1 /* 2 * RAW sockets for IPv6 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 * 8 * Adapted from linux/net/ipv4/raw.c 9 * 10 * Fixes: 11 * Hideaki YOSHIFUJI : sin6_scope_id support 12 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance) 13 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data 14 * 15 * This program is free software; you can redistribute it and/or 16 * modify it under the terms of the GNU General Public License 17 * as published by the Free Software Foundation; either version 18 * 2 of the License, or (at your option) any later version. 19 */ 20 21 #include <linux/errno.h> 22 #include <linux/types.h> 23 #include <linux/socket.h> 24 #include <linux/slab.h> 25 #include <linux/sockios.h> 26 #include <linux/net.h> 27 #include <linux/in6.h> 28 #include <linux/netdevice.h> 29 #include <linux/if_arp.h> 30 #include <linux/icmpv6.h> 31 #include <linux/netfilter.h> 32 #include <linux/netfilter_ipv6.h> 33 #include <linux/skbuff.h> 34 #include <linux/compat.h> 35 #include <linux/uaccess.h> 36 #include <asm/ioctls.h> 37 38 #include <net/net_namespace.h> 39 #include <net/ip.h> 40 #include <net/sock.h> 41 #include <net/snmp.h> 42 43 #include <net/ipv6.h> 44 #include <net/ndisc.h> 45 #include <net/protocol.h> 46 #include <net/ip6_route.h> 47 #include <net/ip6_checksum.h> 48 #include <net/addrconf.h> 49 #include <net/transp_v6.h> 50 #include <net/udp.h> 51 #include <net/inet_common.h> 52 #include <net/tcp_states.h> 53 #if IS_ENABLED(CONFIG_IPV6_MIP6) 54 #include <net/mip6.h> 55 #endif 56 #include <linux/mroute6.h> 57 58 #include <net/raw.h> 59 #include <net/rawv6.h> 60 #include <net/xfrm.h> 61 62 #include <linux/proc_fs.h> 63 #include <linux/seq_file.h> 64 #include <linux/export.h> 65 66 #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */ 67 68 struct raw_hashinfo raw_v6_hashinfo = { 69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock), 70 }; 71 EXPORT_SYMBOL_GPL(raw_v6_hashinfo); 72 73 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk, 74 unsigned short num, const struct in6_addr *loc_addr, 75 const struct in6_addr *rmt_addr, int dif, int sdif) 76 { 77 bool is_multicast = ipv6_addr_is_multicast(loc_addr); 78 79 sk_for_each_from(sk) 80 if (inet_sk(sk)->inet_num == num) { 81 82 if (!net_eq(sock_net(sk), net)) 83 continue; 84 85 if (!ipv6_addr_any(&sk->sk_v6_daddr) && 86 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) 87 continue; 88 89 if (!raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, 90 dif, sdif)) 91 continue; 92 93 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) { 94 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)) 95 goto found; 96 if (is_multicast && 97 inet6_mc_check(sk, loc_addr, rmt_addr)) 98 goto found; 99 continue; 100 } 101 goto found; 102 } 103 sk = NULL; 104 found: 105 return sk; 106 } 107 EXPORT_SYMBOL_GPL(__raw_v6_lookup); 108 109 /* 110 * 0 - deliver 111 * 1 - block 112 */ 113 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb) 114 { 115 struct icmp6hdr _hdr; 116 const struct icmp6hdr *hdr; 117 118 /* We require only the four bytes of the ICMPv6 header, not any 119 * additional bytes of message body in "struct icmp6hdr". 120 */ 121 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 122 ICMPV6_HDRLEN, &_hdr); 123 if (hdr) { 124 const __u32 *data = &raw6_sk(sk)->filter.data[0]; 125 unsigned int type = hdr->icmp6_type; 126 127 return (data[type >> 5] & (1U << (type & 31))) != 0; 128 } 129 return 1; 130 } 131 132 #if IS_ENABLED(CONFIG_IPV6_MIP6) 133 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb); 134 135 static mh_filter_t __rcu *mh_filter __read_mostly; 136 137 int rawv6_mh_filter_register(mh_filter_t filter) 138 { 139 rcu_assign_pointer(mh_filter, filter); 140 return 0; 141 } 142 EXPORT_SYMBOL(rawv6_mh_filter_register); 143 144 int rawv6_mh_filter_unregister(mh_filter_t filter) 145 { 146 RCU_INIT_POINTER(mh_filter, NULL); 147 synchronize_rcu(); 148 return 0; 149 } 150 EXPORT_SYMBOL(rawv6_mh_filter_unregister); 151 152 #endif 153 154 /* 155 * demultiplex raw sockets. 156 * (should consider queueing the skb in the sock receive_queue 157 * without calling rawv6.c) 158 * 159 * Caller owns SKB so we must make clones. 