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
  44 #ifdef CONFIG_SYSCTL
  45 #include <linux/sysctl.h>
  46 #endif
  47
  48 #include <linux/inet.h>
  49 #include <linux/netdevice.h>
  50 #include <linux/icmpv6.h>
  51
  52 #include <net/ip.h>
  53 #include <net/sock.h>
  54
  55 #include <net/ipv6.h>
  56 #include <net/ip6_checksum.h>
  57 #include <net/protocol.h>
  58 #include <net/raw.h>
  59 #include <net/rawv6.h>
  60 #include <net/transp_v6.h>
  61 #include <net/ip6_route.h>
  62 #include <net/addrconf.h>
  63 #include <net/icmp.h>
  64 #include <net/xfrm.h>
  65 #include <net/inet_common.h>
  66
  67 #include <asm/uaccess.h>
  68 #include <asm/system.h>
  69
  70 DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics) __read_mostly;
  71 EXPORT_SYMBOL(icmpv6_statistics);
  72 DEFINE_SNMP_STAT(struct icmpv6msg_mib, icmpv6msg_statistics) __read_mostly;
  73 EXPORT_SYMBOL(icmpv6msg_statistics);
  74
  75 /*
  76  *      The ICMP socket(s). This is the most convenient way to flow control
  77  *      our ICMP output as well as maintain a clean interface throughout
  78  *      all layers. All Socketless IP sends will soon be gone.
  79  *
  80  *      On SMP we have one ICMP socket per-cpu.
  81  */
  82 static inline struct sock *icmpv6_sk(struct net *net)
  83 {
  84         return net->ipv6.icmp_sk[smp_processor_id()];
  85 }
  86
  87 static int icmpv6_rcv(struct sk_buff *skb);
  88
  89 static struct inet6_protocol icmpv6_protocol = {
  90         .handler        =       icmpv6_rcv,
  91         .flags          =       INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
  92 };
  93
  94 static __inline__ int icmpv6_xmit_lock(struct sock *sk)
  95 {
  96         local_bh_disable();
  97
  98         if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
  99                 /* This can happen if the output path (f.e. SIT or
 100                  * ip6ip6 tunnel) signals dst_link_failure() for an
 101                  * outgoing ICMP6 packet.
 102                  */
 103                 local_bh_enable();
 104                 return 1;
 105         }
 106         return 0;
 107 }
 108
 109 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
 110 {
 111         spin_unlock_bh(&sk->sk_lock.slock);
 112 }
 113
 114 /*
 115  * Slightly more convenient version of icmpv6_send.
 116  */
 117 void icmpv6_param_prob(struct sk_buff *skb, int code, int pos)
 118 {
 119         icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos, skb->dev);
 120         kfree_skb(skb);
 121 }
 122
 123 /*
 124  * Figure out, may we reply to this packet with icmp error.
 125  *
 126  * We do not reply, if:
 127  *      - it was icmp error message.
 128  *      - it is truncated, so that it is known, that protocol is ICMPV6
 129  *        (i.e. in the middle of some exthdr)
 130  *
 131  *      --ANK (980726)
 132  */
 133
 134 static int is_ineligible(struct sk_buff *skb)
 135 {
 136         int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
 137         int len = skb->len - ptr;
 138         __u8 nexthdr = ipv6_hdr(skb)->nexthdr;
 139
 140         if (len < 0)
 141                 return 1;
 142
 143         ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr);
 144         if (ptr < 0)
 145                 return 0;
 146         if (nexthdr == IPPROTO_ICMPV6) {
 147                 u8 _type, *tp;
 148                 tp = skb_header_pointer(skb,
 149                         ptr+offsetof(struct icmp6hdr, icmp6_type),
 150                         sizeof(_type), &_type);
 151                 if (tp == NULL ||
 152                     !(*tp & ICMPV6_INFOMSG_MASK))
 153                         return 1;
 154         }
 155         return 0;
 156 }
 157
 158 /*
 159  * Check the ICMP output rate limit
 160  */
 161 static inline int icmpv6_xrlim_allow(struct sock *sk, int type,
 162                                      struct flowi *fl)
 163 {
 164         struct dst_entry *dst;
 165         struct net *net = sock_net(sk);
 166         int res = 0;
 167
 168         /* Informational messages are not limited. */
 169         if (type & ICMPV6_INFOMSG_MASK)
 170                 return 1;
 171
 172         /* Do not limit pmtu discovery, it would break it. */
 173         if (type == ICMPV6_PKT_TOOBIG)
 174                 return 1;
 175
 176         /*
 177          * Look up the output route.
