net/ipv6/ndisc.c

   1 /*
   2  *      Neighbour Discovery for IPv6
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Mike Shaver             <shaver@ingenia.com>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Alexey I. Froloff               :       RFC6106 (DNSSL) support
  19  *      Pierre Ynard                    :       export userland ND options
  20  *                                              through netlink (RDNSS support)
  21  *      Lars Fenneberg                  :       fixed MTU setting on receipt
  22  *                                              of an RA.
  23  *      Janos Farkas                    :       kmalloc failure checks
  24  *      Alexey Kuznetsov                :       state machine reworked
  25  *                                              and moved to net/core.
  26  *      Pekka Savola                    :       RFC2461 validation
  27  *      YOSHIFUJI Hideaki @USAGI        :       Verify ND options properly
  28  */
  29
  30 #define pr_fmt(fmt) "ICMPv6: " fmt
  31
  32 #include <linux/module.h>
  33 #include <linux/errno.h>
  34 #include <linux/types.h>
  35 #include <linux/socket.h>
  36 #include <linux/sockios.h>
  37 #include <linux/sched.h>
  38 #include <linux/net.h>
  39 #include <linux/in6.h>
  40 #include <linux/route.h>
  41 #include <linux/init.h>
  42 #include <linux/rcupdate.h>
  43 #include <linux/slab.h>
  44 #ifdef CONFIG_SYSCTL
  45 #include <linux/sysctl.h>
  46 #endif
  47
  48 #include <linux/if_addr.h>
  49 #include <linux/if_ether.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/ipv6.h>
  52 #include <linux/icmpv6.h>
  53 #include <linux/jhash.h>
  54
  55 #include <net/sock.h>
  56 #include <net/snmp.h>
  57
  58 #include <net/ipv6.h>
  59 #include <net/protocol.h>
  60 #include <net/ndisc.h>
  61 #include <net/ip6_route.h>
  62 #include <net/addrconf.h>
  63 #include <net/icmp.h>
  64
  65 #include <net/netlink.h>
  66 #include <linux/rtnetlink.h>
  67
  68 #include <net/flow.h>
  69 #include <net/ip6_checksum.h>
  70 #include <net/inet_common.h>
  71 #include <linux/proc_fs.h>
  72
  73 #include <linux/netfilter.h>
  74 #include <linux/netfilter_ipv6.h>
  75
  76 static u32 ndisc_hash(const void *pkey,
  77                       const struct net_device *dev,
  78                       __u32 *hash_rnd);
  79 static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
  80 static bool ndisc_allow_add(const struct net_device *dev,
  81                             struct netlink_ext_ack *extack);
  82 static int ndisc_constructor(struct neighbour *neigh);
  83 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
  84 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
  85 static int pndisc_constructor(struct pneigh_entry *n);
  86 static void pndisc_destructor(struct pneigh_entry *n);
  87 static void pndisc_redo(struct sk_buff *skb);
  88
  89 static const struct neigh_ops ndisc_generic_ops = {
  90         .family =               AF_INET6,
  91         .solicit =              ndisc_solicit,
  92         .error_report =         ndisc_error_report,
  93         .output =               neigh_resolve_output,
  94         .connected_output =     neigh_connected_output,
  95 };
  96
  97 static const struct neigh_ops ndisc_hh_ops = {
  98         .family =               AF_INET6,
  99         .solicit =              ndisc_solicit,
 100         .error_report =         ndisc_error_report,
 101         .output =               neigh_resolve_output,
 102         .connected_output =     neigh_resolve_output,
 103 };
 104
 105
 106 static const struct neigh_ops ndisc_direct_ops = {
 107         .family =               AF_INET6,
 108         .output =               neigh_direct_output,
 109         .connected_output =     neigh_direct_output,
 110 };
 111
 112 struct neigh_table nd_tbl = {
 113         .family =       AF_INET6,
 114         .key_len =      sizeof(struct in6_addr),
 115         .protocol =     cpu_to_be16(ETH_P_IPV6),
 116         .hash =         ndisc_hash,
 117         .key_eq =       ndisc_key_eq,
 118         .constructor =  ndisc_constructor,
 119         .pconstructor = pndisc_constructor,
 120         .pdestructor =  pndisc_destructor,
 121         .proxy_redo =   pndisc_redo,
 122         .allow_add  =   ndisc_allow_add,
 123         .id =           "ndisc_cache",
 124         .parms = {
 125                 .tbl                    = &nd_tbl,
 126                 .reachable_time         = ND_REACHABLE_TIME,
 127                 .data = {
 128                         [NEIGH_VAR_MCAST_PROBES] = 3,
 129                         [NEIGH_VAR_UCAST_PROBES] = 3,
 130                         [NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER,
 131                         [NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME,
 132                         [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
 133                         [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
 134                         [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
 135                         [NEIGH_VAR_PROXY_QLEN] = 64,
 136                         [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
 137                         [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
 138                 },
 139         },
 140         .gc_interval =    30 * HZ,
 141         .gc_thresh1 =    128,
 142         .gc_thresh2 =    512,
 143         .gc_thresh3 =   1024,
 144 };
 145 EXPORT_SYMBOL_GPL(nd_tbl);
 146
 147 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, void *data,
 148                               int data_len, int pad)
 149 {
 150         int space = __ndisc_opt_addr_space(data_len, pad);
 151         u8 *opt = skb_put(skb, space);
 152
 153         opt[0] = type;
 154         opt[1] = space>>3;
 155
 156         memset(opt + 2, 0, pad);
 157         opt   += pad;
 158         space -= pad;
 159
 160         memcpy(opt+2, data, data_len);
 161         data_len += 2;
 162         opt += data_len;
 163         space -= data_len;
 164         if (space > 0)
 165                 memset(opt, 0, space);
 166 }
 167 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
 168
 169 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
 170                                           void *data, u8 icmp6_type)
 171 {
 172         __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
 173                                  ndisc_addr_option_pad(skb->dev->type));
 174         ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
 175 }
 176
 177 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
 178                                                    void *ha,
 179                                                    const u8 *ops_data)
 180 {
 181         ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
 182         ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
 183 }
 184
 185 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
 186                                             struct nd_opt_hdr *end)
 187 {
 188         int type;
 189         if (!cur || !end || cur >= end)
 190                 return NULL;
 191         type = cur->nd_opt_type;
 192         do {
 193                 cur = ((void *)cur) + (cur->nd_opt_len << 3);
 194         } while (cur < end && cur->nd_opt_type != type);
 195         return cur <= end && cur->nd_opt_type == type ? cur : NULL;
 196 }
 197
 198 static inline int ndisc_is_useropt(const struct net_device *dev,
 199                                    struct nd_opt_hdr *opt)
 200 {
 201         return opt->nd_opt_type == ND_OPT_RDNSS ||
 202                 opt->nd_opt_type == ND_OPT_DNSSL ||
 203                 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
 204 }
 205
 206 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
 207                                              struct nd_opt_hdr *cur,
 208                                              struct nd_opt_hdr *end)
 209 {
 210         if (!cur || !end || cur >= end)
 211                 return NULL;
 212         do {
 213                 cur = ((void *)cur) + (cur->nd_opt_len << 3);
 214         } while (cur < end && !ndisc_is_useropt(dev, cur));
 215         return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
 216 }
 217
 218 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
 219                                           u8 *opt, int opt_len,
 220                                           struct ndisc_options *ndopts)
 221 {
 222         struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
 223
 224         if (!nd_opt || opt_len < 0 || !ndopts)
 225                 return NULL;
 226         memset(ndopts, 0, sizeof(*ndopts));
 227         while (opt_len) {
 228                 int l;
 229                 if (opt_len < sizeof(struct nd_opt_hdr))
 230                         return NULL;
 231                 l = nd_opt->nd_opt_len << 3;
 232                 if (opt_len < l || l == 0)
 233                         return NULL;
 234                 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
 235                         goto next_opt;
 236                 switch (nd_opt->nd_opt_type) {
 237                 case ND_OPT_SOURCE_LL_ADDR:
 238                 case ND_OPT_TARGET_LL_ADDR:
 239                 case ND_OPT_MTU:
 240                 case ND_OPT_NONCE:
 241                 case ND_OPT_REDIRECT_HDR:
 242                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
 243                                 ND_PRINTK(2, warn,
 244                                           "%s: duplicated ND6 option found: type=%d\n",
 245                                           __func__, nd_opt->nd_opt_type);
 246                         } else {
 247                                 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
 248                         }
 249                         break;
 250                 case ND_OPT_PREFIX_INFO:
 251                         ndopts->nd_opts_pi_end = nd_opt;
 252                         if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
 253                                 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
 254                         break;
 255 #ifdef CONFIG_IPV6_ROUTE_INFO
 256                 case ND_OPT_ROUTE_INFO:
 257                         ndopts->nd_opts_ri_end = nd_opt;
 258                         if (!ndopts->nd_opts_ri)
 259                                 ndopts->nd_opts_ri = nd_opt;
 260                         break;
 261 #endif
 262                 default:
 263                         if (ndisc_is_useropt(dev, nd_opt)) {
 264                                 ndopts->nd_useropts_end = nd_opt;
 265                                 if (!ndopts->nd_useropts)
 266                                         ndopts->nd_useropts = nd_opt;
 267                         } else {
 268                                 /*
 269                                  * Unknown options must be silently ignored,
 270                                  * to accommodate future extension to the
 271                                  * protocol.