160 */ 161 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 162 { 163 const struct in6_addr *saddr; 164 const struct in6_addr *daddr; 165 struct sock *sk; 166 bool delivered = false; 167 __u8 hash; 168 struct net *net; 169 170 saddr = &ipv6_hdr(skb)->saddr; 171 daddr = saddr + 1; 172 173 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 174 175 read_lock(&raw_v6_hashinfo.lock); 176 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 177 178 if (!sk) 179 goto out; 180 181 net = dev_net(skb->dev); 182 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, 183 inet6_iif(skb), inet6_sdif(skb)); 184 185 while (sk) { 186 int filtered; 187 188 delivered = true; 189 switch (nexthdr) { 190 case IPPROTO_ICMPV6: 191 filtered = icmpv6_filter(sk, skb); 192 break; 193 194 #if IS_ENABLED(CONFIG_IPV6_MIP6) 195 case IPPROTO_MH: 196 { 197 /* XXX: To validate MH only once for each packet, 198 * this is placed here. It should be after checking 199 * xfrm policy, however it doesn't. The checking xfrm 200 * policy is placed in rawv6_rcv() because it is 201 * required for each socket. 202 */ 203 mh_filter_t *filter; 204 205 filter = rcu_dereference(mh_filter); 206 filtered = filter ? (*filter)(sk, skb) : 0; 207 break; 208 } 209 #endif 210 default: 211 filtered = 0; 212 break; 213 } 214 215 if (filtered < 0) 216 break; 217 if (filtered == 0) { 218 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 219 220 /* Not releasing hash table! */ 221 if (clone) { 222 nf_reset(clone); 223 rawv6_rcv(sk, clone); 224 } 225 } 226 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr, 227 inet6_iif(skb), inet6_sdif(skb)); 228 } 229 out: 230 read_unlock(&raw_v6_hashinfo.lock); 231 return delivered; 232 } 233 234 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr) 235 { 236 struct sock *raw_sk; 237 238 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]); 239 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr)) 240 raw_sk = NULL; 241 242 return raw_sk != NULL; 243 } 244 245 /* This cleans up af_inet6 a bit. -DaveM */ 246 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 247 { 248 struct inet_sock *inet = inet_sk(sk); 249 struct ipv6_pinfo *np = inet6_sk(sk); 250 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 251 __be32 v4addr = 0; 252 int addr_type; 253 int err; 254 255 if (addr_len < SIN6_LEN_RFC2133) 256 return -EINVAL; 257 258 if (addr->sin6_family != AF_INET6) 259 return -EINVAL; 260 261 addr_type = ipv6_addr_type(&addr->sin6_addr); 262 263 /* Raw sockets are IPv6 only */ 264 if (addr_type == IPV6_ADDR_MAPPED) 265 return -EADDRNOTAVAIL; 266 267 lock_sock(sk); 268 269 err = -EINVAL; 270 if (sk->sk_state != TCP_CLOSE) 271 goto out; 272 273 rcu_read_lock(); 274 /* Check if the address belongs to the host. */ 275 if (addr_type != IPV6_ADDR_ANY) { 276 struct net_device *dev = NULL; 277 278 if (__ipv6_addr_needs_scope_id(addr_type)) { 279 if (addr_len >= sizeof(struct sockaddr_in6) && 280 addr->sin6_scope_id) { 281 /* Override any existing binding, if another 282 * one is supplied by user. 283 */ 284 sk->sk_bound_dev_if = addr->sin6_scope_id; 285 } 286 287 /* Binding to link-local address requires an interface */ 288 if (!sk->sk_bound_dev_if) 289 goto out_unlock; 290 291 err = -ENODEV; 292 dev = dev_get_by_index_rcu(sock_net(sk), 293 sk->sk_bound_dev_if); 294 if (!dev) 295 goto out_unlock; 296 } 297 298 /* ipv4 addr of the socket is invalid. Only the 299 * unspecified and mapped address have a v4 equivalent. 300 */ 301 v4addr = LOOPBACK4_IPV6; 302 if (!(addr_type & IPV6_ADDR_MULTICAST) && 303 !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) { 304 err = -EADDRNOTAVAIL; 305 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr, 306 dev, 0)) { 307 goto out_unlock; 308 } 309 } 310 } 311 312 inet->inet_rcv_saddr = inet->inet_saddr = v4addr; 313 sk->sk_v6_rcv_saddr = addr->sin6_addr; 314 if (!(addr_type & IPV6_ADDR_MULTICAST)) 315 np->saddr = addr->sin6_addr; 316 err = 0; 317 out_unlock: 318 rcu_read_unlock(); 319 out: 320 release_sock(sk); 321 return err; 322 } 323 324 static void rawv6_err(struct sock *sk, struct sk_buff *skb, 325 struct inet6_skb_parm *opt, 326 u8 type, u8 code, int offset, __be32 info) 327 { 328 struct inet_sock *inet = inet_sk(sk); 329 struct ipv6_pinfo *np = inet6_sk(sk); 330 int err; 331 int harderr; 332 333 /* Report error on raw socket, if: 334 1. User requested recverr. 