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 addr_type = ipv6_addr_type(&addr->sin6_addr); 254 255 /* Raw sockets are IPv6 only */ 256 if (addr_type == IPV6_ADDR_MAPPED) 257 return -EADDRNOTAVAIL; 258 259 lock_sock(sk); 260 261 err = -EINVAL; 262 if (sk->sk_state != TCP_CLOSE) 263 goto out; 264 265 rcu_read_lock(); 266 /* Check if the address belongs to the host. */ 267 if (addr_type != IPV6_ADDR_ANY) { 268 struct net_device *dev = NULL; 269 270 if (__ipv6_addr_needs_scope_id(addr_type)) { 271 if (addr_len >= sizeof(struct sockaddr_in6) && 272 addr->sin6_scope_id) { 273 /* Override any existing binding, if another 274 * one is supplied by user. 275 */ 276 sk->sk_bound_dev_if = addr->sin6_scope_id; 277 } 278 279 /* Binding to link-local address requires an interface */ 280 if (!sk->sk_bound_dev_if) 281 goto out_unlock; 282 283 err = -ENODEV; 284 dev = dev_get_by_index_rcu(sock_net(sk), 285 sk->sk_bound_dev_if); 286 if (!dev) 287 goto out_unlock; 288 } 289 290 /* ipv4 addr of the socket is invalid. Only the 291 * unspecified and mapped address have a v4 equivalent. 292 */ 293 v4addr = LOOPBACK4_IPV6; 294 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 295 err = -EADDRNOTAVAIL; 296 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr, 297 dev, 0)) { 298 goto out_unlock; 299 } 300 } 301 } 302 303 inet->inet_rcv_saddr = inet->inet_saddr = v4addr; 304 sk->sk_v6_rcv_saddr = addr->sin6_addr; 305 if (!(addr_type & IPV6_ADDR_MULTICAST)) 306 np->saddr = addr->sin6_addr; 307 err = 0; 308 out_unlock: 309 rcu_read_unlock(); 310 out: 311 release_sock(sk); 312 return err; 313 } 314 315 static void rawv6_err(struct sock *sk, struct sk_buff *skb, 316 struct inet6_skb_parm *opt, 317 u8 type, u8 code, int offset, __be32 info) 318 { 319 struct inet_sock *inet = inet_sk(sk); 320 struct ipv6_pinfo *np = inet6_sk(sk); 321 int err; 322 int harderr; 323 324 /* Report error on raw socket, if: 325 1. User requested recverr. 326 2. Socket is connected (otherwise the error indication 327 is useless without recverr and error is hard. 328 */ 329 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 330 return; 331 332 harderr = icmpv6_err_convert(type, code, &err); 333 if (type == ICMPV6_PKT_TOOBIG) { 334 ip6_sk_update_pmtu(skb, sk, info); 335 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 336 } 337 if (type == NDISC_REDIRECT) { 338 ip6_sk_redirect(skb, sk); 339 return; 340 } 341 if (np->recverr) { 342 u8 *payload = skb->data; 343 if (!inet->hdrincl) 344 payload += offset; 345 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 346 } 347 348 if (np->recverr || harderr) { 349 sk->sk_err = err; 350 sk->sk_error_report(sk); 351 } 352 } 353 354 void raw6_icmp_error(struct sk_buff *skb, int nexthdr, 355 u8 type, u8 code, int inner_offset, __be32 info) 356 { 357 struct sock *sk; 358 int hash; 359 const struct in6_addr *saddr, *daddr; 360 struct net *net; 361 362 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 363 364 read_lock(&raw_v6_hashinfo.lock); 365 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 366 if (sk != NULL) { 367 /* Note: ipv6_hdr(skb) != skb->data */ 368 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data; 369 saddr = &ip6h->saddr; 370 daddr = &ip6h->daddr; 371 net = dev_net(skb->dev); 372 373 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr, 374 IP6CB(skb)->iif))) { 375 rawv6_err(sk, skb, NULL, type, code, 376 inner_offset, info); 377 sk = sk_next(sk); 378 } 379 } 380 read_unlock(&raw_v6_hashinfo.lock); 381 } 382 383 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb) 384 { 385 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) && 386 skb_checksum_complete(skb)) { 387 atomic_inc(&sk->sk_drops); 388 kfree_skb(skb); 389 return NET_RX_DROP; 390 } 391 392 /* Charge it to the socket. */ 393 skb_dst_drop(skb); 394 if (sock_queue_rcv_skb(sk, skb) < 0) { 395 kfree_skb(skb); 396 return NET_RX_DROP; 397 } 398 399 return 0; 400 } 401 402 /* 403 * This is next to useless... 