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 #include <net/l3mdev.h> 72 73 #include <linux/uaccess.h> 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 *this_cpu_ptr(net->ipv6.icmp_sk); 85 } 86 87 static int icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 88 u8 type, u8 code, int offset, __be32 info) 89 { 90 /* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */ 91 struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset); 92 struct net *net = dev_net(skb->dev); 93 94 if (type == ICMPV6_PKT_TOOBIG) 95 ip6_update_pmtu(skb, net, info, skb->dev->ifindex, 0, sock_net_uid(net, NULL)); 96 else if (type == NDISC_REDIRECT) 97 ip6_redirect(skb, net, skb->dev->ifindex, 0, 98 sock_net_uid(net, NULL)); 99 100 if (!(type & ICMPV6_INFOMSG_MASK)) 101 if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST) 102 ping_err(skb, offset, ntohl(info)); 103 104 return 0; 105 } 106 107 static int icmpv6_rcv(struct sk_buff *skb); 108 109 static const struct inet6_protocol icmpv6_protocol = { 110 .handler = icmpv6_rcv, 111 .err_handler = icmpv6_err, 112 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 113 }; 114 115 /* Called with BH disabled */ 116 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net) 117 { 118 struct sock *sk; 119 120 sk = icmpv6_sk(net); 121 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) { 122 /* This can happen if the output path (f.e. SIT or 123 * ip6ip6 tunnel) signals dst_link_failure() for an 124 * outgoing ICMP6 packet. 125 */ 126 return NULL; 127 } 128 return sk; 129 } 130 131 static __inline__ void icmpv6_xmit_unlock(struct sock *sk) 132 { 133 spin_unlock(&sk->sk_lock.slock); 134 } 135 136 /* 137 * Figure out, may we reply to this packet with icmp error. 138 * 139 * We do not reply, if: 140 * - it was icmp error message. 141 * - it is truncated, so that it is known, that protocol is ICMPV6 142 * (i.e. in the middle of some exthdr) 143 * 144 * --ANK (980726) 145 */ 146 147 static bool is_ineligible(const struct sk_buff *skb) 148 { 149 int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data; 150 int len = skb->len - ptr; 151 __u8 nexthdr = ipv6_hdr(skb)->nexthdr; 152 __be16 frag_off; 153 154 if (len < 0) 155 return true; 156 157 ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off); 158 if (ptr < 0) 159 return false; 160 if (nexthdr == IPPROTO_ICMPV6) { 161 u8 _type, *tp; 162 tp = skb_header_pointer(skb, 163 ptr+offsetof(struct icmp6hdr, icmp6_type), 164 sizeof(_type), &_type); 165 if (!tp || !(*tp & ICMPV6_INFOMSG_MASK)) 166 return true; 167 } 168 return false; 169 } 170 171 static bool icmpv6_mask_allow(struct net *net, int type) 172 { 173 if (type > ICMPV6_MSG_MAX) 174 return true; 175 176 /* Limit if icmp type is set in ratemask. */ 177 if (!test_bit(type, net->ipv6.sysctl.icmpv6_ratemask)) 178 return true; 179 180 return false; 181 } 182 183 static bool icmpv6_global_allow(struct net *net, int type) 184 { 185 if (icmpv6_mask_allow(net, type)) 186 return true; 187 188 if (icmp_global_allow()) 189 return true; 190 191 return false; 192 } 193 194 /* 195 * Check the ICMP output rate limit 196 */ 197 static bool icmpv6_xrlim_allow(struct sock *sk, u8 type, 198 struct flowi6 *fl6) 199 { 200 struct net *net = sock_net(sk); 201 struct dst_entry *dst; 202 bool res = false; 203 204 if (icmpv6_mask_allow(net, type)) 205 return true; 206 207 /* 208 * Look up the output route. 209 * XXX: perhaps the expire for routing entries cloned by 210 * this lookup should be more aggressive (not longer than timeout). 