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 <asm/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 static struct raw_hashinfo raw_v6_hashinfo = { 69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock), 70 }; 71 72 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk, 73 unsigned short num, const struct in6_addr *loc_addr, 74 const struct in6_addr *rmt_addr, int dif) 75 { 76 bool is_multicast = ipv6_addr_is_multicast(loc_addr); 77 78 sk_for_each_from(sk) 79 if (inet_sk(sk)->inet_num == num) { 80 81 if (!net_eq(sock_net(sk), net)) 82 continue; 83 84 if (!ipv6_addr_any(&sk->sk_v6_daddr) && 85 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) 86 continue; 87 88 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) 89 continue; 90 91 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) { 92 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)) 93 goto found; 94 if (is_multicast && 95 inet6_mc_check(sk, loc_addr, rmt_addr)) 96 goto found; 97 continue; 98 } 99 goto found; 100 } 101 sk = NULL; 102 found: 103 return sk; 104 } 105 106 /* 107 * 0 - deliver 108 * 1 - block 109 */ 110 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb) 111 { 112 struct icmp6hdr _hdr; 113 const struct icmp6hdr *hdr; 114 115 /* We require only the four bytes of the ICMPv6 header, not any 116 * additional bytes of message body in "struct icmp6hdr". 117 */ 118 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 119 ICMPV6_HDRLEN, &_hdr); 120 if (hdr) { 121 const __u32 *data = &raw6_sk(sk)->filter.data[0]; 122 unsigned int type = hdr->icmp6_type; 123 124 return (data[type >> 5] & (1U << (type & 31))) != 0; 125 } 126 return 1; 127 } 128 129 #if IS_ENABLED(CONFIG_IPV6_MIP6) 130 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb); 131 132 static mh_filter_t __rcu *mh_filter __read_mostly; 133 134 int rawv6_mh_filter_register(mh_filter_t filter) 135 { 136 rcu_assign_pointer(mh_filter, filter); 137 return 0; 138 } 139 EXPORT_SYMBOL(rawv6_mh_filter_register); 140 141 int rawv6_mh_filter_unregister(mh_filter_t filter) 142 { 143 RCU_INIT_POINTER(mh_filter, NULL); 144 synchronize_rcu(); 145 return 0; 146 } 147 EXPORT_SYMBOL(rawv6_mh_filter_unregister); 148 149 #endif 150 151 /* 152 * demultiplex raw sockets. 153 * (should consider queueing the skb in the sock receive_queue 154 * without calling rawv6.c) 155 * 156 * Caller owns SKB so we must make clones. 157 */ 158 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 159 { 160 const struct in6_addr *saddr; 161 const struct in6_addr *daddr; 162 struct sock *sk; 163 bool delivered = false; 164 __u8 hash; 165 struct net *net; 166 167 saddr = &ipv6_hdr(skb)->saddr; 168 daddr = saddr + 1; 169 170 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 171 172 read_lock(&raw_v6_hashinfo.lock); 173 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 174 175 if (sk == NULL) 176 goto out; 177 178 net = dev_net(skb->dev); 179 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif); 180 181 while (sk) { 182 int filtered; 183 184 delivered = true; 185 switch (nexthdr) { 186 case IPPROTO_ICMPV6: 187 filtered = icmpv6_filter(sk, skb); 188 break; 189 190 #if IS_ENABLED(CONFIG_IPV6_MIP6) 191 case IPPROTO_MH: 192 { 193 /* XXX: To validate MH only once for each packet, 194 * this is placed here. It should be after checking 195 * xfrm policy, however it doesn't. The checking xfrm 196 * policy is placed in rawv6_rcv() because it is 197 * required for each socket. 