404 * if we demultiplex in network layer we don't need the extra call 405 * just to queue the skb... 406 * maybe we could have the network decide upon a hint if it 407 * should call raw_rcv for demultiplexing 408 */ 409 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 410 { 411 struct inet_sock *inet = inet_sk(sk); 412 struct raw6_sock *rp = raw6_sk(sk); 413 414 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 415 atomic_inc(&sk->sk_drops); 416 kfree_skb(skb); 417 return NET_RX_DROP; 418 } 419 420 if (!rp->checksum) 421 skb->ip_summed = CHECKSUM_UNNECESSARY; 422 423 if (skb->ip_summed == CHECKSUM_COMPLETE) { 424 skb_postpull_rcsum(skb, skb_network_header(skb), 425 skb_network_header_len(skb)); 426 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 427 &ipv6_hdr(skb)->daddr, 428 skb->len, inet->inet_num, skb->csum)) 429 skb->ip_summed = CHECKSUM_UNNECESSARY; 430 } 431 if (!skb_csum_unnecessary(skb)) 432 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 433 &ipv6_hdr(skb)->daddr, 434 skb->len, 435 inet->inet_num, 0)); 436 437 if (inet->hdrincl) { 438 if (skb_checksum_complete(skb)) { 439 atomic_inc(&sk->sk_drops); 440 kfree_skb(skb); 441 return NET_RX_DROP; 442 } 443 } 444 445 rawv6_rcv_skb(sk, skb); 446 return 0; 447 } 448 449 450 /* 451 * This should be easy, if there is something there 452 * we return it, otherwise we block. 453 */ 454 455 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk, 456 struct msghdr *msg, size_t len, 457 int noblock, int flags, int *addr_len) 458 { 459 struct ipv6_pinfo *np = inet6_sk(sk); 460 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name; 461 struct sk_buff *skb; 462 size_t copied; 463 int err; 464 465 if (flags & MSG_OOB) 466 return -EOPNOTSUPP; 467 468 if (addr_len) 469 *addr_len=sizeof(*sin6); 470 471 if (flags & MSG_ERRQUEUE) 472 return ipv6_recv_error(sk, msg, len); 473 474 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 475 return ipv6_recv_rxpmtu(sk, msg, len); 476 477 skb = skb_recv_datagram(sk, flags, noblock, &err); 478 if (!skb) 479 goto out; 480 481 copied = skb->len; 482 if (copied > len) { 483 copied = len; 484 msg->msg_flags |= MSG_TRUNC; 485 } 486 487 if (skb_csum_unnecessary(skb)) { 488 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 489 } else if (msg->msg_flags&MSG_TRUNC) { 490 if (__skb_checksum_complete(skb)) 491 goto csum_copy_err; 492 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 493 } else { 494 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov); 495 if (err == -EINVAL) 496 goto csum_copy_err; 497 } 498 if (err) 499 goto out_free; 500 501 /* Copy the address. */ 502 if (sin6) { 503 sin6->sin6_family = AF_INET6; 504 sin6->sin6_port = 0; 505 sin6->sin6_addr = ipv6_hdr(skb)->saddr; 506 sin6->sin6_flowinfo = 0; 507 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr, 508 IP6CB(skb)->iif); 509 } 510 511 sock_recv_ts_and_drops(msg, sk, skb); 512 513 if (np->rxopt.all) 514 ip6_datagram_recv_ctl(sk, msg, skb); 515 516 err = copied; 517 if (flags & MSG_TRUNC) 518 err = skb->len; 519 520 out_free: 521 skb_free_datagram(sk, skb); 522 out: 523 return err; 524 525 csum_copy_err: 526 skb_kill_datagram(sk, skb, flags); 527 528 /* Error for blocking case is chosen to masquerade 529 as some normal condition. 