net/ipv6/icmp.c

   1 /*
   2  *      Internet Control Message Protocol (ICMPv6)
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Based on net/ipv4/icmp.c
   9  *
  10  *      RFC 1885
  11  *
  12  *      This program is free software; you can redistribute it and/or
  13  *      modify it under the terms of the GNU General Public License
  14  *      as published by the Free Software Foundation; either version
  15  *      2 of the License, or (at your option) any later version.
  16  */
  17
  18 /*
  19  *      Changes:
  20  *
  21  *      Andi Kleen              :       exception handling
  22  *      Andi Kleen                      add rate limits. never reply to a icmp.
  23  *                                      add more length checks and other fixes.
  24  *      yoshfuji                :       ensure to sent parameter problem for
  25  *                                      fragments.
  26  *      YOSHIFUJI Hideaki @USAGI:       added sysctl for icmp rate limit.
  27  *      Randy Dunlap and
  28  *      YOSHIFUJI Hideaki @USAGI:       Per-interface statistics support
  29  *      Kazunori MIYAZAWA @USAGI:       change output process to use ip6_append_data
  30  */
  31
  32 #include <linux/module.h>
  33 #include <linux/errno.h>
  34 #include <linux/types.h>
  35 #include <linux/socket.h>
  36 #include <linux/in.h>
  37 #include <linux/kernel.h>
  38 #include <linux/sockios.h>
  39 #include <linux/net.h>
  40 #include <linux/skbuff.h>
  41 #include <linux/init.h>
  42 #include <linux/netfilter.h>
  43 #include <linux/slab.h>
  44
  45 #ifdef CONFIG_SYSCTL
  46 #include <linux/sysctl.h>
  47 #endif
  48
  49 #include <linux/inet.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/icmpv6.h>
  52
  53 #include <net/ip.h>
  54 #include <net/sock.h>
  55
  56 #include <net/ipv6.h>
  57 #include <net/ip6_checksum.h>
  58 #include <net/protocol.h>
  59 #include <net/raw.h>
  60 #include <net/rawv6.h>
  61 #include <net/transp_v6.h>
  62 #include <net/ip6_route.h>
  63 #include <net/addrconf.h>
  64 #include <net/icmp.h>
  65 #include <net/xfrm.h>
  66 #include <net/inet_common.h>
  67
  68 #include <asm/uaccess.h>
  69 #include <asm/system.h>
  70
  71 /*
  72  *      The ICMP socket(s). This is the most convenient way to flow control
  73  *      our ICMP output as well as maintain a clean interface throughout
  74  *      all layers. All Socketless IP sends will soon be gone.
  75  *
  76  *      On SMP we have one ICMP socket per-cpu.
  77  */
  78 static inline struct sock *icmpv6_sk(struct net *net)
  79 {
  80         return net->ipv6.icmp_sk[smp_processor_id()];
  81 }
  82
  83 static int icmpv6_rcv(struct sk_buff *skb);
  84
  85 static const struct inet6_protocol icmpv6_protocol = {
  86         .handler        =       icmpv6_rcv,
  87         .flags          =       INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
  88 };
  89
  90 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
  91 {
  92         struct sock *sk;
  93
  94         local_bh_disable();
  95
  96         sk = icmpv6_sk(net);
  97         if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
  98                 /* This can happen if the output path (f.e. SIT or
  99                  * ip6ip6 tunnel) signals dst_link_failure() for an
 100                  * outgoing ICMP6 packet.
 101                  */
 102                 local_bh_enable();
 103                 return NULL;
 104         }
 105         return sk;
 106 }
 107
 108 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
 109 {
 110         spin_unlock_bh(&sk->sk_lock.slock);
 111 }
 112
 113 /*
 114  * Slightly more convenient version of icmpv6_send.
 115  */
 116 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
 117 {
 118         icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos);
 119         kfree_skb(skb);
 120 }
 121
 122 /*
 123  * Figure out, may we reply to this packet with icmp error.
