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 <asm/uaccess.h> 35 #include <asm/ioctls.h> 36 37 #include <net/net_namespace.h> 38 #include <net/ip.h> 39 #include <net/sock.h> 40 #include <net/snmp.h> 41 42 #include <net/ipv6.h> 43 #include <net/ndisc.h> 44 #include <net/protocol.h> 45 #include <net/ip6_route.h> 46 #include <net/ip6_checksum.h> 47 #include <net/addrconf.h> 48 #include <net/transp_v6.h> 49 #include <net/udp.h> 50 #include <net/inet_common.h> 51 #include <net/tcp_states.h> 52 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 53 #include <net/mip6.h> 54 #endif 55 #include <linux/mroute6.h> 56 57 #include <net/raw.h> 58 #include <net/rawv6.h> 59 #include <net/xfrm.h> 60 61 #include <linux/proc_fs.h> 62 #include <linux/seq_file.h> 63 64 static struct raw_hashinfo raw_v6_hashinfo = { 65 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock), 66 }; 67 68 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk, 69 unsigned short num, struct in6_addr *loc_addr, 70 struct in6_addr *rmt_addr, int dif) 71 { 72 struct hlist_node *node; 73 int is_multicast = ipv6_addr_is_multicast(loc_addr); 74 75 sk_for_each_from(sk, node) 76 if (inet_sk(sk)->inet_num == num) { 77 struct ipv6_pinfo *np = inet6_sk(sk); 78 79 if (!net_eq(sock_net(sk), net)) 80 continue; 81 82 if (!ipv6_addr_any(&np->daddr) && 83 !ipv6_addr_equal(&np->daddr, rmt_addr)) 84 continue; 85 86 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) 87 continue; 88 89 if (!ipv6_addr_any(&np->rcv_saddr)) { 90 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr)) 91 goto found; 92 if (is_multicast && 93 inet6_mc_check(sk, loc_addr, rmt_addr)) 94 goto found; 95 continue; 96 } 97 goto found; 98 } 99 sk = NULL; 100 found: 101 return sk; 102 } 103 104 /* 105 * 0 - deliver 106 * 1 - block 107 */ 108 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb) 109 { 110 struct icmp6hdr *icmph; 111 struct raw6_sock *rp = raw6_sk(sk); 112 113 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) { 114 __u32 *data = &rp->filter.data[0]; 115 int bit_nr; 116 117 icmph = (struct icmp6hdr *) skb->data; 118 bit_nr = icmph->icmp6_type; 119 120 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0; 121 } 122 return 0; 123 } 124 125 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 126 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb); 127 128 int rawv6_mh_filter_register(int (*filter)(struct sock *sock, 129 struct sk_buff *skb)) 130 { 131 rcu_assign_pointer(mh_filter, filter); 132 return 0; 133 } 134 EXPORT_SYMBOL(rawv6_mh_filter_register); 135 136 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock, 137 struct sk_buff *skb)) 138 { 139 rcu_assign_pointer(mh_filter, NULL); 140 synchronize_rcu(); 141 return 0; 142 } 143 EXPORT_SYMBOL(rawv6_mh_filter_unregister); 144 145 #endif 146 147 /* 148 * demultiplex raw sockets. 149 * (should consider queueing the skb in the sock receive_queue 150 * without calling rawv6.c) 151 * 152 * Caller owns SKB so we must make clones. 153 */ 154 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 155 { 156 struct in6_addr *saddr; 157 struct in6_addr *daddr; 158 struct sock *sk; 159 int delivered = 0; 160 __u8 hash; 161 struct net *net; 162 163 saddr = &ipv6_hdr(skb)->saddr; 164 daddr = saddr + 1; 165 166 hash = nexthdr & (MAX_INET_PROTOS - 1); 167 168 read_lock(&raw_v6_hashinfo.lock); 169 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 170 171 if (sk == NULL) 172 goto out; 173 174 net = dev_net(skb->dev); 175 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif); 176 177 while (sk) { 178 int filtered; 179 180 delivered = 1; 181 switch (nexthdr) { 182 case IPPROTO_ICMPV6: 183 filtered = icmpv6_filter(sk, skb); 184 break; 185 186 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 187 case IPPROTO_MH: 188 { 189 /* XXX: To validate MH only once for each packet, 190 * this is placed here. It should be after checking 191 * xfrm policy, however it doesn't. The checking xfrm 192 * policy is placed in rawv6_rcv() because it is 193 * required for each socket. 