530 */ 531 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 532 goto out; 533 } 534 535 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 536 struct raw6_sock *rp) 537 { 538 struct sk_buff *skb; 539 int err = 0; 540 int offset; 541 int len; 542 int total_len; 543 __wsum tmp_csum; 544 __sum16 csum; 545 546 if (!rp->checksum) 547 goto send; 548 549 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 550 goto out; 551 552 offset = rp->offset; 553 total_len = inet_sk(sk)->cork.base.length; 554 if (offset >= total_len - 1) { 555 err = -EINVAL; 556 ip6_flush_pending_frames(sk); 557 goto out; 558 } 559 560 /* should be check HW csum miyazawa */ 561 if (skb_queue_len(&sk->sk_write_queue) == 1) { 562 /* 563 * Only one fragment on the socket. 564 */ 565 tmp_csum = skb->csum; 566 } else { 567 struct sk_buff *csum_skb = NULL; 568 tmp_csum = 0; 569 570 skb_queue_walk(&sk->sk_write_queue, skb) { 571 tmp_csum = csum_add(tmp_csum, skb->csum); 572 573 if (csum_skb) 574 continue; 575 576 len = skb->len - skb_transport_offset(skb); 577 if (offset >= len) { 578 offset -= len; 579 continue; 580 } 581 582 csum_skb = skb; 583 } 584 585 skb = csum_skb; 586 } 587 588 offset += skb_transport_offset(skb); 589 if (skb_copy_bits(skb, offset, &csum, 2)) 590 BUG(); 591 592 /* in case cksum was not initialized */ 593 if (unlikely(csum)) 594 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 595 596 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 597 total_len, fl6->flowi6_proto, tmp_csum); 598 599 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP) 600 csum = CSUM_MANGLED_0; 601 602 if (skb_store_bits(skb, offset, &csum, 2)) 603 BUG(); 604 605 send: 606 err = ip6_push_pending_frames(sk); 607 out: 608 return err; 609 } 610 611 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length, 612 struct flowi6 *fl6, struct dst_entry **dstp, 613 unsigned int flags) 614 { 615 struct ipv6_pinfo *np = inet6_sk(sk); 616 struct ipv6hdr *iph; 617 struct sk_buff *skb; 618 int err; 619 struct rt6_info *rt = (struct rt6_info *)*dstp; 620 int hlen = LL_RESERVED_SPACE(rt->dst.dev); 621 int tlen = rt->dst.dev->needed_tailroom; 622 623 if (length > rt->dst.dev->mtu) { 624 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu); 625 return -EMSGSIZE; 626 } 627 if (flags&MSG_PROBE) 628 goto out; 629 630 skb = sock_alloc_send_skb(sk, 631 length + hlen + tlen + 15, 632 flags & MSG_DONTWAIT, &err); 633 if (skb == NULL) 634 goto error; 635 skb_reserve(skb, hlen); 636 637 skb->protocol = htons(ETH_P_IPV6); 638 skb->priority = sk->sk_priority; 639 skb->mark = sk->sk_mark; 640 skb_dst_set(skb, &rt->dst); 641 *dstp = NULL; 642 643 skb_put(skb, length); 644 skb_reset_network_header(skb); 645 iph = ipv6_hdr(skb); 646 647 skb->ip_summed = CHECKSUM_NONE; 648 649 skb->transport_header = skb->network_header; 650 err = memcpy_fromiovecend((void *)iph, from, 0, length); 651 if (err) 652 goto error_fault; 653 654 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); 655 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, 656 rt->dst.dev, dst_output); 657 if (err > 0) 658 err = net_xmit_errno(err); 659 if (err) 660 