 178          * XXX: perhaps the expire for routing entries cloned by
 179          * this lookup should be more aggressive (not longer than timeout).
 180          */
 181         dst = ip6_route_output(net, sk, fl);
 182         if (dst->error) {
 183                 IP6_INC_STATS(ip6_dst_idev(dst),
 184                               IPSTATS_MIB_OUTNOROUTES);
 185         } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
 186                 res = 1;
 187         } else {
 188                 struct rt6_info *rt = (struct rt6_info *)dst;
 189                 int tmo = net->ipv6.sysctl.icmpv6_time;
 190
 191                 /* Give more bandwidth to wider prefixes. */
 192                 if (rt->rt6i_dst.plen < 128)
 193                         tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
 194
 195                 res = xrlim_allow(dst, tmo);
 196         }
 197         dst_release(dst);
 198         return res;
 199 }
 200
 201 /*
 202  *      an inline helper for the "simple" if statement below
 203  *      checks if parameter problem report is caused by an
 204  *      unrecognized IPv6 option that has the Option Type
 205  *      highest-order two bits set to 10
 206  */
 207
 208 static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset)
 209 {
 210         u8 _optval, *op;
 211
 212         offset += skb_network_offset(skb);
 213         op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
 214         if (op == NULL)
 215                 return 1;
 216         return (*op & 0xC0) == 0x80;
 217 }
 218
 219 static int icmpv6_push_pending_frames(struct sock *sk, struct flowi *fl, struct icmp6hdr *thdr, int len)
 220 {
 221         struct sk_buff *skb;
 222         struct icmp6hdr *icmp6h;
 223         int err = 0;
 224
 225         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
 226                 goto out;
 227
 228         icmp6h = icmp6_hdr(skb);
 229         memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
 230         icmp6h->icmp6_cksum = 0;
 231
 232         if (skb_queue_len(&sk->sk_write_queue) == 1) {
 233                 skb->csum = csum_partial((char *)icmp6h,
 234                                         sizeof(struct icmp6hdr), skb->csum);
 235                 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src,
 236                                                       &fl->fl6_dst,
 237                                                       len, fl->proto,
 238                                                       skb->csum);
 239         } else {
 240                 __wsum tmp_csum = 0;
 241
 242                 skb_queue_walk(&sk->sk_write_queue, skb) {
 243                         tmp_csum = csum_add(tmp_csum, skb->csum);
 244                 }
 245
 246                 tmp_csum = csum_partial((char *)icmp6h,
 247                                         sizeof(struct icmp6hdr), tmp_csum);
 248                 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src,
 249                                                       &fl->fl6_dst,
 250                                                       len, fl->proto,
 251                                                       tmp_csum);
 252         }
 253         ip6_push_pending_frames(sk);
 254 out:
 255         return err;
 256 }
 257
 258 struct icmpv6_msg {
 259         struct sk_buff  *skb;
 260         int             offset;
 261         uint8_t         type;
 262 };
 263
 264 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
 265 {
 266         struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
 267         struct sk_buff *org_skb = msg->skb;
 268         __wsum csum = 0;
 269
 270         csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
 271                                       to, len, csum);