211 */ 212 dst = ip6_route_output(net, sk, fl6); 213 if (dst->error) { 214 IP6_INC_STATS(net, ip6_dst_idev(dst), 215 IPSTATS_MIB_OUTNOROUTES); 216 } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) { 217 res = true; 218 } else { 219 struct rt6_info *rt = (struct rt6_info *)dst; 220 int tmo = net->ipv6.sysctl.icmpv6_time; 221 struct inet_peer *peer; 222 223 /* Give more bandwidth to wider prefixes. */ 224 if (rt->rt6i_dst.plen < 128) 225 tmo >>= ((128 - rt->rt6i_dst.plen)>>5); 226 227 peer = inet_getpeer_v6(net->ipv6.peers, &fl6->daddr, 1); 228 res = inet_peer_xrlim_allow(peer, tmo); 229 if (peer) 230 inet_putpeer(peer); 231 } 232 dst_release(dst); 233 return res; 234 } 235 236 /* 237 * an inline helper for the "simple" if statement below 238 * checks if parameter problem report is caused by an 239 * unrecognized IPv6 option that has the Option Type 240 * highest-order two bits set to 10 241 */ 242 243 static bool opt_unrec(struct sk_buff *skb, __u32 offset) 244 { 245 u8 _optval, *op; 246 247 offset += skb_network_offset(skb); 248 op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval); 249 if (!op) 250 return true; 251 return (*op & 0xC0) == 0x80; 252 } 253 254 void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 255 struct icmp6hdr *thdr, int len) 256 { 257 struct sk_buff *skb; 258 struct icmp6hdr *icmp6h; 259 260 skb = skb_peek(&sk->sk_write_queue); 261 if (!skb) 262 return; 263 264 icmp6h = icmp6_hdr(skb); 265 memcpy(icmp6h, thdr, sizeof(struct icmp6hdr)); 266 icmp6h->icmp6_cksum = 0; 267 268 if (skb_queue_len(&sk->sk_write_queue) == 1) { 269 skb->csum = csum_partial(icmp6h, 270 sizeof(struct icmp6hdr), skb->csum); 271 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr, 272 &fl6->daddr, 273 len, fl6->flowi6_proto, 274 skb->csum); 275 } else { 276 __wsum tmp_csum = 0; 277 278 skb_queue_walk(&sk->sk_write_queue, skb) { 279 tmp_csum = csum_add(tmp_csum, skb->csum); 280 } 281 282 tmp_csum = csum_partial(icmp6h, 283 sizeof(struct icmp6hdr), tmp_csum); 284 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr, 285 &fl6->daddr, 286 len, fl6->flowi6_proto, 287 tmp_csum); 288 } 289 ip6_push_pending_frames(sk); 290 } 291 292 struct icmpv6_msg { 293 struct sk_buff *skb; 294 int offset; 295 uint8_t type; 296 }; 297 298 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb) 299 { 300 struct icmpv6_msg *msg = (struct icmpv6_msg *) from; 301 struct sk_buff *org_skb = msg->skb; 302 __wsum csum = 0; 303 304 csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset, 305 to, len, csum); 306 skb->csum = csum_block_add(skb->csum, csum, odd); 307 if (!(msg->type & ICMPV6_INFOMSG_MASK)) 308 nf_ct_attach(skb, org_skb); 309 return 0; 310 } 311 312 #if IS_ENABLED(CONFIG_IPV6_MIP6) 313 static void mip6_addr_swap(struct sk_buff *skb) 314 { 315 struct ipv6hdr *iph = ipv6_hdr(skb); 316 struct inet6_skb_parm *opt = IP6CB(skb); 317 struct ipv6_destopt_hao *hao; 318 struct in6_addr tmp; 319 int off; 320 321 if (opt->dsthao) { 322 off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO); 323 if (likely(off >= 0)) { 324 hao = (struct ipv6_destopt_hao *) 325 (skb_network_header(skb) + off); 326 tmp = iph->saddr; 327 iph->saddr = hao->addr; 328 hao->addr = tmp; 329 } 330 } 331 } 332 #else 333 static inline void mip6_addr_swap(struct sk_buff *skb) {} 334 #endif 335 336 static struct dst_entry *icmpv6_route_lookup(struct net *net, 337 struct sk_buff *skb, 338 struct sock *sk, 339 struct flowi6 *fl6) 340 { 341 struct dst_entry *dst, *dst2; 342 struct flowi6 fl2; 343 int err; 344 345 err = ip6_dst_lookup(net, sk, &dst, fl6); 346 if (err) 347 return ERR_PTR(err); 348 349 /* 350 * We won't send icmp if the destination is known 351 * anycast. 