 272                                  */
 273                                 ND_PRINTK(2, notice,
 274                                           "%s: ignored unsupported option; type=%d, len=%d\n",
 275                                           __func__,
 276                                           nd_opt->nd_opt_type,
 277                                           nd_opt->nd_opt_len);
 278                         }
 279                 }
 280 next_opt:
 281                 opt_len -= l;
 282                 nd_opt = ((void *)nd_opt) + l;
 283         }
 284         return ndopts;
 285 }
 286
 287 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
 288 {
 289         switch (dev->type) {
 290         case ARPHRD_ETHER:
 291         case ARPHRD_IEEE802:    /* Not sure. Check it later. --ANK */
 292         case ARPHRD_FDDI:
 293                 ipv6_eth_mc_map(addr, buf);
 294                 return 0;
 295         case ARPHRD_ARCNET:
 296                 ipv6_arcnet_mc_map(addr, buf);
 297                 return 0;
 298         case ARPHRD_INFINIBAND:
 299                 ipv6_ib_mc_map(addr, dev->broadcast, buf);
 300                 return 0;
 301         case ARPHRD_IPGRE:
 302                 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
 303         default:
 304                 if (dir) {
 305                         memcpy(buf, dev->broadcast, dev->addr_len);
 306                         return 0;
 307                 }
 308         }
 309         return -EINVAL;
 310 }
 311 EXPORT_SYMBOL(ndisc_mc_map);
 312
 313 static u32 ndisc_hash(const void *pkey,
 314                       const struct net_device *dev,
 315                       __u32 *hash_rnd)
 316 {
 317         return ndisc_hashfn(pkey, dev, hash_rnd);
 318 }
 319
 320 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
 321 {
 322         return neigh_key_eq128(n, pkey);
 323 }
 324
 325 static int ndisc_constructor(struct neighbour *neigh)
 326 {
 327         struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
 328         struct net_device *dev = neigh->dev;
 329         struct inet6_dev *in6_dev;
 330         struct neigh_parms *parms;
 331         bool is_multicast = ipv6_addr_is_multicast(addr);
 332
 333         in6_dev = in6_dev_get(dev);
 334         if (!in6_dev) {
 335                 return -EINVAL;
 336         }
 337
 338         parms = in6_dev->nd_parms;
 339         __neigh_parms_put(neigh->parms);
 340         neigh->parms = neigh_parms_clone(parms);
 341
 342         neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
 343         if (!dev->header_ops) {
 344                 neigh->nud_state = NUD_NOARP;
 345                 neigh->ops = &ndisc_direct_ops;
 346                 neigh->output = neigh_direct_output;
 347         } else {
 348                 if (is_multicast) {
 349                         neigh->nud_state = NUD_NOARP;
 350                         ndisc_mc_map(addr, neigh->ha, dev, 1);
 351                 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
 352                         neigh->nud_state = NUD_NOARP;
 353                         memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
 354                         if (dev->flags&IFF_LOOPBACK)
 355                                 neigh->type = RTN_LOCAL;
 356                 } else if (dev->flags&IFF_POINTOPOINT) {
 357                         neigh->nud_state = NUD_NOARP;
 358                         memcpy(neigh->ha, dev->broadcast, dev->addr_len);
 359                 }
 360                 if (dev->header_ops->cache)
 361                         neigh->ops = &ndisc_hh_ops;
 362                 else
 363                         neigh->ops = &ndisc_generic_ops;
 364                 if (neigh->nud_state&NUD_VALID)
 365                         neigh->output = neigh->ops->connected_output;
 366                 else
 367                         neigh->output = neigh->ops->output;
 368         }
 369         in6_dev_put(in6_dev);
 370         return 0;
 371 }
 372
 373 static int pndisc_constructor(struct pneigh_entry *n)
 374 {
 375         struct in6_addr *addr = (struct in6_addr *)&n->key;
 376         struct in6_addr maddr;
 377         struct net_device *dev = n->dev;
 378
 379         if (!dev || !__in6_dev_get(dev))
 380                 return -EINVAL;
 381         addrconf_addr_solict_mult(addr, &maddr);
 382         ipv6_dev_mc_inc(dev, &maddr);
 383         return 0;
 384 }
 385
 386 static void pndisc_destructor(struct pneigh_entry *n)
 387 {
 388         struct in6_addr *addr = (struct in6_addr *)&n->key;
 389         struct in6_addr maddr;
 390         struct net_device *dev = n->dev;
 391
 392         if (!dev || !__in6_dev_get(dev))
 393                 return;
 394         addrconf_addr_solict_mult(addr, &maddr);
 395         ipv6_dev_mc_dec(dev, &maddr);
 396 }
 397
 398 /* called with rtnl held */
 399 static bool ndisc_allow_add(const struct net_device *dev,
 400                             struct netlink_ext_ack *extack)
 401 {
 402         struct inet6_dev *idev = __in6_dev_get(dev);
 403
 404         if (!idev || idev->cnf.disable_ipv6) {
 405                 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
 406                 return false;
 407         }
 408
 409         return true;
 410 }
 411
 412 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
 413                                        int len)
 414 {
 415         int hlen = LL_RESERVED_SPACE(dev);
 416         int tlen = dev->needed_tailroom;
 417         struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
 418         struct sk_buff *skb;
 419
 420         skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
 421         if (!skb) {
 422                 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
 423                           __func__);
 424                 return NULL;
 425         }
 426
 427         skb->protocol = htons(ETH_P_IPV6);
 428         skb->dev = dev;
 429
 430         skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
 431         skb_reset_transport_header(skb);
 432
 433         /* Manually assign socket ownership as we avoid calling
 434          * sock_alloc_send_pskb() to bypass wmem buffer limits
 435          */
 436         skb_set_owner_w(skb, sk);
 437
 438         return skb;
 439 }
 440
 441 static void ip6_nd_hdr(struct sk_buff *skb,
 442                        const struct in6_addr *saddr,
 443                        const struct in6_addr *daddr,
 444                        int hop_limit, int len)
 445 {
 446         struct ipv6hdr *hdr;
 447         struct inet6_dev *idev;
 448         unsigned tclass;
 449
 450         rcu_read_lock();
 451         idev = __in6_dev_get(skb->dev);
 452         tclass = idev ? idev->cnf.ndisc_tclass : 0;
 453         rcu_read_unlock();
 454
 455         skb_push(skb, sizeof(*hdr));
 456         skb_reset_network_header(skb);
 457         hdr = ipv6_hdr(skb);
 458
 459         ip6_flow_hdr(hdr, tclass, 0);
 460
 461         hdr->payload_len = htons(len);
 462         hdr->nexthdr = IPPROTO_ICMPV6;
 463         hdr->hop_limit = hop_limit;
 464
 465         hdr->saddr = *saddr;
 466         hdr->daddr = *daddr;
 467 }
 468
 469 static void ndisc_send_skb(struct sk_buff *skb,
 470                            const struct in6_addr *daddr,
 471                            const struct in6_addr *saddr)
 472 {
 473         struct dst_entry *dst = skb_dst(skb);
 474         struct net *net = dev_net(skb->dev);
