1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Internet Control Message Protocol (ICMPv6) 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * 9 * Based on net/ipv4/icmp.c 10 * 11 * RFC 1885 12 */ 13 14 /* 15 * Changes: 16 * 17 * Andi Kleen : exception handling 18 * Andi Kleen add rate limits. never reply to a icmp. 19 * add more length checks and other fixes. 20 * yoshfuji : ensure to sent parameter problem for 21 * fragments. 22 * YOSHIFUJI Hideaki @USAGI: added sysctl for icmp rate limit. 23 * Randy Dunlap and 24 * YOSHIFUJI Hideaki @USAGI: Per-interface statistics support 25 * Kazunori MIYAZAWA @USAGI: change output process to use ip6_append_data 26 */ 27 28 #define pr_fmt(fmt) "IPv6: " fmt 29 30 #include <linux/module.h> 31 #include <linux/errno.h> 32 #include <linux/types.h> 33 #include <linux/socket.h> 34 #include <linux/in.h> 35 #include <linux/kernel.h> 36 #include <linux/sockios.h> 37 #include <linux/net.h> 38 #include <linux/skbuff.h> 39 #include <linux/init.h> 40 #include <linux/netfilter.h> 41 #include <linux/slab.h> 42 43 #ifdef CONFIG_SYSCTL 44 #include <linux/sysctl.h> 45 #endif 46 47 #include <linux/inet.h> 48 #include <linux/netdevice.h> 49 #include <linux/icmpv6.h> 50 51 #include <net/ip.h> 52 #include <net/sock.h> 53 54 #include <net/ipv6.h> 55 #include <net/ip6_checksum.h> 56 #include <net/ping.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 #include <net/dsfield.h> 67 #include <net/l3mdev.h> 68 69 #include <linux/uaccess.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 struct sock *icmpv6_sk(struct net *net) 79 { 80 return this_cpu_read(*net->ipv6.icmp_sk); 81 } 82 83 static int icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 84 u8 type, u8 code, int offset, __be32 info) 85 { 86 /* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */ 87 struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset); 88 struct net *net = dev_net(skb->dev); 89 90 if (type == ICMPV6_PKT_TOOBIG) 91 ip6_update_pmtu(skb, net, info, skb->dev->ifindex, 0, sock_net_uid(net, NULL)); 92 else if (type == NDISC_REDIRECT) 93 ip6_redirect(skb, net, skb->dev->ifindex, 0, 94 sock_net_uid(net, NULL)); 95 96 if (!(type & ICMPV6_INFOMSG_MASK)) 97 if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST) 98 ping_err(skb, offset, ntohl(info)); 99 100 return 0; 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 /* Called with BH disabled */ 112 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net) 113 { 114 struct sock *sk; 115 116 sk = icmpv6_sk(net); 117 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) { 118 /* This can happen if the output path (f.e. SIT or 119 * ip6ip6 tunnel) signals dst_link_failure() for an 120 * outgoing ICMP6 packet. 121 */ 122 return NULL; 123 } 124 return sk; 125 } 126 127 static __inline__ void icmpv6_xmit_unlock(struct sock *sk) 128 { 129 spin_unlock(&sk->sk_lock.slock); 130 } 131 132 /* 133 * Figure out, may we reply to this packet with icmp error. 134 * 135 * We do not reply, if: 136 * - it was icmp error message. 137 * - it is truncated, so that it is known, that protocol is ICMPV6 138 * (i.e. in the middle of some exthdr) 139 * 140 * --ANK (980726) 141 */ 142 143 static bool is_ineligible(const struct sk_buff *skb) 144 { 145 int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data; 146 int len = skb->len - ptr; 147 __u8 nexthdr = ipv6_hdr(skb)->nexthdr; 148 __be16 frag_off; 149 150 if (len < 0) 151 return true; 152 153 ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off); 154 if (ptr < 0) 155 return false; 156 if (nexthdr == IPPROTO_ICMPV6) { 157 u8 _type, *tp; 158 tp = skb_header_pointer(skb, 159 ptr+offsetof(struct icmp6hdr, icmp6_type), 160 sizeof(_type), &_type); 161 if (!tp || !