goto error; 661 out: 662 return 0; 663 664 error_fault: 665 err = -EFAULT; 666 kfree_skb(skb); 667 error: 668 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 669 if (err == -ENOBUFS && !np->recverr) 670 err = 0; 671 return err; 672 } 673 674 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg) 675 { 676 struct iovec *iov; 677 u8 __user *type = NULL; 678 u8 __user *code = NULL; 679 u8 len = 0; 680 int probed = 0; 681 int i; 682 683 if (!msg->msg_iov) 684 return 0; 685 686 for (i = 0; i < msg->msg_iovlen; i++) { 687 iov = &msg->msg_iov[i]; 688 if (!iov) 689 continue; 690 691 switch (fl6->flowi6_proto) { 692 case IPPROTO_ICMPV6: 693 /* check if one-byte field is readable or not. */ 694 if (iov->iov_base && iov->iov_len < 1) 695 break; 696 697 if (!type) { 698 type = iov->iov_base; 699 /* check if code field is readable or not. */ 700 if (iov->iov_len > 1) 701 code = type + 1; 702 } else if (!code) 703 code = iov->iov_base; 704 705 if (type && code) { 706 if (get_user(fl6->fl6_icmp_type, type) || 707 get_user(fl6->fl6_icmp_code, code)) 708 return -EFAULT; 709 probed = 1; 710 } 711 break; 712 case IPPROTO_MH: 713 if (iov->iov_base && iov->iov_len < 1) 714 break; 715 /* check if type field is readable or not. */ 716 if (iov->iov_len > 2 - len) { 717 u8 __user *p = iov->iov_base; 718 if (get_user(fl6->fl6_mh_type, &p[2 - len])) 719 return -EFAULT; 720 probed = 1; 721 } else 722 len += iov->iov_len; 723 724 break; 725 default: 726 probed = 1; 727 break; 728 } 729 if (probed) 730 break; 731 } 732 return 0; 733 } 734 735 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk, 736 struct msghdr *msg, size_t len) 737 { 738 struct ipv6_txoptions opt_space; 739 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name; 740 struct in6_addr *daddr, *final_p, final; 741 struct inet_sock *inet = inet_sk(sk); 742 struct ipv6_pinfo *np = inet6_sk(sk); 743 struct raw6_sock *rp = raw6_sk(sk); 744 struct ipv6_txoptions *opt = NULL; 745 struct ip6_flowlabel *flowlabel = NULL; 746 struct dst_entry *dst = NULL; 747 struct flowi6 fl6; 748 int addr_len = msg->msg_namelen; 749 int hlimit = -1; 750 int tclass = -1; 751 int dontfrag = -1; 752 u16 proto; 753 int err; 754 755 /* Rough check on arithmetic overflow, 756 better check is made in ip6_append_data(). 757 */ 758 if (len > INT_MAX) 759 return -EMSGSIZE; 760 761 /* Mirror BSD error message compatibility */ 762 if (msg->msg_flags & MSG_OOB) 763 return -EOPNOTSUPP; 764 765 /* 766 * Get and verify the address. 767 */ 768 memset(&fl6, 0, sizeof(fl6)); 769 770 fl6.flowi6_mark = sk->sk_mark; 771 772 if (sin6) { 773 if (addr_len < SIN6_LEN_RFC2133) 774 return -EINVAL; 775 776 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 777 return -EAFNOSUPPORT; 778 779 /* port is the proto value [0..255] carried in nexthdr */ 780 proto = ntohs(sin6->sin6_port); 781 782 if (!proto) 783 proto = inet->inet_num; 784 else if (proto != inet->inet_num) 785 return -EINVAL; 786 787 if (proto > 255) 788 return -EINVAL; 789 790 daddr = &sin6->sin6_addr; 791 if (np->sndflow) { 792 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 793 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 794 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 795 if (flowlabel == NULL) 796 return -EINVAL; 797 daddr = &flowlabel->dst; 798 } 799 } 800 801 /* 802 * Otherwise it will be difficult to maintain 803 * sk->sk_dst_cache. 