194 */ 195 int (*filter)(struct sock *sock, struct sk_buff *skb); 196 197 filter = rcu_dereference(mh_filter); 198 filtered = filter ? filter(sk, skb) : 0; 199 break; 200 } 201 #endif 202 default: 203 filtered = 0; 204 break; 205 } 206 207 if (filtered < 0) 208 break; 209 if (filtered == 0) { 210 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 211 212 /* Not releasing hash table! */ 213 if (clone) { 214 nf_reset(clone); 215 rawv6_rcv(sk, clone); 216 } 217 } 218 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr, 219 IP6CB(skb)->iif); 220 } 221 out: 222 read_unlock(&raw_v6_hashinfo.lock); 223 return delivered; 224 } 225 226 int raw6_local_deliver(struct sk_buff *skb, int nexthdr) 227 { 228 struct sock *raw_sk; 229 230 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]); 231 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr)) 232 raw_sk = NULL; 233 234 return raw_sk != NULL; 235 } 236 237 /* This cleans up af_inet6 a bit. -DaveM */ 238 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 239 { 240 struct inet_sock *inet = inet_sk(sk); 241 struct ipv6_pinfo *np = inet6_sk(sk); 242 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 243 __be32 v4addr = 0; 244 int addr_type; 245 int err; 246 247 if (addr_len < SIN6_LEN_RFC2133) 248 return -EINVAL; 249 addr_type = ipv6_addr_type(&addr->sin6_addr); 250 251 /* Raw sockets are IPv6 only */ 252 if (addr_type == IPV6_ADDR_MAPPED) 253 return -EADDRNOTAVAIL; 254 255 lock_sock(sk); 256 257 err = -EINVAL; 258 if (sk->sk_state != TCP_CLOSE) 259 goto out; 260 261 rcu_read_lock(); 262 /* Check if the address belongs to the host. */ 263 if (addr_type != IPV6_ADDR_ANY) { 264 struct net_device *dev = NULL; 265 266 if (addr_type & IPV6_ADDR_LINKLOCAL) { 267 if (addr_len >= sizeof(struct sockaddr_in6) && 268 addr->sin6_scope_id) { 269 /* Override any existing binding, if another 270 * one is supplied by user. 271 */ 272 sk->sk_bound_dev_if = addr->sin6_scope_id; 273 } 274 275 /* Binding to link-local address requires an interface */ 276 if (!sk->sk_bound_dev_if) 277 goto out_unlock; 278 279 err = -ENODEV; 280 dev = dev_get_by_index_rcu(sock_net(sk), 281 sk->sk_bound_dev_if); 282 if (!dev) 283 goto out_unlock; 284 } 285 286 /* ipv4 addr of the socket is invalid. Only the 287 * unspecified and mapped address have a v4 equivalent. 288 */ 289 v4addr = LOOPBACK4_IPV6; 290 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 291 err = -EADDRNOTAVAIL; 292 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr, 293 dev, 0)) { 294 goto out_unlock; 295 } 296 } 297 } 298 299 inet->inet_rcv_saddr = inet->inet_saddr = v4addr; 300 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr); 301 if (!(addr_type & IPV6_ADDR_MULTICAST)) 302 ipv6_addr_copy(&np->saddr, &addr->sin6_addr); 303 err = 0; 304 out_unlock: 305 rcu_read_unlock(); 306 out: 307 release_sock(sk); 308 return err; 309 } 310 311 static void rawv6_err(struct sock *sk, struct sk_buff *skb, 312 struct inet6_skb_parm *opt, 313 u8 type, u8 code, int offset, __be32 info) 314 { 315 struct inet_sock *inet = inet_sk(sk); 316 struct ipv6_pinfo *np = inet6_sk(sk); 317 int err; 318 int harderr; 319 320 /* Report error on raw socket, if: 321 1. User requested recverr. 322 2. Socket is connected (otherwise the error indication 323 is useless without recverr and error is hard. 324 */ 325 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 326 return; 327 328 harderr = icmpv6_err_convert(type, code, &err); 329 if (type == ICMPV6_PKT_TOOBIG) 330 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 331 332 if (np->recverr) { 333 u8 *payload = skb->data; 334 if (!inet->hdrincl) 335 payload += offset; 336 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 337 } 338 339 if (np->recverr || harderr) { 340 sk->sk_err = err; 341 sk->sk_error_report(sk); 342 } 343 } 344 345 void raw6_icmp_error(struct sk_buff *skb, int nexthdr, 346 u8 type, u8 code, int inner_offset, __be32 info) 347 { 348 struct sock *sk; 349 int hash; 350 struct in6_addr *saddr, *daddr; 351 struct net *net; 352 353 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 354 355 read_lock(&raw_v6_hashinfo.lock); 356 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 357 if (sk != NULL) { 358 /* Note: ipv6_hdr(skb) != skb->data */ 359 struct ipv6hdr *ip6h = (struct ipv6hdr *)skb->data; 360 saddr = &ip6h->saddr; 