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 #define pr_fmt(fmt) "IPv6: " fmt 33 34 #include <linux/module.h> 35 #include <linux/errno.h> 36 #include <linux/types.h> 37 #include <linux/socket.h> 38 #include <linux/in.h> 39 #include <linux/kernel.h> 40 #include <linux/sockios.h> 41 #include <linux/net.h> 42 #include <linux/skbuff.h> 43 #include <linux/init.h> 44 #include <linux/netfilter.h> 45 #include <linux/slab.h> 46 47 #ifdef CONFIG_SYSCTL 48 #include <linux/sysctl.h> 49 #endif 50 51 #include <linux/inet.h> 52 #include <linux/netdevice.h> 53 #include <linux/icmpv6.h> 54 55 #include <net/ip.h> 56 #include <net/sock.h> 57 58 #include <net/ipv6.h> 59 #include <net/ip6_checksum.h> 60 #include <net/ping.h> 61 #include <net/protocol.h> 62 #include <net/raw.h> 63 #include <net/rawv6.h> 64 #include <net/transp_v6.h> 65 #include <net/ip6_route.h> 66 #include <net/addrconf.h> 67 #include <net/icmp.h> 68 #include <net/xfrm.h> 69 #include <net/inet_common.h> 70 #include <net/dsfield.h> 71 72 #include <asm/uaccess.h> 73 74 /* 75 * The ICMP socket(s). This is the most convenient way to flow control 76 * our ICMP output as well as maintain a clean interface throughout 77 * all layers. All Socketless IP sends will soon be gone. 78 * 79 * On SMP we have one ICMP socket per-cpu. 80 */ 81 static inline struct sock *icmpv6_sk(struct net *net) 82 { 83 return net->ipv6.icmp_sk[smp_processor_id()]; 84 } 85 86 static void icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 87 u8 type, u8 code, int offset, __be32 info) 88 { 89 /* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */ 90 struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset); 91 struct net *net = dev_net(skb->dev); 92 93 if (type == ICMPV6_PKT_TOOBIG) 94 ip6_update_pmtu(skb, net, info, 0, 0); 95 else if (type == NDISC_REDIRECT) 96 ip6_redirect(skb, net, skb->dev->ifindex, 0); 97 98 if (!(type & ICMPV6_INFOMSG_MASK)) 99 if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST) 100 ping_err(skb, offset, info); 101 } 102 103 static int icmpv6_rcv(struct sk_buff *skb); 104 105 static const struct inet6_protocol icmpv6_protocol = { 106 .handler = icmpv6_rcv, 107 .err_handler = icmpv6_err, 108 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 109 }; 110 111 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net) 112 { 113 struct sock *sk; 114 115 local_bh_disable(); 116 117 sk = icmpv6_sk(net); 118 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) { 119 /* This can happen if the output path (f.e. SIT or 120 * ip6ip6 tunnel) signals dst_link_failure() for an 121 * outgoing ICMP6 packet. 122 */ 123 local_bh_enable(); 124 return NULL; 125 } 126 return sk; 127 } 128 129 static __inline__ void icmpv6_xmit_unlock(struct sock *sk) 130 { 131 spin_unlock_bh(&sk->sk_lock.slock); 132 } 133 134 /* 135 * Figure out, may we reply to this packet with icmp error. 136 * 137 * We do not reply, if: 138 * - it was icmp error message. 139 * - it is truncated, so that it is known, that protocol is ICMPV6 140 * (i.e. in the middle of some exthdr) 141 * 142 * --ANK (980726) 143 */ 144 145 static bool is_ineligible(const struct sk_buff *skb) 146 { 147 int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data; 148 int len = skb->len - ptr; 149 __u8 nexthdr = ipv6_hdr(skb)->nexthdr; 150 __be16 frag_off; 151 152 if (len < 0) 153 return true; 154 155 ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off); 156 if (ptr < 0) 157 return false; 158 if (nexthdr == IPPROTO_ICMPV6) { 159 u8 _type, *tp; 160 tp = skb_header_pointer(skb, 161 ptr+offsetof(struct icmp6hdr, icmp6_type), 162 sizeof(_type), &_type); 163 if (tp == NULL || 164 !