 124  *
 125  * We do not reply, if:
 126  *      - it was icmp error message.
 127  *      - it is truncated, so that it is known, that protocol is ICMPV6
 128  *        (i.e. in the middle of some exthdr)
 129  *
 130  *      --ANK (980726)
 131  */
 132
 133 static int is_ineligible(struct sk_buff *skb)
 134 {
 135         int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
 136         int len = skb->len - ptr;
 137         __u8 nexthdr = ipv6_hdr(skb)->nexthdr;
 138
 139         if (len < 0)
 140                 return 1;
 141
 142         ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr);
 143         if (ptr < 0)
 144                 return 0;
 145         if (nexthdr == IPPROTO_ICMPV6) {
 146                 u8 _type, *tp;
 147                 tp = skb_header_pointer(skb,
 148                         ptr+offsetof(struct icmp6hdr, icmp6_type),
 149                         sizeof(_type), &_type);
 150                 if (tp == NULL ||
 151                     !(*tp & ICMPV6_INFOMSG_MASK))
 152                         return 1;
 153         }
 154         return 0;
 155 }
 156
 157 /*
 158  * Check the ICMP output rate limit
 159  */
 160 static inline int icmpv6_xrlim_allow(struct sock *sk, u8 type,
 161                                      struct flowi *fl)
 162 {
 163         struct dst_entry *dst;
 164         struct net *net = sock_net(sk);
 165         int res = 0;
 166
 167         /* Informational messages are not limited. */
 168         if (type & ICMPV6_INFOMSG_MASK)
 169                 return 1;
 170
 171         /* Do not limit pmtu discovery, it would break it. */
 172         if (type == ICMPV6_PKT_TOOBIG)
 173                 return 1;
 174
 175         /*
 176          * Look up the output route.
 177          * XXX: perhaps the expire for routing entries cloned by
 178          * this lookup should be more aggressive (not longer than timeout).
 179          */
 180         dst = ip6_route_output(net, sk, fl);
 181         if (dst->error) {
 182                 IP6_INC_STATS(net, ip6_dst_idev(dst),
 183                               IPSTATS_MIB_OUTNOROUTES);
 184         } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
 185                 res = 1;
 186         } else {
 187                 struct rt6_info *rt = (struct rt6_info *)dst;
 188                 int tmo = net->ipv6.sysctl.icmpv6_time;
 189
 190                 /* Give more bandwidth to wider prefixes. */
 191                 if (rt->rt6i_dst.plen < 128)
 192                         tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
 193
 194                 res = xrlim_allow(dst, tmo);
 195         }
 196         dst_release(dst);
 197         return res;
 198 }
 199
 200 /*
 201  *      an inline helper for the "simple" if statement below
 202  *      checks if parameter problem report is caused by an
 203  *      unrecognized IPv6 option that has the Option Type
 204  *      highest-order two bits set to 10
 205  */
 206
 207 static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset)
 208 {
 209         u8 _optval, *op;
 210
 211         offset += skb_network_offset(skb);
 212         op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
 213         if (op == NULL)
 214                 return 1;
 215         return (*op & 0xC0) == 0x80;
 216 }
 217
 218 static int icmpv6_push_pending_frames(struct sock *sk, struct flowi *fl, struct icmp6hdr *thdr, int len)
 219 {
 220         struct sk_buff *skb;
 221         struct icmp6hdr *icmp6h;
 222         int err = 0;
 223
 224         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
 225                 goto out;
 226
 227         icmp6h = icmp6_hdr(skb);
 228         memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
 229         icmp6h->icmp6_cksum = 0;
 230
 231         if (skb_queue_len(&sk->sk_write_queue) == 1) {
 232                 skb->csum = csum_partial(icmp6h,
 233                                         sizeof(struct icmp6hdr), skb->csum);
 234                 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src,
 235                                                       &fl->fl6_dst,
 236                                                       len, fl->proto,
 237                                                       skb->csum);
 238         } else {
 239                 __wsum tmp_csum = 0;
 240
 241                 skb_queue_walk(&sk->sk_write_queue, skb) {