361 daddr = &ip6h->daddr; 362 net = dev_net(skb->dev); 363 364 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr, 365 IP6CB(skb)->iif))) { 366 rawv6_err(sk, skb, NULL, type, code, 367 inner_offset, info); 368 sk = sk_next(sk); 369 } 370 } 371 read_unlock(&raw_v6_hashinfo.lock); 372 } 373 374 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb) 375 { 376 if ((raw6_sk(sk)->checksum || sk->sk_filter) && 377 skb_checksum_complete(skb)) { 378 atomic_inc(&sk->sk_drops); 379 kfree_skb(skb); 380 return NET_RX_DROP; 381 } 382 383 /* Charge it to the socket. */ 384 if (ip_queue_rcv_skb(sk, skb) < 0) { 385 kfree_skb(skb); 386 return NET_RX_DROP; 387 } 388 389 return 0; 390 } 391 392 /* 393 * This is next to useless... 394 * if we demultiplex in network layer we don't need the extra call 395 * just to queue the skb... 396 * maybe we could have the network decide upon a hint if it 397 * should call raw_rcv for demultiplexing 398 */ 399 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 400 { 401 struct inet_sock *inet = inet_sk(sk); 402 struct raw6_sock *rp = raw6_sk(sk); 403 404 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 405 atomic_inc(&sk->sk_drops); 406 kfree_skb(skb); 407 return NET_RX_DROP; 408 } 409 410 if (!rp->checksum) 411 skb->ip_summed = CHECKSUM_UNNECESSARY; 412 413 if (skb->ip_summed == CHECKSUM_COMPLETE) { 414 skb_postpull_rcsum(skb, skb_network_header(skb), 415 skb_network_header_len(skb)); 416 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 417 &ipv6_hdr(skb)->daddr, 418 skb->len, inet->inet_num, skb->csum)) 419 skb->ip_summed = CHECKSUM_UNNECESSARY; 420 } 421 if (!skb_csum_unnecessary(skb)) 422 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 423 &ipv6_hdr(skb)->daddr, 424 skb->len, 425 inet->inet_num, 0)); 426 427 if (inet->hdrincl) { 428 if (skb_checksum_complete(skb)) { 429 atomic_inc(&sk->sk_drops); 430 kfree_skb(skb); 431 return NET_RX_DROP; 432 } 433 } 434 435 rawv6_rcv_skb(sk, skb); 436 return 0; 437 } 438 439 440 /* 441 * This should be easy, if there is something there 442 * we return it, otherwise we block. 443 */ 444 445 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk, 446 struct msghdr *msg, size_t len, 447 int noblock, int flags, int *addr_len) 448 { 449 struct ipv6_pinfo *np = inet6_sk(sk); 450 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name; 451 struct sk_buff *skb; 452 size_t copied; 453 int err; 454 455 if (flags & MSG_OOB) 456 return -EOPNOTSUPP; 457 458 if (addr_len) 459 *addr_len=sizeof(*sin6); 460 461 if (flags & MSG_ERRQUEUE) 462 return ipv6_recv_error(sk, msg, len); 463 464 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 465 return ipv6_recv_rxpmtu(sk, msg, len); 466 467 skb = skb_recv_datagram(sk, flags, noblock, &err); 468 if (!skb) 469 goto out; 470 471 copied = skb->len; 472 if (copied > len) { 473 copied = len; 474 msg->msg_flags |= MSG_TRUNC; 475 } 476 477 if (skb_csum_unnecessary(skb)) { 478 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 479 } else if (msg->msg_flags&MSG_TRUNC) { 480 if (__skb_checksum_complete(skb)) 481 goto csum_copy_err; 482 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 483 } else { 484 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov); 485 if (err == -EINVAL) 486 goto csum_copy_err; 487 } 488 if (err) 489 goto out_free; 490 491 /* Copy the address. */ 492 if (sin6) { 493 sin6->sin6_family = AF_INET6; 494 sin6->sin6_port = 0; 495 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr); 496 sin6->sin6_flowinfo = 0; 497 sin6->sin6_scope_id = 0; 498 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) 499 sin6->sin6_scope_id = IP6CB(skb)->iif; 500 } 501 502 sock_recv_ts_and_drops(msg, sk, skb); 503 504 if (np->rxopt.all) 505 datagram_recv_ctl(sk, msg, skb); 506 507 err = copied; 508 if (flags & MSG_TRUNC) 509 err = skb->len; 510 511 out_free: 512 skb_free_datagram(sk, skb); 513 out: 514 return err; 515 516 csum_copy_err: 517 skb_kill_datagram(sk, skb, flags); 518 519 /* Error for blocking case is chosen to masquerade 520 as some normal condition. 