 272         skb->csum = csum_block_add(skb->csum, csum, odd);
 273         if (!(msg->type & ICMPV6_INFOMSG_MASK))
 274                 nf_ct_attach(skb, org_skb);
 275         return 0;
 276 }
 277
 278 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 279 static void mip6_addr_swap(struct sk_buff *skb)
 280 {
 281         struct ipv6hdr *iph = ipv6_hdr(skb);
 282         struct inet6_skb_parm *opt = IP6CB(skb);
 283         struct ipv6_destopt_hao *hao;
 284         struct in6_addr tmp;
 285         int off;
 286
 287         if (opt->dsthao) {
 288                 off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
 289                 if (likely(off >= 0)) {
 290                         hao = (struct ipv6_destopt_hao *)
 291                                         (skb_network_header(skb) + off);
 292                         ipv6_addr_copy(&tmp, &iph->saddr);
 293                         ipv6_addr_copy(&iph->saddr, &hao->addr);
 294                         ipv6_addr_copy(&hao->addr, &tmp);
 295                 }
 296         }
 297 }
 298 #else
 299 static inline void mip6_addr_swap(struct sk_buff *skb) {}
 300 #endif
 301
 302 /*
 303  *      Send an ICMP message in response to a packet in error
 304  */
 305 void icmpv6_send(struct sk_buff *skb, int type, int code, __u32 info,
 306                  struct net_device *dev)
 307 {
 308         struct net *net = dev_net(skb->dev);
 309         struct inet6_dev *idev = NULL;
 310         struct ipv6hdr *hdr = ipv6_hdr(skb);
 311         struct sock *sk;
 312         struct ipv6_pinfo *np;
 313         struct in6_addr *saddr = NULL;
 314         struct dst_entry *dst;
 315         struct dst_entry *dst2;
 316         struct icmp6hdr tmp_hdr;
 317         struct flowi fl;
 318         struct flowi fl2;
 319         struct icmpv6_msg msg;
 320         int iif = 0;
 321         int addr_type = 0;
 322         int len;
 323         int hlimit, tclass;
 324         int err = 0;
 325
 326         if ((u8 *)hdr < skb->head ||
 327             (skb->network_header + sizeof(*hdr)) > skb->tail)
 328                 return;
 329
 330         /*
 331          *      Make sure we respect the rules
 332          *      i.e. RFC 1885 2.4(e)
 333          *      Rule (e.1) is enforced by not using icmpv6_send
 334          *      in any code that processes icmp errors.
 335          */
 336         addr_type = ipv6_addr_type(&hdr->daddr);
 337
 338         if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0))
 339                 saddr = &hdr->daddr;
 340
 341         /*
 342          *      Dest addr check
 343          */
 344
 345         if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) {
 346                 if (type != ICMPV6_PKT_TOOBIG &&
 347                     !(type == ICMPV6_PARAMPROB &&
 348                       code == ICMPV6_UNK_OPTION &&
 349                       (opt_unrec(skb, info))))
 350                         return;
 351
 352                 saddr = NULL;
 353         }
 354
 355         addr_type = ipv6_addr_type(&hdr->saddr);
 356
 357         /*
 358          *      Source addr check
 359          */
 360
 361         if (addr_type & IPV6_ADDR_LINKLOCAL)
 362                 iif = skb->dev->ifindex;
 363
 364         /*
 365          *      Must not send error if the source does not uniquely
 366          *      identify a single node (RFC2463 Section 2.4).
 367          *      We check unspecified / multicast addresses here,
 368          *      and anycast addresses will be checked later.
 369          */
 370         if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
 371                 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n");
 372                 return;
 373         }
 374
 375         /*
 376          *      Never answer to a ICMP packet.