 242                         tmp_csum = csum_add(tmp_csum, skb->csum);
 243                 }
 244
 245                 tmp_csum = csum_partial(icmp6h,
 246                                         sizeof(struct icmp6hdr), tmp_csum);
 247                 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src,
 248                                                       &fl->fl6_dst,
 249                                                       len, fl->proto,
 250                                                       tmp_csum);
 251         }
 252         ip6_push_pending_frames(sk);
 253 out:
 254         return err;
 255 }
 256
 257 struct icmpv6_msg {
 258         struct sk_buff  *skb;
 259         int             offset;
 260         uint8_t         type;
 261 };
 262
 263 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
 264 {
 265         struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
 266         struct sk_buff *org_skb = msg->skb;
 267         __wsum csum = 0;
 268
 269         csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
 270                                       to, len, csum);
 271         skb->csum = csum_block_add(skb->csum, csum, odd);
 272         if (!(msg->type & ICMPV6_INFOMSG_MASK))
 273                 nf_ct_attach(skb, org_skb);
 274         return 0;
 275 }
 276
 277 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 278 static void mip6_addr_swap(struct sk_buff *skb)
 279 {
 280         struct ipv6hdr *iph = ipv6_hdr(skb);
 281         struct inet6_skb_parm *opt = IP6CB(skb);
 282         struct ipv6_destopt_hao *hao;
 283         struct in6_addr tmp;
 284         int off;
 285
 286         if (opt->dsthao) {
 287                 off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
 288                 if (likely(off >= 0)) {
 289                         hao = (struct ipv6_destopt_hao *)
 290                                         (skb_network_header(skb) + off);
 291                         ipv6_addr_copy(&tmp, &iph->saddr);
 292                         ipv6_addr_copy(&iph->saddr, &hao->addr);
 293                         ipv6_addr_copy(&hao->addr, &tmp);
 294                 }
 295         }
 296 }
 297 #else
 298 static inline void mip6_addr_swap(struct sk_buff *skb) {}
 299 #endif
 300
 301 /*
 302  *      Send an ICMP message in response to a packet in error
 303  */
 304 void icmpv6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
 305 {
 306         struct net *net = dev_net(skb->dev);
 307         struct inet6_dev *idev = NULL;
 308         struct ipv6hdr *hdr = ipv6_hdr(skb);
 309         struct sock *sk;
 310         struct ipv6_pinfo *np;
 311         struct in6_addr *saddr = NULL;
 312         struct dst_entry *dst;
 313         struct dst_entry *dst2;
 314         struct icmp6hdr tmp_hdr;
 315         struct flowi fl;
 316         struct flowi fl2;
 317         struct icmpv6_msg msg;
 318         int iif = 0;
 319         int addr_type = 0;
 320         int len;
 321         int hlimit;
 322         int err = 0;
 323
 324         if ((u8 *)hdr < skb->head ||
 325             (skb->network_header + sizeof(*hdr)) > skb->tail)
 326                 return;
 327
 328         /*
 329          *      Make sure we respect the rules
 330          *      i.e. RFC 1885 2.4(e)
 331          *      Rule (e.1) is enforced by not using icmpv6_send
 332          *      in any code that processes icmp errors.
 333          */
 334         addr_type = ipv6_addr_type(&hdr->daddr);
 335
 336         if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0))
 337                 saddr = &hdr->daddr;
 338
 339         /*
 340          *      Dest addr check
 341          */
 342
 343         if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) {
 344                 if (type != ICMPV6_PKT_TOOBIG &&
 345                     !(type == ICMPV6_PARAMPROB &&
 346                       code == ICMPV6_UNK_OPTION &&
 347                       (opt_unrec(skb, info))))
 348                         return;
 349
 350                 saddr = NULL;
 351         }
 352
 353         addr_type = ipv6_addr_type(&hdr->saddr);
 354
 355         /*
 356          *      Source addr check
 357          */
 358
 359         if (addr_type & IPV6_ADDR_LINKLOCAL)
 360                 iif = skb->dev->ifindex;
 361
 362         /*
 363          *      Must not send error if the source does not uniquely
 364          *      identify a single node (RFC2463 Section 2.4).