198 */ 199 mh_filter_t *filter; 200 201 filter = rcu_dereference(mh_filter); 202 filtered = filter ? (*filter)(sk, skb) : 0; 203 break; 204 } 205 #endif 206 default: 207 filtered = 0; 208 break; 209 } 210 211 if (filtered < 0) 212 break; 213 if (filtered == 0) { 214 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 215 216 /* Not releasing hash table! */ 217 if (clone) { 218 nf_reset(clone); 219 rawv6_rcv(sk, clone); 220 } 221 } 222 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr, 223 IP6CB(skb)->iif); 224 } 225 out: 226 read_unlock(&raw_v6_hashinfo.lock); 227 return delivered; 228 } 229 230 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr) 231 { 232 struct sock *raw_sk; 233 234 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]); 235 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr)) 236 raw_sk = NULL; 237 238 return raw_sk != NULL; 239 } 240 241 /* This cleans up af_inet6 a bit. -DaveM */ 242 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 243 { 244 struct inet_sock *inet = inet_sk(sk); 245 struct ipv6_pinfo *np = inet6_sk(sk); 246 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 247 __be32 v4addr = 0; 248 int addr_type; 249 int err; 250 251 if (addr_len < SIN6_LEN_RFC2133) 252 return -EINVAL; 253 254 if (addr->sin6_family != AF_INET6) 255 return -EINVAL; 256 257 addr_type = ipv6_addr_type(&addr->sin6_addr); 258 259 /* Raw sockets are IPv6 only */ 260 if (addr_type == IPV6_ADDR_MAPPED) 261 return -EADDRNOTAVAIL; 262 263 lock_sock(sk); 264 265 err = -EINVAL; 266 if (sk->sk_state != TCP_CLOSE) 267 goto out; 268 269 rcu_read_lock(); 270 /* Check if the address belongs to the host. */ 271 if (addr_type != IPV6_ADDR_ANY) { 272 struct net_device *dev = NULL; 273 274 if (__ipv6_addr_needs_scope_id(addr_type)) { 275 if (addr_len >= sizeof(struct sockaddr_in6) && 276 addr->sin6_scope_id) { 277 /* Override any existing binding, if another 278 * one is supplied by user. 279 */ 280 sk->sk_bound_dev_if = addr->sin6_scope_id; 281 } 282 283 /* Binding to link-local address requires an interface */ 284 if (!sk->sk_bound_dev_if) 285 goto out_unlock; 286 287 err = -ENODEV; 288 dev = dev_get_by_index_rcu(sock_net(sk), 289 sk->sk_bound_dev_if); 290 if (!dev) 291 goto out_unlock; 292 } 293 294 /* ipv4 addr of the socket is invalid. Only the 295 * unspecified and mapped address have a v4 equivalent. 296 */ 297 v4addr = LOOPBACK4_IPV6; 298 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 299 err = -EADDRNOTAVAIL; 300 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr, 301 dev, 0)) { 302 goto out_unlock; 303 } 304 } 305 } 306 307 inet->inet_rcv_saddr = inet->inet_saddr = v4addr; 308 sk->sk_v6_rcv_saddr = addr->sin6_addr; 309 if (!(addr_type & IPV6_ADDR_MULTICAST)) 310 np->saddr = addr->sin6_addr; 311 err = 0; 312 out_unlock: 313 rcu_read_unlock(); 314 out: 315 release_sock(sk); 316 return err; 317 } 318 319 static void rawv6_err(struct sock *sk, struct sk_buff *skb, 320 struct inet6_skb_parm *opt, 321 u8 type, u8 code, int offset, __be32 info) 322 { 323 struct inet_sock *inet = inet_sk(sk); 324 struct ipv6_pinfo *np = inet6_sk(sk); 325 int err; 326 int harderr; 327 328 /* Report error on raw socket, if: 329 1. User requested recverr. 330 2. Socket is connected (otherwise the error indication 331 is useless without recverr and error is hard. 332 */ 333 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 334 return; 335 336 harderr = icmpv6_err_convert(type, code, &err); 337 if (type == ICMPV6_PKT_TOOBIG) { 338 ip6_sk_update_pmtu(skb, sk, info); 339 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 340 } 341 if (type == NDISC_REDIRECT) { 342 ip6_sk_redirect(skb, sk); 343 return; 344 } 345 if (np->recverr) { 346 u8 *payload = skb->data; 347 if (!inet->hdrincl) 348 payload += offset; 349 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 350 } 351 352 if (np->recverr || harderr) { 353 sk->sk_err = err; 354 sk->sk_error_report(sk); 355 } 356 } 357 358 void raw6_icmp_error(struct sk_buff *skb, int nexthdr, 359 u8 type, u8 code, int inner_offset, __be32 info) 360 { 361 struct sock *sk; 362 int hash; 363 const struct in6_addr *saddr, *daddr; 364 struct net *net; 365 366 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 367 368 read_lock(&raw_v6_hashinfo.lock); 369 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 370 if (sk != NULL) { 371 /* Note: ipv6_hdr(skb) != skb->data */ 372 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data; 373 saddr = &ip6h->saddr; 374 daddr = &ip6h->daddr; 375 net = dev_net(skb->dev); 376 377 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr, 