521 */ 522 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 523 goto out; 524 } 525 526 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl, 527 struct raw6_sock *rp) 528 { 529 struct sk_buff *skb; 530 int err = 0; 531 int offset; 532 int len; 533 int total_len; 534 __wsum tmp_csum; 535 __sum16 csum; 536 537 if (!rp->checksum) 538 goto send; 539 540 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 541 goto out; 542 543 offset = rp->offset; 544 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) - 545 skb->data); 546 if (offset >= total_len - 1) { 547 err = -EINVAL; 548 ip6_flush_pending_frames(sk); 549 goto out; 550 } 551 552 /* should be check HW csum miyazawa */ 553 if (skb_queue_len(&sk->sk_write_queue) == 1) { 554 /* 555 * Only one fragment on the socket. 556 */ 557 tmp_csum = skb->csum; 558 } else { 559 struct sk_buff *csum_skb = NULL; 560 tmp_csum = 0; 561 562 skb_queue_walk(&sk->sk_write_queue, skb) { 563 tmp_csum = csum_add(tmp_csum, skb->csum); 564 565 if (csum_skb) 566 continue; 567 568 len = skb->len - skb_transport_offset(skb); 569 if (offset >= len) { 570 offset -= len; 571 continue; 572 } 573 574 csum_skb = skb; 575 } 576 577 skb = csum_skb; 578 } 579 580 offset += skb_transport_offset(skb); 581 if (skb_copy_bits(skb, offset, &csum, 2)) 582 BUG(); 583 584 /* in case cksum was not initialized */ 585 if (unlikely(csum)) 586 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 587 588 csum = csum_ipv6_magic(&fl->fl6_src, 589 &fl->fl6_dst, 590 total_len, fl->proto, tmp_csum); 591 592 if (csum == 0 && fl->proto == IPPROTO_UDP) 593 csum = CSUM_MANGLED_0; 594 595 if (skb_store_bits(skb, offset, &csum, 2)) 596 BUG(); 597 598 send: 599 err = ip6_push_pending_frames(sk); 600 out: 601 return err; 602 } 603 604 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length, 605 struct flowi *fl, struct rt6_info *rt, 606 unsigned int flags) 607 { 608 struct ipv6_pinfo *np = inet6_sk(sk); 609 struct ipv6hdr *iph; 610 struct sk_buff *skb; 611 int err; 612 613 if (length > rt->u.dst.dev->mtu) { 614 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu); 615 return -EMSGSIZE; 616 } 617 if (flags&MSG_PROBE) 618 goto out; 619 620 skb = sock_alloc_send_skb(sk, 621 length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15, 622 flags & MSG_DONTWAIT, &err); 623 if (skb == NULL) 624 goto error; 625 skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev)); 626 627 skb->priority = sk->sk_priority; 628 skb->mark = sk->sk_mark; 629 skb_dst_set(skb, dst_clone(&rt->u.dst)); 630 631 skb_put(skb, length); 632 skb_reset_network_header(skb); 633 iph = ipv6_hdr(skb); 634 635 skb->ip_summed = CHECKSUM_NONE; 636 637 skb->transport_header = skb->network_header; 638 err = memcpy_fromiovecend((void *)iph, from, 0, length); 639 if (err) 640 goto error_fault; 641 642 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); 643 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, 644 rt->u.dst.dev, dst_output); 645 if (err > 0) 646 err = net_xmit_errno(err); 647 if (err) 648 goto error; 649 out: 650 return 0; 651 652 error_fault: 653 err = -EFAULT; 654 kfree_skb(skb); 655 error: 656 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 657 if (err == -ENOBUFS && !np->recverr) 658 err = 0; 659 return err; 660 } 661 662 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg) 663 { 664 struct iovec *iov; 665 u8 __user *type = NULL; 666 u8 __user *code = NULL; 667 u8 len = 0; 668 int probed = 0; 669 int i; 670 671 if (!msg->msg_iov) 672 return 0; 673 674 for (i = 0; i < msg->msg_iovlen; i++) { 675 iov = &msg->msg_iov[i]; 676 if (!iov) 677 continue; 678 679 switch (fl->proto) { 680 case IPPROTO_ICMPV6: 681 /* check if one-byte field is readable or not. */ 682 if (iov->iov_base && iov->iov_len < 1) 683 break; 684 685 if (!type) { 686 type = iov->iov_base; 687 /* check if code field is readable or not. */ 688 if (iov->iov_len > 1) 689 code = type + 1; 690 } else if (!code) 691 code = iov->iov_base; 692 693 if (type && code) { 694 if (get_user(fl->fl_icmp_type, type) || 695 get_user(fl->fl_icmp_code, code)) 696 return -EFAULT; 697 probed = 1; 698 } 699 break; 700 case IPPROTO_MH: 701 if (iov->iov_base && iov->iov_len < 1) 702 break; 703 /* check if type field is readable or not. */ 704 if (iov->iov_len > 2 - len) { 705 u8 __user *p = iov->iov_base; 706 if (get_user(fl->fl_mh_type, &p[2 - len])) 707 return -EFAULT; 708 probed = 1; 709 } else 710 len += iov->iov_len; 711 712 break; 713 default: 714 probed = 1; 715 break; 716 } 717 if (probed) 718 break; 719 } 720 return 0; 721 } 722 723 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk, 724 struct msghdr *msg, size_t len) 725 { 726 struct ipv6_txoptions opt_space; 727 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name; 728 struct in6_addr *daddr, *final_p = NULL, final; 729 struct inet_sock *inet = inet_sk(sk); 730 struct ipv6_pinfo *np = inet6_sk(sk); 731 struct raw6_sock *rp = raw6_sk(sk); 732 struct ipv6_txoptions *opt = NULL; 733 struct ip6_flowlabel *flowlabel = NULL; 734 struct dst_entry *dst = NULL; 735 struct flowi fl; 736 int addr_len = msg->msg_namelen; 737 int hlimit = -1; 738 int tclass = -1; 739 int dontfrag = -1; 740 u16 proto; 741 int err; 742 743 /* Rough check on arithmetic overflow, 744 better check is made in ip6_append_data(). 745 */ 746 if (len > INT_MAX) 747 return -EMSGSIZE; 748 749 /* Mirror BSD error message compatibility */ 750 if (msg->msg_flags & MSG_OOB) 751 return -EOPNOTSUPP; 752 753 /* 754 * Get and verify the address. 755 */ 756 memset(&fl, 0, sizeof(fl)); 757 758 fl.mark = sk->sk_mark; 759 760 if (sin6) { 761 if (addr_len < SIN6_LEN_RFC2133) 762 return -EINVAL; 763 764 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 765 return(-EAFNOSUPPORT); 766 767 /* port is the proto value [0..255] carried in nexthdr */ 768 proto = ntohs(sin6->sin6_port); 769 770 if (!proto) 771 proto = inet->inet_num; 772 else if (proto != inet->inet_num) 773 return(-EINVAL); 774 775 if (proto > 255) 776 return(-EINVAL); 777 778 daddr = &sin6->sin6_addr; 779 if (np->sndflow) { 780 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 781 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { 782 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 783 if (flowlabel == NULL) 784 return -EINVAL; 785 daddr = &flowlabel->dst; 786 } 787 } 788 789 /* 790 * Otherwise it will be difficult to maintain 791 * sk->sk_dst_cache. 792 */ 793 if (sk->sk_state == TCP_ESTABLISHED && 794 ipv6_addr_equal(daddr, &np->daddr)) 795 daddr = &np->daddr; 796 797 if (addr_len >= sizeof(struct sockaddr_in6) && 798 sin6->sin6_scope_id && 799 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) 800 fl.oif = sin6->sin6_scope_id; 801 } else { 802 if (sk->sk_state != TCP_ESTABLISHED) 803 return -EDESTADDRREQ; 804 805 proto = inet->inet_num; 806 daddr = &np->daddr; 807 fl.fl6_flowlabel = np->flow_label; 808 } 809 810 if (fl.oif == 0) 811 fl.oif = sk->sk_bound_dev_if; 812 813 if (msg->msg_controllen) { 814 opt = &opt_space; 815 memset(opt, 0, sizeof(struct ipv6_txoptions)); 816 opt->tot_len = sizeof(struct ipv6_txoptions); 817 818 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit, 819 &tclass, &dontfrag); 820 if (err < 0) { 821 fl6_sock_release(flowlabel); 822 return err; 823 } 824 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 825 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 826 if (flowlabel == NULL) 827 return -EINVAL; 828 } 829 if (!(opt->opt_nflen|opt->opt_flen)) 830 opt = NULL; 831 } 832 if (opt == NULL) 833 opt = np->opt; 834 if (flowlabel) 835 opt = fl6_merge_options(&opt_space, flowlabel, opt); 836 opt = ipv6_fixup_options(&opt_space, opt); 837 838 fl.proto = proto; 839 err = rawv6_probe_proto_opt(&fl, msg); 840 if (err) 841 goto out; 842 843 if (!ipv6_addr_any(daddr)) 844 ipv6_addr_copy(&fl.fl6_dst, daddr); 845 else 846 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 847 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr)) 848 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 849 850 /* merge ip6_build_xmit from ip6_output */ 851 if (opt && opt->srcrt) { 852 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt; 853 ipv6_addr_copy(&final, &fl.fl6_dst); 854 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 855 final_p = &final; 856 } 857 858 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) 859 fl.oif = np->mcast_oif; 860 security_sk_classify_flow(sk, &fl); 861 862 err = ip6_dst_lookup(sk, &dst, &fl); 863 if (err) 864 goto out; 865 if (final_p) 866 