(*tp & ICMPV6_INFOMSG_MASK)) 165 return true; 166 } 167 return false; 168 } 169 170 /* 171 * Check the ICMP output rate limit 172 */ 173 static inline bool icmpv6_xrlim_allow(struct sock *sk, u8 type, 174 struct flowi6 *fl6) 175 { 176 struct dst_entry *dst; 177 struct net *net = sock_net(sk); 178 bool res = false; 179 180 /* Informational messages are not limited. */ 181 if (type & ICMPV6_INFOMSG_MASK) 182 return true; 183 184 /* Do not limit pmtu discovery, it would break it. */ 185 if (type == ICMPV6_PKT_TOOBIG) 186 return true; 187 188 /* 189 * Look up the output route. 190 * XXX: perhaps the expire for routing entries cloned by 191 * this lookup should be more aggressive (not longer than timeout). 192 */ 193 dst = ip6_route_output(net, sk, fl6); 194 if (dst->error) { 195 IP6_INC_STATS(net, ip6_dst_idev(dst), 196 IPSTATS_MIB_OUTNOROUTES); 197 } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) { 198 res = true; 199 } else { 200 struct rt6_info *rt = (struct rt6_info *)dst; 201 int tmo = net->ipv6.sysctl.icmpv6_time; 202 struct inet_peer *peer; 203 204 /* Give more bandwidth to wider prefixes. */ 205 if (rt->rt6i_dst.plen < 128) 206 tmo >>= ((128 - rt->rt6i_dst.plen)>>5); 207 208 peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1); 209 res = inet_peer_xrlim_allow(peer, tmo); 210 if (peer) 211 inet_putpeer(peer); 212 } 213 dst_release(dst); 214 return res; 215 } 216 217 /* 218 * an inline helper for the "simple" if statement below 219 * checks if parameter problem report is caused by an 220 * unrecognized IPv6 option that has the Option Type 221 * highest-order two bits set to 10 222 */ 223 224 static bool opt_unrec(struct sk_buff *skb, __u32 offset) 225 { 226 u8 _optval, *op; 227 228 offset += skb_network_offset(skb); 229 op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval); 230 if (op == NULL) 231 return true; 232 return (*op & 0xC0) == 0x80; 233 } 234 235 int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 236 struct icmp6hdr *thdr, int len) 237 { 238 struct sk_buff *skb; 239 struct icmp6hdr *icmp6h; 240 int err = 0; 241 242 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 243 goto out; 244 245 icmp6h = icmp6_hdr(skb); 246 memcpy(icmp6h, thdr, sizeof(struct icmp6hdr)); 247 icmp6h->icmp6_cksum = 0; 248 249 if (skb_queue_len(&sk->sk_write_queue) == 1) { 250 skb->csum = csum_partial(icmp6h, 251 sizeof(struct icmp6hdr), skb->csum); 252 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr, 253 &fl6->daddr, 254 len, fl6->flowi6_proto, 255 skb->csum); 256 } else { 257 __wsum tmp_csum = 0; 258 259 skb_queue_walk(&sk->sk_write_queue, skb) { 260 tmp_csum = csum_add(tmp_csum, skb->csum); 261 } 262 263 tmp_csum = csum_partial(icmp6h, 264 sizeof(struct icmp6hdr), tmp_csum); 265 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr, 266 &fl6->daddr, 267 len, fl6->flowi6_proto, 268 tmp_csum); 269 } 270 ip6_push_pending_frames(sk); 271 out: 272 return err; 273 } 274 275 struct icmpv6_msg { 276 struct sk_buff *skb; 277 int offset; 278 uint8_t type; 279 }; 280 281 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb) 282 { 283 struct icmpv6_msg *msg = (struct icmpv6_msg *) from; 284 struct sk_buff *org_skb = msg->skb; 285 __wsum csum = 0; 286 287 csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset, 288 to, len, csum); 289 skb->csum = csum_block_add(skb->csum, csum, odd); 290 if (!