 377          */
 378         if (is_ineligible(skb)) {
 379                 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n");
 380                 return;
 381         }
 382
 383         mip6_addr_swap(skb);
 384
 385         memset(&fl, 0, sizeof(fl));
 386         fl.proto = IPPROTO_ICMPV6;
 387         ipv6_addr_copy(&fl.fl6_dst, &hdr->saddr);
 388         if (saddr)
 389                 ipv6_addr_copy(&fl.fl6_src, saddr);
 390         fl.oif = iif;
 391         fl.fl_icmp_type = type;
 392         fl.fl_icmp_code = code;
 393         security_skb_classify_flow(skb, &fl);
 394
 395         sk = icmpv6_sk(net);
 396         np = inet6_sk(sk);
 397
 398         if (icmpv6_xmit_lock(sk))
 399                 return;
 400
 401         if (!icmpv6_xrlim_allow(sk, type, &fl))
 402                 goto out;
 403
 404         tmp_hdr.icmp6_type = type;
 405         tmp_hdr.icmp6_code = code;
 406         tmp_hdr.icmp6_cksum = 0;
 407         tmp_hdr.icmp6_pointer = htonl(info);
 408
 409         if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
 410                 fl.oif = np->mcast_oif;
 411
 412         err = ip6_dst_lookup(sk, &dst, &fl);
 413         if (err)
 414                 goto out;
 415
 416         /*
 417          * We won't send icmp if the destination is known
 418          * anycast.
 419          */
 420         if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) {
 421                 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n");
 422                 goto out_dst_release;
 423         }
 424
 425         /* No need to clone since we're just using its address. */
 426         dst2 = dst;
 427
 428         err = xfrm_lookup(&dst, &fl, sk, 0);
 429         switch (err) {
 430         case 0:
 431                 if (dst != dst2)
 432                         goto route_done;
 433                 break;
 434         case -EPERM:
 435                 dst = NULL;
 436                 break;
 437         default:
 438                 goto out;
 439         }
 440
 441         if (xfrm_decode_session_reverse(skb, &fl2, AF_INET6))
 442                 goto relookup_failed;
 443
 444         if (ip6_dst_lookup(sk, &dst2, &fl))
 445                 goto relookup_failed;
 446
 447         err = xfrm_lookup(&dst2, &fl, sk, XFRM_LOOKUP_ICMP);
 448         switch (err) {
 449         case 0:
 450                 dst_release(dst);
 451                 dst = dst2;
 452                 break;
 453         case -EPERM:
 454                 goto out_dst_release;
 455         default:
 456 relookup_failed:
 457                 if (!dst)
 458                         goto out;
 459                 break;
 460         }
 461
 462 route_done:
 463         if (ipv6_addr_is_multicast(&fl.fl6_dst))
 464                 hlimit = np->mcast_hops;
 465         else
 466                 hlimit = np->hop_limit;
 467         if (hlimit < 0)
 468                 hlimit = ip6_dst_hoplimit(dst);
 469
 470         tclass = np->tclass;
 471         if (tclass < 0)
 472                 tclass = 0;
 473
 474         msg.skb = skb;
 475         msg.offset = skb_network_offset(skb);
 476         msg.type = type;
 477
 478         len = skb->len - msg.offset;
 479         len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr));
 480         if (len < 0) {
 481                 LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n");
 482                 goto out_dst_release;
 483         }
 484
 485         idev = in6_dev_get(skb->dev);
 486
 487         err = ip6_append_data(sk, icmpv6_getfrag, &msg,
 488                               len + sizeof(struct icmp6hdr),
 489                               sizeof(struct icmp6hdr),
 490                               hlimit, tclass, NULL, &fl, (struct rt6_info*)dst,
 491                               MSG_DONTWAIT);
 492         if (err) {
 493                 ip6_flush_pending_frames(sk);
 494                 goto out_put;
 495         }
 496         err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, len + sizeof(struct icmp6hdr));
 497
 498 out_put:
 499         if (likely(idev != NULL))