ipv6_addr_copy(&fl.fl6_dst, final_p); 867 868 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT); 869 if (err < 0) { 870 if (err == -EREMOTE) 871 err = ip6_dst_blackhole(sk, &dst, &fl); 872 if (err < 0) 873 goto out; 874 } 875 876 if (hlimit < 0) { 877 if (ipv6_addr_is_multicast(&fl.fl6_dst)) 878 hlimit = np->mcast_hops; 879 else 880 hlimit = np->hop_limit; 881 if (hlimit < 0) 882 hlimit = ip6_dst_hoplimit(dst); 883 } 884 885 if (tclass < 0) 886 tclass = np->tclass; 887 888 if (dontfrag < 0) 889 dontfrag = np->dontfrag; 890 891 if (msg->msg_flags&MSG_CONFIRM) 892 goto do_confirm; 893 894 back_from_confirm: 895 if (inet->hdrincl) { 896 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags); 897 } else { 898 lock_sock(sk); 899 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, 900 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst, 901 msg->msg_flags, dontfrag); 902 903 if (err) 904 ip6_flush_pending_frames(sk); 905 else if (!(msg->msg_flags & MSG_MORE)) 906 err = rawv6_push_pending_frames(sk, &fl, rp); 907 release_sock(sk); 908 } 909 done: 910 dst_release(dst); 911 out: 912 fl6_sock_release(flowlabel); 913 return err<0?err:len; 914 do_confirm: 915 dst_confirm(dst); 916 if (!(msg->msg_flags & MSG_PROBE) || len) 917 goto back_from_confirm; 918 err = 0; 919 goto done; 920 } 921 922 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 923 char __user *optval, int optlen) 924 { 925 switch (optname) { 926 case ICMPV6_FILTER: 927 if (optlen > sizeof(struct icmp6_filter)) 928 optlen = sizeof(struct icmp6_filter); 929 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 930 return -EFAULT; 931 return 0; 932 default: 933 return -ENOPROTOOPT; 934 } 935 936 return 0; 937 } 938 939 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 940 char __user *optval, int __user *optlen) 941 { 942 int len; 943 944 switch (optname) { 945 case ICMPV6_FILTER: 946 if (get_user(len, optlen)) 947 return -EFAULT; 948 if (len < 0) 949 return -EINVAL; 950 if (len > sizeof(struct icmp6_filter)) 951 len = sizeof(struct icmp6_filter); 952 if (put_user(len, optlen)) 953 return -EFAULT; 954 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 955 return -EFAULT; 956 return 0; 957 default: 958 return -ENOPROTOOPT; 959 } 960 961 return 0; 962 } 963 964 965 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 966 char __user *optval, unsigned int optlen) 967 { 968 struct raw6_sock *rp = raw6_sk(sk); 969 int val; 970 971 if (get_user(val, (int __user *)optval)) 972 return -EFAULT; 973 974 switch (optname) { 975 case IPV6_CHECKSUM: 976 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && 977 level == IPPROTO_IPV6) { 978 /* 979 * RFC3542 tells that IPV6_CHECKSUM socket 980 * option in the IPPROTO_IPV6 level is not 981 * allowed on ICMPv6 sockets. 982 * If you want to set it, use IPPROTO_RAW 983 * level IPV6_CHECKSUM socket option 984 * (Linux extension). 985 */ 986 return -EINVAL; 987 } 988 989 /* You may get strange result with a positive odd offset; 990 RFC2292bis agrees with me. */ 991 if (val > 0 && (val&1)) 992 return(-EINVAL); 993 if (val < 0) { 994 rp->checksum = 0; 995 } else { 996 rp->checksum = 1; 997 rp->offset = val; 998 } 999 1000 return 0; 1001 break; 1002 1003 default: 1004 return(-ENOPROTOOPT); 1005 } 1006 } 1007 1008 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 1009 char __user *optval, unsigned int optlen) 1010 { 1011 switch(level) { 1012 case SOL_RAW: 1013 break; 1014 1015 case SOL_ICMPV6: 1016 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1017 return -EOPNOTSUPP; 1018 return rawv6_seticmpfilter(sk, level, optname, optval, 1019 optlen); 1020 case SOL_IPV6: 1021 if (optname == IPV6_CHECKSUM) 1022 break; 1023 default: 1024 return ipv6_setsockopt(sk, level, optname, optval, 1025 optlen); 1026 } 1027 1028 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1029 } 1030 1031 #ifdef CONFIG_COMPAT 1032 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 1033 char __user *optval, unsigned int optlen) 1034 { 1035 switch (level) { 1036 case SOL_RAW: 1037 break; 1038 case SOL_ICMPV6: 1039 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1040 return -EOPNOTSUPP; 1041 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1042 