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 * $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $ 11 * 12 * Fixes: 13 * Hideaki YOSHIFUJI : sin6_scope_id support 14 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance) 15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data 16 * 17 * This program is free software; you can redistribute it and/or 18 * modify it under the terms of the GNU General Public License 19 * as published by the Free Software Foundation; either version 20 * 2 of the License, or (at your option) any later version. 21 */ 22 23 #include <linux/errno.h> 24 #include <linux/types.h> 25 #include <linux/socket.h> 26 #include <linux/sockios.h> 27 #include <linux/net.h> 28 #include <linux/in6.h> 29 #include <linux/netdevice.h> 30 #include <linux/if_arp.h> 31 #include <linux/icmpv6.h> 32 #include <linux/netfilter.h> 33 #include <linux/netfilter_ipv6.h> 34 #include <linux/skbuff.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 defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 54 #include <net/mip6.h> 55 #endif 56 57 #include <net/rawv6.h> 58 #include <net/xfrm.h> 59 60 #include <linux/proc_fs.h> 61 #include <linux/seq_file.h> 62 63 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE]; 64 DEFINE_RWLOCK(raw_v6_lock); 65 66 static void raw_v6_hash(struct sock *sk) 67 { 68 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num & 69 (RAWV6_HTABLE_SIZE - 1)]; 70 71 write_lock_bh(&raw_v6_lock); 72 sk_add_node(sk, list); 73 sock_prot_inc_use(sk->sk_prot); 74 write_unlock_bh(&raw_v6_lock); 75 } 76 77 static void raw_v6_unhash(struct sock *sk) 78 { 79 write_lock_bh(&raw_v6_lock); 80 if (sk_del_node_init(sk)) 81 sock_prot_dec_use(sk->sk_prot); 82 write_unlock_bh(&raw_v6_lock); 83 } 84 85 86 /* Grumble... icmp and ip_input want to get at this... */ 87 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num, 88 struct in6_addr *loc_addr, struct in6_addr *rmt_addr, 89 int dif) 90 { 91 struct hlist_node *node; 92 int is_multicast = ipv6_addr_is_multicast(loc_addr); 93 94 sk_for_each_from(sk, node) 95 if (inet_sk(sk)->num == num) { 96 struct ipv6_pinfo *np = inet6_sk(sk); 97 98 if (!ipv6_addr_any(&np->daddr) && 99 !ipv6_addr_equal(&np->daddr, rmt_addr)) 100 continue; 101 102 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) 103 continue; 104 105 if (!ipv6_addr_any(&np->rcv_saddr)) { 106 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr)) 107 goto found; 108 if (is_multicast && 109 inet6_mc_check(sk, loc_addr, rmt_addr)) 110 goto found; 111 continue; 112 } 113 goto found; 114 } 115 sk = NULL; 116 found: 117 return sk; 118 } 119 120 /* 121 * 0 - deliver 122 * 1 - block 123 */ 124 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb) 125 { 126 struct icmp6hdr *icmph; 127 struct raw6_sock *rp = raw6_sk(sk); 128 129 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) { 130 __u32 *data = &rp->filter.data[0]; 131 int bit_nr; 132 133 icmph = (struct icmp6hdr *) skb->data; 134 bit_nr = icmph->icmp6_type; 135 136 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0; 137 } 138 return 0; 139 } 140 141 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 142 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb); 143 144 int rawv6_mh_filter_register(int (*filter)(struct sock *sock, 145 struct sk_buff *skb)) 146 { 147 rcu_assign_pointer(mh_filter, filter); 148 return 0; 149 } 150 EXPORT_SYMBOL(rawv6_mh_filter_register); 151 152 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock, 153 struct sk_buff *skb)) 154 { 155 rcu_assign_pointer(mh_filter, NULL); 156 synchronize_rcu(); 157 return 0; 158 } 159 EXPORT_SYMBOL(rawv6_mh_filter_unregister); 160 161 #endif 162 163 /* 164 * demultiplex raw sockets. 165 * (should consider queueing the skb in the sock receive_queue 166 * without calling rawv6.c) 167 * 168 * Caller owns SKB so we must make clones. 169 */ 170 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 171 { 172 struct in6_addr *saddr; 173 struct in6_addr *daddr; 174 struct sock *sk; 175 int delivered = 0; 176 __u8 hash; 177 178 saddr = &ipv6_hdr(skb)->saddr; 179 daddr = saddr + 1; 180 181 hash = nexthdr & (MAX_INET_PROTOS - 1); 182 183 read_lock(&raw_v6_lock); 184 sk = sk_head(&raw_v6_htable[hash]); 185 186 /* 187 * The first socket found will be delivered after 188 * delivery to transport protocols. 189 */ 190 191 if (sk == NULL) 192 goto out; 193 194 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif); 195 196 while (sk) { 197 int filtered; 198 199 delivered = 1; 200 switch (nexthdr) { 201 case IPPROTO_ICMPV6: 202 filtered = icmpv6_filter(sk, skb); 203 break; 204 205 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 206 case IPPROTO_MH: 207 { 208 /* XXX: To validate MH only once for each packet, 209 * this is placed here. It should be after checking 210 * xfrm policy, however it doesn't. The checking xfrm 211 * policy is placed in rawv6_rcv() because it is 212 * required for each socket. 