335 2. Socket is connected (otherwise the error indication 336 is useless without recverr and error is hard. 337 */ 338 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 339 return; 340 341 harderr = icmpv6_err_convert(type, code, &err); 342 if (type == ICMPV6_PKT_TOOBIG) { 343 ip6_sk_update_pmtu(skb, sk, info); 344 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 345 } 346 if (type == NDISC_REDIRECT) { 347 ip6_sk_redirect(skb, sk); 348 return; 349 } 350 if (np->recverr) { 351 u8 *payload = skb->data; 352 if (!inet->hdrincl) 353 payload += offset; 354 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 355 } 356 357 if (np->recverr || harderr) { 358 sk->sk_err = err; 359 sk->sk_error_report(sk); 360 } 361 } 362 363 void raw6_icmp_error(struct sk_buff *skb, int nexthdr, 364 u8 type, u8 code, int inner_offset, __be32 info) 365 { 366 struct sock *sk; 367 int hash; 368 const struct in6_addr *saddr, *daddr; 369 struct net *net; 370 371 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 372 373 read_lock(&raw_v6_hashinfo.lock); 374 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 375 if (sk) { 376 /* Note: ipv6_hdr(skb) != skb->data */ 377 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data; 378 saddr = &ip6h->saddr; 379 daddr = &ip6h->daddr; 380 net = dev_net(skb->dev); 381 382 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr, 383 inet6_iif(skb), inet6_iif(skb)))) { 384 rawv6_err(sk, skb, NULL, type, code, 385 inner_offset, info); 386 sk = sk_next(sk); 387 } 388 } 389 read_unlock(&raw_v6_hashinfo.lock); 390 } 391 392 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb) 393 { 394 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) && 395 skb_checksum_complete(skb)) { 396 atomic_inc(&sk->sk_drops); 397 kfree_skb(skb); 398 return NET_RX_DROP; 399 } 400 401 /* Charge it to the socket. */ 402 skb_dst_drop(skb); 403 if (sock_queue_rcv_skb(sk, skb) < 0) { 404 kfree_skb(skb); 405 return NET_RX_DROP; 406 } 407 408 return 0; 409 } 410 411 /* 412 * This is next to useless... 413 * if we demultiplex in network layer we don't need the extra call 414 * just to queue the skb... 415 * maybe we could have the network decide upon a hint if it 416 * should call raw_rcv for demultiplexing 417 */ 418 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 419 { 420 struct inet_sock *inet = inet_sk(sk); 421 struct raw6_sock *rp = raw6_sk(sk); 422 423 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 424 atomic_inc(&sk->sk_drops); 425 kfree_skb(skb); 426 return NET_RX_DROP; 427 } 428 429 if (!rp->checksum) 430 skb->ip_summed = CHECKSUM_UNNECESSARY; 431 432 if (skb->ip_summed == CHECKSUM_COMPLETE) { 433 skb_postpull_rcsum(skb, skb_network_header(skb), 434 skb_network_header_len(skb)); 435 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 436 &ipv6_hdr(skb)->daddr, 437 skb->len, inet->inet_num, skb->csum)) 438 skb->ip_summed = CHECKSUM_UNNECESSARY; 439 } 440 if (!skb_csum_unnecessary(skb)) 441 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 442 &ipv6_hdr(skb)->daddr, 443 skb->len, 444 inet->inet_num, 0)); 445 446 if (inet->hdrincl) { 447 if (skb_checksum_complete(skb)) { 448 atomic_inc(&sk->sk_drops); 449 kfree_skb(skb); 450 return NET_RX_DROP; 451 } 452 } 453 454 rawv6_rcv_skb(sk, skb); 455 return 0; 456 } 457 458 459 /* 460 * This should be easy, if there is something there 461 * we return it, otherwise we block. 462 */ 463 464 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 465 int noblock, int flags, int *addr_len) 466 { 467 struct ipv6_pinfo *np = inet6_sk(sk); 468 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 469 struct sk_buff *skb; 470 size_t copied; 471 int err; 472 473 if (flags & MSG_OOB) 474 return -EOPNOTSUPP; 475 476 if (flags & MSG_ERRQUEUE) 477 return ipv6_recv_error(sk, msg, len, addr_len); 478 479 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 480 return ipv6_recv_rxpmtu(sk, msg, len, addr_len); 481 482 skb = skb_recv_datagram(sk, flags, noblock, &err); 483 if (!skb) 484 goto out; 485 486 copied = skb->len; 487 if (copied > len) { 488 copied = len; 489 msg->msg_flags |= MSG_TRUNC; 490 } 491 492 if (skb_csum_unnecessary(skb)) { 493 err = skb_copy_datagram_msg(skb, 0, msg, copied); 494 } else if (msg->msg_flags&MSG_TRUNC) { 495 if (__skb_checksum_complete(skb)) 496 goto csum_copy_err; 497 err = skb_copy_datagram_msg(skb, 0, msg, copied); 498 } else { 499 err = skb_copy_and_csum_datagram_msg(skb, 0, msg); 500 if (err == -EINVAL) 501 goto csum_copy_err; 502 } 503 if (err) 504 goto out_free; 505 506 /* Copy the address. */ 507 if (sin6) { 508 sin6->sin6_family = AF_INET6; 509 sin6->sin6_port = 0; 510 sin6->sin6_addr = ipv6_hdr(skb)->saddr; 511 sin6->sin6_flowinfo = 0; 512 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr, 513 inet6_iif(skb)); 514 *addr_len = sizeof(*sin6); 515 } 516 517 sock_recv_ts_and_drops(msg, sk, skb); 518 519 if (np->rxopt.all) 520 ip6_datagram_recv_ctl(sk, msg, skb); 521 522 err = copied; 523 if (flags & MSG_TRUNC) 524 err = skb->len; 525 526 out_free: 527 skb_free_datagram(sk, skb); 528 out: 529 return err; 530 531 csum_copy_err: 532 skb_kill_datagram(sk, skb, flags); 533 534 /* Error for blocking case is chosen to masquerade 535 as some normal condition. 