 365          *      We check unspecified / multicast addresses here,
 366          *      and anycast addresses will be checked later.
 367          */
 368         if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
 369                 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n");
 370                 return;
 371         }
 372
 373         /*
 374          *      Never answer to a ICMP packet.
 375          */
 376         if (is_ineligible(skb)) {
 377                 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n");
 378                 return;
 379         }
 380
 381         mip6_addr_swap(skb);
 382
 383         memset(&fl, 0, sizeof(fl));
 384         fl.proto = IPPROTO_ICMPV6;
 385         ipv6_addr_copy(&fl.fl6_dst, &hdr->saddr);
 386         if (saddr)
 387                 ipv6_addr_copy(&fl.fl6_src, saddr);
 388         fl.oif = iif;
 389         fl.fl_icmp_type = type;
 390         fl.fl_icmp_code = code;
 391         security_skb_classify_flow(skb, &fl);
 392
 393         sk = icmpv6_xmit_lock(net);
 394         if (sk == NULL)
 395                 return;
 396         np = inet6_sk(sk);
 397
 398         if (!icmpv6_xrlim_allow(sk, type, &fl))
 399                 goto out;
 400
 401         tmp_hdr.icmp6_type = type;
 402         tmp_hdr.icmp6_code = code;
 403         tmp_hdr.icmp6_cksum = 0;
 404         tmp_hdr.icmp6_pointer = htonl(info);
 405
 406         if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
 407                 fl.oif = np->mcast_oif;
 408
 409         err = ip6_dst_lookup(sk, &dst, &fl);
 410         if (err)
 411                 goto out;
 412
 413         /*
 414          * We won't send icmp if the destination is known
 415          * anycast.
 416          */
 417         if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) {
 418                 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n");
 419                 goto out_dst_release;
 420         }
 421
 422         /* No need to clone since we're just using its address. */
 423         dst2 = dst;
 424
 425         err = xfrm_lookup(net, &dst, &fl, sk, 0);
 426         switch (err) {
 427         case 0:
 428                 if (dst != dst2)
 429                         goto route_done;
 430                 break;
 431         case -EPERM:
 432                 dst = NULL;
 433                 break;
 434         default:
 435                 goto out;
 436         }
 437
 438         if (xfrm_decode_session_reverse(skb, &fl2, AF_INET6))
 439                 goto relookup_failed;
 440
 441         if (ip6_dst_lookup(sk, &dst2, &fl2))
 442                 goto relookup_failed;
 443
 444         err = xfrm_lookup(net, &dst2, &fl2, sk, XFRM_LOOKUP_ICMP);
 445         switch (err) {
 446         case 0:
 447                 dst_release(dst);
 448                 dst = dst2;
 449                 break;
 450         case -EPERM:
 451                 goto out_dst_release;
 452         default:
 453 relookup_failed:
 454                 if (!dst)
 455                         goto out;
 456                 break;
 457         }
 458
 459 route_done:
 460         if (ipv6_addr_is_multicast(&fl.fl6_dst))
 461                 hlimit = np->mcast_hops;
 462         else
 463                 hlimit = np->hop_limit;
 464         if (hlimit < 0)
 465                 hlimit = ip6_dst_hoplimit(dst);
 466
 467         msg.skb = skb;
 468         msg.offset = skb_network_offset(skb);
 469         msg.type = type;
 470
 471         len = skb->len - msg.offset;
 472         len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr));
 473         if (len < 0) {
 474                 LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n");
 475                 goto out_dst_release;
 476         }
 477
 478         idev = in6_dev_get(skb->dev);
 479
 480         err = ip6_append_data(sk, icmpv6_getfrag, &msg,
 481                               len + sizeof(struct icmp6hdr),
 482                               sizeof(struct icmp6hdr), hlimit,
 483                               np->tclass, NULL, &fl, (struct rt6_info*)dst,
 484                               MSG_DONTWAIT, np->dontfrag);
 485         if (err) {
 486                 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS);
 487                 ip6_flush_pending_frames(sk);