case SOL_IPV6: 1043 if (optname == IPV6_CHECKSUM) 1044 break; 1045 default: 1046 return compat_ipv6_setsockopt(sk, level, optname, 1047 optval, optlen); 1048 } 1049 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1050 } 1051 #endif 1052 1053 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1054 char __user *optval, int __user *optlen) 1055 { 1056 struct raw6_sock *rp = raw6_sk(sk); 1057 int val, len; 1058 1059 if (get_user(len,optlen)) 1060 return -EFAULT; 1061 1062 switch (optname) { 1063 case IPV6_CHECKSUM: 1064 /* 1065 * We allow getsockopt() for IPPROTO_IPV6-level 1066 * IPV6_CHECKSUM socket option on ICMPv6 sockets 1067 * since RFC3542 is silent about it. 1068 */ 1069 if (rp->checksum == 0) 1070 val = -1; 1071 else 1072 val = rp->offset; 1073 break; 1074 1075 default: 1076 return -ENOPROTOOPT; 1077 } 1078 1079 len = min_t(unsigned int, sizeof(int), len); 1080 1081 if (put_user(len, optlen)) 1082 return -EFAULT; 1083 if (copy_to_user(optval,&val,len)) 1084 return -EFAULT; 1085 return 0; 1086 } 1087 1088 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1089 char __user *optval, int __user *optlen) 1090 { 1091 switch(level) { 1092 case SOL_RAW: 1093 break; 1094 1095 case SOL_ICMPV6: 1096 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1097 return -EOPNOTSUPP; 1098 return rawv6_geticmpfilter(sk, level, optname, optval, 1099 optlen); 1100 case SOL_IPV6: 1101 if (optname == IPV6_CHECKSUM) 1102 break; 1103 default: 1104 return ipv6_getsockopt(sk, level, optname, optval, 1105 optlen); 1106 } 1107 1108 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1109 } 1110 1111 #ifdef CONFIG_COMPAT 1112 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1113 char __user *optval, int __user *optlen) 1114 { 1115 switch (level) { 1116 case SOL_RAW: 1117 break; 1118 case SOL_ICMPV6: 1119 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1120 return -EOPNOTSUPP; 1121 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1122 case SOL_IPV6: 1123 if (optname == IPV6_CHECKSUM) 1124 break; 1125 default: 1126 return compat_ipv6_getsockopt(sk, level, optname, 1127 optval, optlen); 1128 } 1129 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1130 } 1131 #endif 1132 1133 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1134 { 1135 switch(cmd) { 1136 case SIOCOUTQ: 1137 { 1138 int amount = sk_wmem_alloc_get(sk); 1139 1140 return put_user(amount, (int __user *)arg); 1141 } 1142 case SIOCINQ: 1143 { 1144 struct sk_buff *skb; 1145 int amount = 0; 1146 1147 spin_lock_bh(&sk->sk_receive_queue.lock); 1148 skb = skb_peek(&sk->sk_receive_queue); 1149 if (skb != NULL) 1150 amount = skb->tail - skb->transport_header; 1151 spin_unlock_bh(&sk->sk_receive_queue.lock); 1152 return put_user(amount, (int __user *)arg); 1153 } 1154 1155 default: 1156 #ifdef CONFIG_IPV6_MROUTE 1157 return ip6mr_ioctl(sk, cmd, (void __user *)arg); 1158 #else 1159 return -ENOIOCTLCMD; 1160 #endif 1161 } 1162 } 1163 1164 static void rawv6_close(struct sock *sk, long timeout) 1165 { 1166 if (inet_sk(sk)->inet_num == IPPROTO_RAW) 1167 ip6_ra_control(sk, -1); 1168 ip6mr_sk_done(sk); 1169 sk_common_release(sk); 1170 } 1171 1172 static void raw6_destroy(struct sock *sk) 1173 { 1174 lock_sock(sk); 1175 ip6_flush_pending_frames(sk); 1176 release_sock(sk); 1177 1178 inet6_destroy_sock(sk); 1179 } 1180 1181 static int rawv6_init_sk(struct sock *sk) 1182 { 1183 struct raw6_sock *rp = raw6_sk(sk); 1184 1185 switch (inet_sk(sk)->inet_num) { 1186 case IPPROTO_ICMPV6: 1187 rp->checksum = 1; 1188 rp->offset = 2; 1189 break; 1190 case IPPROTO_MH: 1191 rp->checksum = 1; 1192 rp->offset = 4; 1193 break; 1194 default: 1195 break; 1196 } 1197 return(0); 1198 } 1199 1200 struct proto rawv6_prot = { 1201 .name = "RAWv6", 1202 .owner = THIS_MODULE, 1203 .close = rawv6_close, 1204 .destroy = raw6_destroy, 1205 .connect = ip6_datagram_connect, 1206 .disconnect = udp_disconnect, 1207 .ioctl = rawv6_ioctl, 1208 .init = rawv6_init_sk, 1209 .setsockopt = rawv6_setsockopt, 1210 .getsockopt = rawv6_getsockopt, 1211 .sendmsg = rawv6_sendmsg, 1212 .recvmsg = rawv6_recvmsg, 1213 .bind = rawv6_bind, 1214 .backlog_rcv = rawv6_rcv_skb, 1215 .hash = raw_hash_sk, 1216 .unhash = raw_unhash_sk, 1217 .obj_size = sizeof(struct raw6_sock), 1218 .h.raw_hash = &raw_v6_hashinfo, 1219 #ifdef CONFIG_COMPAT 1220 .compat_setsockopt = compat_rawv6_setsockopt, 1221 .compat_getsockopt = compat_rawv6_getsockopt, 1222 #endif 1223 }; 1224 1225 #ifdef CONFIG_PROC_FS 1226 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1227 { 1228 struct ipv6_pinfo *np = inet6_sk(sp); 1229 struct in6_addr *dest, *src; 1230 __u16 destp, srcp; 1231 1232 dest = &np->daddr; 1233 src = &np->rcv_saddr; 1234 destp = 0; 1235 srcp = inet_sk(sp)->inet_num; 1236 seq_printf(seq, 1237 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1238 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n", 1239 i, 1240 src->s6_addr32[0], src->s6_addr32[1], 1241 src->s6_addr32[2], src->s6_addr32[3], srcp, 1242 dest->s6_addr32[0], dest->s6_addr32[1], 1243 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1244 sp->sk_state, 1245 sk_wmem_alloc_get(sp), 1246 sk_rmem_alloc_get(sp), 1247 0, 0L, 0, 1248 sock_i_uid(sp), 0, 1249 sock_i_ino(sp), 1250 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); 1251 } 1252 1253 static int raw6_seq_show(struct seq_file *seq, void *v) 1254 { 1255 if (v == SEQ_START_TOKEN) 1256 seq_printf(seq, 1257 " sl " 1258 "local_address " 1259 "remote_address " 1260 "st tx_queue rx_queue tr tm->when retrnsmt" 1261 " uid timeout inode ref pointer drops\n"); 1262 else 1263 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket); 1264 return 0; 1265 } 1266 1267 static const struct seq_operations raw6_seq_ops = { 1268 .start = raw_seq_start, 1269 .next = raw_seq_next, 1270 .stop = raw_seq_stop, 1271 .show = raw6_seq_show, 1272 }; 1273 1274 static int raw6_seq_open(struct inode *inode, struct file *file) 1275 { 1276 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops); 1277 } 1278 1279 static const struct file_operations raw6_seq_fops = { 1280 .owner = THIS_MODULE, 1281 .open = raw6_seq_open, 1282 .read = seq_read, 1283 .llseek = seq_lseek, 1284 .release = seq_release_net, 1285 }; 1286 1287 static int __net_init raw6_init_net(struct net *net) 1288 { 1289 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops)) 1290 return -ENOMEM; 1291 1292 return 0; 1293 } 1294 1295 static void __net_exit raw6_exit_net(struct net *net) 1296 { 1297 proc_net_remove(net, "raw6"); 1298 } 1299 1300 static struct pernet_operations raw6_net_ops = { 1301 .init = raw6_init_net, 1302 .exit = raw6_exit_net, 1303 }; 1304 1305 int __init raw6_proc_init(void) 1306 { 1307 return register_pernet_subsys(&raw6_net_ops); 1308 } 1309 1310 void raw6_proc_exit(void) 1311 { 1312 unregister_pernet_subsys(&raw6_net_ops); 1313 } 1314 #endif /* CONFIG_PROC_FS */ 1315 1316 /* Same as inet6_dgram_ops, sans udp_poll. */ 1317 static const struct proto_ops inet6_sockraw_ops = { 1318 .family = PF_INET6, 1319 .owner = THIS_MODULE, 1320 .release = inet6_release, 1321 .bind = inet6_bind, 1322 .connect = inet_dgram_connect, /* ok */ 1323 .socketpair = sock_no_socketpair, /* a do nothing */ 1324 .accept = sock_no_accept, /* a do nothing */ 1325 .getname = inet6_getname, 1326 .poll = datagram_poll, /* ok */ 1327 .ioctl = inet6_ioctl, /* must change */ 1328 .listen = sock_no_listen, /* ok */ 1329 .shutdown = inet_shutdown, /* ok */ 1330 .setsockopt = sock_common_setsockopt, /* ok */ 1331 .getsockopt = sock_common_getsockopt, /* ok */ 1332 .sendmsg = inet_sendmsg, /* ok */ 1333 .recvmsg = sock_common_recvmsg, /* ok */ 1334 .mmap = sock_no_mmap, 1335 .sendpage = sock_no_sendpage, 1336 #ifdef CONFIG_COMPAT 1337 .compat_setsockopt = compat_sock_common_setsockopt, 1338 .compat_getsockopt = compat_sock_common_getsockopt, 1339 #endif 1340 }; 1341 1342 static struct inet_protosw rawv6_protosw = { 1343 .type = SOCK_RAW, 1344 .protocol = IPPROTO_IP, /* wild card */ 1345 .prot = &rawv6_prot, 1346 .ops = &inet6_sockraw_ops, 1347 .no_check = UDP_CSUM_DEFAULT, 1348 .flags = INET_PROTOSW_REUSE, 1349 }; 1350 1351 int __init rawv6_init(void) 1352 { 1353 int ret; 1354 1355 ret = inet6_register_protosw(&rawv6_protosw); 1356 if (ret) 1357 goto out; 1358 out: 1359 return ret; 1360 } 1361 1362 void rawv6_exit(void) 1363 { 1364 inet6_unregister_protosw(&rawv6_protosw); 1365 } 1366