352 */ 353 if (ipv6_anycast_destination(dst, &fl6->daddr)) { 354 net_dbg_ratelimited("icmp6_send: acast source\n"); 355 dst_release(dst); 356 return ERR_PTR(-EINVAL); 357 } 358 359 /* No need to clone since we're just using its address. */ 360 dst2 = dst; 361 362 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0); 363 if (!IS_ERR(dst)) { 364 if (dst != dst2) 365 return dst; 366 } else { 367 if (PTR_ERR(dst) == -EPERM) 368 dst = NULL; 369 else 370 return dst; 371 } 372 373 err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6); 374 if (err) 375 goto relookup_failed; 376 377 err = ip6_dst_lookup(net, sk, &dst2, &fl2); 378 if (err) 379 goto relookup_failed; 380 381 dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP); 382 if (!IS_ERR(dst2)) { 383 dst_release(dst); 384 dst = dst2; 385 } else { 386 err = PTR_ERR(dst2); 387 if (err == -EPERM) { 388 dst_release(dst); 389 return dst2; 390 } else 391 goto relookup_failed; 392 } 393 394 relookup_failed: 395 if (dst) 396 return dst; 397 return ERR_PTR(err); 398 } 399 400 static int icmp6_iif(const struct sk_buff *skb) 401 { 402 int iif = skb->dev->ifindex; 403 404 /* for local traffic to local address, skb dev is the loopback 405 * device. Check if there is a dst attached to the skb and if so 406 * get the real device index. Same is needed for replies to a link 407 * local address on a device enslaved to an L3 master device 408 */ 409 if (unlikely(iif == LOOPBACK_IFINDEX || netif_is_l3_master(skb->dev))) { 410 const struct rt6_info *rt6 = skb_rt6_info(skb); 411 412 if (rt6) 413 iif = rt6->rt6i_idev->dev->ifindex; 414 } 415 416 return iif; 417 } 418 419 /* 420 * Send an ICMP message in response to a packet in error 421 */ 422 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info, 423 const struct in6_addr *force_saddr) 424 { 425 struct inet6_dev *idev = NULL; 426 struct ipv6hdr *hdr = ipv6_hdr(skb); 427 struct sock *sk; 428 struct net *net; 429 struct ipv6_pinfo *np; 430 const struct in6_addr *saddr = NULL; 431 struct dst_entry *dst; 432 struct icmp6hdr tmp_hdr; 433 struct flowi6 fl6; 434 struct icmpv6_msg msg; 435 struct ipcm6_cookie ipc6; 436 int iif = 0; 437 int addr_type = 0; 438 int len; 439 u32 mark; 440 441 if ((u8 *)hdr < skb->head || 442 (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb)) 443 return; 444 445 if (!skb->dev) 446 return; 447 net = dev_net(skb->dev); 448 mark = IP6_REPLY_MARK(net, skb->mark); 449 /* 450 * Make sure we respect the rules 451 * i.e. RFC 1885 2.4(e) 452 * Rule (e.1) is enforced by not using icmp6_send 453 * in any code that processes icmp errors. 454 */ 455 addr_type = ipv6_addr_type(&hdr->daddr); 456 457 if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) || 458 ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr)) 459 saddr = &hdr->daddr; 460 461 /* 462 * Dest addr check 463 */ 464 465 if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) { 466 if (type != ICMPV6_PKT_TOOBIG && 467 !(type == ICMPV6_PARAMPROB && 468 code == ICMPV6_UNK_OPTION && 469 (opt_unrec(skb, info)))) 470 return; 471 472 saddr = NULL; 473 } 474 475 addr_type = ipv6_addr_type(&hdr->saddr); 476 477 /* 478 * Source addr check 479 */ 480 481 if (__ipv6_addr_needs_scope_id(addr_type)) { 482 iif = icmp6_iif(skb); 483 } else { 484 dst = skb_dst(skb); 485 iif = l3mdev_master_ifindex(dst ? dst->dev : skb->dev); 486 } 487 488 /* 489 * Must not send error if the source does not uniquely 490 * identify a single node (RFC2463 Section 2.4). 491 * We check unspecified / multicast addresses here, 492 * and anycast addresses will be checked later. 