(*tp & ICMPV6_INFOMSG_MASK)) 162 return true; 163 } 164 return false; 165 } 166 167 static bool icmpv6_mask_allow(struct net *net, int type) 168 { 169 if (type > ICMPV6_MSG_MAX) 170 return true; 171 172 /* Limit if icmp type is set in ratemask. */ 173 if (!test_bit(type, net->ipv6.sysctl.icmpv6_ratemask)) 174 return true; 175 176 return false; 177 } 178 179 static bool icmpv6_global_allow(struct net *net, int type) 180 { 181 if (icmpv6_mask_allow(net, type)) 182 return true; 183 184 if (icmp_global_allow()) 185 return true; 186 187 return false; 188 } 189 190 /* 191 * Check the ICMP output rate limit 192 */ 193 static bool icmpv6_xrlim_allow(struct sock *sk, u8 type, 194 struct flowi6 *fl6) 195 { 196 struct net *net = sock_net(sk); 197 struct dst_entry *dst; 198 bool res = false; 199 200 if (icmpv6_mask_allow(net, type)) 201 return true; 202 203 /* 204 * Look up the output route. 205 * XXX: perhaps the expire for routing entries cloned by 206 * this lookup should be more aggressive (not longer than timeout). 207 */ 208 dst = ip6_route_output(net, sk, fl6); 209 if (dst->error) { 210 IP6_INC_STATS(net, ip6_dst_idev(dst), 211 IPSTATS_MIB_OUTNOROUTES); 212 } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) { 213 res = true; 214 } else { 215 struct rt6_info *rt = (struct rt6_info *)dst; 216 int tmo = net->ipv6.sysctl.icmpv6_time; 217 struct inet_peer *peer; 218 219 /* Give more bandwidth to wider prefixes. */ 220 if (rt->rt6i_dst.plen < 128) 221 tmo >>= ((128 - rt->rt6i_dst.plen)>>5); 222 223 peer = inet_getpeer_v6(net->ipv6.peers, &fl6->daddr, 1); 224 res = inet_peer_xrlim_allow(peer, tmo); 225 if (peer) 226 inet_putpeer(peer); 227 } 228 dst_release(dst); 229 return res; 230 } 231 232 /* 233 * an inline helper for the "simple" if statement below 234 * checks if parameter problem report is caused by an 235 * unrecognized IPv6 option that has the Option Type 236 * highest-order two bits set to 10 237 */ 238 239 static bool opt_unrec(struct sk_buff *skb, __u32 offset) 240 { 241 u8 _optval, *op; 242 243 offset += skb_network_offset(skb); 244 op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval); 245 if (!op) 246 return true; 247 return (*op & 0xC0) == 0x80; 248 } 249 250 void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 251 struct icmp6hdr *thdr, int len) 252 { 253 struct sk_buff *skb; 254 struct icmp6hdr *icmp6h; 255 256 skb = skb_peek(&sk->sk_write_queue); 257 if (!skb) 258 return; 259 260 icmp6h = icmp6_hdr(skb); 261 memcpy(icmp6h, thdr, sizeof(struct icmp6hdr)); 262 icmp6h->icmp6_cksum = 0; 263 264 if (skb_queue_len(&sk->sk_write_queue) == 1) { 265 skb->csum = csum_partial(icmp6h, 266 sizeof(struct icmp6hdr), skb->csum); 267 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr, 268 &fl6->daddr, 269 len, fl6->flowi6_proto, 270 skb->csum); 271 } else { 272 __wsum tmp_csum = 0; 273 274 skb_queue_walk(&sk->sk_write_queue, skb) { 275 tmp_csum = csum_add(tmp_csum, skb->csum); 276 } 277 278 tmp_csum = csum_partial(icmp6h, 279 sizeof(struct icmp6hdr), tmp_csum); 280 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr, 281 &fl6->daddr, 282 len, fl6->flowi6_proto, 283 tmp_csum); 284 } 285 ip6_push_pending_frames(sk); 286 } 287 288 struct icmpv6_msg { 289 struct sk_buff *skb; 290 int offset; 291 uint8_t type; 292 }; 293 294 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb) 295 { 296 struct icmpv6_msg *msg = (struct icmpv6_msg *) from; 297 struct sk_buff *org_skb = msg->skb; 298 __wsum csum = 0; 299 300 csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset, 301 to, len, csum); 302 skb->csum = csum_block_add(skb->csum, csum, odd); 303 if (!