 488                 goto out_put;
 489         }
 490         err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, len + sizeof(struct icmp6hdr));
 491
 492 out_put:
 493         if (likely(idev != NULL))
 494                 in6_dev_put(idev);
 495 out_dst_release:
 496         dst_release(dst);
 497 out:
 498         icmpv6_xmit_unlock(sk);
 499 }
 500
 501 EXPORT_SYMBOL(icmpv6_send);
 502
 503 static void icmpv6_echo_reply(struct sk_buff *skb)
 504 {
 505         struct net *net = dev_net(skb->dev);
 506         struct sock *sk;
 507         struct inet6_dev *idev;
 508         struct ipv6_pinfo *np;
 509         struct in6_addr *saddr = NULL;
 510         struct icmp6hdr *icmph = icmp6_hdr(skb);
 511         struct icmp6hdr tmp_hdr;
 512         struct flowi fl;
 513         struct icmpv6_msg msg;
 514         struct dst_entry *dst;
 515         int err = 0;
 516         int hlimit;
 517
 518         saddr = &ipv6_hdr(skb)->daddr;
 519
 520         if (!ipv6_unicast_destination(skb))
 521                 saddr = NULL;
 522
 523         memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
 524         tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
 525
 526         memset(&fl, 0, sizeof(fl));
 527         fl.proto = IPPROTO_ICMPV6;
 528         ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
 529         if (saddr)
 530                 ipv6_addr_copy(&fl.fl6_src, saddr);
 531         fl.oif = skb->dev->ifindex;
 532         fl.fl_icmp_type = ICMPV6_ECHO_REPLY;
 533         security_skb_classify_flow(skb, &fl);
 534
 535         sk = icmpv6_xmit_lock(net);
 536         if (sk == NULL)
 537                 return;
 538         np = inet6_sk(sk);
 539
 540         if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
 541                 fl.oif = np->mcast_oif;
 542
 543         err = ip6_dst_lookup(sk, &dst, &fl);
 544         if (err)
 545                 goto out;
 546         if ((err = xfrm_lookup(net, &dst, &fl, sk, 0)) < 0)
 547                 goto out;
 548
 549         if (ipv6_addr_is_multicast(&fl.fl6_dst))
 550                 hlimit = np->mcast_hops;
 551         else
 552                 hlimit = np->hop_limit;
 553         if (hlimit < 0)
 554                 hlimit = ip6_dst_hoplimit(dst);
 555
 556         idev = in6_dev_get(skb->dev);
 557
 558         msg.skb = skb;
 559         msg.offset = 0;
 560         msg.type = ICMPV6_ECHO_REPLY;
 561
 562         err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
 563                                 sizeof(struct icmp6hdr), hlimit, np->tclass, NULL, &fl,
 564                                 (struct rt6_info*)dst, MSG_DONTWAIT,
 565                                 np->dontfrag);
 566
 567         if (err) {
 568                 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS);
 569                 ip6_flush_pending_frames(sk);
 570                 goto out_put;
 571         }
 572         err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, skb->len + sizeof(struct icmp6hdr));
 573
 574 out_put:
 575         if (likely(idev != NULL))
 576                 in6_dev_put(idev);
 577         dst_release(dst);
 578 out:
 579         icmpv6_xmit_unlock(sk);
 580 }
 581
 582 static void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
 583 {
 584         const struct inet6_protocol *ipprot;
 585         int inner_offset;
 586         int hash;
 587         u8 nexthdr;
 588
 589         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 590                 return;
 591
 592         nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
 593         if (ipv6_ext_hdr(nexthdr)) {
 594                 /* now skip over extension headers */
 595                 inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr);
 596                 if (inner_offset<0)
 597                         return;
 598         } else {
 599                 inner_offset = sizeof(struct ipv6hdr);
 600         }
 601
 602         /* Checkin header including 8 bytes of inner protocol header. */
 603         if (!pskb_may_pull(skb, inner_offset+8))
 604                 return;
 605
 606         /* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
 607            Without this we will not able f.e. to make source routed
 608            pmtu discovery.