536 */ 537 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 538 goto out; 539 } 540 541 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 542 struct raw6_sock *rp) 543 { 544 struct sk_buff *skb; 545 int err = 0; 546 int offset; 547 int len; 548 int total_len; 549 __wsum tmp_csum; 550 __sum16 csum; 551 552 if (!rp->checksum) 553 goto send; 554 555 skb = skb_peek(&sk->sk_write_queue); 556 if (!skb) 557 goto out; 558 559 offset = rp->offset; 560 total_len = inet_sk(sk)->cork.base.length; 561 if (offset >= total_len - 1) { 562 err = -EINVAL; 563 ip6_flush_pending_frames(sk); 564 goto out; 565 } 566 567 /* should be check HW csum miyazawa */ 568 if (skb_queue_len(&sk->sk_write_queue) == 1) { 569 /* 570 * Only one fragment on the socket. 571 */ 572 tmp_csum = skb->csum; 573 } else { 574 struct sk_buff *csum_skb = NULL; 575 tmp_csum = 0; 576 577 skb_queue_walk(&sk->sk_write_queue, skb) { 578 tmp_csum = csum_add(tmp_csum, skb->csum); 579 580 if (csum_skb) 581 continue; 582 583 len = skb->len - skb_transport_offset(skb); 584 if (offset >= len) { 585 offset -= len; 586 continue; 587 } 588 589 csum_skb = skb; 590 } 591 592 skb = csum_skb; 593 } 594 595 offset += skb_transport_offset(skb); 596 err = skb_copy_bits(skb, offset, &csum, 2); 597 if (err < 0) { 598 ip6_flush_pending_frames(sk); 599 goto out; 600 } 601 602 /* in case cksum was not initialized */ 603 if (unlikely(csum)) 604 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 605 606 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 607 total_len, fl6->flowi6_proto, tmp_csum); 608 609 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP) 610 csum = CSUM_MANGLED_0; 611 612 BUG_ON(skb_store_bits(skb, offset, &csum, 2)); 613 614 send: 615 err = ip6_push_pending_frames(sk); 616 out: 617 return err; 618 } 619 620 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length, 621 struct flowi6 *fl6, struct dst_entry **dstp, 622 unsigned int flags, const struct sockcm_cookie *sockc) 623 { 624 struct ipv6_pinfo *np = inet6_sk(sk); 625 struct net *net = sock_net(sk); 626 struct ipv6hdr *iph; 627 struct sk_buff *skb; 628 int err; 629 struct rt6_info *rt = (struct rt6_info *)*dstp; 630 int hlen = LL_RESERVED_SPACE(rt->dst.dev); 631 int tlen = rt->dst.dev->needed_tailroom; 632 633 if (length > rt->dst.dev->mtu) { 634 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu); 635 return -EMSGSIZE; 636 } 637 if (length < sizeof(struct ipv6hdr)) 638 return -EINVAL; 639 if (flags&MSG_PROBE) 640 goto out; 641 642 skb = sock_alloc_send_skb(sk, 643 length + hlen + tlen + 15, 644 flags & MSG_DONTWAIT, &err); 645 if (!skb) 646 goto error; 647 skb_reserve(skb, hlen); 648 649 skb->protocol = htons(ETH_P_IPV6); 650 skb->priority = sk->sk_priority; 651 skb->mark = sk->sk_mark; 652 skb->tstamp = sockc->transmit_time; 653 654 skb_put(skb, length); 655 skb_reset_network_header(skb); 656 iph = ipv6_hdr(skb); 657 658 skb->ip_summed = CHECKSUM_NONE; 659 660 skb_setup_tx_timestamp(skb, sockc->tsflags); 661 662 if (flags & MSG_CONFIRM) 663 skb_set_dst_pending_confirm(skb, 1); 664 665 skb->transport_header = skb->network_header; 666 err = memcpy_from_msg(iph, msg, length); 667 if (err) { 668 err = -EFAULT; 669 kfree_skb(skb); 670 goto error; 671 } 672 673 skb_dst_set(skb, &rt->dst); 674 *dstp = NULL; 675 676 /* if egress device is enslaved to an L3 master device pass the 677 * skb to its handler for processing 678 */ 679 skb = l3mdev_ip6_out(sk, skb); 680 if (unlikely(!skb)) 681 return 0; 682 683 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev 684 * in the error path. Since skb has been freed, the dst could 685 * have been queued for deletion. 