(msg->type & ICMPV6_INFOMSG_MASK)) 304 nf_ct_attach(skb, org_skb); 305 return 0; 306 } 307 308 #if IS_ENABLED(CONFIG_IPV6_MIP6) 309 static void mip6_addr_swap(struct sk_buff *skb) 310 { 311 struct ipv6hdr *iph = ipv6_hdr(skb); 312 struct inet6_skb_parm *opt = IP6CB(skb); 313 struct ipv6_destopt_hao *hao; 314 struct in6_addr tmp; 315 int off; 316 317 if (opt->dsthao) { 318 off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO); 319 if (likely(off >= 0)) { 320 hao = (struct ipv6_destopt_hao *) 321 (skb_network_header(skb) + off); 322 tmp = iph->saddr; 323 iph->saddr = hao->addr; 324 hao->addr = tmp; 325 } 326 } 327 } 328 #else 329 static inline void mip6_addr_swap(struct sk_buff *skb) {} 330 #endif 331 332 static struct dst_entry *icmpv6_route_lookup(struct net *net, 333 struct sk_buff *skb, 334 struct sock *sk, 335 struct flowi6 *fl6) 336 { 337 struct dst_entry *dst, *dst2; 338 struct flowi6 fl2; 339 int err; 340 341 err = ip6_dst_lookup(net, sk, &dst, fl6); 342 if (err) 343 return ERR_PTR(err); 344 345 /* 346 * We won't send icmp if the destination is known 347 * anycast. 348 */ 349 if (ipv6_anycast_destination(dst, &fl6->daddr)) { 350 net_dbg_ratelimited("icmp6_send: acast source\n"); 351 dst_release(dst); 352 return ERR_PTR(-EINVAL); 353 } 354 355 /* No need to clone since we're just using its address. */ 356 dst2 = dst; 357 358 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0); 359 if (!IS_ERR(dst)) { 360 if (dst != dst2) 361 return dst; 362 } else { 363 if (PTR_ERR(dst) == -EPERM) 364 dst = NULL; 365 else 366 return dst; 367 } 368 369 err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6); 370 if (err) 371 goto relookup_failed; 372 373 err = ip6_dst_lookup(net, sk, &dst2, &fl2); 374 if (err) 375 goto relookup_failed; 376 377 dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP); 378 if (!IS_ERR(dst2)) { 379 dst_release(dst); 380 dst = dst2; 381 } else { 382 err = PTR_ERR(dst2); 383 if (err == -EPERM) { 384 dst_release(dst); 385 return dst2; 386 } else 387 goto relookup_failed; 388 } 389 390 relookup_failed: 391 if (dst) 392 return dst; 393 return ERR_PTR(err); 394 } 395 396 static struct net_device *icmp6_dev(const struct sk_buff *skb) 397 { 398 struct net_device *dev = skb->dev; 399 400 /* for local traffic to local address, skb dev is the loopback 401 * device. Check if there is a dst attached to the skb and if so 402 * get the real device index. Same is needed for replies to a link 403 * local address on a device enslaved to an L3 master device 404 */ 405 if (unlikely(dev->ifindex == LOOPBACK_IFINDEX || netif_is_l3_master(skb->dev))) { 406 const struct rt6_info *rt6 = skb_rt6_info(skb); 407 408 if (rt6) 409 dev = rt6->rt6i_idev->dev; 410 } 411 412 return dev; 413 } 414 415 static int icmp6_iif(const struct sk_buff *skb) 416 { 417 return icmp6_dev(skb)->ifindex; 418 } 419 420 /* 421 * Send an ICMP message in response to a packet in error 422 */ 423 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info, 424 const struct in6_addr *force_saddr) 425 { 426 struct inet6_dev *idev = NULL; 427 struct ipv6hdr *hdr = ipv6_hdr(skb); 428 struct sock *sk; 429 struct net *net; 430 struct ipv6_pinfo *np; 431 const struct in6_addr *saddr = NULL; 432 struct dst_entry *dst; 433 struct icmp6hdr tmp_hdr; 434 struct flowi6 fl6; 435 struct icmpv6_msg msg; 436 struct ipcm6_cookie ipc6; 437 int iif = 0; 438 int addr_type = 0; 439 int len; 440 u32 mark; 441 442 if ((u8 *)hdr < skb->head || 443 (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb)) 444 return; 445 446 if (!skb->dev) 447 return; 448 net = dev_net(skb->dev); 449 mark = IP6_REPLY_MARK(net, skb->mark); 450 /* 451 * Make sure we respect the rules 452 * i.e. RFC 1885 2.4(e) 453 * Rule (e.1) is enforced by not using icmp6_send 454 * in any code that processes icmp errors. 