 609            Corresponding argument (opt) to notifiers is already added.
 610            --ANK (980726)
 611          */
 612
 613         hash = nexthdr & (MAX_INET_PROTOS - 1);
 614
 615         rcu_read_lock();
 616         ipprot = rcu_dereference(inet6_protos[hash]);
 617         if (ipprot && ipprot->err_handler)
 618                 ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
 619         rcu_read_unlock();
 620
 621         raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
 622 }
 623
 624 /*
 625  *      Handle icmp messages
 626  */
 627
 628 static int icmpv6_rcv(struct sk_buff *skb)
 629 {
 630         struct net_device *dev = skb->dev;
 631         struct inet6_dev *idev = __in6_dev_get(dev);
 632         struct in6_addr *saddr, *daddr;
 633         struct ipv6hdr *orig_hdr;
 634         struct icmp6hdr *hdr;
 635         u8 type;
 636
 637         if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 638                 struct sec_path *sp = skb_sec_path(skb);
 639                 int nh;
 640
 641                 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
 642                                  XFRM_STATE_ICMP))
 643                         goto drop_no_count;
 644
 645                 if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(*orig_hdr)))
 646                         goto drop_no_count;
 647
 648                 nh = skb_network_offset(skb);
 649                 skb_set_network_header(skb, sizeof(*hdr));
 650
 651                 if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
 652                         goto drop_no_count;
 653
 654                 skb_set_network_header(skb, nh);
 655         }
 656
 657         ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INMSGS);
 658
 659         saddr = &ipv6_hdr(skb)->saddr;
 660         daddr = &ipv6_hdr(skb)->daddr;
 661
 662         /* Perform checksum. */
 663         switch (skb->ip_summed) {
 664         case CHECKSUM_COMPLETE:
 665                 if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6,
 666                                      skb->csum))
 667                         break;
 668                 /* fall through */
 669         case CHECKSUM_NONE:
 670                 skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len,
 671                                              IPPROTO_ICMPV6, 0));
 672                 if (__skb_checksum_complete(skb)) {
 673                         LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [%pI6 > %pI6]\n",
 674                                        saddr, daddr);
 675                         goto discard_it;
 676                 }
 677         }
 678
 679         if (!pskb_pull(skb, sizeof(*hdr)))
 680                 goto discard_it;
 681
 682         hdr = icmp6_hdr(skb);
 683
 684         type = hdr->icmp6_type;
 685
 686         ICMP6MSGIN_INC_STATS_BH(dev_net(dev), idev, type);
 687
 688         switch (type) {
 689         case ICMPV6_ECHO_REQUEST:
 690                 icmpv6_echo_reply(skb);
 691                 break;
 692
 693         case ICMPV6_ECHO_REPLY:
 694                 /* we couldn't care less */
 695                 break;
 696
 697         case ICMPV6_PKT_TOOBIG:
 698                 /* BUGGG_FUTURE: if packet contains rthdr, we cannot update
 699                    standard destination cache. Seems, only "advanced"
 700                    destination cache will allow to solve this problem
 701                    --ANK (980726)
 702                  */
 703                 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 704                         goto discard_it;
 705                 hdr = icmp6_hdr(skb);
 706                 orig_hdr = (struct ipv6hdr *) (hdr + 1);
 707                 rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev,
 708                                    ntohl(hdr->icmp6_mtu));
 709
 710                 /*
 711                  *      Drop through to notify
 712                  */
 713
 714         case ICMPV6_DEST_UNREACH:
 715         case ICMPV6_TIME_EXCEED:
 716         case ICMPV6_PARAMPROB:
 717                 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
 718                 break;
 719
 720         case NDISC_ROUTER_SOLICITATION:
 721         case NDISC_ROUTER_ADVERTISEMENT:
 722         case NDISC_NEIGHBOUR_SOLICITATION:
 723         case NDISC_NEIGHBOUR_ADVERTISEMENT:
 724         case NDISC_REDIRECT:
 725                 ndisc_rcv(skb);
 726                 break;
 727
 728         case ICMPV6_MGM_QUERY:
 729                 igmp6_event_query(skb);
 730                 break;
 731
 732         case ICMPV6_MGM_REPORT:
 733                 igmp6_event_report(skb);
 734                 break;
 735
 736         case ICMPV6_MGM_REDUCTION:
 737         case ICMPV6_NI_QUERY:
 738         case ICMPV6_NI_REPLY:
 739         case ICMPV6_MLD2_REPORT:
 740         case ICMPV6_DHAAD_REQUEST:
 741         case ICMPV6_DHAAD_REPLY:
 742         case ICMPV6_MOBILE_PREFIX_SOL:
 743         case ICMPV6_MOBILE_PREFIX_ADV:
 744                 break;
 745
 746         default:
 747                 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n");
 748
 749                 /* informational */
 750                 if (type & ICMPV6_INFOMSG_MASK)
 751                         break;
 752
 753                 /*
 754                  * error of unknown type.
 755                  * must pass to upper level
 756                  */
 757
 758                 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
 759         }
 760
 761         kfree_skb(skb);
 762         return 0;
 763
 764 discard_it:
 765         ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INERRORS);
 766 drop_no_count:
 767         kfree_skb(skb);
 768         return 0;
 769 }
 770
 771 void icmpv6_flow_init(struct sock *sk, struct flowi *fl,
 772                       u8 type,
 773                       const struct in6_addr *saddr,
 774                       const struct in6_addr *daddr,
 775                       int oif)
 776 {
 777         memset(fl, 0, sizeof(*fl));
 778         ipv6_addr_copy(&fl->fl6_src, saddr);
 779         ipv6_addr_copy(&fl->fl6_dst, daddr);
 780         fl->proto               = IPPROTO_ICMPV6;
 781         fl->fl_icmp_type        = type;
 782         fl->fl_icmp_code        = 0;
 783         fl->oif                 = oif;
 784         security_sk_classify_flow(sk, fl);
 785 }
 786
 787 /*
 788  * Special lock-class for __icmpv6_sk:
 789  */
 790 static struct lock_class_key icmpv6_socket_sk_dst_lock_key;
 791
 792 static int __net_init icmpv6_sk_init(struct net *net)
 793 {
 794         struct sock *sk;
 795         int err, i, j;
 796
 797         net->ipv6.icmp_sk =
 798                 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
 799         if (net->ipv6.icmp_sk == NULL)
 800                 return -ENOMEM;
 801
 802         for_each_possible_cpu(i) {
 803                 err = inet_ctl_sock_create(&sk, PF_INET6,
 804                                            SOCK_RAW, IPPROTO_ICMPV6, net);
 805                 if (err < 0) {
 806                         printk(KERN_ERR
 807                                "Failed to initialize the ICMP6 control socket "
 808                                "(err %d).\n",
 809                                err);
 810                         goto fail;
 811                 }
 812
 813                 net->ipv6.icmp_sk[i] = sk;
 814
 815                 /*
 816                  * Split off their lock-class, because sk->sk_dst_lock
 817                  * gets used from softirqs, which is safe for
 818                  * __icmpv6_sk (because those never get directly used
 819                  * via userspace syscalls), but unsafe for normal sockets.
 820                  */
 821                 lockdep_set_class(&sk->sk_dst_lock,
 822                                   &icmpv6_socket_sk_dst_lock_key);
 823
 824                 /* Enough space for 2 64K ICMP packets, including
 825                  * sk_buff struct overhead.