493 */ 494 if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) { 495 net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n", 496 &hdr->saddr, &hdr->daddr); 497 return; 498 } 499 500 /* 501 * Never answer to a ICMP packet. 502 */ 503 if (is_ineligible(skb)) { 504 net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n", 505 &hdr->saddr, &hdr->daddr); 506 return; 507 } 508 509 /* Needed by both icmp_global_allow and icmpv6_xmit_lock */ 510 local_bh_disable(); 511 512 /* Check global sysctl_icmp_msgs_per_sec ratelimit */ 513 if (!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, type)) 514 goto out_bh_enable; 515 516 mip6_addr_swap(skb); 517 518 memset(&fl6, 0, sizeof(fl6)); 519 fl6.flowi6_proto = IPPROTO_ICMPV6; 520 fl6.daddr = hdr->saddr; 521 if (force_saddr) 522 saddr = force_saddr; 523 if (saddr) 524 fl6.saddr = *saddr; 525 fl6.flowi6_mark = mark; 526 fl6.flowi6_oif = iif; 527 fl6.fl6_icmp_type = type; 528 fl6.fl6_icmp_code = code; 529 fl6.flowi6_uid = sock_net_uid(net, NULL); 530 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, NULL); 531 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); 532 533 sk = icmpv6_xmit_lock(net); 534 if (!sk) 535 goto out_bh_enable; 536 537 sk->sk_mark = mark; 538 np = inet6_sk(sk); 539 540 if (!icmpv6_xrlim_allow(sk, type, &fl6)) 541 goto out; 542 543 tmp_hdr.icmp6_type = type; 544 tmp_hdr.icmp6_code = code; 545 tmp_hdr.icmp6_cksum = 0; 546 tmp_hdr.icmp6_pointer = htonl(info); 547 548 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 549 fl6.flowi6_oif = np->mcast_oif; 550 else if (!fl6.flowi6_oif) 551 fl6.flowi6_oif = np->ucast_oif; 552 553 ipcm6_init_sk(&ipc6, np); 554 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); 555 556 dst = icmpv6_route_lookup(net, skb, sk, &fl6); 557 if (IS_ERR(dst)) 558 goto out; 559 560 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 561 562 msg.skb = skb; 563 msg.offset = skb_network_offset(skb); 564 msg.type = type; 565 566 len = skb->len - msg.offset; 567 len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr)); 568 if (len < 0) { 569 net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n", 570 &hdr->saddr, &hdr->daddr); 571 goto out_dst_release; 572 } 573 574 rcu_read_lock(); 575 idev = __in6_dev_get(skb->dev); 576 577 if (ip6_append_data(sk, icmpv6_getfrag, &msg, 578 len + sizeof(struct icmp6hdr), 579 sizeof(struct icmp6hdr), 580 &ipc6, &fl6, (struct rt6_info *)dst, 581 MSG_DONTWAIT)) { 582 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS); 583 ip6_flush_pending_frames(sk); 584 } else { 585 icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, 586 len + sizeof(struct icmp6hdr)); 587 } 588 rcu_read_unlock(); 589 out_dst_release: 590 dst_release(dst); 591 out: 592 icmpv6_xmit_unlock(sk); 593 out_bh_enable: 594 local_bh_enable(); 595 } 596 597 /* Slightly more convenient version of icmp6_send. 598 */ 599 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos) 600 { 601 icmp6_send(skb, ICMPV6_PARAMPROB, code, pos, NULL); 602 kfree_skb(skb); 603 } 604 605 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH 606 * if sufficient data bytes are available 607 * @nhs is the size of the tunnel header(s) : 608 * Either an IPv4 header for SIT encap 609 * an IPv4 header + GRE header for GRE encap 610 */ 611 int ip6_err_gen_icmpv6_unreach(struct sk_buff *skb, int nhs, int type, 612 unsigned int data_len) 613 { 614 struct in6_addr temp_saddr; 615 struct rt6_info *rt; 616 struct sk_buff *skb2; 617 u32 info = 0; 618 619 if (!pskb_may_pull(skb, nhs + sizeof(struct ipv6hdr) + 8)) 620 return 