378 IP6CB(skb)->iif))) { 379 rawv6_err(sk, skb, NULL, type, code, 380 inner_offset, info); 381 sk = sk_next(sk); 382 } 383 } 384 read_unlock(&raw_v6_hashinfo.lock); 385 } 386 387 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb) 388 { 389 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) && 390 skb_checksum_complete(skb)) { 391 atomic_inc(&sk->sk_drops); 392 kfree_skb(skb); 393 return NET_RX_DROP; 394 } 395 396 /* Charge it to the socket. */ 397 skb_dst_drop(skb); 398 if (sock_queue_rcv_skb(sk, skb) < 0) { 399 kfree_skb(skb); 400 return NET_RX_DROP; 401 } 402 403 return 0; 404 } 405 406 /* 407 * This is next to useless... 408 * if we demultiplex in network layer we don't need the extra call 409 * just to queue the skb... 410 * maybe we could have the network decide upon a hint if it 411 * should call raw_rcv for demultiplexing 412 */ 413 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 414 { 415 struct inet_sock *inet = inet_sk(sk); 416 struct raw6_sock *rp = raw6_sk(sk); 417 418 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 419 atomic_inc(&sk->sk_drops); 420 kfree_skb(skb); 421 return NET_RX_DROP; 422 } 423 424 if (!rp->checksum) 425 skb->ip_summed = CHECKSUM_UNNECESSARY; 426 427 if (skb->ip_summed == CHECKSUM_COMPLETE) { 428 skb_postpull_rcsum(skb, skb_network_header(skb), 429 skb_network_header_len(skb)); 430 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 431 &ipv6_hdr(skb)->daddr, 432 skb->len, inet->inet_num, skb->csum)) 433 skb->ip_summed = CHECKSUM_UNNECESSARY; 434 } 435 if (!skb_csum_unnecessary(skb)) 436 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 437 &ipv6_hdr(skb)->daddr, 438 skb->len, 439 inet->inet_num, 0)); 440 441 if (inet->hdrincl) { 442 if (skb_checksum_complete(skb)) { 443 atomic_inc(&sk->sk_drops); 444 kfree_skb(skb); 445 return NET_RX_DROP; 446 } 447 } 448 449 rawv6_rcv_skb(sk, skb); 450 return 0; 451 } 452 453 454 /* 455 * This should be easy, if there is something there 456 * we return it, otherwise we block. 457 */ 458 459 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk, 460 struct msghdr *msg, size_t len, 461 int noblock, int flags, int *addr_len) 462 { 463 struct ipv6_pinfo *np = inet6_sk(sk); 464 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 465 struct sk_buff *skb; 466 size_t copied; 467 int err; 468 469 if (flags & MSG_OOB) 470 return -EOPNOTSUPP; 471 472 if (flags & MSG_ERRQUEUE) 473 return ipv6_recv_error(sk, msg, len, addr_len); 474 475 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 476 return ipv6_recv_rxpmtu(sk, msg, len, addr_len); 477 478 skb = skb_recv_datagram(sk, flags, noblock, &err); 479 if (!skb) 480 goto out; 481 482 copied = skb->len; 483 if (copied > len) { 484 copied = len; 485 msg->msg_flags |= MSG_TRUNC; 486 } 487 488 if (skb_csum_unnecessary(skb)) { 489 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 490 } else if (msg->msg_flags&MSG_TRUNC) { 491 if (__skb_checksum_complete(skb)) 492 goto csum_copy_err; 493 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 494 } else { 495 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov); 496 if (err == -EINVAL) 497 goto csum_copy_err; 498 } 499 if (err) 500 goto out_free; 501 502 /* Copy the address. */ 503 if (sin6) { 504 sin6->sin6_family = AF_INET6; 505 sin6->sin6_port = 0; 506 sin6->sin6_addr = ipv6_hdr(skb)->saddr; 507 sin6->sin6_flowinfo = 0; 508 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr, 509 IP6CB(skb)->iif); 510 *addr_len = sizeof(*sin6); 511 } 512 513 sock_recv_ts_and_drops(msg, sk, skb); 514 515 if (np->rxopt.all) 516 ip6_datagram_recv_ctl(sk, msg, skb); 517 518 err = copied; 519 if (flags & MSG_TRUNC) 520 err = skb->len; 521 522 out_free: 523 skb_free_datagram(sk, skb); 524 out: 525 return err; 526 527 csum_copy_err: 528 skb_kill_datagram(sk, skb, flags); 529 530 /* Error for blocking case is chosen to masquerade 531 as some normal condition. 