(msg->type & ICMPV6_INFOMSG_MASK)) 291 nf_ct_attach(skb, org_skb); 292 return 0; 293 } 294 295 #if IS_ENABLED(CONFIG_IPV6_MIP6) 296 static void mip6_addr_swap(struct sk_buff *skb) 297 { 298 struct ipv6hdr *iph = ipv6_hdr(skb); 299 struct inet6_skb_parm *opt = IP6CB(skb); 300 struct ipv6_destopt_hao *hao; 301 struct in6_addr tmp; 302 int off; 303 304 if (opt->dsthao) { 305 off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO); 306 if (likely(off >= 0)) { 307 hao = (struct ipv6_destopt_hao *) 308 (skb_network_header(skb) + off); 309 tmp = iph->saddr; 310 iph->saddr = hao->addr; 311 hao->addr = tmp; 312 } 313 } 314 } 315 #else 316 static inline void mip6_addr_swap(struct sk_buff *skb) {} 317 #endif 318 319 static struct dst_entry *icmpv6_route_lookup(struct net *net, 320 struct sk_buff *skb, 321 struct sock *sk, 322 struct flowi6 *fl6) 323 { 324 struct dst_entry *dst, *dst2; 325 struct flowi6 fl2; 326 int err; 327 328 err = ip6_dst_lookup(sk, &dst, fl6); 329 if (err) 330 return ERR_PTR(err); 331 332 /* 333 * We won't send icmp if the destination is known 334 * anycast. 335 */ 336 if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) { 337 LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: acast source\n"); 338 dst_release(dst); 339 return ERR_PTR(-EINVAL); 340 } 341 342 /* No need to clone since we're just using its address. */ 343 dst2 = dst; 344 345 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0); 346 if (!IS_ERR(dst)) { 347 if (dst != dst2) 348 return dst; 349 } else { 350 if (PTR_ERR(dst) == -EPERM) 351 dst = NULL; 352 else 353 return dst; 354 } 355 356 err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6); 357 if (err) 358 goto relookup_failed; 359 360 err = ip6_dst_lookup(sk, &dst2, &fl2); 361 if (err) 362 goto relookup_failed; 363 364 dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP); 365 if (!IS_ERR(dst2)) { 366 dst_release(dst); 367 dst = dst2; 368 } else { 369 err = PTR_ERR(dst2); 370 if (err == -EPERM) { 371 dst_release(dst); 372 return dst2; 373 } else 374 goto relookup_failed; 375 } 376 377 relookup_failed: 378 if (dst) 379 return dst; 380 return ERR_PTR(err); 381 } 382 383 /* 384 * Send an ICMP message in response to a packet in error 385 */ 386 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info) 387 { 388 struct net *net = dev_net(skb->dev); 389 struct inet6_dev *idev = NULL; 390 struct ipv6hdr *hdr = ipv6_hdr(skb); 391 struct sock *sk; 392 struct ipv6_pinfo *np; 393 const struct in6_addr *saddr = NULL; 394 struct dst_entry *dst; 395 struct icmp6hdr tmp_hdr; 396 struct flowi6 fl6; 397 struct icmpv6_msg msg; 398 int iif = 0; 399 int addr_type = 0; 400 int len; 401 int hlimit; 402 int err = 0; 403 404 if ((u8 *)hdr < skb->head || 405 (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb)) 406 return; 407 408 /* 409 * Make sure we respect the rules 410 * i.e. RFC 1885 2.4(e) 411 * Rule (e.1) is enforced by not using icmp6_send 412 * in any code that processes icmp errors. 413 */ 414 addr_type = ipv6_addr_type(&hdr->daddr); 415 416 if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) || 417 ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr)) 418 saddr = &hdr->daddr; 419 420 /* 421 * Dest addr check 422 */ 423 424 if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) { 425 if (type != ICMPV6_PKT_TOOBIG && 426 !(type == ICMPV6_PARAMPROB && 427 code == ICMPV6_UNK_OPTION && 428 (opt_unrec(skb, info)))) 429 return; 430 431 saddr = NULL; 432 } 433 434 addr_type = ipv6_addr_type(&hdr->saddr); 435 436 /* 437 * Source addr check 438 */ 439 440 if (__ipv6_addr_needs_scope_id(addr_type)) 441 iif = skb->dev->ifindex; 442 443 /* 444 * Must not send error if the source does not uniquely 445 * identify a single node (RFC2463 Section 2.4). 