1; 621 622 /* RFC 4884 (partial) support for ICMP extensions */ 623 if (data_len < 128 || (data_len & 7) || skb->len < data_len) 624 data_len = 0; 625 626 skb2 = data_len ? skb_copy(skb, GFP_ATOMIC) : skb_clone(skb, GFP_ATOMIC); 627 628 if (!skb2) 629 return 1; 630 631 skb_dst_drop(skb2); 632 skb_pull(skb2, nhs); 633 skb_reset_network_header(skb2); 634 635 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 636 skb, 0); 637 638 if (rt && rt->dst.dev) 639 skb2->dev = rt->dst.dev; 640 641 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr); 642 643 if (data_len) { 644 /* RFC 4884 (partial) support : 645 * insert 0 padding at the end, before the extensions 646 */ 647 __skb_push(skb2, nhs); 648 skb_reset_network_header(skb2); 649 memmove(skb2->data, skb2->data + nhs, data_len - nhs); 650 memset(skb2->data + data_len - nhs, 0, nhs); 651 /* RFC 4884 4.5 : Length is measured in 64-bit words, 652 * and stored in reserved[0] 653 */ 654 info = (data_len/8) << 24; 655 } 656 if (type == ICMP_TIME_EXCEEDED) 657 icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 658 info, &temp_saddr); 659 else 660 icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 661 info, &temp_saddr); 662 if (rt) 663 ip6_rt_put(rt); 664 665 kfree_skb(skb2); 666 667 return 0; 668 } 669 EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach); 670 671 static void icmpv6_echo_reply(struct sk_buff *skb) 672 { 673 struct net *net = dev_net(skb->dev); 674 struct sock *sk; 675 struct inet6_dev *idev; 676 struct ipv6_pinfo *np; 677 const struct in6_addr *saddr = NULL; 678 struct icmp6hdr *icmph = icmp6_hdr(skb); 679 struct icmp6hdr tmp_hdr; 680 struct flowi6 fl6; 681 struct icmpv6_msg msg; 682 struct dst_entry *dst; 683 struct ipcm6_cookie ipc6; 684 u32 mark = IP6_REPLY_MARK(net, skb->mark); 685 bool acast; 686 687 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) && 688 net->ipv6.sysctl.icmpv6_echo_ignore_multicast) 689 return; 690 691 saddr = &ipv6_hdr(skb)->daddr; 692 693 acast = ipv6_anycast_destination(skb_dst(skb), saddr); 694 if (acast && net->ipv6.sysctl.icmpv6_echo_ignore_anycast) 695 return; 696 697 if (!ipv6_unicast_destination(skb) && 698 !(net->ipv6.sysctl.anycast_src_echo_reply && acast)) 699 saddr = NULL; 700 701 memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr)); 702 tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY; 703 704 memset(&fl6, 0, sizeof(fl6)); 705 fl6.flowi6_proto = IPPROTO_ICMPV6; 706 fl6.daddr = ipv6_hdr(skb)->saddr; 707 if (saddr) 708 fl6.saddr = *saddr; 709 fl6.flowi6_oif = icmp6_iif(skb); 710 fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY; 711 fl6.flowi6_mark = mark; 712 fl6.flowi6_uid = sock_net_uid(net, NULL); 713 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); 714 715 local_bh_disable(); 716 sk = icmpv6_xmit_lock(net); 717 if (!sk) 718 goto out_bh_enable; 719 sk->sk_mark = mark; 720 np = inet6_sk(sk); 721 722 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 723 fl6.flowi6_oif = np->mcast_oif; 724 else if (!fl6.flowi6_oif) 725 fl6.flowi6_oif = np->ucast_oif; 726 727 if (ip6_dst_lookup(net, sk, &dst, &fl6)) 728 goto out; 729 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0); 730 if (IS_ERR(dst)) 731 goto out; 732 733 /* Check the ratelimit */ 734 if ((!