455 */ 456 addr_type = ipv6_addr_type(&hdr->daddr); 457 458 if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) || 459 ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr)) 460 saddr = &hdr->daddr; 461 462 /* 463 * Dest addr check 464 */ 465 466 if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) { 467 if (type != ICMPV6_PKT_TOOBIG && 468 !(type == ICMPV6_PARAMPROB && 469 code == ICMPV6_UNK_OPTION && 470 (opt_unrec(skb, info)))) 471 return; 472 473 saddr = NULL; 474 } 475 476 addr_type = ipv6_addr_type(&hdr->saddr); 477 478 /* 479 * Source addr check 480 */ 481 482 if (__ipv6_addr_needs_scope_id(addr_type)) { 483 iif = icmp6_iif(skb); 484 } else { 485 dst = skb_dst(skb); 486 iif = l3mdev_master_ifindex(dst ? dst->dev : skb->dev); 487 } 488 489 /* 490 * Must not send error if the source does not uniquely 491 * identify a single node (RFC2463 Section 2.4). 492 * We check unspecified / multicast addresses here, 493 * and anycast addresses will be checked later. 494 */ 495 if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) { 496 net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n", 497 &hdr->saddr, &hdr->daddr); 498 return; 499 } 500 501 /* 502 * Never answer to a ICMP packet. 503 */ 504 if (is_ineligible(skb)) { 505 net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n", 506 &hdr->saddr, &hdr->daddr); 507 return; 508 } 509 510 /* Needed by both icmp_global_allow and icmpv6_xmit_lock */ 511 local_bh_disable(); 512 513 /* Check global sysctl_icmp_msgs_per_sec ratelimit */ 514 if (!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, type)) 515 goto out_bh_enable; 516 517 mip6_addr_swap(skb); 518 519 memset(&fl6, 0, sizeof(fl6)); 520 fl6.flowi6_proto = IPPROTO_ICMPV6; 521 fl6.daddr = hdr->saddr; 522 if (force_saddr) 523 saddr = force_saddr; 524 if (saddr) 525 fl6.saddr = *saddr; 526 fl6.flowi6_mark = mark; 527 fl6.flowi6_oif = iif; 528 fl6.fl6_icmp_type = type; 529 fl6.fl6_icmp_code = code; 530 fl6.flowi6_uid = sock_net_uid(net, NULL); 531 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, NULL); 532 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); 533 534 sk = icmpv6_xmit_lock(net); 535 if (!sk) 536 goto out_bh_enable; 537 538 sk->sk_mark = mark; 539 np = inet6_sk(sk); 540 541 if (!icmpv6_xrlim_allow(sk, type, &fl6)) 542 goto out; 543 544 tmp_hdr.icmp6_type = type; 545 tmp_hdr.icmp6_code = code; 546 tmp_hdr.icmp6_cksum = 0; 547 tmp_hdr.icmp6_pointer = htonl(info); 548 549 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 550 fl6.flowi6_oif = np->mcast_oif; 551 else if (!fl6.flowi6_oif) 552 fl6.flowi6_oif = np->ucast_oif; 553 554 ipcm6_init_sk(&ipc6, np); 555 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); 556 557 dst = icmpv6_route_lookup(net, skb, sk, &fl6); 558 if (IS_ERR(dst)) 559 goto out; 560 561 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 562 563 msg.skb = skb; 564 msg.offset = skb_network_offset(skb); 565 msg.type = type; 566 567 len = skb->len - msg.offset; 568 len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr)); 569 if (len < 0) { 570 net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n", 571 &hdr->saddr, &hdr->daddr); 572 goto out_dst_release; 573 } 574 575 rcu_read_lock(); 576 idev = __in6_dev_get(skb->dev); 577 578 if (ip6_append_data(sk, icmpv6_getfrag, &msg, 579 len + sizeof(struct icmp6hdr), 580 sizeof(struct icmp6hdr), 581 &ipc6, &fl6, (struct rt6_info *)dst, 582 MSG_DONTWAIT)) { 583 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS); 584 ip6_flush_pending_frames(sk); 585 } else { 586 icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, 587 len + sizeof(struct icmp6hdr)); 588 } 589 rcu_read_unlock(); 590 out_dst_release: 591 dst_release(dst); 592 out: 593 icmpv6_xmit_unlock(sk); 594 out_bh_enable: 595 local_bh_enable(); 596 } 597 598 /* Slightly more convenient version of icmp6_send. 599 */ 600 