446 * We check unspecified / multicast addresses here, 447 * and anycast addresses will be checked later. 448 */ 449 if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) { 450 LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: addr_any/mcast source\n"); 451 return; 452 } 453 454 /* 455 * Never answer to a ICMP packet. 456 */ 457 if (is_ineligible(skb)) { 458 LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: no reply to icmp error\n"); 459 return; 460 } 461 462 mip6_addr_swap(skb); 463 464 memset(&fl6, 0, sizeof(fl6)); 465 fl6.flowi6_proto = IPPROTO_ICMPV6; 466 fl6.daddr = hdr->saddr; 467 if (saddr) 468 fl6.saddr = *saddr; 469 fl6.flowi6_oif = iif; 470 fl6.fl6_icmp_type = type; 471 fl6.fl6_icmp_code = code; 472 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); 473 474 sk = icmpv6_xmit_lock(net); 475 if (sk == NULL) 476 return; 477 np = inet6_sk(sk); 478 479 if (!icmpv6_xrlim_allow(sk, type, &fl6)) 480 goto out; 481 482 tmp_hdr.icmp6_type = type; 483 tmp_hdr.icmp6_code = code; 484 tmp_hdr.icmp6_cksum = 0; 485 tmp_hdr.icmp6_pointer = htonl(info); 486 487 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 488 fl6.flowi6_oif = np->mcast_oif; 489 else if (!fl6.flowi6_oif) 490 fl6.flowi6_oif = np->ucast_oif; 491 492 dst = icmpv6_route_lookup(net, skb, sk, &fl6); 493 if (IS_ERR(dst)) 494 goto out; 495 496 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 497 498 msg.skb = skb; 499 msg.offset = skb_network_offset(skb); 500 msg.type = type; 501 502 len = skb->len - msg.offset; 503 len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr)); 504 if (len < 0) { 505 LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n"); 506 goto out_dst_release; 507 } 508 509 rcu_read_lock(); 510 idev = __in6_dev_get(skb->dev); 511 512 err = ip6_append_data(sk, icmpv6_getfrag, &msg, 513 len + sizeof(struct icmp6hdr), 514 sizeof(struct icmp6hdr), hlimit, 515 np->tclass, NULL, &fl6, (struct rt6_info *)dst, 516 MSG_DONTWAIT, np->dontfrag); 517 if (err) { 518 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS); 519 ip6_flush_pending_frames(sk); 520 } else { 521 err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, 522 len + sizeof(struct icmp6hdr)); 523 } 524 rcu_read_unlock(); 525 out_dst_release: 526 dst_release(dst); 527 out: 528 icmpv6_xmit_unlock(sk); 529 } 530 531 /* Slightly more convenient version of icmp6_send. 532 */ 533 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos) 534 { 535 icmp6_send(skb, ICMPV6_PARAMPROB, code, pos); 536 kfree_skb(skb); 537 } 538 539 static void icmpv6_echo_reply(struct sk_buff *skb) 540 { 541 struct net *net = dev_net(skb->dev); 542 struct sock *sk; 543 struct inet6_dev *idev; 544 struct ipv6_pinfo *np; 545 const struct in6_addr *saddr = NULL; 546 struct icmp6hdr *icmph = icmp6_hdr(skb); 547 struct icmp6hdr tmp_hdr; 548 struct flowi6 fl6; 549 struct icmpv6_msg msg; 550 struct dst_entry *dst; 551 int err = 0; 552 int hlimit; 553 u8 tclass; 554 555 saddr = &ipv6_hdr(skb)->daddr; 556 557 if (!ipv6_unicast_destination(skb) && 558 !