804 */ 805 if (sk->sk_state == TCP_ESTABLISHED && 806 ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) 807 daddr = &sk->sk_v6_daddr; 808 809 if (addr_len >= sizeof(struct sockaddr_in6) && 810 sin6->sin6_scope_id && 811 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) 812 fl6.flowi6_oif = sin6->sin6_scope_id; 813 } else { 814 if (sk->sk_state != TCP_ESTABLISHED) 815 return -EDESTADDRREQ; 816 817 proto = inet->inet_num; 818 daddr = &sk->sk_v6_daddr; 819 fl6.flowlabel = np->flow_label; 820 } 821 822 if (fl6.flowi6_oif == 0) 823 fl6.flowi6_oif = sk->sk_bound_dev_if; 824 825 if (msg->msg_controllen) { 826 opt = &opt_space; 827 memset(opt, 0, sizeof(struct ipv6_txoptions)); 828 opt->tot_len = sizeof(struct ipv6_txoptions); 829 830 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt, 831 &hlimit, &tclass, &dontfrag); 832 if (err < 0) { 833 fl6_sock_release(flowlabel); 834 return err; 835 } 836 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 837 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 838 if (flowlabel == NULL) 839 return -EINVAL; 840 } 841 if (!(opt->opt_nflen|opt->opt_flen)) 842 opt = NULL; 843 } 844 if (opt == NULL) 845 opt = np->opt; 846 if (flowlabel) 847 opt = fl6_merge_options(&opt_space, flowlabel, opt); 848 opt = ipv6_fixup_options(&opt_space, opt); 849 850 fl6.flowi6_proto = proto; 851 err = rawv6_probe_proto_opt(&fl6, msg); 852 if (err) 853 goto out; 854 855 if (!ipv6_addr_any(daddr)) 856 fl6.daddr = *daddr; 857 else 858 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 859 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 860 fl6.saddr = np->saddr; 861 862 final_p = fl6_update_dst(&fl6, opt, &final); 863 864 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 865 fl6.flowi6_oif = np->mcast_oif; 866 else if (!fl6.flowi6_oif) 867 fl6.flowi6_oif = np->ucast_oif; 868 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 869 870 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true); 871 if (IS_ERR(dst)) { 872 err = PTR_ERR(dst); 873 goto out; 874 } 875 if (hlimit < 0) { 876 if (ipv6_addr_is_multicast(&fl6.daddr)) 877 hlimit = np->mcast_hops; 878 else 879 hlimit = np->hop_limit; 880 if (hlimit < 0) 881 hlimit = ip6_dst_hoplimit(dst); 882 } 883 884 if (tclass < 0) 885 tclass = np->tclass; 886 887 if (dontfrag < 0) 888 dontfrag = np->dontfrag; 889 890 if (msg->msg_flags&MSG_CONFIRM) 891 goto do_confirm; 892 893 back_from_confirm: 894 if (inet->hdrincl) 895 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags); 896 else { 897 lock_sock(sk); 898 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, 899 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst, 900 msg->msg_flags, dontfrag); 901 902 if (err) 903 ip6_flush_pending_frames(sk); 904 else if (!(msg->msg_flags & MSG_MORE)) 905 err = rawv6_push_pending_frames(sk, &fl6, rp); 906 release_sock(sk); 907 } 908 done: 909 dst_release(dst); 910 out: 911 fl6_sock_release(flowlabel); 912 return err<0?err:len; 913 do_confirm: 914 dst_confirm(dst); 915 if (!(msg->msg_flags & MSG_PROBE) || len) 916 goto back_from_confirm; 917 err = 0; 918 goto done; 919 } 920 921 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 922 char __user *optval, int optlen) 923 { 924 switch (optname) { 925 case ICMPV6_FILTER: 926 if (optlen > sizeof(struct icmp6_filter)) 927 optlen = sizeof(struct icmp6_filter); 928 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 929 return -EFAULT; 930 return 0; 931 default: 932 return -ENOPROTOOPT; 933 } 934 935 return 0; 936 } 937 938 