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos) 601 { 602 icmp6_send(skb, ICMPV6_PARAMPROB, code, pos, NULL); 603 kfree_skb(skb); 604 } 605 606 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH 607 * if sufficient data bytes are available 608 * @nhs is the size of the tunnel header(s) : 609 * Either an IPv4 header for SIT encap 610 * an IPv4 header + GRE header for GRE encap 611 */ 612 int ip6_err_gen_icmpv6_unreach(struct sk_buff *skb, int nhs, int type, 613 unsigned int data_len) 614 { 615 struct in6_addr temp_saddr; 616 struct rt6_info *rt; 617 struct sk_buff *skb2; 618 u32 info = 0; 619 620 if (!pskb_may_pull(skb, nhs + sizeof(struct ipv6hdr) + 8)) 621 return 1; 622 623 /* RFC 4884 (partial) support for ICMP extensions */ 624 if (data_len < 128 || (data_len & 7) || skb->len < data_len) 625 data_len = 0; 626 627 skb2 = data_len ? skb_copy(skb, GFP_ATOMIC) : skb_clone(skb, GFP_ATOMIC); 628 629 if (!skb2) 630 return 1; 631 632 skb_dst_drop(skb2); 633 skb_pull(skb2, nhs); 634 skb_reset_network_header(skb2); 635 636 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 637 skb, 0); 638 639 if (rt && rt->dst.dev) 640 skb2->dev = rt->dst.dev; 641 642 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr); 643 644 if (data_len) { 645 /* RFC 4884 (partial) support : 646 * insert 0 padding at the end, before the extensions 647 */ 648 __skb_push(skb2, nhs); 649 skb_reset_network_header(skb2); 650 memmove(skb2->data, skb2->data + nhs, data_len - nhs); 651 memset(skb2->data + data_len - nhs, 0, nhs); 652 /* RFC 4884 4.5 : Length is measured in 64-bit words, 653 * and stored in reserved[0] 654 */ 655 info = (data_len/8) << 24; 656 } 657 if (type == ICMP_TIME_EXCEEDED) 658 icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 659 info, &temp_saddr); 660 else 661 icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 662 info, &temp_saddr); 663 if (rt) 664 ip6_rt_put(rt); 665 666 kfree_skb(skb2); 667 668 return 0; 669 } 670 EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach); 671 672 static void icmpv6_echo_reply(struct sk_buff *skb) 673 { 674 struct net *net = dev_net(skb->dev); 675 struct sock *sk; 676 struct inet6_dev *idev; 677 struct ipv6_pinfo *np; 678 const struct in6_addr *saddr = NULL; 679 struct icmp6hdr *icmph = icmp6_hdr(skb); 680 struct icmp6hdr tmp_hdr; 681 struct flowi6 fl6; 682 struct icmpv6_msg msg; 683 struct dst_entry *dst; 684 struct ipcm6_cookie ipc6; 685 u32 mark = IP6_REPLY_MARK(net, skb->mark); 686 bool acast; 687 688 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) && 689 net->ipv6.sysctl.icmpv6_echo_ignore_multicast) 690 return; 691 692 saddr = &ipv6_hdr(skb)->daddr; 693 694 acast = ipv6_anycast_destination(skb_dst(skb), saddr); 695 if (acast && net->ipv6.sysctl.icmpv6_echo_ignore_anycast) 696 return; 697 698 if (!ipv6_unicast_destination(skb) && 699 !(net->ipv6.sysctl.anycast_src_echo_reply && acast)) 700 saddr = NULL; 701 702 memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr)); 703 tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY; 704 705 memset(&fl6, 0, sizeof(fl6)); 706 fl6.flowi6_proto = IPPROTO_ICMPV6; 707 fl6.daddr = ipv6_hdr(skb)->saddr; 708 if (saddr) 709 fl6.saddr = *saddr; 710 fl6.flowi6_oif = icmp6_iif(skb); 711 fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY; 712 fl6.flowi6_mark = mark; 713 fl6.flowi6_uid = sock_net_uid(net, NULL); 714 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); 715 716 local_bh_disable(); 717 sk = icmpv6_xmit_lock(net); 718 if (!sk) 719 goto out_bh_enable; 720 sk->sk_mark = mark; 721 np = inet6_sk(sk); 722 723 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 724 fl6.flowi6_oif = np->mcast_oif; 725 else if (!fl6.flowi6_oif) 726 fl6.flowi6_oif = np->ucast_oif; 727 728 if (ip6_dst_lookup(net, sk, &dst, &fl6)) 729 goto out; 730 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0); 731 if (IS_ERR(dst)) 732 goto out; 733 734 /* Check the ratelimit */ 735 if ((!