532 */ 533 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 534 goto out; 535 } 536 537 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 538 struct raw6_sock *rp) 539 { 540 struct sk_buff *skb; 541 int err = 0; 542 int offset; 543 int len; 544 int total_len; 545 __wsum tmp_csum; 546 __sum16 csum; 547 548 if (!rp->checksum) 549 goto send; 550 551 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 552 goto out; 553 554 offset = rp->offset; 555 total_len = inet_sk(sk)->cork.base.length; 556 if (offset >= total_len - 1) { 557 err = -EINVAL; 558 ip6_flush_pending_frames(sk); 559 goto out; 560 } 561 562 /* should be check HW csum miyazawa */ 563 if (skb_queue_len(&sk->sk_write_queue) == 1) { 564 /* 565 * Only one fragment on the socket. 566 */ 567 tmp_csum = skb->csum; 568 } else { 569 struct sk_buff *csum_skb = NULL; 570 tmp_csum = 0; 571 572 skb_queue_walk(&sk->sk_write_queue, skb) { 573 tmp_csum = csum_add(tmp_csum, skb->csum); 574 575 if (csum_skb) 576 continue; 577 578 len = skb->len - skb_transport_offset(skb); 579 if (offset >= len) { 580 offset -= len; 581 continue; 582 } 583 584 csum_skb = skb; 585 } 586 587 skb = csum_skb; 588 } 589 590 offset += skb_transport_offset(skb); 591 if (skb_copy_bits(skb, offset, &csum, 2)) 592 BUG(); 593 594 /* in case cksum was not initialized */ 595 if (unlikely(csum)) 596 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 597 598 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 599 total_len, fl6->flowi6_proto, tmp_csum); 600 601 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP) 602 csum = CSUM_MANGLED_0; 603 604 if (skb_store_bits(skb, offset, &csum, 2)) 605 BUG(); 606 607 send: 608 err = ip6_push_pending_frames(sk); 609 out: 610 return err; 611 } 612 613 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length, 614 struct flowi6 *fl6, struct dst_entry **dstp, 615 unsigned int flags) 616 { 617 struct ipv6_pinfo *np = inet6_sk(sk); 618 struct ipv6hdr *iph; 619 struct sk_buff *skb; 620 int err; 621 struct rt6_info *rt = (struct rt6_info *)*dstp; 622 int hlen = LL_RESERVED_SPACE(rt->dst.dev); 623 int tlen = rt->dst.dev->needed_tailroom; 624 625 if (length > rt->dst.dev->mtu) { 626 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu); 627 return -EMSGSIZE; 628 } 629 if (flags&MSG_PROBE) 630 goto out; 631 632 skb = sock_alloc_send_skb(sk, 633 length + hlen + tlen + 15, 634 flags & MSG_DONTWAIT, &err); 635 if (skb == NULL) 636 goto error; 637 skb_reserve(skb, hlen); 638 639 skb->protocol = htons(ETH_P_IPV6); 640 skb->priority = sk->sk_priority; 641 skb->mark = sk->sk_mark; 642 skb_dst_set(skb, &rt->dst); 643 *dstp = NULL; 644 645 skb_put(skb, length); 646 skb_reset_network_header(skb); 647 iph = ipv6_hdr(skb); 648 649 skb->ip_summed = CHECKSUM_NONE; 650 651 skb->transport_header = skb->network_header; 652 err = memcpy_fromiovecend((void *)iph, from, 0, length); 653 if (err) 654 goto error_fault; 655 656 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); 657 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, 658 rt->dst.dev, dst_output); 659 if (err > 0) 660 err = net_xmit_errno(err); 661 if (err) 662 goto error; 663 out: 664 return 0; 665 666 error_fault: 667 err = -EFAULT; 668 kfree_skb(skb); 669 error: 670 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 671 if (err == -ENOBUFS && !np->recverr) 672 err = 0; 673 return err; 674 } 675 676 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg) 677 { 678 struct iovec *iov; 679 u8 __user *type = NULL; 680 u8 __user *code = NULL; 681 u8 len = 0; 682 int probed = 0; 683 int i; 684 685 if (!msg->msg_iov) 686 return 0; 687 688 for (i = 0; i < msg->msg_iovlen; i++) { 689 iov = &msg->msg_iov[i]; 690 if (!iov) 691 continue; 692 693 switch (fl6->flowi6_proto) { 694 case IPPROTO_ICMPV6: 695 /* check if one-byte field is readable or not. */ 696 if (iov->iov_base && iov->iov_len < 1) 697 break; 698 699 if (!type) { 700 type = iov->iov_base; 701 /* check if code field is readable or not. */ 702 if (iov->iov_len > 1) 703 code = type + 1; 704 } else if (!code) 705 code = iov->iov_base; 706 707 if (type && code) { 708 if (get_user(fl6->fl6_icmp_type, type) || 