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, ICMPV6_ECHO_REPLY)) || 735 !icmpv6_xrlim_allow(sk, ICMPV6_ECHO_REPLY, &fl6)) 736 goto out_dst_release; 737 738 idev = __in6_dev_get(skb->dev); 739 740 msg.skb = skb; 741 msg.offset = 0; 742 msg.type = ICMPV6_ECHO_REPLY; 743 744 ipcm6_init_sk(&ipc6, np); 745 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 746 ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb)); 747 748 if (ip6_append_data(sk, icmpv6_getfrag, &msg, 749 skb->len + sizeof(struct icmp6hdr), 750 sizeof(struct icmp6hdr), &ipc6, &fl6, 751 (struct rt6_info *)dst, MSG_DONTWAIT)) { 752 __ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS); 753 ip6_flush_pending_frames(sk); 754 } else { 755 icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, 756 skb->len + sizeof(struct icmp6hdr)); 757 } 758 out_dst_release: 759 dst_release(dst); 760 out: 761 icmpv6_xmit_unlock(sk); 762 out_bh_enable: 763 local_bh_enable(); 764 } 765 766 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info) 767 { 768 const struct inet6_protocol *ipprot; 769 int inner_offset; 770 __be16 frag_off; 771 u8 nexthdr; 772 struct net *net = dev_net(skb->dev); 773 774 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 775 goto out; 776 777 nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr; 778 if (ipv6_ext_hdr(nexthdr)) { 779 /* now skip over extension headers */ 780 inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), 781 &nexthdr, &frag_off); 782 if (inner_offset < 0) 783 goto out; 784 } else { 785 inner_offset = sizeof(struct ipv6hdr); 786 } 787 788 /* Checkin header including 8 bytes of inner protocol header. */ 789 if (!pskb_may_pull(skb, inner_offset+8)) 790 goto out; 791 792 /* BUGGG_FUTURE: we should try to parse exthdrs in this packet. 793 Without this we will not able f.e. to make source routed 794 pmtu discovery. 795 Corresponding argument (opt) to notifiers is already added. 796 --ANK (980726) 797 */ 798 799 ipprot = rcu_dereference(inet6_protos[nexthdr]); 800 if (ipprot && ipprot->err_handler) 801 ipprot->err_handler(skb, NULL, type, code, inner_offset, info); 802 803 raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info); 804 return; 805 806 out: 807 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS); 808 } 809 810 /* 811 * Handle icmp messages 812 */ 813 814 static int icmpv6_rcv(struct sk_buff *skb) 815 { 816 struct net *net = dev_net(skb->dev); 817 struct net_device *dev = skb->dev; 818 struct inet6_dev *idev = __in6_dev_get(dev); 819 const struct in6_addr *saddr, *daddr; 820 struct icmp6hdr *hdr; 821 u8 type; 822 bool success = false; 823 824 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 825 struct sec_path *sp = skb_sec_path(skb); 826 int nh; 827 828 if (!(sp && sp->xvec[sp->len - 1]->props.flags & 829 XFRM_STATE_ICMP)) 830 goto drop_no_count; 831 832 if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr))) 833 goto drop_no_count; 834 835 nh = skb_network_offset(skb); 836 skb_set_network_header(skb, sizeof(*hdr)); 837 838 if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) 839 goto drop_no_count; 840 841 skb_set_network_header(skb, nh); 842 } 843 844 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS); 845 846 saddr = &ipv6_hdr(skb)->saddr; 847 daddr = &ipv6_hdr(skb)->daddr; 848 849 if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) { 850 net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n", 851 saddr, daddr); 852 goto csum_error; 853 } 854 855 if (!pskb_pull(skb, sizeof(*hdr))) 856 goto discard_it; 857 858 hdr = icmp6_hdr(skb); 859 860 type = hdr->icmp6_type; 861 862 ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type); 863 864 switch (type) { 865 case ICMPV6_ECHO_REQUEST: 866 if (!net->ipv6.sysctl.icmpv6_echo_ignore_all) 867 icmpv6_echo_reply(skb); 868 break; 869 870 case ICMPV6_ECHO_REPLY: 871 success = ping_rcv(skb); 872 break; 873 874 case ICMPV6_PKT_TOOBIG: 875 /* BUGGG_FUTURE: if packet contains rthdr, we cannot update 876 standard destination cache. Seems, only "advanced" 877 destination cache will allow to solve this problem 878 --ANK (980726) 879 */ 880 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 881 goto discard_it; 882 hdr = icmp6_hdr(skb); 883 884 /* to notify */ 885 /* fall through */ 886 case ICMPV6_DEST_UNREACH: 887 case ICMPV6_TIME_EXCEED: 888 case ICMPV6_PARAMPROB: 889 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); 890 break; 891 892 case NDISC_ROUTER_SOLICITATION: 893 case NDISC_ROUTER_ADVERTISEMENT: 894 case NDISC_NEIGHBOUR_SOLICITATION: 895 case NDISC_NEIGHBOUR_ADVERTISEMENT: 896 case NDISC_REDIRECT: 897 ndisc_rcv(skb); 898 break; 899 900 case ICMPV6_MGM_QUERY: 901 igmp6_event_query(skb); 902 break; 903 904 case ICMPV6_MGM_REPORT: 905 igmp6_event_report(skb); 906 break; 907 908 case ICMPV6_MGM_REDUCTION: 909 case ICMPV6_NI_QUERY: 910 case ICMPV6_NI_REPLY: 911 case ICMPV6_MLD2_REPORT: 912 case ICMPV6_DHAAD_REQUEST: 913 case ICMPV6_DHAAD_REPLY: 914 case ICMPV6_MOBILE_PREFIX_SOL: 915 case ICMPV6_MOBILE_PREFIX_ADV: 916 break; 917 918 default: 919 /* informational */ 920 if (type & ICMPV6_INFOMSG_MASK) 921 break; 922 923 net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n", 924 saddr, daddr); 925 926 /* 927 * error of unknown type. 928 * must pass to upper level 929 */ 930 931 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); 932 } 933 934 /* until the v6 path can be better sorted assume failure and 935 * preserve the status quo behaviour for the rest of the paths to here 936 */ 937 if (success) 938 consume_skb(skb); 939 else 940 kfree_skb(skb); 941 942 return 0; 943 944 csum_error: 945 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS); 946 discard_it: 947 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS); 948 drop_no_count: 949 kfree_skb(skb); 950 return 0; 951 } 952 953 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6, 954 u8 type, 955 const struct in6_addr *saddr, 956 const struct in6_addr *daddr, 957 int oif) 958 { 959 memset(fl6, 0, sizeof(*fl6)); 960 fl6->saddr = *saddr; 961 fl6->daddr = *daddr; 962 fl6->flowi6_proto = IPPROTO_ICMPV6; 963 fl6->fl6_icmp_type = type; 964 fl6->fl6_icmp_code = 0; 965 fl6->flowi6_oif = oif; 966 security_sk_classify_flow(sk, flowi6_to_flowi(fl6)); 967 } 968 969 static void __net_exit icmpv6_sk_exit(struct net *net) 970 { 971 int i; 972 973 for_each_possible_cpu(i) 974 inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv6.icmp_sk, i)); 975 free_percpu(net->ipv6.icmp_sk); 976 } 977 978 static int __net_init icmpv6_sk_init(struct net *net) 979 { 980 struct sock *sk; 981 int err, i; 982 983 net->ipv6.icmp_sk = alloc_percpu(struct sock *); 984 if (!net->ipv6.icmp_sk) 985 return -ENOMEM; 986 987 for_each_possible_cpu(i) { 988 err = inet_ctl_sock_create(&sk, PF_INET6, 989 SOCK_RAW, IPPROTO_ICMPV6, net); 990 if (err < 0) { 991 pr_err("Failed to initialize the ICMP6 control socket (err %d)\n", 992 err); 993 goto fail; 994 } 995 996 *per_cpu_ptr(net->ipv6.icmp_sk, i) = sk; 997 998 /* Enough space for 2 64K ICMP packets, including 999 * sk_buff struct overhead. 