(net->ipv6.sysctl.anycast_src_echo_reply && 559 ipv6_anycast_destination(skb))) 560 saddr = NULL; 561 562 memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr)); 563 tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY; 564 565 memset(&fl6, 0, sizeof(fl6)); 566 fl6.flowi6_proto = IPPROTO_ICMPV6; 567 fl6.daddr = ipv6_hdr(skb)->saddr; 568 if (saddr) 569 fl6.saddr = *saddr; 570 fl6.flowi6_oif = skb->dev->ifindex; 571 fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY; 572 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); 573 574 sk = icmpv6_xmit_lock(net); 575 if (sk == NULL) 576 return; 577 np = inet6_sk(sk); 578 579 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 580 fl6.flowi6_oif = np->mcast_oif; 581 else if (!fl6.flowi6_oif) 582 fl6.flowi6_oif = np->ucast_oif; 583 584 err = ip6_dst_lookup(sk, &dst, &fl6); 585 if (err) 586 goto out; 587 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0); 588 if (IS_ERR(dst)) 589 goto out; 590 591 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 592 593 idev = __in6_dev_get(skb->dev); 594 595 msg.skb = skb; 596 msg.offset = 0; 597 msg.type = ICMPV6_ECHO_REPLY; 598 599 tclass = ipv6_get_dsfield(ipv6_hdr(skb)); 600 err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr), 601 sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl6, 602 (struct rt6_info *)dst, MSG_DONTWAIT, 603 np->dontfrag); 604 605 if (err) { 606 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS); 607 ip6_flush_pending_frames(sk); 608 } else { 609 err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, 610 skb->len + sizeof(struct icmp6hdr)); 611 } 612 dst_release(dst); 613 out: 614 icmpv6_xmit_unlock(sk); 615 } 616 617 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info) 618 { 619 const struct inet6_protocol *ipprot; 620 int inner_offset; 621 __be16 frag_off; 622 u8 nexthdr; 623 624 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 625 return; 626 627 nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr; 628 if (ipv6_ext_hdr(nexthdr)) { 629 /* now skip over extension headers */ 630 inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), 631 &nexthdr, &frag_off); 632 if (inner_offset<0) 633 return; 634 } else { 635 inner_offset = sizeof(struct ipv6hdr); 636 } 637 638 /* Checkin header including 8 bytes of inner protocol header. */ 639 if (!pskb_may_pull(skb, inner_offset+8)) 640 return; 641 642 /* BUGGG_FUTURE: we should try to parse exthdrs in this packet. 643 Without this we will not able f.e. to make source routed 644 pmtu discovery. 645 Corresponding argument (opt) to notifiers is already added. 646 --ANK (980726) 647 */ 648 649 rcu_read_lock(); 650 ipprot = rcu_dereference(inet6_protos[nexthdr]); 651 if (ipprot && ipprot->err_handler) 652 ipprot->err_handler(skb, NULL, type, code, inner_offset, info); 653 rcu_read_unlock(); 654 655 raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info); 656 } 657 658 /* 659 * Handle icmp messages 660 */ 661 662 static int icmpv6_rcv(struct sk_buff *skb) 663 { 664 struct net_device *dev = skb->dev; 665 struct inet6_dev *idev = __in6_dev_get(dev); 666 const struct in6_addr *saddr, *daddr; 667 struct icmp6hdr *hdr; 668 u8 type; 669 670 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 671 struct sec_path *sp = skb_sec_path(skb); 672 int nh; 673 674 if (!(sp && sp->xvec[sp->len - 1]->props.flags & 675 XFRM_STATE_ICMP)) 676 goto drop_no_count; 677 678 if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr))) 679 goto drop_no_count; 680 681 nh = skb_network_offset(skb); 682 skb_set_network_header(skb, sizeof(*hdr)); 683 684 if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) 685 goto drop_no_count; 686 687 skb_set_network_header(skb, nh); 688 } 689 690 ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INMSGS); 691 692 saddr = &ipv6_hdr(skb)->saddr; 693 daddr = &ipv6_hdr(skb)->daddr; 694 695 /* Perform checksum. */ 696 switch (skb->ip_summed) { 697 case CHECKSUM_COMPLETE: 698 