686 */ 687 rcu_read_lock(); 688 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); 689 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb, 690 NULL, rt->dst.dev, dst_output); 691 if (err > 0) 692 err = net_xmit_errno(err); 693 if (err) { 694 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 695 rcu_read_unlock(); 696 goto error_check; 697 } 698 rcu_read_unlock(); 699 out: 700 return 0; 701 702 error: 703 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 704 error_check: 705 if (err == -ENOBUFS && !np->recverr) 706 err = 0; 707 return err; 708 } 709 710 struct raw6_frag_vec { 711 struct msghdr *msg; 712 int hlen; 713 char c[4]; 714 }; 715 716 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6) 717 { 718 int err = 0; 719 switch (fl6->flowi6_proto) { 720 case IPPROTO_ICMPV6: 721 rfv->hlen = 2; 722 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); 723 if (!err) { 724 fl6->fl6_icmp_type = rfv->c[0]; 725 fl6->fl6_icmp_code = rfv->c[1]; 726 } 727 break; 728 case IPPROTO_MH: 729 rfv->hlen = 4; 730 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); 731 if (!err) 732 fl6->fl6_mh_type = rfv->c[2]; 733 } 734 return err; 735 } 736 737 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd, 738 struct sk_buff *skb) 739 { 740 struct raw6_frag_vec *rfv = from; 741 742 if (offset < rfv->hlen) { 743 int copy = min(rfv->hlen - offset, len); 744 745 if (skb->ip_summed == CHECKSUM_PARTIAL) 746 memcpy(to, rfv->c + offset, copy); 747 else 748 skb->csum = csum_block_add( 749 skb->csum, 750 csum_partial_copy_nocheck(rfv->c + offset, 751 to, copy, 0), 752 odd); 753 754 odd = 0; 755 offset += copy; 756 to += copy; 757 len -= copy; 758 759 if (!len) 760 return 0; 761 } 762 763 offset -= rfv->hlen; 764 765 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); 766 } 767 768 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 769 { 770 struct ipv6_txoptions *opt_to_free = NULL; 771 struct ipv6_txoptions opt_space; 772 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 773 struct in6_addr *daddr, *final_p, final; 774 struct inet_sock *inet = inet_sk(sk); 775 struct ipv6_pinfo *np = inet6_sk(sk); 776 struct raw6_sock *rp = raw6_sk(sk); 777 struct ipv6_txoptions *opt = NULL; 778 struct ip6_flowlabel *flowlabel = NULL; 779 struct dst_entry *dst = NULL; 780 struct raw6_frag_vec rfv; 781 struct flowi6 fl6; 782 struct ipcm6_cookie ipc6; 783 int addr_len = msg->msg_namelen; 784 u16 proto; 785 int err; 786 787 /* Rough check on arithmetic overflow, 788 better check is made in ip6_append_data(). 789 */ 790 if (len > INT_MAX) 791 return -EMSGSIZE; 792 793 /* Mirror BSD error message compatibility */ 794 if (msg->msg_flags & MSG_OOB) 795 return -EOPNOTSUPP; 796 797 /* 798 * Get and verify the address. 799 */ 800 memset(&fl6, 0, sizeof(fl6)); 801 802 fl6.flowi6_mark = sk->sk_mark; 803 fl6.flowi6_uid = sk->sk_uid; 804 805 ipcm6_init(&ipc6); 806 ipc6.sockc.tsflags = sk->sk_tsflags; 807 808 if (sin6) { 809 if (addr_len < SIN6_LEN_RFC2133) 810 return -EINVAL; 811 812 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 813 return -EAFNOSUPPORT; 814 815 /* port is the proto value [0..255] carried in nexthdr */ 816 proto = ntohs(sin6->sin6_port); 817 818 if (!proto) 819 proto = inet->inet_num; 820 else if (proto != inet->inet_num) 821 return -EINVAL; 822 823 if (proto > 255) 824 return -EINVAL; 825 826 daddr = &sin6->sin6_addr; 827 if (np->sndflow) { 828 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 829 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 830 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 831 if (!flowlabel) 832 return -EINVAL; 833 } 834 } 835 836 /* 837 * Otherwise it will be difficult to maintain 838 * sk->sk_dst_cache. 839 */ 840 if (sk->sk_state == TCP_ESTABLISHED && 841 ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) 842 daddr = &sk->sk_v6_daddr; 843 844 if (addr_len >= sizeof(struct sockaddr_in6) && 845 sin6->sin6_scope_id && 846 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) 847 fl6.flowi6_oif = sin6->sin6_scope_id; 848 } else { 849 if (sk->sk_state != TCP_ESTABLISHED) 850 return -EDESTADDRREQ; 851 852 proto = inet->inet_num; 853 daddr = &sk->sk_v6_daddr; 854 fl6.flowlabel = np->flow_label; 855 } 856 857 if (fl6.flowi6_oif == 0) 858 fl6.flowi6_oif = sk->sk_bound_dev_if; 859 860 if (msg->msg_controllen) { 861 opt = &opt_space; 862 memset(opt, 0, sizeof(struct ipv6_txoptions)); 863 opt->tot_len = sizeof(struct ipv6_txoptions); 864 ipc6.opt = opt; 865 866 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6); 867 if (err < 0) { 868 fl6_sock_release(flowlabel); 869 return err; 870 } 871 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 872 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 873 if (!flowlabel) 874 return -EINVAL; 875 } 876 if (!