1000 */ 1001 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); 1002 } 1003 return 0; 1004 1005 fail: 1006 icmpv6_sk_exit(net); 1007 return err; 1008 } 1009 1010 static struct pernet_operations icmpv6_sk_ops = { 1011 .init = icmpv6_sk_init, 1012 .exit = icmpv6_sk_exit, 1013 }; 1014 1015 int __init icmpv6_init(void) 1016 { 1017 int err; 1018 1019 err = register_pernet_subsys(&icmpv6_sk_ops); 1020 if (err < 0) 1021 return err; 1022 1023 err = -EAGAIN; 1024 if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0) 1025 goto fail; 1026 1027 err = inet6_register_icmp_sender(icmp6_send); 1028 if (err) 1029 goto sender_reg_err; 1030 return 0; 1031 1032 sender_reg_err: 1033 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); 1034 fail: 1035 pr_err("Failed to register ICMP6 protocol\n"); 1036 unregister_pernet_subsys(&icmpv6_sk_ops); 1037 return err; 1038 } 1039 1040 void icmpv6_cleanup(void) 1041 { 1042 inet6_unregister_icmp_sender(icmp6_send); 1043 unregister_pernet_subsys(&icmpv6_sk_ops); 1044 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); 1045 } 1046 1047 1048 static const struct icmp6_err { 1049 int err; 1050 int fatal; 1051 } tab_unreach[] = { 1052 { /* NOROUTE */ 1053 .err = ENETUNREACH, 1054 .fatal = 0, 1055 }, 1056 { /* ADM_PROHIBITED */ 1057 .err = EACCES, 1058 .fatal = 1, 1059 }, 1060 { /* Was NOT_NEIGHBOUR, now reserved */ 1061 .err = EHOSTUNREACH, 1062 .fatal = 0, 1063 }, 1064 { /* ADDR_UNREACH */ 1065 .err = EHOSTUNREACH, 1066 .fatal = 0, 1067 }, 1068 { /* PORT_UNREACH */ 1069 .err = ECONNREFUSED, 1070 .fatal = 1, 1071 }, 1072 { /* POLICY_FAIL */ 1073 .err = EACCES, 1074 .fatal = 1, 1075 }, 1076 { /* REJECT_ROUTE */ 1077 .err = EACCES, 1078 .fatal = 1, 1079 }, 1080 }; 1081 1082 int icmpv6_err_convert(u8 type, u8 code, int *err) 1083 { 1084 int fatal = 0; 1085 1086 *err = EPROTO; 1087 1088 switch (type) { 1089 case ICMPV6_DEST_UNREACH: 1090 fatal = 1; 1091 if (code < ARRAY_SIZE(tab_unreach)) { 1092 *err = tab_unreach[code].err; 1093 fatal = tab_unreach[code].fatal; 1094 } 1095 break; 1096 1097 case ICMPV6_PKT_TOOBIG: 1098 *err = EMSGSIZE; 1099 break; 1100 1101 case ICMPV6_PARAMPROB: 1102 *err = EPROTO; 1103 fatal = 1; 1104 break; 1105 1106 case ICMPV6_TIME_EXCEED: 1107 *err = EHOSTUNREACH; 1108 break; 1109 } 1110 1111 return fatal; 1112 } 1113 EXPORT_SYMBOL(icmpv6_err_convert); 1114 1115 #ifdef CONFIG_SYSCTL 1116 static struct ctl_table ipv6_icmp_table_template[] = { 1117 { 1118 .procname = "ratelimit", 1119 .data = &init_net.ipv6.sysctl.icmpv6_time, 1120 .maxlen = sizeof(int), 1121 .mode = 0644, 1122 .proc_handler = proc_dointvec_ms_jiffies, 1123 }, 1124 { 1125 .procname = "echo_ignore_all", 1126 .data = &init_net.ipv6.sysctl.icmpv6_echo_ignore_all, 1127 .maxlen = sizeof(int), 1128 .mode = 0644, 1129 .proc_handler = proc_dointvec, 1130 }, 1131 { 1132 .procname = "echo_ignore_multicast", 1133 .data = &init_net.ipv6.sysctl.icmpv6_echo_ignore_multicast, 1134 .maxlen = sizeof(int), 1135 .mode = 0644, 1136 .proc_handler = proc_dointvec, 1137 }, 1138 { 1139 .procname = "echo_ignore_anycast", 1140 .data = &init_net.ipv6.sysctl.icmpv6_echo_ignore_anycast, 1141 .maxlen = sizeof(int), 1142 .mode = 0644, 1143 .proc_handler = proc_dointvec, 1144 }, 1145 { 1146 .procname = "ratemask", 1147 .data = &init_net.ipv6.sysctl.icmpv6_ratemask_ptr, 1148 .maxlen = ICMPV6_MSG_MAX + 1, 1149 .mode = 0644, 1150 .proc_handler = proc_do_large_bitmap, 1151 }, 1152 { }, 1153 }; 1154 1155 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net) 1156 { 1157 struct ctl_table *table; 1158 1159 table = kmemdup(ipv6_icmp_table_template, 1160 sizeof(ipv6_icmp_table_template), 1161 GFP_KERNEL); 1162 1163 if (table) { 1164 table[0].data = &net->ipv6.sysctl.icmpv6_time; 1165 table[1].data = &net->ipv6.sysctl.icmpv6_echo_ignore_all; 1166 table[2].data = &net->ipv6.sysctl.icmpv6_echo_ignore_multicast; 1167 table[3].data = &net->ipv6.sysctl.icmpv6_echo_ignore_anycast; 1168 table[4].data = &net->ipv6.sysctl.icmpv6_ratemask_ptr; 1169 } 1170 return table; 1171 } 1172 #endif 1173