 500                 in6_dev_put(idev);
 501 out_dst_release:
 502         dst_release(dst);
 503 out:
 504         icmpv6_xmit_unlock(sk);
 505 }
 506
 507 EXPORT_SYMBOL(icmpv6_send);
 508
 509 static void icmpv6_echo_reply(struct sk_buff *skb)
 510 {
 511         struct net *net = dev_net(skb->dev);
 512         struct sock *sk;
 513         struct inet6_dev *idev;
 514         struct ipv6_pinfo *np;
 515         struct in6_addr *saddr = NULL;
 516         struct icmp6hdr *icmph = icmp6_hdr(skb);
 517         struct icmp6hdr tmp_hdr;
 518         struct flowi fl;
 519         struct icmpv6_msg msg;
 520         struct dst_entry *dst;
 521         int err = 0;
 522         int hlimit;
 523         int tclass;
 524
 525         saddr = &ipv6_hdr(skb)->daddr;
 526
 527         if (!ipv6_unicast_destination(skb))
 528                 saddr = NULL;
 529
 530         memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
 531         tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
 532
 533         memset(&fl, 0, sizeof(fl));
 534         fl.proto = IPPROTO_ICMPV6;
 535         ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
 536         if (saddr)
 537                 ipv6_addr_copy(&fl.fl6_src, saddr);
 538         fl.oif = skb->dev->ifindex;
 539         fl.fl_icmp_type = ICMPV6_ECHO_REPLY;
 540         security_skb_classify_flow(skb, &fl);
 541
 542         sk = icmpv6_sk(net);
 543         np = inet6_sk(sk);
 544
 545         if (icmpv6_xmit_lock(sk))
 546                 return;
 547
 548         if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
 549                 fl.oif = np->mcast_oif;
 550
 551         err = ip6_dst_lookup(sk, &dst, &fl);
 552         if (err)
 553                 goto out;
 554         if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
 555                 goto out;
 556
 557         if (ipv6_addr_is_multicast(&fl.fl6_dst))
 558                 hlimit = np->mcast_hops;
 559         else
 560                 hlimit = np->hop_limit;
 561         if (hlimit < 0)
 562                 hlimit = ip6_dst_hoplimit(dst);
 563
 564         tclass = np->tclass;
 565         if (tclass < 0)
 566                 tclass = 0;
 567
 568         idev = in6_dev_get(skb->dev);
 569
 570         msg.skb = skb;
 571         msg.offset = 0;
 572         msg.type = ICMPV6_ECHO_REPLY;
 573
 574         err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
 575                                 sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl,
 576                                 (struct rt6_info*)dst, MSG_DONTWAIT);
 577
 578         if (err) {
 579                 ip6_flush_pending_frames(sk);
 580                 goto out_put;
 581         }
 582         err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, skb->len + sizeof(struct icmp6hdr));
 583
 584 out_put:
 585         if (likely(idev != NULL))
 586                 in6_dev_put(idev);
 587         dst_release(dst);
 588 out:
 589         icmpv6_xmit_unlock(sk);
 590 }
 591
 592 static void icmpv6_notify(struct sk_buff *skb, int type, int code, __be32 info)
 593 {
 594         struct inet6_protocol *ipprot;
 595         int inner_offset;
 596         int hash;
 597         u8 nexthdr;
 598
 599         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 600                 return;
 601
 602         nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
 603         if (ipv6_ext_hdr(nexthdr)) {
 604                 /* now skip over extension headers */
 605                 inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr);
 606                 if (inner_offset<0)
 607                         return;
 608         } else {
 609                 inner_offset = sizeof(struct ipv6hdr);
 610         }
 611
 612         /* Checkin header including 8 bytes of inner protocol header. */
 613         if (!pskb_may_pull(skb, inner_offset+8))
 614                 return;
 615
 616         /* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
 617            Without this we will not able f.e. to make source routed
 618            pmtu discovery.