 826                  */
 827                 sk->sk_sndbuf =
 828                         (2 * ((64 * 1024) + sizeof(struct sk_buff)));
 829         }
 830         return 0;
 831
 832  fail:
 833         for (j = 0; j < i; j++)
 834                 inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
 835         kfree(net->ipv6.icmp_sk);
 836         return err;
 837 }
 838
 839 static void __net_exit icmpv6_sk_exit(struct net *net)
 840 {
 841         int i;
 842
 843         for_each_possible_cpu(i) {
 844                 inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
 845         }
 846         kfree(net->ipv6.icmp_sk);
 847 }
 848
 849 static struct pernet_operations icmpv6_sk_ops = {
 850        .init = icmpv6_sk_init,
 851        .exit = icmpv6_sk_exit,
 852 };
 853
 854 int __init icmpv6_init(void)
 855 {
 856         int err;
 857
 858         err = register_pernet_subsys(&icmpv6_sk_ops);
 859         if (err < 0)
 860                 return err;
 861
 862         err = -EAGAIN;
 863         if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
 864                 goto fail;
 865         return 0;
 866
 867 fail:
 868         printk(KERN_ERR "Failed to register ICMP6 protocol\n");
 869         unregister_pernet_subsys(&icmpv6_sk_ops);
 870         return err;
 871 }
 872
 873 void icmpv6_cleanup(void)
 874 {
 875         unregister_pernet_subsys(&icmpv6_sk_ops);
 876         inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
 877 }
 878
 879
 880 static const struct icmp6_err {
 881         int err;
 882         int fatal;
 883 } tab_unreach[] = {
 884         {       /* NOROUTE */
 885                 .err    = ENETUNREACH,
 886                 .fatal  = 0,
 887         },
 888         {       /* ADM_PROHIBITED */
 889                 .err    = EACCES,
 890                 .fatal  = 1,
 891         },
 892         {       /* Was NOT_NEIGHBOUR, now reserved */
 893                 .err    = EHOSTUNREACH,
 894                 .fatal  = 0,
 895         },
 896         {       /* ADDR_UNREACH */
 897                 .err    = EHOSTUNREACH,
 898                 .fatal  = 0,
 899         },
 900         {       /* PORT_UNREACH */
 901                 .err    = ECONNREFUSED,
 902                 .fatal  = 1,
 903         },
 904 };
 905
 906 int icmpv6_err_convert(u8 type, u8 code, int *err)
 907 {
 908         int fatal = 0;
 909
 910         *err = EPROTO;
 911
 912         switch (type) {
 913         case ICMPV6_DEST_UNREACH:
 914                 fatal = 1;
 915                 if (code <= ICMPV6_PORT_UNREACH) {
 916                         *err  = tab_unreach[code].err;
 917                         fatal = tab_unreach[code].fatal;
 918                 }
 919                 break;
 920
 921         case ICMPV6_PKT_TOOBIG:
 922                 *err = EMSGSIZE;
 923                 break;
 924
 925         case ICMPV6_PARAMPROB:
 926                 *err = EPROTO;
 927                 fatal = 1;
 928                 break;
 929
 930         case ICMPV6_TIME_EXCEED:
 931                 *err = EHOSTUNREACH;
 932                 break;
 933         }
 934
 935         return fatal;
 936 }
 937
 938 EXPORT_SYMBOL(icmpv6_err_convert);
 939
 940 #ifdef CONFIG_SYSCTL
 941 ctl_table ipv6_icmp_table_template[] = {
 942         {
 943                 .procname       = "ratelimit",
 944                 .data           = &init_net.ipv6.sysctl.icmpv6_time,
 945                 .maxlen         = sizeof(int),
 946                 .mode           = 0644,
 947                 .proc_handler   = proc_dointvec_ms_jiffies,
 948         },
 949         { },
 950 };
 951
 952 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
 953 {
 954         struct ctl_table *table;
 955
 956         table = kmemdup(ipv6_icmp_table_template,
 957                         sizeof(ipv6_icmp_table_template),
 958                         GFP_KERNEL);
 959
 960         if (table)
 961                 table[0].data = &net->ipv6.sysctl.icmpv6_time;
 962
 963         return table;
 964 }
 965 #endif
 966