709 get_user(fl6->fl6_icmp_code, code)) 710 return -EFAULT; 711 probed = 1; 712 } 713 break; 714 case IPPROTO_MH: 715 if (iov->iov_base && iov->iov_len < 1) 716 break; 717 /* check if type field is readable or not. */ 718 if (iov->iov_len > 2 - len) { 719 u8 __user *p = iov->iov_base; 720 if (get_user(fl6->fl6_mh_type, &p[2 - len])) 721 return -EFAULT; 722 probed = 1; 723 } else 724 len += iov->iov_len; 725 726 break; 727 default: 728 probed = 1; 729 break; 730 } 731 if (probed) 732 break; 733 } 734 return 0; 735 } 736 737 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk, 738 struct msghdr *msg, size_t len) 739 { 740 struct ipv6_txoptions opt_space; 741 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 742 struct in6_addr *daddr, *final_p, final; 743 struct inet_sock *inet = inet_sk(sk); 744 struct ipv6_pinfo *np = inet6_sk(sk); 745 struct raw6_sock *rp = raw6_sk(sk); 746 struct ipv6_txoptions *opt = NULL; 747 struct ip6_flowlabel *flowlabel = NULL; 748 struct dst_entry *dst = NULL; 749 struct flowi6 fl6; 750 int addr_len = msg->msg_namelen; 751 int hlimit = -1; 752 int tclass = -1; 753 int dontfrag = -1; 754 u16 proto; 755 int err; 756 757 /* Rough check on arithmetic overflow, 758 better check is made in ip6_append_data(). 759 */ 760 if (len > INT_MAX) 761 return -EMSGSIZE; 762 763 /* Mirror BSD error message compatibility */ 764 if (msg->msg_flags & MSG_OOB) 765 return -EOPNOTSUPP; 766 767 /* 768 * Get and verify the address. 769 */ 770 memset(&fl6, 0, sizeof(fl6)); 771 772 fl6.flowi6_mark = sk->sk_mark; 773 774 if (sin6) { 775 if (addr_len < SIN6_LEN_RFC2133) 776 return -EINVAL; 777 778 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 779 return -EAFNOSUPPORT; 780 781 /* port is the proto value [0..255] carried in nexthdr */ 782 proto = ntohs(sin6->sin6_port); 783 784 if (!proto) 785 proto = inet->inet_num; 786 else if (proto != inet->inet_num) 787 return -EINVAL; 788 789 if (proto > 255) 790 return -EINVAL; 791 792 daddr = &sin6->sin6_addr; 793 if (np->sndflow) { 794 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 795 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 796 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 797 if (flowlabel == NULL) 798 return -EINVAL; 799 } 800 } 801 802 /* 803 * Otherwise it will be difficult to maintain 804 * sk->sk_dst_cache. 805 */ 806 if (sk->sk_state == TCP_ESTABLISHED && 807 ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) 808 daddr = &sk->sk_v6_daddr; 809 810 if (addr_len >= sizeof(struct sockaddr_in6) && 811 sin6->sin6_scope_id && 812 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) 813 fl6.flowi6_oif = sin6->sin6_scope_id; 814 } else { 815 if (sk->sk_state != TCP_ESTABLISHED) 816 return -EDESTADDRREQ; 817 818 proto = inet->inet_num; 819 daddr = &sk->sk_v6_daddr; 820 fl6.flowlabel = np->flow_label; 821 } 822 823 if (fl6.flowi6_oif == 0) 824 fl6.flowi6_oif = sk->sk_bound_dev_if; 825 826 if (msg->msg_controllen) { 827 opt = &opt_space; 828 memset(opt, 0, sizeof(struct ipv6_txoptions)); 829 opt->tot_len = sizeof(struct ipv6_txoptions); 830 831 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt, 832 &hlimit, &tclass, &dontfrag); 833 if (err < 0) { 834 fl6_sock_release(flowlabel); 835 return err; 836 } 837 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 838 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 839 if (flowlabel == NULL) 840 return -EINVAL; 841 } 842 if (!