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 939 char __user *optval, int __user *optlen) 940 { 941 int len; 942 943 switch (optname) { 944 case ICMPV6_FILTER: 945 if (get_user(len, optlen)) 946 return -EFAULT; 947 if (len < 0) 948 return -EINVAL; 949 if (len > sizeof(struct icmp6_filter)) 950 len = sizeof(struct icmp6_filter); 951 if (put_user(len, optlen)) 952 return -EFAULT; 953 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 954 return -EFAULT; 955 return 0; 956 default: 957 return -ENOPROTOOPT; 958 } 959 960 return 0; 961 } 962 963 964 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 965 char __user *optval, unsigned int optlen) 966 { 967 struct raw6_sock *rp = raw6_sk(sk); 968 int val; 969 970 if (get_user(val, (int __user *)optval)) 971 return -EFAULT; 972 973 switch (optname) { 974 case IPV6_CHECKSUM: 975 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && 976 level == IPPROTO_IPV6) { 977 /* 978 * RFC3542 tells that IPV6_CHECKSUM socket 979 * option in the IPPROTO_IPV6 level is not 980 * allowed on ICMPv6 sockets. 981 * If you want to set it, use IPPROTO_RAW 982 * level IPV6_CHECKSUM socket option 983 * (Linux extension). 984 */ 985 return -EINVAL; 986 } 987 988 /* You may get strange result with a positive odd offset; 989 RFC2292bis agrees with me. */ 990 if (val > 0 && (val&1)) 991 return -EINVAL; 992 if (val < 0) { 993 rp->checksum = 0; 994 } else { 995 rp->checksum = 1; 996 rp->offset = val; 997 } 998 999 return 0; 1000 1001 default: 1002 return -ENOPROTOOPT; 1003 } 1004 } 1005 1006 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 1007 char __user *optval, unsigned int optlen) 1008 { 1009 switch (level) { 1010 case SOL_RAW: 1011 break; 1012 1013 case SOL_ICMPV6: 1014 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1015 return -EOPNOTSUPP; 1016 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1017 case SOL_IPV6: 1018 if (optname == IPV6_CHECKSUM) 1019 break; 1020 default: 1021 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1022 } 1023 1024 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1025 } 1026 1027 #ifdef CONFIG_COMPAT 1028 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 1029 char __user *optval, unsigned int optlen) 1030 { 1031 switch (level) { 1032 case SOL_RAW: 1033 break; 1034 case SOL_ICMPV6: 1035 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1036 return -EOPNOTSUPP; 1037 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1038 case SOL_IPV6: 1039 if (optname == IPV6_CHECKSUM) 1040 break; 1041 default: 1042 return compat_ipv6_setsockopt(sk, level, optname, 1043 optval, optlen); 1044 } 1045 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1046 } 1047 #endif 1048 1049 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1050 char __user *optval, int __user *optlen) 1051 { 1052 struct raw6_sock *rp = raw6_sk(sk); 1053 int val, len; 1054 1055 if (get_user(len,optlen)) 1056 return -EFAULT; 1057 1058 switch (optname) { 1059 case IPV6_CHECKSUM: 1060 /* 1061 * We allow getsockopt() for IPPROTO_IPV6-level 1062 * IPV6_CHECKSUM socket option on ICMPv6 sockets 1063 * since RFC3542 is silent about it. 1064 */ 1065 if (rp->checksum == 0) 1066 val = -1; 1067 else 1068 val = rp->offset; 1069 break; 1070 1071 default: 1072 return -ENOPROTOOPT; 1073 } 1074 1075 len = min_t(unsigned int, sizeof(int), len); 1076 1077 if (put_user(len, optlen)) 1078 return -EFAULT; 1079 if (copy_to_user(optval,&val,len)) 1080 return -EFAULT; 1081 return 0; 