(opt->opt_nflen|opt->opt_flen)) 877 opt = NULL; 878 } 879 if (!opt) { 880 opt = txopt_get(np); 881 opt_to_free = opt; 882 } 883 if (flowlabel) 884 opt = fl6_merge_options(&opt_space, flowlabel, opt); 885 opt = ipv6_fixup_options(&opt_space, opt); 886 887 fl6.flowi6_proto = proto; 888 rfv.msg = msg; 889 rfv.hlen = 0; 890 err = rawv6_probe_proto_opt(&rfv, &fl6); 891 if (err) 892 goto out; 893 894 if (!ipv6_addr_any(daddr)) 895 fl6.daddr = *daddr; 896 else 897 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 898 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 899 fl6.saddr = np->saddr; 900 901 final_p = fl6_update_dst(&fl6, opt, &final); 902 903 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 904 fl6.flowi6_oif = np->mcast_oif; 905 else if (!fl6.flowi6_oif) 906 fl6.flowi6_oif = np->ucast_oif; 907 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 908 909 if (inet->hdrincl) 910 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH; 911 912 if (ipc6.tclass < 0) 913 ipc6.tclass = np->tclass; 914 915 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); 916 917 dst = ip6_dst_lookup_flow(sk, &fl6, final_p); 918 if (IS_ERR(dst)) { 919 err = PTR_ERR(dst); 920 goto out; 921 } 922 if (ipc6.hlimit < 0) 923 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 924 925 if (ipc6.dontfrag < 0) 926 ipc6.dontfrag = np->dontfrag; 927 928 if (msg->msg_flags&MSG_CONFIRM) 929 goto do_confirm; 930 931 back_from_confirm: 932 if (inet->hdrincl) 933 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, 934 msg->msg_flags, &ipc6.sockc); 935 else { 936 ipc6.opt = opt; 937 lock_sock(sk); 938 err = ip6_append_data(sk, raw6_getfrag, &rfv, 939 len, 0, &ipc6, &fl6, (struct rt6_info *)dst, 940 msg->msg_flags); 941 942 if (err) 943 ip6_flush_pending_frames(sk); 944 else if (!(msg->msg_flags & MSG_MORE)) 945 err = rawv6_push_pending_frames(sk, &fl6, rp); 946 release_sock(sk); 947 } 948 done: 949 dst_release(dst); 950 out: 951 fl6_sock_release(flowlabel); 952 txopt_put(opt_to_free); 953 return err < 0 ? err : len; 954 do_confirm: 955 if (msg->msg_flags & MSG_PROBE) 956 dst_confirm_neigh(dst, &fl6.daddr); 957 if (!(msg->msg_flags & MSG_PROBE) || len) 958 goto back_from_confirm; 959 err = 0; 960 goto done; 961 } 962 963 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 964 char __user *optval, int optlen) 965 { 966 switch (optname) { 967 case ICMPV6_FILTER: 968 if (optlen > sizeof(struct icmp6_filter)) 969 optlen = sizeof(struct icmp6_filter); 970 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 971 return -EFAULT; 972 return 0; 973 default: 974 return -ENOPROTOOPT; 975 } 976 977 return 0; 978 } 979 980 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 981 char __user *optval, int __user *optlen) 982 { 983 int len; 984 985 switch (optname) { 986 case ICMPV6_FILTER: 987 if (get_user(len, optlen)) 988 return -EFAULT; 989 if (len < 0) 990 return -EINVAL; 991 if (len > sizeof(struct icmp6_filter)) 992 len = sizeof(struct icmp6_filter); 993 if (put_user(len, optlen)) 994 return -EFAULT; 995 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 996 return -EFAULT; 997 return 0; 998 default: 999 return -ENOPROTOOPT; 1000 } 1001 1002 return 0; 1003 } 1004 1005 1006 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 1007 char __user *optval, unsigned int optlen) 1008 { 1009 struct raw6_sock *rp = raw6_sk(sk); 1010 int val; 1011 1012 if (get_user(val, (int __user *)optval)) 1013 return -EFAULT; 1014 1015 switch (optname) { 1016 case IPV6_HDRINCL: 1017 if (sk->sk_type != SOCK_RAW) 1018 return -EINVAL; 1019 inet_sk(sk)->hdrincl = !!val; 1020 return 0; 1021 case IPV6_CHECKSUM: 1022 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && 1023 level == IPPROTO_IPV6) { 1024 /* 1025 * RFC3542 tells that IPV6_CHECKSUM socket 1026 * option in the IPPROTO_IPV6 level is not 1027 * allowed on ICMPv6 sockets. 1028 * If you want to set it, use IPPROTO_RAW 1029 * level IPV6_CHECKSUM socket option 1030 * (Linux extension). 