(opt->opt_nflen|opt->opt_flen)) 843 opt = NULL; 844 } 845 if (opt == NULL) 846 opt = np->opt; 847 if (flowlabel) 848 opt = fl6_merge_options(&opt_space, flowlabel, opt); 849 opt = ipv6_fixup_options(&opt_space, opt); 850 851 fl6.flowi6_proto = proto; 852 err = rawv6_probe_proto_opt(&fl6, msg); 853 if (err) 854 goto out; 855 856 if (!ipv6_addr_any(daddr)) 857 fl6.daddr = *daddr; 858 else 859 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 860 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 861 fl6.saddr = np->saddr; 862 863 final_p = fl6_update_dst(&fl6, opt, &final); 864 865 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 866 fl6.flowi6_oif = np->mcast_oif; 867 else if (!fl6.flowi6_oif) 868 fl6.flowi6_oif = np->ucast_oif; 869 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 870 871 dst = ip6_dst_lookup_flow(sk, &fl6, final_p); 872 if (IS_ERR(dst)) { 873 err = PTR_ERR(dst); 874 goto out; 875 } 876 if (hlimit < 0) 877 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 878 879 if (tclass < 0) 880 tclass = np->tclass; 881 882 if (dontfrag < 0) 883 dontfrag = np->dontfrag; 884 885 if (msg->msg_flags&MSG_CONFIRM) 886 goto do_confirm; 887 888 back_from_confirm: 889 if (inet->hdrincl) 890 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags); 891 else { 892 lock_sock(sk); 893 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, 894 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst, 895 msg->msg_flags, dontfrag); 896 897 if (err) 898 ip6_flush_pending_frames(sk); 899 else if (!(msg->msg_flags & MSG_MORE)) 900 err = rawv6_push_pending_frames(sk, &fl6, rp); 901 release_sock(sk); 902 } 903 done: 904 dst_release(dst); 905 out: 906 fl6_sock_release(flowlabel); 907 return err<0?err:len; 908 do_confirm: 909 dst_confirm(dst); 910 if (!(msg->msg_flags & MSG_PROBE) || len) 911 goto back_from_confirm; 912 err = 0; 913 goto done; 914 } 915 916 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 917 char __user *optval, int optlen) 918 { 919 switch (optname) { 920 case ICMPV6_FILTER: 921 if (optlen > sizeof(struct icmp6_filter)) 922 optlen = sizeof(struct icmp6_filter); 923 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 924 return -EFAULT; 925 return 0; 926 default: 927 return -ENOPROTOOPT; 928 } 929 930 return 0; 931 } 932 933 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 934 char __user *optval, int __user *optlen) 935 { 936 int len; 937 938 switch (optname) { 939 case ICMPV6_FILTER: 940 if (get_user(len, optlen)) 941 return -EFAULT; 942 if (len < 0) 943 return -EINVAL; 944 if (len > sizeof(struct icmp6_filter)) 945 len = sizeof(struct icmp6_filter); 946 if (put_user(len, optlen)) 947 return -EFAULT; 948 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 949 return -EFAULT; 950 return 0; 951 default: 952 return -ENOPROTOOPT; 953 } 954 955 return 0; 956 } 957 958 959 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 960 char __user *optval, unsigned int optlen) 961 { 962 struct raw6_sock *rp = raw6_sk(sk); 963 int val; 964 965 if (get_user(val, (int __user *)optval)) 966 return -EFAULT; 967 968 switch (optname) { 969 case IPV6_CHECKSUM: 970 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && 971 level == IPPROTO_IPV6) { 972 /* 973 * RFC3542 tells that IPV6_CHECKSUM socket 974 * option in the IPPROTO_IPV6 level is not 975 * allowed on ICMPv6 sockets. 976 * If you want to set it, use IPPROTO_RAW 977 * level IPV6_CHECKSUM socket option 978 * (Linux extension). 979 */ 980 return -EINVAL; 981 } 982 983 /* You may get strange result with a positive odd offset; 984 RFC2292bis agrees with me. */ 985 if (val > 0 && (val&1)) 986 return -EINVAL; 987 if (val < 0) { 988 rp->checksum = 0; 989 } else { 990 rp->checksum = 1; 991 rp->offset = val; 992 } 993 994 return 0; 995 996 default: 997 return -ENOPROTOOPT; 998 } 999 } 1000 1001 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 1002 char __user *optval, unsigned int optlen) 1003 { 1004 switch (level) { 1005 case SOL_RAW: 1006 break; 1007 1008 case SOL_ICMPV6: 1009 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1010 return -EOPNOTSUPP; 1011 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1012 case SOL_IPV6: 1013 if (optname == IPV6_CHECKSUM) 1014 break; 1015 default: 1016 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1017 } 1018 1019 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1020 } 1021 1022 #ifdef CONFIG_COMPAT 1023 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 1024 char __user *optval, unsigned int optlen) 1025 { 1026 switch (level) { 1027 case SOL_RAW: 1028 break; 1029 case SOL_ICMPV6: 1030 