 475         struct sock *sk = net->ipv6.ndisc_sk;
 476         struct inet6_dev *idev;
 477         int err;
 478         struct icmp6hdr *icmp6h = icmp6_hdr(skb);
 479         u8 type;
 480
 481         type = icmp6h->icmp6_type;
 482
 483         if (!dst) {
 484                 struct flowi6 fl6;
 485                 int oif = skb->dev->ifindex;
 486
 487                 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
 488                 dst = icmp6_dst_alloc(skb->dev, &fl6);
 489                 if (IS_ERR(dst)) {
 490                         kfree_skb(skb);
 491                         return;
 492                 }
 493
 494                 skb_dst_set(skb, dst);
 495         }
 496
 497         icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
 498                                               IPPROTO_ICMPV6,
 499                                               csum_partial(icmp6h,
 500                                                            skb->len, 0));
 501
 502         ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
 503
 504         rcu_read_lock();
 505         idev = __in6_dev_get(dst->dev);
 506         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
 507
 508         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
 509                       net, sk, skb, NULL, dst->dev,
 510                       dst_output);
 511         if (!err) {
 512                 ICMP6MSGOUT_INC_STATS(net, idev, type);
 513                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
 514         }
 515
 516         rcu_read_unlock();
 517 }
 518
 519 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
 520                    const struct in6_addr *solicited_addr,
 521                    bool router, bool solicited, bool override, bool inc_opt)
 522 {
 523         struct sk_buff *skb;
 524         struct in6_addr tmpaddr;
 525         struct inet6_ifaddr *ifp;
 526         const struct in6_addr *src_addr;
 527         struct nd_msg *msg;
 528         int optlen = 0;
 529
 530         /* for anycast or proxy, solicited_addr != src_addr */
 531         ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
 532         if (ifp) {
 533                 src_addr = solicited_addr;
 534                 if (ifp->flags & IFA_F_OPTIMISTIC)
 535                         override = false;
 536                 inc_opt |= ifp->idev->cnf.force_tllao;
 537                 in6_ifa_put(ifp);
 538         } else {
 539                 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
 540                                        inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
 541                                        &tmpaddr))
 542                         return;
 543                 src_addr = &tmpaddr;
 544         }
 545
 546         if (!dev->addr_len)
 547                 inc_opt = false;
 548         if (inc_opt)
 549                 optlen += ndisc_opt_addr_space(dev,
 550                                                NDISC_NEIGHBOUR_ADVERTISEMENT);
 551
 552         skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
 553         if (!skb)
 554                 return;
 555
 556         msg = skb_put(skb, sizeof(*msg));
 557         *msg = (struct nd_msg) {
 558                 .icmph = {
 559                         .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
 560                         .icmp6_router = router,
 561                         .icmp6_solicited = solicited,
 562                         .icmp6_override = override,
 563                 },
 564                 .target = *solicited_addr,
 565         };
 566
 567         if (inc_opt)
 568                 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
 569                                        dev->dev_addr,
 570                                        NDISC_NEIGHBOUR_ADVERTISEMENT);
 571
 572         ndisc_send_skb(skb, daddr, src_addr);
 573 }
 574
 575 static void ndisc_send_unsol_na(struct net_device *dev)
 576 {
 577         struct inet6_dev *idev;
 578         struct inet6_ifaddr *ifa;
 579
 580         idev = in6_dev_get(dev);
 581         if (!idev)
 582                 return;
 583
 584         read_lock_bh(&idev->lock);
 585         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 586                 /* skip tentative addresses until dad completes */
 587                 if (ifa->flags & IFA_F_TENTATIVE &&
 588                     !(ifa->flags & IFA_F_OPTIMISTIC))
 589                         continue;
 590
 591                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
 592                               /*router=*/ !!idev->cnf.forwarding,
 593                               /*solicited=*/ false, /*override=*/ true,
 594                               /*inc_opt=*/ true);
 595         }
 596         read_unlock_bh(&idev->lock);
 597
 598         in6_dev_put(idev);
 599 }
 600
 601 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
 602                    const struct in6_addr *daddr, const struct in6_addr *saddr,
 603                    u64 nonce)
 604 {
 605         struct sk_buff *skb;
 606         struct in6_addr addr_buf;
 607         int inc_opt = dev->addr_len;
 608         int optlen = 0;
 609         struct nd_msg *msg;
 610
 611         if (!saddr) {
 612                 if (ipv6_get_lladdr(dev, &addr_buf,
 613                                    (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
 614                         return;
 615                 saddr = &addr_buf;
 616         }
 617
 618         if (ipv6_addr_any(saddr))
 619                 inc_opt = false;
 620         if (inc_opt)
 621                 optlen += ndisc_opt_addr_space(dev,
 622                                                NDISC_NEIGHBOUR_SOLICITATION);
 623         if (nonce != 0)
 624                 optlen += 8;
 625
 626         skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
 627         if (!skb)
 628                 return;
 629
 630         msg = skb_put(skb, sizeof(*msg));
 631         *msg = (struct nd_msg) {
 632                 .icmph = {
 633                         .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
 634                 },
 635                 .target = *solicit,
 636         };
 637
 638         if (inc_opt)
 639                 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
 640                                        dev->dev_addr,
 641                                        NDISC_NEIGHBOUR_SOLICITATION);
 642         if (nonce != 0) {
 643                 u8 *opt = skb_put(skb, 8);
 644
 645                 opt[0] = ND_OPT_NONCE;
 646                 opt[1] = 8 >> 3;
 647                 memcpy(opt + 2, &nonce, 6);
 648         }
 649
 650         ndisc_send_skb(skb, daddr, saddr);
 651 }
 652
 653 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
 654                    const struct in6_addr *daddr)
 655 {
 656         struct sk_buff *skb;
 657         struct rs_msg *msg;
 658         int send_sllao = dev->addr_len;
 659         int optlen = 0;
 660
 661 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
 662         /*
 663          * According to section 2.2 of RFC 4429, we must not
 664          * send router solicitations with a sllao from
 665          * optimistic addresses, but we may send the solicitation
 666          * if we don't include the sllao.  So here we check
 667          * if our address is optimistic, and if so, we
 668          * suppress the inclusion of the sllao.