 619            Corresponding argument (opt) to notifiers is already added.
 620            --ANK (980726)
 621          */
 622
 623         hash = nexthdr & (MAX_INET_PROTOS - 1);
 624
 625         rcu_read_lock();
 626         ipprot = rcu_dereference(inet6_protos[hash]);
 627         if (ipprot && ipprot->err_handler)
 628                 ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
 629         rcu_read_unlock();
 630
 631         raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
 632 }
 633
 634 /*
 635  *      Handle icmp messages
 636  */
 637
 638 static int icmpv6_rcv(struct sk_buff *skb)
 639 {
 640         struct net_device *dev = skb->dev;
 641         struct inet6_dev *idev = __in6_dev_get(dev);
 642         struct in6_addr *saddr, *daddr;
 643         struct ipv6hdr *orig_hdr;
 644         struct icmp6hdr *hdr;
 645         int type;
 646
 647         if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 648                 int nh;
 649
 650                 if (!(skb->sp && skb->sp->xvec[skb->sp->len - 1]->props.flags &
 651                                  XFRM_STATE_ICMP))
 652                         goto drop_no_count;
 653
 654                 if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(*orig_hdr)))
 655                         goto drop_no_count;
 656
 657                 nh = skb_network_offset(skb);
 658                 skb_set_network_header(skb, sizeof(*hdr));
 659
 660                 if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
 661                         goto drop_no_count;
 662
 663                 skb_set_network_header(skb, nh);
 664         }
 665
 666         ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INMSGS);
 667
 668         saddr = &ipv6_hdr(skb)->saddr;
 669         daddr = &ipv6_hdr(skb)->daddr;
 670
 671         /* Perform checksum. */
 672         switch (skb->ip_summed) {
 673         case CHECKSUM_COMPLETE:
 674                 if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6,
 675                                      skb->csum))
 676                         break;
 677                 /* fall through */
 678         case CHECKSUM_NONE:
 679                 skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len,
 680                                              IPPROTO_ICMPV6, 0));
 681                 if (__skb_checksum_complete(skb)) {
 682                         LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [" NIP6_FMT " > " NIP6_FMT "]\n",
 683                                        NIP6(*saddr), NIP6(*daddr));
 684                         goto discard_it;
 685                 }
 686         }
 687
 688         if (!pskb_pull(skb, sizeof(*hdr)))
 689                 goto discard_it;
 690
 691         hdr = icmp6_hdr(skb);
 692
 693         type = hdr->icmp6_type;
 694
 695         ICMP6MSGIN_INC_STATS_BH(idev, type);
 696
 697         switch (type) {
 698         case ICMPV6_ECHO_REQUEST:
 699                 icmpv6_echo_reply(skb);
 700                 break;
 701
 702         case ICMPV6_ECHO_REPLY:
 703                 /* we couldn't care less */
 704                 break;
 705
 706         case ICMPV6_PKT_TOOBIG:
 707                 /* BUGGG_FUTURE: if packet contains rthdr, we cannot update
 708                    standard destination cache. Seems, only "advanced"
 709                    destination cache will allow to solve this problem
 710                    --ANK (980726)
 711                  */
 712                 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 713                         goto discard_it;
 714                 hdr = icmp6_hdr(skb);
 715                 orig_hdr = (struct ipv6hdr *) (hdr + 1);
 716                 rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev,
 717                                    ntohl(hdr->icmp6_mtu));
 718
 719                 /*
 720                  *      Drop through to notify
 721                  */
 722
 723         case ICMPV6_DEST_UNREACH:
 724         case ICMPV6_TIME_EXCEED:
 725         case ICMPV6_PARAMPROB:
 726                 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
 727                 break;
 728
 729         case NDISC_ROUTER_SOLICITATION:
 730         case NDISC_ROUTER_ADVERTISEMENT:
 731         case NDISC_NEIGHBOUR_SOLICITATION:
 732         case NDISC_NEIGHBOUR_ADVERTISEMENT:
 733         case NDISC_REDIRECT:
 734                 ndisc_rcv(skb);
 735                 break;
 736
 737         case ICMPV6_MGM_QUERY:
 738                 igmp6_event_query(skb);
 739                 break;
 740
 741         case ICMPV6_MGM_REPORT:
 742                 igmp6_event_report(skb);
 743                 break;
 744
 745         case ICMPV6_MGM_REDUCTION:
 746         case ICMPV6_NI_QUERY:
 747         case ICMPV6_NI_REPLY:
 748         case ICMPV6_MLD2_REPORT:
 749         case ICMPV6_DHAAD_REQUEST:
 750         case ICMPV6_DHAAD_REPLY:
 751         case ICMPV6_MOBILE_PREFIX_SOL:
 752         case ICMPV6_MOBILE_PREFIX_ADV:
 753                 break;
 754
 755         default:
 756                 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n");
 757
 758                 /* informational */
 759                 if (type & ICMPV6_INFOMSG_MASK)
 760                         break;
 761
 762                 /*
 763                  * error of unknown type.