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1031 return -EOPNOTSUPP; 1032 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1033 case SOL_IPV6: 1034 if (optname == IPV6_CHECKSUM) 1035 break; 1036 default: 1037 return compat_ipv6_setsockopt(sk, level, optname, 1038 optval, optlen); 1039 } 1040 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1041 } 1042 #endif 1043 1044 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1045 char __user *optval, int __user *optlen) 1046 { 1047 struct raw6_sock *rp = raw6_sk(sk); 1048 int val, len; 1049 1050 if (get_user(len,optlen)) 1051 return -EFAULT; 1052 1053 switch (optname) { 1054 case IPV6_CHECKSUM: 1055 /* 1056 * We allow getsockopt() for IPPROTO_IPV6-level 1057 * IPV6_CHECKSUM socket option on ICMPv6 sockets 1058 * since RFC3542 is silent about it. 1059 */ 1060 if (rp->checksum == 0) 1061 val = -1; 1062 else 1063 val = rp->offset; 1064 break; 1065 1066 default: 1067 return -ENOPROTOOPT; 1068 } 1069 1070 len = min_t(unsigned int, sizeof(int), len); 1071 1072 if (put_user(len, optlen)) 1073 return -EFAULT; 1074 if (copy_to_user(optval,&val,len)) 1075 return -EFAULT; 1076 return 0; 1077 } 1078 1079 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1080 char __user *optval, int __user *optlen) 1081 { 1082 switch (level) { 1083 case SOL_RAW: 1084 break; 1085 1086 case SOL_ICMPV6: 1087 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1088 return -EOPNOTSUPP; 1089 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1090 case SOL_IPV6: 1091 if (optname == IPV6_CHECKSUM) 1092 break; 1093 default: 1094 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1095 } 1096 1097 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1098 } 1099 1100 #ifdef CONFIG_COMPAT 1101 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1102 char __user *optval, int __user *optlen) 1103 { 1104 switch (level) { 1105 case SOL_RAW: 1106 break; 1107 case SOL_ICMPV6: 1108 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1109 return -EOPNOTSUPP; 1110 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1111 case SOL_IPV6: 1112 if (optname == IPV6_CHECKSUM) 1113 break; 1114 default: 1115 return compat_ipv6_getsockopt(sk, level, optname, 1116 optval, optlen); 1117 } 1118 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1119 } 1120 #endif 1121 1122 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1123 { 1124 switch (cmd) { 1125 case SIOCOUTQ: { 1126 int amount = sk_wmem_alloc_get(sk); 1127 1128 return put_user(amount, (int __user *)arg); 1129 } 1130 case SIOCINQ: { 1131 struct sk_buff *skb; 1132 int amount = 0; 1133 1134 spin_lock_bh(&sk->sk_receive_queue.lock); 1135 skb = skb_peek(&sk->sk_receive_queue); 1136 if (skb != NULL) 1137 amount = skb_tail_pointer(skb) - 1138 skb_transport_header(skb); 1139 spin_unlock_bh(&sk->sk_receive_queue.lock); 1140 return put_user(amount, (int __user *)arg); 1141 } 1142 1143 default: 1144 #ifdef CONFIG_IPV6_MROUTE 1145 return ip6mr_ioctl(sk, cmd, (void __user *)arg); 1146 #else 1147 return -ENOIOCTLCMD; 1148 #endif 1149 } 1150 } 1151 1152 #ifdef CONFIG_COMPAT 1153 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 1154 { 1155 switch (cmd) { 1156 case SIOCOUTQ: 1157 case SIOCINQ: 1158 return -ENOIOCTLCMD; 1159 default: 1160 #ifdef CONFIG_IPV6_MROUTE 1161 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg)); 1162 #else 1163 return -ENOIOCTLCMD; 1164 #endif 1165 } 1166 } 1167 #endif 1168 1169 static void rawv6_close(struct sock *sk, long timeout) 1170 { 1171 if (inet_sk(sk)->inet_num == IPPROTO_RAW) 1172 ip6_ra_control(sk, -1); 1173 ip6mr_sk_done(sk); 1174 sk_common_release(sk); 1175 } 1176 1177 static void raw6_destroy(struct sock *sk) 1178 { 1179 lock_sock(sk); 1180 ip6_flush_pending_frames(sk); 1181 release_sock(sk); 1182 1183 inet6_destroy_sock(sk); 1184 } 1185 1186 static int rawv6_init_sk(struct sock *sk) 1187 { 1188 struct raw6_sock *rp = raw6_sk(sk); 1189 1190 switch (inet_sk(sk)->inet_num) { 1191 case IPPROTO_ICMPV6: 1192 rp->checksum = 1; 1193 rp->offset = 