 669          */
 670         if (send_sllao) {
 671                 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
 672                                                            dev, 1);
 673                 if (ifp) {
 674                         if (ifp->flags & IFA_F_OPTIMISTIC)  {
 675                                 send_sllao = 0;
 676                         }
 677                         in6_ifa_put(ifp);
 678                 } else {
 679                         send_sllao = 0;
 680                 }
 681         }
 682 #endif
 683         if (send_sllao)
 684                 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
 685
 686         skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
 687         if (!skb)
 688                 return;
 689
 690         msg = skb_put(skb, sizeof(*msg));
 691         *msg = (struct rs_msg) {
 692                 .icmph = {
 693                         .icmp6_type = NDISC_ROUTER_SOLICITATION,
 694                 },
 695         };
 696
 697         if (send_sllao)
 698                 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
 699                                        dev->dev_addr,
 700                                        NDISC_ROUTER_SOLICITATION);
 701
 702         ndisc_send_skb(skb, daddr, saddr);
 703 }
 704
 705
 706 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
 707 {
 708         /*
 709          *      "The sender MUST return an ICMP
 710          *       destination unreachable"
 711          */
 712         dst_link_failure(skb);
 713         kfree_skb(skb);
 714 }
 715
 716 /* Called with locked neigh: either read or both */
 717
 718 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
 719 {
 720         struct in6_addr *saddr = NULL;
 721         struct in6_addr mcaddr;
 722         struct net_device *dev = neigh->dev;
 723         struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
 724         int probes = atomic_read(&neigh->probes);
 725
 726         if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
 727                                            dev, false, 1,
 728                                            IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
 729                 saddr = &ipv6_hdr(skb)->saddr;
 730         probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
 731         if (probes < 0) {
 732                 if (!(neigh->nud_state & NUD_VALID)) {
 733                         ND_PRINTK(1, dbg,
 734                                   "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
 735                                   __func__, target);
 736                 }
 737                 ndisc_send_ns(dev, target, target, saddr, 0);
 738         } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
 739                 neigh_app_ns(neigh);
 740         } else {
 741                 addrconf_addr_solict_mult(target, &mcaddr);
 742                 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
 743         }
 744 }
 745
 746 static int pndisc_is_router(const void *pkey,
 747                             struct net_device *dev)
 748 {
 749         struct pneigh_entry *n;
 750         int ret = -1;
 751
 752         read_lock_bh(&nd_tbl.lock);
 753         n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
 754         if (n)
 755                 ret = !!(n->flags & NTF_ROUTER);
 756         read_unlock_bh(&nd_tbl.lock);
 757
 758         return ret;
 759 }
 760
 761 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
 762                   const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
 763                   struct ndisc_options *ndopts)
 764 {
 765         neigh_update(neigh, lladdr, new, flags, 0);
 766         /* report ndisc ops about neighbour update */
 767         ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
 768 }
 769
 770 static void ndisc_recv_ns(struct sk_buff *skb)
 771 {
 772         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
 773         const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
 774         const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
 775         u8 *lladdr = NULL;
 776         u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
 777                                     offsetof(struct nd_msg, opt));
 778         struct ndisc_options ndopts;
 779         struct net_device *dev = skb->dev;
 780         struct inet6_ifaddr *ifp;
 781         struct inet6_dev *idev = NULL;
 782         struct neighbour *neigh;
 783         int dad = ipv6_addr_any(saddr);
 784         bool inc;
 785         int is_router = -1;
 786         u64 nonce = 0;
 787
 788         if (skb->len < sizeof(struct nd_msg)) {
 789                 ND_PRINTK(2, warn, "NS: packet too short\n");
 790                 return;
 791         }
 792
 793         if (ipv6_addr_is_multicast(&msg->target)) {
 794                 ND_PRINTK(2, warn, "NS: multicast target address\n");
 795                 return;
 796         }
 797
 798         /*
 799          * RFC2461 7.1.1:
 800          * DAD has to be destined for solicited node multicast address.
 801          */
 802         if (dad && !ipv6_addr_is_solict_mult(daddr)) {
 803                 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
 804                 return;
 805         }
 806
 807         if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
 808                 ND_PRINTK(2, warn, "NS: invalid ND options\n");
 809                 return;
 810         }
 811
 812         if (ndopts.nd_opts_src_lladdr) {
 813                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
 814                 if (!lladdr) {
 815                         ND_PRINTK(2, warn,
 816                                   "NS: invalid link-layer address length\n");
 817                         return;
 818                 }
 819
 820                 /* RFC2461 7.1.1:
 821                  *      If the IP source address is the unspecified address,
 822                  *      there MUST NOT be source link-layer address option
 823                  *      in the message.
 824                  */
 825                 if (dad) {
 826                         ND_PRINTK(2, warn,
 827                                   "NS: bad DAD packet (link-layer address option)\n");
 828                         return;
 829                 }
 830         }
 831         if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
 832                 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
 833
 834         inc = ipv6_addr_is_multicast(daddr);
 835
 836         ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
 837         if (ifp) {
 838 have_ifp:
 839                 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
 840                         if (dad) {
 841                                 if (nonce != 0 && ifp->dad_nonce == nonce) {
 842                                         u8 *np = (u8 *)&nonce;
 843                                         /* Matching nonce if looped back */
 844                                         ND_PRINTK(2, notice,
 845                                                   "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
 846                                                   ifp->idev->dev->name,
 847                                                   &ifp->addr, np);
 848                                         goto out;
 849                                 }
 850                                 /*
 851                                  * We are colliding with another node
 852                                  * who is doing DAD
 853                                  * so fail our DAD process
 854                                  */
 855                                 addrconf_dad_failure(skb, ifp);
 856                                 return;
 857                         } else {
 858                                 /*
 859                                  * This is not a dad solicitation.
 860                                  * If we are an optimistic node,
 861                                  * we should respond.
 862                                  * Otherwise, we should ignore it.
 863                                  */
 864                                 if (!(ifp->flags & IFA_F_OPTIMISTIC))
 865                                         goto out;
 866                         }
 867                 }
 868
 869                 idev = ifp->idev;
 870         } else {
 871                 struct net *net = dev_net(dev);
 872
 873                 /* perhaps an address on the master device */
 874                 if (netif_is_l3_slave(dev)) {
 875                         struct net_device *mdev;
 876
 877                         mdev = netdev_master_upper_dev_get_rcu(dev);
 878                         if (mdev) {
 879                                 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
 880                                 if (ifp)
 881                                         goto have_ifp;
 882                         }
 883                 }
 884
 885                 idev = in6_dev_get(dev);
 886                 if (!idev) {
 887                         /* XXX: count this drop? */
 888                         return;
 889                 }
 890
 891                 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
 892                     (idev->cnf.forwarding &&
 893                      (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
 894                      (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
 895                         if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
 896                             skb->pkt_type != PACKET_HOST &&
 897                             inc &&
 898                             NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
 899                                 /*
 900                                  * for anycast or proxy,
 901                                  * sender should delay its response
 902                                  * by a random time between 0 and
 903                                  * MAX_ANYCAST_DELAY_TIME seconds.