1082 } 1083 1084 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1085 char __user *optval, int __user *optlen) 1086 { 1087 switch (level) { 1088 case SOL_RAW: 1089 break; 1090 1091 case SOL_ICMPV6: 1092 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1093 return -EOPNOTSUPP; 1094 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1095 case SOL_IPV6: 1096 if (optname == IPV6_CHECKSUM) 1097 break; 1098 default: 1099 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1100 } 1101 1102 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1103 } 1104 1105 #ifdef CONFIG_COMPAT 1106 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1107 char __user *optval, int __user *optlen) 1108 { 1109 switch (level) { 1110 case SOL_RAW: 1111 break; 1112 case SOL_ICMPV6: 1113 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1114 return -EOPNOTSUPP; 1115 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1116 case SOL_IPV6: 1117 if (optname == IPV6_CHECKSUM) 1118 break; 1119 default: 1120 return compat_ipv6_getsockopt(sk, level, optname, 1121 optval, optlen); 1122 } 1123 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1124 } 1125 #endif 1126 1127 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1128 { 1129 switch (cmd) { 1130 case SIOCOUTQ: { 1131 int amount = sk_wmem_alloc_get(sk); 1132 1133 return put_user(amount, (int __user *)arg); 1134 } 1135 case SIOCINQ: { 1136 struct sk_buff *skb; 1137 int amount = 0; 1138 1139 spin_lock_bh(&sk->sk_receive_queue.lock); 1140 skb = skb_peek(&sk->sk_receive_queue); 1141 if (skb != NULL) 1142 amount = skb_tail_pointer(skb) - 1143 skb_transport_header(skb); 1144 spin_unlock_bh(&sk->sk_receive_queue.lock); 1145 return put_user(amount, (int __user *)arg); 1146 } 1147 1148 default: 1149 #ifdef CONFIG_IPV6_MROUTE 1150 return ip6mr_ioctl(sk, cmd, (void __user *)arg); 1151 #else 1152 return -ENOIOCTLCMD; 1153 #endif 1154 } 1155 } 1156 1157 #ifdef CONFIG_COMPAT 1158 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 1159 { 1160 switch (cmd) { 1161 case SIOCOUTQ: 1162 case SIOCINQ: 1163 return -ENOIOCTLCMD; 1164 default: 1165 #ifdef CONFIG_IPV6_MROUTE 1166 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg)); 1167 #else 1168 return -ENOIOCTLCMD; 1169 #endif 1170 } 1171 } 1172 #endif 1173 1174 static void rawv6_close(struct sock *sk, long timeout) 1175 { 1176 if (inet_sk(sk)->inet_num == IPPROTO_RAW) 1177 ip6_ra_control(sk, -1); 1178 ip6mr_sk_done(sk); 1179 sk_common_release(sk); 1180 } 1181 1182 static void raw6_destroy(struct sock *sk) 1183 { 1184 lock_sock(sk); 1185 ip6_flush_pending_frames(sk); 1186 release_sock(sk); 1187 1188 inet6_destroy_sock(sk); 1189 } 1190 1191 static int rawv6_init_sk(struct sock *sk) 1192 { 1193 struct raw6_sock *rp = raw6_sk(sk); 1194 1195 switch (inet_sk(sk)->inet_num) { 1196 case IPPROTO_ICMPV6: 1197 rp->checksum = 1; 1198 rp->offset = 2; 1199 break; 1200 case IPPROTO_MH: 1201 rp->checksum = 1; 1202 rp->offset = 4; 1203 break; 1204 default: 1205 break; 1206 } 1207 return 0; 1208 } 1209 1210 struct proto rawv6_prot = { 1211 .name = "RAWv6", 1212 .owner = THIS_MODULE, 1213 .close = rawv6_close, 1214 .destroy = raw6_destroy, 1215 .connect = ip6_datagram_connect, 1216 .disconnect = udp_disconnect, 1217 .ioctl = rawv6_ioctl, 1218 .init = rawv6_init_sk, 1219 .setsockopt = rawv6_setsockopt, 1220 .getsockopt = rawv6_getsockopt, 1221 .sendmsg = rawv6_sendmsg, 1222 .recvmsg = rawv6_recvmsg, 1223 .bind = rawv6_bind, 1224 .backlog_rcv = rawv6_rcv_skb, 1225 .hash = raw_hash_sk, 1226 .unhash = raw_unhash_sk, 1227 .obj_size = sizeof(struct raw6_sock), 1228 .h.raw_hash = &raw_v6_hashinfo, 1229 #ifdef CONFIG_COMPAT 1230 .compat_setsockopt = compat_rawv6_setsockopt, 1231 .compat_getsockopt = compat_rawv6_getsockopt, 1232 .compat_ioctl = compat_rawv6_ioctl, 1233 #endif 1234 }; 1235 1236 #ifdef CONFIG_PROC_FS 1237 static int raw6_seq_show(struct seq_file *seq, void *v) 1238 { 1239 if (v == SEQ_START_TOKEN) { 1240 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); 1241 } else { 1242 struct sock *sp = v; 1243 __u16 srcp = inet_sk(sp)->inet_num; 1244 ip6_dgram_sock_seq_show(seq, v, srcp, 0, 1245 raw_seq_private(seq)->bucket); 1246 } 1247 return 0; 1248 } 1249 1250 static const struct seq_operations raw6_seq_ops = { 1251 .start = raw_seq_start, 1252 .next = raw_seq_next, 1253 .stop = raw_seq_stop, 1254 .show = raw6_seq_show, 1255 }; 1256 1257 static int raw6_seq_open(struct inode *inode, struct file *file) 1258 { 1259 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops); 1260 } 1261 1262 static const struct file_operations raw6_seq_fops = { 1263 .owner = THIS_MODULE, 1264 .open = raw6_seq_open, 1265 .read = seq_read, 1266 .llseek = seq_lseek, 1267 .release = seq_release_net, 1268 }; 1269 1270 static int __net_init raw6_init_net(struct net *net) 1271 { 1272 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops)) 1273 return -ENOMEM; 1274 1275 return 0; 1276 } 1277 1278 static void __net_exit raw6_exit_net(struct net *net) 1279 { 1280 remove_proc_entry("raw6", net->proc_net); 1281 } 1282 1283 static struct pernet_operations raw6_net_ops = { 1284 .init = raw6_init_net, 1285 .exit = raw6_exit_net, 1286 }; 1287 1288 int __init raw6_proc_init(void) 1289 { 1290 return register_pernet_subsys(&raw6_net_ops); 1291 } 1292 1293 void raw6_proc_exit(void) 1294 { 1295 unregister_pernet_subsys(&raw6_net_ops); 1296 } 1297 #endif /* CONFIG_PROC_FS */ 1298 1299 /* Same as inet6_dgram_ops, sans udp_poll. */ 1300 static const struct proto_ops inet6_sockraw_ops = { 1301 .family = PF_INET6, 1302 .owner = THIS_MODULE, 1303 .release = inet6_release, 1304 .bind = inet6_bind, 1305 .connect = inet_dgram_connect, /* ok */ 1306 .socketpair = sock_no_socketpair, /* a do nothing */ 1307 .accept = sock_no_accept, /* a do nothing */ 1308 .getname = inet6_getname, 1309 .poll = datagram_poll, /* ok */ 1310 .ioctl = inet6_ioctl, /* must change */ 1311 .listen = sock_no_listen, /* ok */ 1312 .shutdown = inet_shutdown, /* ok */ 1313 .setsockopt = sock_common_setsockopt, /* ok */ 1314 .getsockopt = sock_common_getsockopt, /* ok */ 1315 .sendmsg = inet_sendmsg, /* ok */ 1316 .recvmsg = sock_common_recvmsg, /* ok */ 1317 .mmap = sock_no_mmap, 1318 .sendpage = sock_no_sendpage, 1319 #ifdef CONFIG_COMPAT 1320 .compat_setsockopt = compat_sock_common_setsockopt, 1321 .compat_getsockopt = compat_sock_common_getsockopt, 1322 #endif 1323 }; 1324 1325 static struct inet_protosw rawv6_protosw = { 1326 .type = SOCK_RAW, 1327 .protocol = IPPROTO_IP, /* wild card */ 1328 .prot = &rawv6_prot, 1329 .ops = &inet6_sockraw_ops, 1330 .no_check = UDP_CSUM_DEFAULT, 1331 .flags = INET_PROTOSW_REUSE, 1332 }; 1333 1334 int __init rawv6_init(void) 1335 { 1336 int ret; 1337 1338 ret = inet6_register_protosw(&rawv6_protosw); 1339 if (ret) 1340 goto out; 1341 out: 1342 return ret; 1343 } 1344 1345 void rawv6_exit(void) 1346 { 1347 inet6_unregister_protosw(&rawv6_protosw); 1348 } 1349