2; 1194 break; 1195 case IPPROTO_MH: 1196 rp->checksum = 1; 1197 rp->offset = 4; 1198 break; 1199 default: 1200 break; 1201 } 1202 return 0; 1203 } 1204 1205 struct proto rawv6_prot = { 1206 .name = "RAWv6", 1207 .owner = THIS_MODULE, 1208 .close = rawv6_close, 1209 .destroy = raw6_destroy, 1210 .connect = ip6_datagram_connect_v6_only, 1211 .disconnect = udp_disconnect, 1212 .ioctl = rawv6_ioctl, 1213 .init = rawv6_init_sk, 1214 .setsockopt = rawv6_setsockopt, 1215 .getsockopt = rawv6_getsockopt, 1216 .sendmsg = rawv6_sendmsg, 1217 .recvmsg = rawv6_recvmsg, 1218 .bind = rawv6_bind, 1219 .backlog_rcv = rawv6_rcv_skb, 1220 .hash = raw_hash_sk, 1221 .unhash = raw_unhash_sk, 1222 .obj_size = sizeof(struct raw6_sock), 1223 .h.raw_hash = &raw_v6_hashinfo, 1224 #ifdef CONFIG_COMPAT 1225 .compat_setsockopt = compat_rawv6_setsockopt, 1226 .compat_getsockopt = compat_rawv6_getsockopt, 1227 .compat_ioctl = compat_rawv6_ioctl, 1228 #endif 1229 }; 1230 1231 #ifdef CONFIG_PROC_FS 1232 static int raw6_seq_show(struct seq_file *seq, void *v) 1233 { 1234 if (v == SEQ_START_TOKEN) { 1235 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); 1236 } else { 1237 struct sock *sp = v; 1238 __u16 srcp = inet_sk(sp)->inet_num; 1239 ip6_dgram_sock_seq_show(seq, v, srcp, 0, 1240 raw_seq_private(seq)->bucket); 1241 } 1242 return 0; 1243 } 1244 1245 static const struct seq_operations raw6_seq_ops = { 1246 .start = raw_seq_start, 1247 .next = raw_seq_next, 1248 .stop = raw_seq_stop, 1249 .show = raw6_seq_show, 1250 }; 1251 1252 static int raw6_seq_open(struct inode *inode, struct file *file) 1253 { 1254 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops); 1255 } 1256 1257 static const struct file_operations raw6_seq_fops = { 1258 .owner = THIS_MODULE, 1259 .open = raw6_seq_open, 1260 .read = seq_read, 1261 .llseek = seq_lseek, 1262 .release = seq_release_net, 1263 }; 1264 1265 static int __net_init raw6_init_net(struct net *net) 1266 { 1267 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops)) 1268 return -ENOMEM; 1269 1270 return 0; 1271 } 1272 1273 static void __net_exit raw6_exit_net(struct net *net) 1274 { 1275 remove_proc_entry("raw6", net->proc_net); 1276 } 1277 1278 static struct pernet_operations raw6_net_ops = { 1279 .init = raw6_init_net, 1280 .exit = raw6_exit_net, 1281 }; 1282 1283 int __init raw6_proc_init(void) 1284 { 1285 return register_pernet_subsys(&raw6_net_ops); 1286 } 1287 1288 void raw6_proc_exit(void) 1289 { 1290 unregister_pernet_subsys(&raw6_net_ops); 1291 } 1292 #endif /* CONFIG_PROC_FS */ 1293 1294 /* Same as inet6_dgram_ops, sans udp_poll. */ 1295 static const struct proto_ops inet6_sockraw_ops = { 1296 .family = PF_INET6, 1297 .owner = THIS_MODULE, 1298 .release = inet6_release, 1299 .bind = inet6_bind, 1300 .connect = inet_dgram_connect, /* ok */ 1301 .socketpair = sock_no_socketpair, /* a do nothing */ 1302 .accept = sock_no_accept, /* a do nothing */ 1303 .getname = inet6_getname, 1304 .poll = datagram_poll, /* ok */ 1305 .ioctl = inet6_ioctl, /* must change */ 1306 .listen = sock_no_listen, /* ok */ 1307 .shutdown = inet_shutdown, /* ok */ 1308 .setsockopt = sock_common_setsockopt, /* ok */ 1309 .getsockopt = sock_common_getsockopt, /* ok */ 1310 .sendmsg = inet_sendmsg, /* ok */ 1311 .recvmsg = sock_common_recvmsg, /* ok */ 1312 .mmap = sock_no_mmap, 1313 .sendpage = sock_no_sendpage, 1314 #ifdef CONFIG_COMPAT 1315 .compat_setsockopt = compat_sock_common_setsockopt, 1316 .compat_getsockopt = compat_sock_common_getsockopt, 1317 #endif 1318 }; 1319 1320 static struct inet_protosw rawv6_protosw = { 1321 .type = SOCK_RAW, 1322 .protocol = IPPROTO_IP, /* wild card */ 1323 .prot = &rawv6_prot, 1324 .ops = &inet6_sockraw_ops, 1325 .flags = INET_PROTOSW_REUSE, 1326 }; 1327 1328 int __init rawv6_init(void) 1329 { 1330 int ret; 1331 1332 ret = inet6_register_protosw(&rawv6_protosw); 1333 if (ret) 1334 goto out; 1335 out: 1336 return ret; 1337 } 1338 1339 void rawv6_exit(void) 1340 { 1341 inet6_unregister_protosw(&rawv6_protosw); 1342 } 1343