 904                                  * (RFC2461) -- yoshfuji
 905                                  */
 906                                 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
 907                                 if (n)
 908                                         pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
 909                                 goto out;
 910                         }
 911                 } else
 912                         goto out;
 913         }
 914
 915         if (is_router < 0)
 916                 is_router = idev->cnf.forwarding;
 917
 918         if (dad) {
 919                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
 920                               !!is_router, false, (ifp != NULL), true);
 921                 goto out;
 922         }
 923
 924         if (inc)
 925                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
 926         else
 927                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
 928
 929         /*
 930          *      update / create cache entry
 931          *      for the source address
 932          */
 933         neigh = __neigh_lookup(&nd_tbl, saddr, dev,
 934                                !inc || lladdr || !dev->addr_len);
 935         if (neigh)
 936                 ndisc_update(dev, neigh, lladdr, NUD_STALE,
 937                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
 938                              NEIGH_UPDATE_F_OVERRIDE,
 939                              NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
 940         if (neigh || !dev->header_ops) {
 941                 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
 942                               true, (ifp != NULL && inc), inc);
 943                 if (neigh)
 944                         neigh_release(neigh);
 945         }
 946
 947 out:
 948         if (ifp)
 949                 in6_ifa_put(ifp);
 950         else
 951                 in6_dev_put(idev);
 952 }
 953
 954 static void ndisc_recv_na(struct sk_buff *skb)
 955 {
 956         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
 957         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
 958         const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
 959         u8 *lladdr = NULL;
 960         u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
 961                                     offsetof(struct nd_msg, opt));
 962         struct ndisc_options ndopts;
 963         struct net_device *dev = skb->dev;
 964         struct inet6_dev *idev = __in6_dev_get(dev);
 965         struct inet6_ifaddr *ifp;
 966         struct neighbour *neigh;
 967
 968         if (skb->len < sizeof(struct nd_msg)) {
 969                 ND_PRINTK(2, warn, "NA: packet too short\n");
 970                 return;
 971         }
 972
 973         if (ipv6_addr_is_multicast(&msg->target)) {
 974                 ND_PRINTK(2, warn, "NA: target address is multicast\n");
 975                 return;
 976         }
 977
 978         if (ipv6_addr_is_multicast(daddr) &&
 979             msg->icmph.icmp6_solicited) {
 980                 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
 981                 return;
 982         }
 983
 984         /* For some 802.11 wireless deployments (and possibly other networks),
 985          * there will be a NA proxy and unsolicitd packets are attacks
 986          * and thus should not be accepted.
 987          */
 988         if (!msg->icmph.icmp6_solicited && idev &&
 989             idev->cnf.drop_unsolicited_na)
 990                 return;
 991
 992         if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
 993                 ND_PRINTK(2, warn, "NS: invalid ND option\n");
 994                 return;
 995         }
 996         if (ndopts.nd_opts_tgt_lladdr) {
 997                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
 998                 if (!lladdr) {
 999                         ND_PRINTK(2, warn,
1000                                   "NA: invalid link-layer address length\n");
1001                         return;
1002                 }
1003         }
1004         ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1005         if (ifp) {
1006                 if (skb->pkt_type != PACKET_LOOPBACK
1007                     && (ifp->flags & IFA_F_TENTATIVE)) {
1008                                 addrconf_dad_failure(skb, ifp);
1009                                 return;
1010                 }
1011                 /* What should we make now? The advertisement
1012                    is invalid, but ndisc specs say nothing
1013                    about it. It could be misconfiguration, or
1014                    an smart proxy agent tries to help us :-)
1015
1016                    We should not print the error if NA has been
1017                    received from loopback - it is just our own
1018                    unsolicited advertisement.
1019                  */
1020                 if (skb->pkt_type != PACKET_LOOPBACK)
1021                         ND_PRINTK(1, warn,
1022                                   "NA: %pM advertised our address %pI6c on %s!\n",
1023                                   eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1024                 in6_ifa_put(ifp);
1025                 return;
1026         }
1027         neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1028
1029         if (neigh) {
1030                 u8 old_flags = neigh->flags;
1031                 struct net *net = dev_net(dev);
1032
1033                 if (neigh->nud_state & NUD_FAILED)
1034                         goto out;
1035
1036                 /*
1037                  * Don't update the neighbor cache entry on a proxy NA from
1038                  * ourselves because either the proxied node is off link or it
1039                  * has already sent a NA to us.
1040                  */
1041                 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1042                     net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1043                     pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1044                         /* XXX: idev->cnf.proxy_ndp */
1045                         goto out;
1046                 }
1047
1048                 ndisc_update(dev, neigh, lladdr,
1049                              msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1050                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1051                              (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1052                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1053                              (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1054                              NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1055
1056                 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1057                         /*
1058                          * Change: router to host
1059                          */
1060                         rt6_clean_tohost(dev_net(dev),  saddr);
1061                 }
1062
1063 out:
1064                 neigh_release(neigh);
1065         }
1066 }
1067
1068 static void ndisc_recv_rs(struct sk_buff *skb)
1069 {
1070         struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1071         unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1072         struct neighbour *neigh;
1073         struct inet6_dev *idev;
1074         const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1075         struct ndisc_options ndopts;
1076         u8 *lladdr = NULL;
1077
1078         if (skb->len < sizeof(*rs_msg))
1079                 return;
1080
1081         idev = __in6_dev_get(skb->dev);
1082         if (!idev) {
1083                 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1084                 return;
1085         }
1086
1087         /* Don't accept RS if we're not in router mode */
1088         if (!idev->cnf.forwarding)
1089                 goto out;
1090
1091         /*
1092          * Don't update NCE if src = ::;
1093          * this implies that the source node has no ip address assigned yet.
1094          */
1095         if (ipv6_addr_any(saddr))
1096                 goto out;
1097
1098         /* Parse ND options */
1099         if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1100                 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1101                 goto out;
1102         }
1103
1104         if (ndopts.nd_opts_src_lladdr) {
1105                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1106                                              skb->dev);
1107                 if (!lladdr)
1108                         goto out;
1109         }
1110
1111         neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1112         if (neigh) {
1113                 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1114                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1115                              NEIGH_UPDATE_F_OVERRIDE|
1116                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1117                              NDISC_ROUTER_SOLICITATION, &ndopts);
1118                 neigh_release(neigh);
1119         }
1120 out:
1121         return;
1122 }
1123
1124 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1125 {
1126         struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1127         struct sk_buff *skb;
1128         struct nlmsghdr *nlh;
1129         struct nduseroptmsg *ndmsg;
1130         struct net *net = dev_net(ra->dev);
1131         int err;
1132         int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1133                                     + (opt->nd_opt_len << 3));
1134         size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1135
1136         skb = nlmsg_new(msg_size, GFP_ATOMIC);
1137         if (!skb) {
1138                 err = -ENOBUFS;
1139                 goto errout;
1140         }
1141
1142         nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1143         if (!nlh) {
1144                 goto nla_put_failure;
1145         }
1146
1147         ndmsg = nlmsg_data(nlh);
1148         ndmsg->nduseropt_family = AF_INET6;
1149         ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1150         ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1151         ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1152         ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1153
1154         memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1155
1156         if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1157                 goto nla_put_failure;
1158         nlmsg_end(skb, nlh);
1159
1160         rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1161         return;
1162
1163 nla_put_failure:
1164         nlmsg_free(skb);
1165         err = -EMSGSIZE;
1166 errout:
1167         rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1168 }
1169
1170 static void ndisc_router_discovery(struct sk_buff *skb)
1171 {
1172         struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1173         struct neighbour *neigh = NULL;
1174         struct inet6_dev *in6_dev;
1175         struct fib6_info *rt = NULL;
1176         struct net *net;
1177         int lifetime;
1178         struct ndisc_options ndopts;
1179         int optlen;
1180         unsigned int pref = 0;
1181         __u32 old_if_flags;
1182         bool send_ifinfo_notify = false;
1183
1184         __u8 *opt = (__u8 *)(ra_msg + 1);
1185
1186         optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1187                 sizeof(struct ra_msg);
1188
1189         ND_PRINTK(2, info,
1190                   "RA: %s, dev: %s\n",
1191                   __func__, skb->dev->name);
1192         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1193                 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1194                 return;
1195         }
1196         if (optlen < 0) {
1197                 ND_PRINTK(2, warn, "RA: packet too short\n");
1198                 return;
1199         }
1200
1201 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1202         if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1203                 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1204                 return;
1205         }
1206 #endif
1207
1208         /*
1209          *      set the RA_RECV flag in the interface
1210          */
1211
1212         in6_dev = __in6_dev_get(skb->dev);
1213         if (!in6_dev) {
1214                 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1215                           skb->dev->name);
1216                 return;
1217         }
1218
1219         if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1220                 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1221                 return;
1222         }
1223
1224         if (!ipv6_accept_ra(in6_dev)) {
1225                 ND_PRINTK(2, info,
1226                           "RA: %s, did not accept ra for dev: %s\n",
1227                           __func__, skb->dev->name);
1228                 goto skip_linkparms;
1229         }
1230
1231 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1232         /* skip link-specific parameters from interior routers */
1233         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1234                 ND_PRINTK(2, info,
1235                           "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1236                           __func__, skb->dev->name);
1237                 goto skip_linkparms;
1238         }
1239 #endif
1240
1241         if (in6_dev->if_flags & IF_RS_SENT) {
1242                 /*
1243                  *      flag that an RA was received after an RS was sent
1244                  *      out on this interface.
