_rtnl_register();
7154         if (err < 0)
7155                 goto errout;
7156
7157         return 0;
7158 errout:
7159         rtnl_unregister_all(PF_INET6);
7160         rtnl_af_unregister(&inet6_ops);
7161         unregister_netdevice_notifier(&ipv6_dev_notf);
7162 errlo:
7163         destroy_workqueue(addrconf_wq);
7164 out_nowq:
7165         unregister_pernet_subsys(&addrconf_ops);
7166 out_addrlabel:
7167         ipv6_addr_label_cleanup();
7168 out:
7169         return err;
7170 }
7171
7172 void addrconf_cleanup(void)
7173 {
7174         struct net_device *dev;
7175         int i;
7176
7177         unregister_netdevice_notifier(&ipv6_dev_notf);
7178         unregister_pernet_subsys(&addrconf_ops);
7179         ipv6_addr_label_cleanup();
7180
7181         rtnl_af_unregister(&inet6_ops);
7182
7183         rtnl_lock();
7184
7185         /* clean dev list */
7186         for_each_netdev(&init_net, dev) {
7187                 if (__in6_dev_get(dev) == NULL)
7188                         continue;
7189                 addrconf_ifdown(dev, 1);
7190         }
7191         addrconf_ifdown(init_net.loopback_dev, 2);
7192
7193         /*
7194          *      Check hash table.
7195          */
7196         spin_lock_bh(&addrconf_hash_lock);
7197         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7198                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7199         spin_unlock_bh(&addrconf_hash_lock);
7200         cancel_delayed_work(&addr_chk_work);
7201         rtnl_unlock();
7202
7203         destroy_workqueue(addrconf_wq);
7204 }