if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6, 699 skb->csum)) 700 break; 701 /* fall through */ 702 case CHECKSUM_NONE: 703 skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len, 704 IPPROTO_ICMPV6, 0)); 705 if (__skb_checksum_complete(skb)) { 706 LIMIT_NETDEBUG(KERN_DEBUG 707 "ICMPv6 checksum failed [%pI6c > %pI6c]\n", 708 saddr, daddr); 709 goto csum_error; 710 } 711 } 712 713 if (!pskb_pull(skb, sizeof(*hdr))) 714 goto discard_it; 715 716 hdr = icmp6_hdr(skb); 717 718 type = hdr->icmp6_type; 719 720 ICMP6MSGIN_INC_STATS_BH(dev_net(dev), idev, type); 721 722 switch (type) { 723 case ICMPV6_ECHO_REQUEST: 724 icmpv6_echo_reply(skb); 725 break; 726 727 case ICMPV6_ECHO_REPLY: 728 ping_rcv(skb); 729 break; 730 731 case ICMPV6_PKT_TOOBIG: 732 /* BUGGG_FUTURE: if packet contains rthdr, we cannot update 733 standard destination cache. Seems, only "advanced" 734 destination cache will allow to solve this problem 735 --ANK (980726) 736 */ 737 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 738 goto discard_it; 739 hdr = icmp6_hdr(skb); 740 741 /* 742 * Drop through to notify 743 */ 744 745 case ICMPV6_DEST_UNREACH: 746 case ICMPV6_TIME_EXCEED: 747 case ICMPV6_PARAMPROB: 748 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); 749 break; 750 751 case NDISC_ROUTER_SOLICITATION: 752 case NDISC_ROUTER_ADVERTISEMENT: 753 case NDISC_NEIGHBOUR_SOLICITATION: 754 case NDISC_NEIGHBOUR_ADVERTISEMENT: 755 case NDISC_REDIRECT: 756 ndisc_rcv(skb); 757 break; 758 759 case ICMPV6_MGM_QUERY: 760 igmp6_event_query(skb); 761 break; 762 763 case ICMPV6_MGM_REPORT: 764 igmp6_event_report(skb); 765 break; 766 767 case ICMPV6_MGM_REDUCTION: 768 case ICMPV6_NI_QUERY: 769 case ICMPV6_NI_REPLY: 770 case ICMPV6_MLD2_REPORT: 771 case ICMPV6_DHAAD_REQUEST: 772 case ICMPV6_DHAAD_REPLY: 773 case ICMPV6_MOBILE_PREFIX_SOL: 774 case ICMPV6_MOBILE_PREFIX_ADV: 775 break; 776 777 default: 778 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n"); 779 780 /* informational */ 781 if (type & ICMPV6_INFOMSG_MASK) 782 break; 783 784 /* 785 * error of unknown type. 786 * must pass to upper level 787 */ 788 789 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); 790 } 791 792 kfree_skb(skb); 793 return 0; 794 795 csum_error: 796 ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS); 797 discard_it: 798 ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INERRORS); 799 drop_no_count: 800 kfree_skb(skb); 801 return 0; 802 } 803 804 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6, 805 u8 type, 806 const struct in6_addr *saddr, 807 const struct in6_addr *daddr, 808 int oif) 809 { 810 memset(fl6, 0, sizeof(*fl6)); 811 fl6->saddr = *saddr; 812 fl6->daddr = *daddr; 813 fl6->flowi6_proto = IPPROTO_ICMPV6; 814 fl6->fl6_icmp_type = type; 815 fl6->fl6_icmp_code = 0; 816 fl6->flowi6_oif = oif; 817 security_sk_classify_flow(sk, flowi6_to_flowi(fl6)); 818 } 819 820 /* 821 * Special lock-class for __icmpv6_sk: 822 */ 823 static struct lock_class_key icmpv6_socket_sk_dst_lock_key; 824 825 static int __net_init icmpv6_sk_init(struct net *net) 826 { 827 struct sock *sk; 828 int err, i, j; 829 830 net->ipv6.icmp_sk = 831 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL); 832 if (net->ipv6.icmp_sk == NULL) 833 return -ENOMEM; 834 835 for_each_possible_cpu(i) { 836 err = inet_ctl_sock_create(&sk, PF_INET6, 837 SOCK_RAW, IPPROTO_ICMPV6, net); 838 if (err < 0) { 839 pr_err("Failed to initialize the ICMP6 control socket (err %d)\n", 840 err); 841 goto fail; 842 } 843 844 net->ipv6.icmp_sk[i] = sk; 845 846 /* 847 * Split off their lock-class, because sk->sk_dst_lock 848 * gets used from softirqs, which is safe for 849 * __icmpv6_sk (because those never get directly used 850 * via userspace syscalls), but unsafe for normal sockets. 