1245                  */
1246                 in6_dev->if_flags |= IF_RA_RCVD;
1247         }
1248
1249         /*
1250          * Remember the managed/otherconf flags from most recently
1251          * received RA message (RFC 2462) -- yoshfuji
1252          */
1253         old_if_flags = in6_dev->if_flags;
1254         in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1255                                 IF_RA_OTHERCONF)) |
1256                                 (ra_msg->icmph.icmp6_addrconf_managed ?
1257                                         IF_RA_MANAGED : 0) |
1258                                 (ra_msg->icmph.icmp6_addrconf_other ?
1259                                         IF_RA_OTHERCONF : 0);
1260
1261         if (old_if_flags != in6_dev->if_flags)
1262                 send_ifinfo_notify = true;
1263
1264         if (!in6_dev->cnf.accept_ra_defrtr) {
1265                 ND_PRINTK(2, info,
1266                           "RA: %s, defrtr is false for dev: %s\n",
1267                           __func__, skb->dev->name);
1268                 goto skip_defrtr;
1269         }
1270
1271         /* Do not accept RA with source-addr found on local machine unless
1272          * accept_ra_from_local is set to true.
1273          */
1274         net = dev_net(in6_dev->dev);
1275         if (!in6_dev->cnf.accept_ra_from_local &&
1276             ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1277                 ND_PRINTK(2, info,
1278                           "RA from local address detected on dev: %s: default router ignored\n",
1279                           skb->dev->name);
1280                 goto skip_defrtr;
1281         }
1282
1283         lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1284
1285 #ifdef CONFIG_IPV6_ROUTER_PREF
1286         pref = ra_msg->icmph.icmp6_router_pref;
1287         /* 10b is handled as if it were 00b (medium) */
1288         if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1289             !in6_dev->cnf.accept_ra_rtr_pref)
1290                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1291 #endif
1292
1293         rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1294
1295         if (rt) {
1296                 neigh = ip6_neigh_lookup(&rt->fib6_nh.fib_nh_gw6,
1297                                          rt->fib6_nh.fib_nh_dev, NULL,
1298                                           &ipv6_hdr(skb)->saddr);
1299                 if (!neigh) {
1300                         ND_PRINTK(0, err,
1301                                   "RA: %s got default router without neighbour\n",
1302                                   __func__);
1303                         fib6_info_release(rt);
1304                         return;
1305                 }
1306         }
1307         if (rt && lifetime == 0) {
1308                 ip6_del_rt(net, rt);
1309                 rt = NULL;
1310         }
1311
1312         ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, for dev: %s\n",
1313                   rt, lifetime, skb->dev->name);
1314         if (!rt && lifetime) {
1315                 ND_PRINTK(3, info, "RA: adding default router\n");
1316
1317                 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1318                                          skb->dev, pref);
1319                 if (!rt) {
1320                         ND_PRINTK(0, err,
1321                                   "RA: %s failed to add default route\n",
1322                                   __func__);
1323                         return;
1324                 }
1325
1326                 neigh = ip6_neigh_lookup(&rt->fib6_nh.fib_nh_gw6,
1327                                          rt->fib6_nh.fib_nh_dev, NULL,
1328                                           &ipv6_hdr(skb)->saddr);
1329                 if (!neigh) {
1330                         ND_PRINTK(0, err,
1331                                   "RA: %s got default router without neighbour\n",
1332                                   __func__);
1333                         fib6_info_release(rt);
1334                         return;
1335                 }
1336                 neigh->flags |= NTF_ROUTER;
1337         } else if (rt) {
1338                 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1339         }
1340
1341         if (rt)
1342                 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1343         if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1344             ra_msg->icmph.icmp6_hop_limit) {
1345                 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1346                         in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1347                         fib6_metric_set(rt, RTAX_HOPLIMIT,
1348                                         ra_msg->icmph.icmp6_hop_limit);
1349                 } else {
1350                         ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1351                 }
1352         }
1353
1354 skip_defrtr:
1355
1356         /*
1357          *      Update Reachable Time and Retrans Timer
1358          */
1359
1360         if (in6_dev->nd_parms) {
1361                 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1362
1363                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1364                         rtime = (rtime*HZ)/1000;
1365                         if (rtime < HZ/10)
1366                                 rtime = HZ/10;
1367                         NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1368                         in6_dev->tstamp = jiffies;
1369                         send_ifinfo_notify = true;
1370                 }
1371
1372                 rtime = ntohl(ra_msg->reachable_time);
1373                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1374                         rtime = (rtime*HZ)/1000;
1375
1376                         if (rtime < HZ/10)
1377                                 rtime = HZ/10;
1378
1379                         if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1380                                 NEIGH_VAR_SET(in6_dev->nd_parms,
1381                                               BASE_REACHABLE_TIME, rtime);
1382                                 NEIGH_VAR_SET(in6_dev->nd_parms,
1383                                               GC_STALETIME, 3 * rtime);
1384                                 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1385                                 in6_dev->tstamp = jiffies;
1386                                 send_ifinfo_notify = true;
1387                         }
1388                 }
1389         }
1390
1391         /*
1392          *      Send a notify if RA changed managed/otherconf flags or timer settings
1393          */
1394         if (send_ifinfo_notify)
1395                 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1396
1397 skip_linkparms:
1398
1399         /*
1400          *      Process options.