 764                  * must pass to upper level
 765                  */
 766
 767                 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
 768         }
 769
 770         kfree_skb(skb);
 771         return 0;
 772
 773 discard_it:
 774         ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INERRORS);
 775 drop_no_count:
 776         kfree_skb(skb);
 777         return 0;
 778 }
 779
 780 void icmpv6_flow_init(struct sock *sk, struct flowi *fl,
 781                       u8 type,
 782                       const struct in6_addr *saddr,
 783                       const struct in6_addr *daddr,
 784                       int oif)
 785 {
 786         memset(fl, 0, sizeof(*fl));
 787         ipv6_addr_copy(&fl->fl6_src, saddr);
 788         ipv6_addr_copy(&fl->fl6_dst, daddr);
 789         fl->proto               = IPPROTO_ICMPV6;
 790         fl->fl_icmp_type        = type;
 791         fl->fl_icmp_code        = 0;
 792         fl->oif                 = oif;
 793         security_sk_classify_flow(sk, fl);
 794 }
 795
 796 /*
 797  * Special lock-class for __icmpv6_sk:
 798  */
 799 static struct lock_class_key icmpv6_socket_sk_dst_lock_key;
 800
 801 static int __net_init icmpv6_sk_init(struct net *net)
 802 {
 803         struct sock *sk;
 804         int err, i, j;
 805
 806         net->ipv6.icmp_sk =
 807                 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
 808         if (net->ipv6.icmp_sk == NULL)
 809                 return -ENOMEM;
 810
 811         for_each_possible_cpu(i) {
 812                 err = inet_ctl_sock_create(&sk, PF_INET6,
 813                                            SOCK_RAW, IPPROTO_ICMPV6, net);
 814                 if (err < 0) {
 815                         printk(KERN_ERR
 816                                "Failed to initialize the ICMP6 control socket "
 817                                "(err %d).\n",
 818                                err);
 819                         goto fail;
 820                 }
 821
 822                 net->ipv6.icmp_sk[i] = sk;
 823
 824                 /*
 825                  * Split off their lock-class, because sk->sk_dst_lock
 826                  * gets used from softirqs, which is safe for
 827                  * __icmpv6_sk (because those never get directly used
 828                  * via userspace syscalls), but unsafe for normal sockets.
 829                  */
 830                 lockdep_set_class(&sk->sk_dst_lock,
 831                                   &icmpv6_socket_sk_dst_lock_key);
 832
 833                 /* Enough space for 2 64K ICMP packets, including
 834                  * sk_buff struct overhead.