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, ICMPV6_ECHO_REPLY)) || 736 !icmpv6_xrlim_allow(sk, ICMPV6_ECHO_REPLY, &fl6)) 737 goto out_dst_release; 738 739 idev = __in6_dev_get(skb->dev); 740 741 msg.skb = skb; 742 msg.offset = 0; 743 msg.type = ICMPV6_ECHO_REPLY; 744 745 ipcm6_init_sk(&ipc6, np); 746 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 747 ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb)); 748 749 if (ip6_append_data(sk, icmpv6_getfrag, &msg, 750 skb->len + sizeof(struct icmp6hdr), 751 sizeof(struct icmp6hdr), &ipc6, &fl6, 752 (struct rt6_info *)dst, MSG_DONTWAIT)) { 753 __ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS); 754 ip6_flush_pending_frames(sk); 755 } else { 756 icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, 757 skb->len + sizeof(struct icmp6hdr)); 758 } 759 out_dst_release: 760 dst_release(dst); 761 out: 762 icmpv6_xmit_unlock(sk); 763 out_bh_enable: 764 local_bh_enable(); 765 } 766 767 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info) 768 { 769 const struct inet6_protocol *ipprot; 770 int inner_offset; 771 __be16 frag_off; 772 u8 nexthdr; 773 struct net *net = dev_net(skb->dev); 774 775 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 776 goto out; 777 778 nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr; 779 if (ipv6_ext_hdr(nexthdr)) { 780 /* now skip over extension headers */ 781 inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), 782 &nexthdr, &frag_off); 783 if (inner_offset < 0) 784 goto out; 785 } else { 786 inner_offset = sizeof(struct ipv6hdr); 787 } 788 789 /* Checkin header including 8 bytes of inner protocol header. */ 790 if (!pskb_may_pull(skb, inner_offset+8)) 791 goto out; 792 793 /* BUGGG_FUTURE: we should try to parse exthdrs in this packet. 794 Without this we will not able f.e. to make source routed 795 pmtu discovery. 796 Corresponding argument (opt) to notifiers is already added. 797 --ANK (980726) 798 */ 799 800 ipprot = rcu_dereference(inet6_protos[nexthdr]); 801 if (ipprot && ipprot->err_handler) 802 ipprot->err_handler(skb, NULL, type, code, inner_offset, info); 803 804 raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info); 805 return; 806 807 out: 808 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS); 809 } 810 811 /* 812 * Handle icmp messages 813 */ 814 815 static int icmpv6_rcv(struct sk_buff *skb) 816 { 817 struct net *net = dev_net(skb->dev); 818 struct net_device *dev = icmp6_dev(skb); 819 struct inet6_dev *idev = __in6_dev_get(dev); 820 const struct in6_addr *saddr, *daddr; 821 struct icmp6hdr *hdr; 822 u8 type; 823 bool success = false; 824 825 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 826 struct sec_path *sp = skb_sec_path(skb); 827 int nh; 828 829 if (!(sp && sp->xvec[sp->len - 1]->props.flags & 830 XFRM_STATE_ICMP)) 831 goto drop_no_count; 832 833 if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr))) 834 goto drop_no_count; 835 836 nh = skb_network_offset(skb); 837 skb_set_network_header(skb, sizeof(*hdr)); 838 839 if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) 840 goto drop_no_count; 841 842 skb_set_network_header(skb, nh); 843 } 844 845 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS); 846 847 saddr = &ipv6_hdr(skb)->saddr; 848 daddr = &ipv6_hdr(skb)->daddr; 849 850 if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) { 851 net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n", 852 saddr, daddr); 853 goto csum_error; 854 } 855 856 if (!pskb_pull(skb, sizeof(*hdr))) 857 goto discard_it; 858 859 hdr = icmp6_hdr(skb); 860 861 type = hdr->icmp6_type; 862 863 ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type); 864 865 switch (type) { 866 case ICMPV6_ECHO_REQUEST: 867 if (!net->ipv6.sysctl.icmpv6_echo_ignore_all) 868 icmpv6_echo_reply(skb); 869 break; 870 871 case ICMPV6_ECHO_REPLY: 872 success = ping_rcv(skb); 873 break; 874 875 case ICMPV6_PKT_TOOBIG: 876 /* BUGGG_FUTURE: if packet contains rthdr, we cannot update 877 standard destination cache. Seems, only "advanced" 878 destination cache will allow to solve this problem 879 --ANK (980726) 880 */ 881 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 882 goto discard_it; 883 hdr = icmp6_hdr(skb); 884 885 /* to notify */ 886 /* fall through */ 887 case ICMPV6_DEST_UNREACH: 888 case ICMPV6_TIME_EXCEED: 889 case ICMPV6_PARAMPROB: 890 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); 891 break; 892 893 case NDISC_ROUTER_SOLICITATION: 894 case NDISC_ROUTER_ADVERTISEMENT: 895 case NDISC_NEIGHBOUR_SOLICITATION: 896 case NDISC_NEIGHBOUR_ADVERTISEMENT: 897 case NDISC_REDIRECT: 898 ndisc_rcv(skb); 899 break; 900 901 case ICMPV6_MGM_QUERY: 902 igmp6_event_query(skb); 903 break; 904 905 case ICMPV6_MGM_REPORT: 906 igmp6_event_report(skb); 907 break; 908 909 case ICMPV6_MGM_REDUCTION: 910 case ICMPV6_NI_QUERY: 911 case ICMPV6_NI_REPLY: 912 case ICMPV6_MLD2_REPORT: 913 case ICMPV6_DHAAD_REQUEST: 914 case ICMPV6_DHAAD_REPLY: 915 case ICMPV6_MOBILE_PREFIX_SOL: 916 case ICMPV6_MOBILE_PREFIX_ADV: 917 break; 918 919 default: 920 /* informational */ 921 if (type & ICMPV6_INFOMSG_MASK) 922 break; 923 924 net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n", 925 saddr, daddr); 926 927 /* 928 * error of unknown type. 929 * must pass to upper level 930 */ 931 932 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); 933 } 934 935 /* until the v6 path can be better sorted assume failure and 936 * preserve the status quo behaviour for the rest of the paths to here 937 */ 938 if (success) 939 consume_skb(skb); 940 else 941 kfree_skb(skb); 942 943 return 0; 944 945 csum_error: 946 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS); 947 discard_it: 948 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS); 949 drop_no_count: 950 kfree_skb(skb); 951 return 0; 952 } 953 954 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6, 955 u8 type, 956 const struct in6_addr *saddr, 957 const struct in6_addr *daddr, 958 int oif) 959 { 960 memset(fl6, 0, sizeof(*fl6)); 961 fl6->saddr = *saddr; 962 fl6->daddr = *daddr; 963 fl6->flowi6_proto = IPPROTO_ICMPV6; 964 fl6->fl6_icmp_type = type; 965 fl6->fl6_icmp_code = 0; 966 fl6->flowi6_oif = oif; 967 security_sk_classify_flow(sk, flowi6_to_flowi(fl6)); 968 } 969 970 static void __net_exit icmpv6_sk_exit(struct net *net) 971 { 972 int i; 973 974 for_each_possible_cpu(i) 975 inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv6.icmp_sk, i)); 976 free_percpu(net->ipv6.icmp_sk); 977 } 978 979 static int __net_init icmpv6_sk_init(struct net *net) 980 { 981 struct sock *sk; 982 int err, i; 983 984 net->ipv6.icmp_sk = alloc_percpu(struct sock *); 985 if (!net->ipv6.icmp_sk) 986 return -ENOMEM; 987 988 for_each_possible_cpu(i) { 989 err = inet_ctl_sock_create(&sk, PF_INET6, 990 SOCK_RAW, IPPROTO_ICMPV6, net); 991 if (err < 0) { 992 pr_err("Failed to initialize the ICMP6 control socket (err %d)\n", 993 err); 994 goto fail; 995 } 996 997 *per_cpu_ptr(net->ipv6.icmp_sk, i) = sk; 998 999 /* Enough space for 2 64K ICMP packets, including 1000 * sk_buff struct overhead. 