1401          */
1402
1403         if (!neigh)
1404                 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1405                                        skb->dev, 1);
1406         if (neigh) {
1407                 u8 *lladdr = NULL;
1408                 if (ndopts.nd_opts_src_lladdr) {
1409                         lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1410                                                      skb->dev);
1411                         if (!lladdr) {
1412                                 ND_PRINTK(2, warn,
1413                                           "RA: invalid link-layer address length\n");
1414                                 goto out;
1415                         }
1416                 }
1417                 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1418                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1419                              NEIGH_UPDATE_F_OVERRIDE|
1420                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1421                              NEIGH_UPDATE_F_ISROUTER,
1422                              NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1423         }
1424
1425         if (!ipv6_accept_ra(in6_dev)) {
1426                 ND_PRINTK(2, info,
1427                           "RA: %s, accept_ra is false for dev: %s\n",
1428                           __func__, skb->dev->name);
1429                 goto out;
1430         }
1431
1432 #ifdef CONFIG_IPV6_ROUTE_INFO
1433         if (!in6_dev->cnf.accept_ra_from_local &&
1434             ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1435                           in6_dev->dev, 0)) {
1436                 ND_PRINTK(2, info,
1437                           "RA from local address detected on dev: %s: router info ignored.\n",
1438                           skb->dev->name);
1439                 goto skip_routeinfo;
1440         }
1441
1442         if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1443                 struct nd_opt_hdr *p;
1444                 for (p = ndopts.nd_opts_ri;
1445                      p;
1446                      p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1447                         struct route_info *ri = (struct route_info *)p;
1448 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1449                         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1450                             ri->prefix_len == 0)
1451                                 continue;
1452 #endif
1453                         if (ri->prefix_len == 0 &&
1454                             !in6_dev->cnf.accept_ra_defrtr)
1455                                 continue;
1456                         if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1457                                 continue;
1458                         if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1459                                 continue;
1460                         rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1461                                       &ipv6_hdr(skb)->saddr);
1462                 }
1463         }
1464
1465 skip_routeinfo:
1466 #endif
1467
1468 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1469         /* skip link-specific ndopts from interior routers */
1470         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1471                 ND_PRINTK(2, info,
1472                           "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1473                           __func__, skb->dev->name);
1474                 goto out;
1475         }
1476 #endif
1477
1478         if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1479                 struct nd_opt_hdr *p;
1480                 for (p = ndopts.nd_opts_pi;
1481                      p;
1482                      p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1483                         addrconf_prefix_rcv(skb->dev, (u8 *)p,
1484                                             (p->nd_opt_len) << 3,
1485                                             ndopts.nd_opts_src_lladdr != NULL);
1486                 }
1487         }
1488
1489         if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1490                 __be32 n;
1491                 u32 mtu;
1492
1493                 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1494                 mtu = ntohl(n);
1495
1496                 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1497                         ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1498                 } else if (in6_dev->cnf.mtu6 != mtu) {
1499                         in6_dev->cnf.mtu6 = mtu;
1500                         fib6_metric_set(rt, RTAX_MTU, mtu);
1501                         rt6_mtu_change(skb->dev, mtu);
1502                 }
1503         }
1504
1505         if (ndopts.nd_useropts) {
1506                 struct nd_opt_hdr *p;
1507                 for (p = ndopts.nd_useropts;
1508                      p;
1509                      p = ndisc_next_useropt(skb->dev, p,
1510                                             ndopts.nd_useropts_end)) {
1511                         ndisc_ra_useropt(skb, p);
1512                 }
1513         }
1514
1515         if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1516                 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1517         }
1518 out:
1519         fib6_info_release(rt);
1520         if (neigh)
1521                 neigh_release(neigh);
1522 }
1523
1524 static void ndisc_redirect_rcv(struct sk_buff *skb)
1525 {
1526         u8 *hdr;
1527         struct ndisc_options ndopts;
1528         struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1529         u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1530                                     offsetof(struct rd_msg, opt));
1531
1532 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1533         switch (skb->ndisc_nodetype) {
1534         case NDISC_NODETYPE_HOST:
1535         case NDISC_NODETYPE_NODEFAULT:
1536                 ND_PRINTK(2, warn,
1537                           "Redirect: from host or unauthorized router\n");
1538                 return;
1539         }
1540 #endif
1541
1542         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1543                 ND_PRINTK(2, warn,
1544                           "Redirect: source address is not link-local\n");
1545                 return;
1546         }
1547
1548         if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1549                 return;
1550
1551         if (!ndopts.nd_opts_rh) {
1552                 ip6_redirect_no_header(skb, dev_net(skb->dev),
1553                                         skb->dev->ifindex);
1554                 return;
1555         }
1556
1557         hdr = (u8 *)ndopts.nd_opts_rh;
1558         hdr += 8;
1559         if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1560                 return;
1561
1562         icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1563 }
1564
1565 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1566                                            struct sk_buff *orig_skb,
1567                                            int rd_len)
1568 {
1569         u8 *opt = skb_put(skb, rd_len);
1570
1571         memset(opt, 0, 8);
1572         *(opt++) = ND_OPT_REDIRECT_HDR;
1573         *(opt++) = (rd_len >> 3);
1574         opt += 6;
1575
1576         skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1577                       rd_len - 8);
1578 }
1579
1580 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1581 {
1582         struct net_device *dev = skb->dev;
1583         struct net *net = dev_net(dev);
1584         struct sock *sk = net->ipv6.ndisc_sk;
1585         int optlen = 0;
1586         struct inet_peer *peer;
1587         struct sk_buff *buff;
1588         struct rd_msg *msg;
1589         struct in6_addr saddr_buf;
1590         struct rt6_info *rt;
1591         struct dst_entry *dst;
1592         struct flowi6 fl6;
1593         int rd_len;
1594         u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1595            ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1596         bool ret;
1597
1598         if (netif_is_l3_master(skb->dev)) {
1599                 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1600                 if (!dev)
1601                         return;
1602         }
1603
1604         if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1605                 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1606                           dev->name);
1607                 return;
1608         }
1609
1610         if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1611             ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1612                 ND_PRINTK(2, warn,
1613                           "Redirect: target address is not link-local unicast\n");
1614                 return;
1615         }
1616
1617         icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1618                          &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1619
1620         dst = ip6_route_output(net, NULL, &fl6);
1621         if (dst->error) {
1622                 dst_release(dst);
1623                 return;
1624         }
1625         dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1626         if (IS_ERR(dst))
1627                 return;
1628
1629         rt = (struct rt6_info *) dst;
1630
1631         if (rt->rt6i_flags & RTF_GATEWAY) {
1632                 ND_PRINTK(2, warn,
1633                           "Redirect: destination is not a neighbour\n");
1634                 goto release;
1635         }
1636         peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1637         ret = inet_peer_xrlim_allow(peer, 1*HZ);
1638         if (peer)
1639                 inet_putpeer(peer);
1640         if (!ret)
1641                 goto release;
1642
1643         if (dev->addr_len) {
1644                 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1645                 if (!neigh) {
1646                         ND_PRINTK(2, warn,
1647                                   "Redirect: no neigh for target address\n");
1648                         goto release;
1649                 }
1650
1651                 read_lock_bh(&neigh->lock);
1652                 if (neigh->nud_state & NUD_VALID) {
1653                         memcpy(ha_buf, neigh->ha, dev->addr_len);
1654                         read_unlock_bh(&neigh->lock);
1655                         ha = ha_buf;
1656                         optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1657                                                                 ops_data_buf,
1658                                                                 &ops_data);
1659                 } else
1660                         read_unlock_bh(&neigh->lock);
1661
1662                 neigh_release(neigh);
1663         }
1664
1665         rd_len = min_t(unsigned int,
1666                        IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1667                        skb->len + 8);
1668         rd_len &= ~0x7;
1669         optlen += rd_len;
1670
1671         buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1672         if (!buff)
1673                 goto release;
1674
1675         msg = skb_put(buff, sizeof(*msg));
1676         *msg = (struct rd_msg) {
1677                 .icmph = {
1678                         .icmp6_type = NDISC_REDIRECT,
1679                 },
1680                 .target = *target,
1681                 .dest = ipv6_hdr(skb)->daddr,
1682         };
1683
1684         /*
1685          *      include target_address option
1686          */
1687
1688         if (ha)
1689                 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1690
1691         /*
1692          *      build redirect option and copy skb over to the new packet.