851 */ 852 lockdep_set_class(&sk->sk_dst_lock, 853 &icmpv6_socket_sk_dst_lock_key); 854 855 /* Enough space for 2 64K ICMP packets, including 856 * sk_buff struct overhead. 857 */ 858 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); 859 } 860 return 0; 861 862 fail: 863 for (j = 0; j < i; j++) 864 inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]); 865 kfree(net->ipv6.icmp_sk); 866 return err; 867 } 868 869 static void __net_exit icmpv6_sk_exit(struct net *net) 870 { 871 int i; 872 873 for_each_possible_cpu(i) { 874 inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]); 875 } 876 kfree(net->ipv6.icmp_sk); 877 } 878 879 static struct pernet_operations icmpv6_sk_ops = { 880 .init = icmpv6_sk_init, 881 .exit = icmpv6_sk_exit, 882 }; 883 884 int __init icmpv6_init(void) 885 { 886 int err; 887 888 err = register_pernet_subsys(&icmpv6_sk_ops); 889 if (err < 0) 890 return err; 891 892 err = -EAGAIN; 893 if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0) 894 goto fail; 895 896 err = inet6_register_icmp_sender(icmp6_send); 897 if (err) 898 goto sender_reg_err; 899 return 0; 900 901 sender_reg_err: 902 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); 903 fail: 904 pr_err("Failed to register ICMP6 protocol\n"); 905 unregister_pernet_subsys(&icmpv6_sk_ops); 906 return err; 907 } 908 909 void icmpv6_cleanup(void) 910 { 911 inet6_unregister_icmp_sender(icmp6_send); 912 unregister_pernet_subsys(&icmpv6_sk_ops); 913 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); 914 } 915 916 917 static const struct icmp6_err { 918 int err; 919 int fatal; 920 } tab_unreach[] = { 921 { /* NOROUTE */ 922 .err = ENETUNREACH, 923 .fatal = 0, 924 }, 925 { /* ADM_PROHIBITED */ 926 .err = EACCES, 927 .fatal = 1, 928 }, 929 { /* Was NOT_NEIGHBOUR, now reserved */ 930 .err = EHOSTUNREACH, 931 .fatal = 0, 932 }, 933 { /* ADDR_UNREACH */ 934 .err = EHOSTUNREACH, 935 .fatal = 0, 936 }, 937 { /* PORT_UNREACH */ 938 .err = ECONNREFUSED, 939 .fatal = 1, 940 }, 941 { /* POLICY_FAIL */ 942 .err = EACCES, 943 .fatal = 1, 944 }, 945 { /* REJECT_ROUTE */ 946 .err = EACCES, 947 .fatal = 1, 948 }, 949 }; 950 951 int icmpv6_err_convert(u8 type, u8 code, int *err) 952 { 953 int fatal = 0; 954 955 *err = EPROTO; 956 957 switch (type) { 958 case ICMPV6_DEST_UNREACH: 959 fatal = 1; 960 if (code < ARRAY_SIZE(tab_unreach)) { 961 *err = tab_unreach[code].err; 962 fatal = tab_unreach[code].fatal; 963 } 964 break; 965 966 case ICMPV6_PKT_TOOBIG: 967 *err = EMSGSIZE; 968 break; 969 970 case ICMPV6_PARAMPROB: 971 *err = EPROTO; 972 fatal = 1; 973 break; 974 975 case ICMPV6_TIME_EXCEED: 976 *err = EHOSTUNREACH; 977 break; 978 } 979 980 return fatal; 981 } 982 EXPORT_SYMBOL(icmpv6_err_convert); 983 984 #ifdef CONFIG_SYSCTL 985 static struct ctl_table ipv6_icmp_table_template[] = { 986 { 987 .procname = "ratelimit", 988 .data = &init_net.ipv6.sysctl.icmpv6_time, 989 .maxlen = sizeof(int), 990 .mode = 0644, 991 .proc_handler = proc_dointvec_ms_jiffies, 992 }, 993 { }, 994 }; 995 996 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net) 997 { 998 struct ctl_table *table; 999 1000 table = kmemdup(ipv6_icmp_table_template, 1001 sizeof(ipv6_icmp_table_template), 1002 GFP_KERNEL); 1003 1004 if (table) 1005 table[0].data = &net->ipv6.sysctl.icmpv6_time; 1006 1007 return table; 1008 } 1009 #endif 1010 1011