1001 */ 1002 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024); 1003 } 1004 return 0; 1005 1006 fail: 1007 icmpv6_sk_exit(net); 1008 return err; 1009 } 1010 1011 static struct pernet_operations icmpv6_sk_ops = { 1012 .init = icmpv6_sk_init, 1013 .exit = icmpv6_sk_exit, 1014 }; 1015 1016 int __init icmpv6_init(void) 1017 { 1018 int err; 1019 1020 err = register_pernet_subsys(&icmpv6_sk_ops); 1021 if (err < 0) 1022 return err; 1023 1024 err = -EAGAIN; 1025 if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0) 1026 goto fail; 1027 1028 err = inet6_register_icmp_sender(icmp6_send); 1029 if (err) 1030 goto sender_reg_err; 1031 return 0; 1032 1033 sender_reg_err: 1034 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); 1035 fail: 1036 pr_err("Failed to register ICMP6 protocol\n"); 1037 unregister_pernet_subsys(&icmpv6_sk_ops); 1038 return err; 1039 } 1040 1041 void icmpv6_cleanup(void) 1042 { 1043 inet6_unregister_icmp_sender(icmp6_send); 1044 unregister_pernet_subsys(&icmpv6_sk_ops); 1045 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); 1046 } 1047 1048 1049 static const struct icmp6_err { 1050 int err; 1051 int fatal; 1052 } tab_unreach[] = { 1053 { /* NOROUTE */ 1054 .err = ENETUNREACH, 1055 .fatal = 0, 1056 }, 1057 { /* ADM_PROHIBITED */ 1058 .err = EACCES, 1059 .fatal = 1, 1060 }, 1061 { /* Was NOT_NEIGHBOUR, now reserved */ 1062 .err = EHOSTUNREACH, 1063 .fatal = 0, 1064 }, 1065 { /* ADDR_UNREACH */ 1066 .err = EHOSTUNREACH, 1067 .fatal = 0, 1068 }, 1069 { /* PORT_UNREACH */ 1070 .err = ECONNREFUSED, 1071 .fatal = 1, 1072 }, 1073 { /* POLICY_FAIL */ 1074 .err = EACCES, 1075 .fatal = 1, 1076 }, 1077 { /* REJECT_ROUTE */ 1078 .err = EACCES, 1079 .fatal = 1, 1080 }, 1081 }; 1082 1083 int icmpv6_err_convert(u8 type, u8 code, int *err) 1084 { 1085 int fatal = 0; 1086 1087 *err = EPROTO; 1088 1089 switch (type) { 1090 case ICMPV6_DEST_UNREACH: 1091 fatal = 1; 1092 if (code < ARRAY_SIZE(tab_unreach)) { 1093 *err = tab_unreach[code].err; 1094 fatal = tab_unreach[code].fatal; 1095 } 1096 break; 1097 1098 case ICMPV6_PKT_TOOBIG: 1099 *err = EMSGSIZE; 1100 break; 1101 1102 case ICMPV6_PARAMPROB: 1103 *err = EPROTO; 1104 fatal = 1; 1105 break; 1106 1107 case ICMPV6_TIME_EXCEED: 1108 *err = EHOSTUNREACH; 1109 break; 1110 } 1111 1112 return fatal; 1113 } 1114 EXPORT_SYMBOL(icmpv6_err_convert); 1115 1116 #ifdef CONFIG_SYSCTL 1117 static struct ctl_table ipv6_icmp_table_template[] = { 1118 { 1119 .procname = "ratelimit", 1120 .data = &init_net.ipv6.sysctl.icmpv6_time, 1121 .maxlen = sizeof(int), 1122 .mode = 0644, 1123 .proc_handler = proc_dointvec_ms_jiffies, 1124 }, 1125 { 1126 .procname = "echo_ignore_all", 1127 .data = &init_net.ipv6.sysctl.icmpv6_echo_ignore_all, 1128 .maxlen = sizeof(int), 1129 .mode = 0644, 1130 .proc_handler = proc_dointvec, 1131 }, 1132 { 1133 .procname = "echo_ignore_multicast", 1134 .data = &init_net.ipv6.sysctl.icmpv6_echo_ignore_multicast, 1135 .maxlen = sizeof(int), 1136 .mode = 0644, 1137 .proc_handler = proc_dointvec, 1138 }, 1139 { 1140 .procname = "echo_ignore_anycast", 1141 .data = &init_net.ipv6.sysctl.icmpv6_echo_ignore_anycast, 1142 .maxlen = sizeof(int), 1143 .mode = 0644, 1144 .proc_handler = proc_dointvec, 1145 }, 1146 { 1147 .procname = "ratemask", 1148 .data = &init_net.ipv6.sysctl.icmpv6_ratemask_ptr, 1149 .maxlen = ICMPV6_MSG_MAX + 1, 1150 .mode = 0644, 1151 .proc_handler = proc_do_large_bitmap, 1152 }, 1153 { }, 1154 }; 1155 1156 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net) 1157 { 1158 struct ctl_table *table; 1159 1160 table = kmemdup(ipv6_icmp_table_template, 1161 sizeof(ipv6_icmp_table_template), 1162 GFP_KERNEL); 1163 1164 if (table) { 1165 table[0].data = &net->ipv6.sysctl.icmpv6_time; 1166 table[1].data = &net->ipv6.sysctl.icmpv6_echo_ignore_all; 1167 table[2].data = &net->ipv6.sysctl.icmpv6_echo_ignore_multicast; 1168 table[3].data = &net->ipv6.sysctl.icmpv6_echo_ignore_anycast; 1169 table[4].data = &net->ipv6.sysctl.icmpv6_ratemask_ptr; 1170 } 1171 return table; 1172 } 1173 #endif 1174