1693          */
1694
1695         if (rd_len)
1696                 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1697
1698         skb_dst_set(buff, dst);
1699         ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1700         return;
1701
1702 release:
1703         dst_release(dst);
1704 }
1705
1706 static void pndisc_redo(struct sk_buff *skb)
1707 {
1708         ndisc_recv_ns(skb);
1709         kfree_skb(skb);
1710 }
1711
1712 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1713 {
1714         struct inet6_dev *idev = __in6_dev_get(skb->dev);
1715
1716         if (!idev)
1717                 return true;
1718         if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1719             idev->cnf.suppress_frag_ndisc) {
1720                 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1721                 return true;
1722         }
1723         return false;
1724 }
1725
1726 int ndisc_rcv(struct sk_buff *skb)
1727 {
1728         struct nd_msg *msg;
1729
1730         if (ndisc_suppress_frag_ndisc(skb))
1731                 return 0;
1732
1733         if (skb_linearize(skb))
1734                 return 0;
1735
1736         msg = (struct nd_msg *)skb_transport_header(skb);
1737
1738         __skb_push(skb, skb->data - skb_transport_header(skb));
1739
1740         if (ipv6_hdr(skb)->hop_limit != 255) {
1741                 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1742                           ipv6_hdr(skb)->hop_limit);
1743                 return 0;
1744         }
1745
1746         if (msg->icmph.icmp6_code != 0) {
1747                 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1748                           msg->icmph.icmp6_code);
1749                 return 0;
1750         }
1751
1752         switch (msg->icmph.icmp6_type) {
1753         case NDISC_NEIGHBOUR_SOLICITATION:
1754                 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1755                 ndisc_recv_ns(skb);
1756                 break;
1757
1758         case NDISC_NEIGHBOUR_ADVERTISEMENT:
1759                 ndisc_recv_na(skb);
1760                 break;
1761
1762         case NDISC_ROUTER_SOLICITATION:
1763                 ndisc_recv_rs(skb);
1764                 break;
1765
1766         case NDISC_ROUTER_ADVERTISEMENT:
1767                 ndisc_router_discovery(skb);
1768                 break;
1769
1770         case NDISC_REDIRECT:
1771                 ndisc_redirect_rcv(skb);
1772                 break;
1773         }
1774
1775         return 0;
1776 }
1777
1778 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1779 {
1780         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1781         struct netdev_notifier_change_info *change_info;
1782         struct net *net = dev_net(dev);
1783         struct inet6_dev *idev;
1784
1785         switch (event) {
1786         case NETDEV_CHANGEADDR:
1787                 neigh_changeaddr(&nd_tbl, dev);
1788                 fib6_run_gc(0, net, false);
1789                 /* fallthrough */
1790         case NETDEV_UP:
1791                 idev = in6_dev_get(dev);
1792                 if (!idev)
1793                         break;
1794                 if (idev->cnf.ndisc_notify ||
1795                     net->ipv6.devconf_all->ndisc_notify)
1796                         ndisc_send_unsol_na(dev);
1797                 in6_dev_put(idev);
1798                 break;
1799         case NETDEV_CHANGE:
1800                 change_info = ptr;
1801                 if (change_info->flags_changed & IFF_NOARP)
1802                         neigh_changeaddr(&nd_tbl, dev);
1803                 if (!netif_carrier_ok(dev))
1804                         neigh_carrier_down(&nd_tbl, dev);
1805                 break;
1806         case NETDEV_DOWN:
1807                 neigh_ifdown(&nd_tbl, dev);
1808                 fib6_run_gc(0, net, false);
1809                 break;
1810         case NETDEV_NOTIFY_PEERS:
1811                 ndisc_send_unsol_na(dev);
1812                 break;
1813         default:
1814                 break;
1815         }
1816
1817         return NOTIFY_DONE;
1818 }
1819
1820 static struct notifier_block ndisc_netdev_notifier = {
1821         .notifier_call = ndisc_netdev_event,
1822         .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1823 };
1824
1825 #ifdef CONFIG_SYSCTL
1826 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1827                                          const char *func, const char *dev_name)
1828 {
1829         static char warncomm[TASK_COMM_LEN];
1830         static int warned;
1831         if (strcmp(warncomm, current->comm) && warned < 5) {
1832                 strcpy(warncomm, current->comm);
1833                 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1834                         warncomm, func,
1835                         dev_name, ctl->procname,
1836                         dev_name, ctl->procname);
1837                 warned++;
1838         }
1839 }
1840
1841 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1842 {
1843         struct net_device *dev = ctl->extra1;
1844         struct inet6_dev *idev;
1845         int ret;
1846
1847         if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1848             (strcmp(ctl->procname, "base_reachable_time") == 0))
1849                 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1850
1851         if (strcmp(ctl->procname, "retrans_time") == 0)
1852                 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1853
1854         else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1855                 ret = neigh_proc_dointvec_jiffies(ctl, write,
1856                                                   buffer, lenp, ppos);
1857
1858         else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1859                  (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1860                 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1861                                                      buffer, lenp, ppos);
1862         else
1863                 ret = -1;
1864
1865         if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1866                 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1867                         idev->nd_parms->reachable_time =
1868                                         neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1869                 idev->tstamp = jiffies;
1870                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1871                 in6_dev_put(idev);
1872         }
1873         return ret;
1874 }
1875
1876
1877 #endif
1878
1879 static int __net_init ndisc_net_init(struct net *net)
1880 {
1881         struct ipv6_pinfo *np;
1882         struct sock *sk;
1883         int err;
1884
1885         err = inet_ctl_sock_create(&sk, PF_INET6,
1886                                    SOCK_RAW, IPPROTO_ICMPV6, net);
1887         if (err < 0) {
1888                 ND_PRINTK(0, err,
1889                           "NDISC: Failed to initialize the control socket (err %d)\n",
1890                           err);
1891                 return err;
1892         }
1893
1894         net->ipv6.ndisc_sk = sk;
1895
1896         np = inet6_sk(sk);
1897         np->hop_limit = 255;
1898         /* Do not loopback ndisc messages */
1899         np->mc_loop = 0;
1900
1901         return 0;
1902 }
1903
1904 static void __net_exit ndisc_net_exit(struct net *net)
1905 {
1906         inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1907 }
1908
1909 static struct pernet_operations ndisc_net_ops = {
1910         .init = ndisc_net_init,
1911         .exit = ndisc_net_exit,
1912 };
1913
1914 int __init ndisc_init(void)
1915 {
1916         int err;
1917
1918         err = register_pernet_subsys(&ndisc_net_ops);
1919         if (err)
1920                 return err;
1921         /*
1922          * Initialize the neighbour table
1923          */
1924         neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1925
1926 #ifdef CONFIG_SYSCTL
1927         err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1928                                     ndisc_ifinfo_sysctl_change);
1929         if (err)
1930                 goto out_unregister_pernet;
1931 out:
1932 #endif
1933         return err;
1934
1935 #ifdef CONFIG_SYSCTL
1936 out_unregister_pernet:
1937         unregister_pernet_subsys(&ndisc_net_ops);
1938         goto out;
1939 #endif
1940 }
1941
1942 int __init ndisc_late_init(void)
1943 {
1944         return register_netdevice_notifier(&ndisc_netdev_notifier);
1945 }
1946
1947 void ndisc_late_cleanup(void)
1948 {
1949         unregister_netdevice_notifier(&ndisc_netdev_notifier);
1950 }
1951
1952 void ndisc_cleanup(void)
1953 {
1954 #ifdef CONFIG_SYSCTL
1955         neigh_sysctl_unregister(&nd_tbl.parms);
1956 #endif
1957         neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1958         unregister_pernet_subsys(&ndisc_net_ops);
1959 }