 835                  */
 836                 sk->sk_sndbuf =
 837                         (2 * ((64 * 1024) + sizeof(struct sk_buff)));
 838         }
 839         return 0;
 840
 841  fail:
 842         for (j = 0; j < i; j++)
 843                 inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
 844         kfree(net->ipv6.icmp_sk);
 845         return err;
 846 }
 847
 848 static void __net_exit icmpv6_sk_exit(struct net *net)
 849 {
 850         int i;
 851
 852         for_each_possible_cpu(i) {
 853                 inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
 854         }
 855         kfree(net->ipv6.icmp_sk);
 856 }
 857
 858 static struct pernet_operations icmpv6_sk_ops = {
 859        .init = icmpv6_sk_init,
 860        .exit = icmpv6_sk_exit,
 861 };
 862
 863 int __init icmpv6_init(void)
 864 {
 865         int err;
 866
 867         err = register_pernet_subsys(&icmpv6_sk_ops);
 868         if (err < 0)
 869                 return err;
 870
 871         err = -EAGAIN;
 872         if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
 873                 goto fail;
 874         return 0;
 875
 876 fail:
 877         printk(KERN_ERR "Failed to register ICMP6 protocol\n");
 878         unregister_pernet_subsys(&icmpv6_sk_ops);
 879         return err;
 880 }
 881
 882 void icmpv6_cleanup(void)
 883 {
 884         unregister_pernet_subsys(&icmpv6_sk_ops);
 885         inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
 886 }
 887
 888
 889 static const struct icmp6_err {
 890         int err;
 891         int fatal;
 892 } tab_unreach[] = {
 893         {       /* NOROUTE */
 894                 .err    = ENETUNREACH,
 895                 .fatal  = 0,
 896         },
 897         {       /* ADM_PROHIBITED */
 898                 .err    = EACCES,
 899                 .fatal  = 1,
 900         },
 901         {       /* Was NOT_NEIGHBOUR, now reserved */
 902                 .err    = EHOSTUNREACH,
 903                 .fatal  = 0,
 904         },
 905         {       /* ADDR_UNREACH */
 906                 .err    = EHOSTUNREACH,
 907                 .fatal  = 0,
 908         },
 909         {       /* PORT_UNREACH */
 910                 .err    = ECONNREFUSED,
 911                 .fatal  = 1,
 912         },
 913 };
 914
 915 int icmpv6_err_convert(int type, int code, int *err)
 916 {
 917         int fatal = 0;
 918
 919         *err = EPROTO;
 920
 921         switch (type) {
 922         case ICMPV6_DEST_UNREACH:
 923                 fatal = 1;
 924                 if (code <= ICMPV6_PORT_UNREACH) {
 925                         *err  = tab_unreach[code].err;
 926                         fatal = tab_unreach[code].fatal;
 927                 }
 928                 break;
 929
 930         case ICMPV6_PKT_TOOBIG:
 931                 *err = EMSGSIZE;
 932                 break;
 933
 934         case ICMPV6_PARAMPROB:
 935                 *err = EPROTO;
 936                 fatal = 1;
 937                 break;
 938
 939         case ICMPV6_TIME_EXCEED:
 940                 *err = EHOSTUNREACH;
 941                 break;
 942         }
 943
 944         return fatal;
 945 }
 946
 947 EXPORT_SYMBOL(icmpv6_err_convert);
 948
 949 #ifdef CONFIG_SYSCTL
 950 ctl_table ipv6_icmp_table_template[] = {
 951         {
 952                 .ctl_name       = NET_IPV6_ICMP_RATELIMIT,
 953                 .procname       = "ratelimit",
 954                 .data           = &init_net.ipv6.sysctl.icmpv6_time,
 955                 .maxlen         = sizeof(int),
 956                 .mode           = 0644,
 957                 .proc_handler   = &proc_dointvec_ms_jiffies,
 958                 .strategy       = &sysctl_ms_jiffies
 959         },
 960         { .ctl_name = 0 },
 961 };
 962
 963 struct ctl_table *ipv6_icmp_sysctl_init(struct net *net)
 964 {
 965         struct ctl_table *table;
 966
 967         table = kmemdup(ipv6_icmp_table_template,
 968                         sizeof(ipv6_icmp_table_template),
 969                         GFP_KERNEL);
 970
 971         if (table)
 972                 table[0].data = &net->ipv6.sysctl.icmpv6_time;
 973
 974         return table;
 975 }
 976 #endif
 977