1031 */ 1032 return -EINVAL; 1033 } 1034 1035 /* You may get strange result with a positive odd offset; 1036 RFC2292bis agrees with me. */ 1037 if (val > 0 && (val&1)) 1038 return -EINVAL; 1039 if (val < 0) { 1040 rp->checksum = 0; 1041 } else { 1042 rp->checksum = 1; 1043 rp->offset = val; 1044 } 1045 1046 return 0; 1047 1048 default: 1049 return -ENOPROTOOPT; 1050 } 1051 } 1052 1053 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 1054 char __user *optval, unsigned int optlen) 1055 { 1056 switch (level) { 1057 case SOL_RAW: 1058 break; 1059 1060 case SOL_ICMPV6: 1061 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1062 return -EOPNOTSUPP; 1063 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1064 case SOL_IPV6: 1065 if (optname == IPV6_CHECKSUM || 1066 optname == IPV6_HDRINCL) 1067 break; 1068 /* fall through */ 1069 default: 1070 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1071 } 1072 1073 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1074 } 1075 1076 #ifdef CONFIG_COMPAT 1077 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 1078 char __user *optval, unsigned int optlen) 1079 { 1080 switch (level) { 1081 case SOL_RAW: 1082 break; 1083 case SOL_ICMPV6: 1084 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1085 return -EOPNOTSUPP; 1086 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1087 case SOL_IPV6: 1088 if (optname == IPV6_CHECKSUM || 1089 optname == IPV6_HDRINCL) 1090 break; 1091 /* fall through */ 1092 default: 1093 return compat_ipv6_setsockopt(sk, level, optname, 1094 optval, optlen); 1095 } 1096 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1097 } 1098 #endif 1099 1100 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1101 char __user *optval, int __user *optlen) 1102 { 1103 struct raw6_sock *rp = raw6_sk(sk); 1104 int val, len; 1105 1106 if (get_user(len, optlen)) 1107 return -EFAULT; 1108 1109 switch (optname) { 1110 case IPV6_HDRINCL: 1111 val = inet_sk(sk)->hdrincl; 1112 break; 1113 case IPV6_CHECKSUM: 1114 /* 1115 * We allow getsockopt() for IPPROTO_IPV6-level 1116 * IPV6_CHECKSUM socket option on ICMPv6 sockets 1117 * since RFC3542 is silent about it. 1118 */ 1119 if (rp->checksum == 0) 1120 val = -1; 1121 else 1122 val = rp->offset; 1123 break; 1124 1125 default: 1126 return -ENOPROTOOPT; 1127 } 1128 1129 len = min_t(unsigned int, sizeof(int), len); 1130 1131 if (put_user(len, optlen)) 1132 return -EFAULT; 1133 if (copy_to_user(optval, &val, len)) 1134 return -EFAULT; 1135 return 0; 1136 } 1137 1138 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1139 char __user *optval, int __user *optlen) 1140 { 1141 switch (level) { 1142 case SOL_RAW: 1143 break; 1144 1145 case SOL_ICMPV6: 1146 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1147 return -EOPNOTSUPP; 1148 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1149 case SOL_IPV6: 1150 if (optname == IPV6_CHECKSUM || 1151 optname == IPV6_HDRINCL) 1152 break; 1153 /* fall through */ 1154 default: 1155 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1156 } 1157 1158 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1159 } 1160 1161 #ifdef CONFIG_COMPAT 1162 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1163 char __user *optval, int __user *optlen) 1164 { 1165 switch (level) { 1166 case SOL_RAW: 1167 break; 1168 case SOL_ICMPV6: 1169 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1170 return -EOPNOTSUPP; 1171 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1172 case SOL_IPV6: 1173 if (optname == IPV6_CHECKSUM || 1174 optname == IPV6_HDRINCL) 1175 break; 1176 /* fall through */ 1177 default: 1178 return compat_ipv6_getsockopt(sk, level, optname, 1179 optval, optlen); 1180 } 1181 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1182 } 1183 #endif 1184 1185 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1186 { 1187 switch (cmd) { 1188 case SIOCOUTQ: { 1189 int amount = sk_wmem_alloc_get(sk); 1190 1191 return put_user(amount, (int __user *)arg); 1192 } 1193 case SIOCINQ: { 1194 struct sk_buff *skb; 1195 int amount = 0; 1196 1197 spin_lock_bh(&sk->sk_receive_queue.lock); 1198 skb = skb_peek(&sk->sk_receive_queue); 1199 if (skb) 1200 amount = skb->len; 1201 spin_unlock_bh(&sk->sk_receive_queue.lock); 1202 return put_user(amount, (int __user *)arg); 1203 } 1204 1205 default: 1206 #ifdef CONFIG_IPV6_MROUTE 1207 return ip6mr_ioctl(sk, cmd, (void __user *)arg); 1208 #else 1209 return -ENOIOCTLCMD; 1210 #endif 1211 } 1212 } 1213 1214 #ifdef CONFIG_COMPAT 1215 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 1216 { 1217 switch (cmd) { 1218 case SIOCOUTQ: 1219 case SIOCINQ: 1220 return -ENOIOCTLCMD; 1221 default: 1222 #ifdef CONFIG_IPV6_MROUTE 1223 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg)); 1224 #else 1225 return -ENOIOCTLCMD; 1226 #endif 1227 } 1228 } 1229 #endif 1230 1231 static void rawv6_close(struct sock *sk, long timeout) 1232 { 1233 if (inet_sk(sk)->inet_num == IPPROTO_RAW) 1234 ip6_ra_control(sk, -1); 1235 ip6mr_sk_done(sk); 1236 sk_common_release(sk); 1237 } 1238 1239 static void raw6_destroy(struct sock *sk) 1240 { 1241 lock_sock(sk); 1242 ip6_flush_pending_frames(sk); 1243 release_sock(sk); 1244 1245 inet6_destroy_sock(sk); 1246 } 1247 1248 static int rawv6_init_sk(struct sock *sk) 1249 { 1250 struct raw6_sock *rp = raw6_sk(sk); 1251 1252 switch (inet_sk(sk)->inet_num) { 1253 case IPPROTO_ICMPV6: 1254 rp->checksum = 1; 1255 rp->offset = 2; 1256 break; 1257 case IPPROTO_MH: 1258 rp->checksum = 1; 1259 rp->offset = 4; 1260 break; 1261 default: 1262 break; 1263 } 1264 return 0; 1265 } 1266 1267 struct proto rawv6_prot = { 1268 .name = "RAWv6", 1269 .owner = THIS_MODULE, 1270 .close = rawv6_close, 1271 .destroy = raw6_destroy, 1272 .connect = ip6_datagram_connect_v6_only, 1273 .disconnect = __udp_disconnect, 1274 .ioctl = rawv6_ioctl, 1275 .init = rawv6_init_sk, 1276 .setsockopt = rawv6_setsockopt, 1277 .getsockopt = rawv6_getsockopt, 1278 .sendmsg = rawv6_sendmsg, 1279 .recvmsg = rawv6_recvmsg, 1280 .bind = rawv6_bind, 1281 .backlog_rcv = rawv6_rcv_skb, 1282 .hash = raw_hash_sk, 1283 .unhash = raw_unhash_sk, 1284 .obj_size = sizeof(struct raw6_sock), 1285 .useroffset = offsetof(struct raw6_sock, filter), 1286 .usersize = sizeof_field(struct raw6_sock, filter), 1287 .h.raw_hash = &raw_v6_hashinfo, 1288 #ifdef CONFIG_COMPAT 1289 .compat_setsockopt = compat_rawv6_setsockopt, 1290 .compat_getsockopt = compat_rawv6_getsockopt, 1291 .compat_ioctl = compat_rawv6_ioctl, 1292 #endif 1293 .diag_destroy = raw_abort, 1294 }; 1295 1296 #ifdef CONFIG_PROC_FS 1297 static int raw6_seq_show(struct seq_file *seq, void *v) 1298 { 1299 if (v == SEQ_START_TOKEN) { 1300 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); 1301 } else { 1302 struct sock *sp = v; 1303 __u16 srcp = inet_sk(sp)->inet_num; 1304 ip6_dgram_sock_seq_show(seq, v, srcp, 0, 1305 raw_seq_private(seq)->bucket); 1306 } 1307 return 0; 1308 } 1309 1310 static const struct seq_operations raw6_seq_ops = { 1311 .start = raw_seq_start, 1312 .next = raw_seq_next, 1313 .stop = raw_seq_stop, 1314 .show = raw6_seq_show, 1315 }; 1316 1317 static int __net_init raw6_init_net(struct net *net) 1318 { 1319 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops, 1320 sizeof(struct raw_iter_state), &raw_v6_hashinfo)) 1321 return -ENOMEM; 1322 1323 return 0; 1324 } 1325 1326 static void __net_exit raw6_exit_net(struct net *net) 1327 { 1328 remove_proc_entry("raw6", net->proc_net); 1329 } 1330 1331 static struct pernet_operations raw6_net_ops = { 1332 .init = raw6_init_net, 1333 .exit = raw6_exit_net, 1334 }; 1335 1336 int __init raw6_proc_init(void) 1337 { 1338 return register_pernet_subsys(&raw6_net_ops); 1339 } 1340 1341 void raw6_proc_exit(void) 1342 { 1343 unregister_pernet_subsys(&raw6_net_ops); 1344 } 1345 #endif /* CONFIG_PROC_FS */ 1346 1347 /* Same as inet6_dgram_ops, sans udp_poll. */ 1348 const struct proto_ops inet6_sockraw_ops = { 1349 .family = PF_INET6, 1350 .owner = THIS_MODULE, 1351 .release = inet6_release, 1352 .bind = inet6_bind, 1353 .connect = inet_dgram_connect, /* ok */ 1354 .socketpair = sock_no_socketpair, /* a do nothing */ 1355 .accept = sock_no_accept, /* a do nothing */ 1356 .getname = inet6_getname, 1357 .poll = datagram_poll, /* ok */ 1358 .ioctl = inet6_ioctl, /* must change */ 1359 .gettstamp = sock_gettstamp, 1360 .listen = sock_no_listen, /* ok */ 1361 .shutdown = inet_shutdown, /* ok */ 1362 .setsockopt = sock_common_setsockopt, /* ok */ 1363 .getsockopt = sock_common_getsockopt, /* ok */ 1364 .sendmsg = inet_sendmsg, /* ok */ 1365 .recvmsg = sock_common_recvmsg, /* ok */ 1366 .mmap = sock_no_mmap, 1367 .sendpage = sock_no_sendpage, 1368 #ifdef CONFIG_COMPAT 1369 .compat_setsockopt = compat_sock_common_setsockopt, 1370 .compat_getsockopt = compat_sock_common_getsockopt, 1371 #endif 1372 }; 1373 1374 static struct inet_protosw rawv6_protosw = { 1375 .type = SOCK_RAW, 1376 .protocol = IPPROTO_IP, /* wild card */ 1377 .prot = &rawv6_prot, 1378 .ops = &inet6_sockraw_ops, 1379 .flags = INET_PROTOSW_REUSE, 1380 }; 1381 1382 int __init rawv6_init(void) 1383 { 1384 return inet6_register_protosw(&rawv6_protosw); 1385 } 1386 1387 void rawv6_exit(void) 1388 { 1389 inet6_unregister_protosw(&rawv6_protosw); 1390 } 1391