213 */ 214 int (*filter)(struct sock *sock, struct sk_buff *skb); 215 216 filter = rcu_dereference(mh_filter); 217 filtered = filter ? filter(sk, skb) : 0; 218 break; 219 } 220 #endif 221 default: 222 filtered = 0; 223 break; 224 } 225 226 if (filtered < 0) 227 break; 228 if (filtered == 0) { 229 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 230 231 /* Not releasing hash table! */ 232 if (clone) { 233 nf_reset(clone); 234 rawv6_rcv(sk, clone); 235 } 236 } 237 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr, 238 IP6CB(skb)->iif); 239 } 240 out: 241 read_unlock(&raw_v6_lock); 242 return delivered; 243 } 244 245 /* This cleans up af_inet6 a bit. -DaveM */ 246 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 247 { 248 struct inet_sock *inet = inet_sk(sk); 249 struct ipv6_pinfo *np = inet6_sk(sk); 250 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 251 __be32 v4addr = 0; 252 int addr_type; 253 int err; 254 255 if (addr_len < SIN6_LEN_RFC2133) 256 return -EINVAL; 257 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 /* Check if the address belongs to the host. */ 270 if (addr_type != IPV6_ADDR_ANY) { 271 struct net_device *dev = NULL; 272 273 if (addr_type & IPV6_ADDR_LINKLOCAL) { 274 if (addr_len >= sizeof(struct sockaddr_in6) && 275 addr->sin6_scope_id) { 276 /* Override any existing binding, if another 277 * one is supplied by user. 278 */ 279 sk->sk_bound_dev_if = addr->sin6_scope_id; 280 } 281 282 /* Binding to link-local address requires an interface */ 283 if (!sk->sk_bound_dev_if) 284 goto out; 285 286 dev = dev_get_by_index(&init_net, sk->sk_bound_dev_if); 287 if (!dev) { 288 err = -ENODEV; 289 goto out; 290 } 291 } 292 293 /* ipv4 addr of the socket is invalid. Only the 294 * unspecified and mapped address have a v4 equivalent. 295 */ 296 v4addr = LOOPBACK4_IPV6; 297 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 298 err = -EADDRNOTAVAIL; 299 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) { 300 if (dev) 301 dev_put(dev); 302 goto out; 303 } 304 } 305 if (dev) 306 dev_put(dev); 307 } 308 309 inet->rcv_saddr = inet->saddr = v4addr; 310 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr); 311 if (!(addr_type & IPV6_ADDR_MULTICAST)) 312 ipv6_addr_copy(&np->saddr, &addr->sin6_addr); 313 err = 0; 314 out: 315 release_sock(sk); 316 return err; 317 } 318 319 void rawv6_err(struct sock *sk, struct sk_buff *skb, 320 struct inet6_skb_parm *opt, 321 int type, int 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 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 339 340 if (np->recverr) { 341 u8 *payload = skb->data; 342 if (!inet->hdrincl) 343 payload += offset; 344 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 345 } 346 347 if (np->recverr || harderr) { 348 sk->sk_err = err; 349 sk->sk_error_report(sk); 350 } 351 } 352 353 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb) 354 { 355 if ((raw6_sk(sk)->checksum || sk->sk_filter) && 356 skb_checksum_complete(skb)) { 357 /* FIXME: increment a raw6 drops counter here */ 358 kfree_skb(skb); 359 return 0; 360 } 361 362 /* Charge it to the socket. */ 363 if (sock_queue_rcv_skb(sk,skb)<0) { 364 /* FIXME: increment a raw6 drops counter here */ 365 kfree_skb(skb); 366 return 0; 367 } 368 369 return 0; 370 } 371 372 /* 373 * This is next to useless... 374 * if we demultiplex in network layer we don't need the extra call 375 * just to queue the skb... 376 * maybe we could have the network decide upon a hint if it 377 * should call raw_rcv for demultiplexing 378 */ 379 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 380 { 381 struct inet_sock *inet = inet_sk(sk); 382 struct raw6_sock *rp = raw6_sk(sk); 383 384 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 385 kfree_skb(skb); 386 return NET_RX_DROP; 387 } 388 389 if (!rp->checksum) 390 skb->ip_summed = CHECKSUM_UNNECESSARY; 391 392 if (skb->ip_summed == CHECKSUM_COMPLETE) { 393 skb_postpull_rcsum(skb, skb_network_header(skb), 394 skb_network_header_len(skb)); 395 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 396 &ipv6_hdr(skb)->daddr, 397 skb->len, inet->num, skb->csum)) 398 skb->ip_summed = CHECKSUM_UNNECESSARY; 399 } 400 if (!skb_csum_unnecessary(skb)) 401 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 402 &ipv6_hdr(skb)->daddr, 403 skb->len, 404 inet->num, 0)); 405 406 if (inet->hdrincl) { 407 if (skb_checksum_complete(skb)) { 408 /* FIXME: increment a raw6 drops counter here */ 409 kfree_skb(skb); 410 return 0; 411 } 412 } 413 414 rawv6_rcv_skb(sk, skb); 415 return 0; 416 } 417 418 419 /* 420 * This should be easy, if there is something there 421 * we return it, otherwise we block. 422 */ 423 424 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk, 425 struct msghdr *msg, size_t len, 426 int noblock, int flags, int *addr_len) 427 { 428 struct ipv6_pinfo *np = inet6_sk(sk); 429 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name; 430 struct sk_buff *skb; 431 size_t copied; 432 int err; 433 434 if (flags & MSG_OOB) 435 return -EOPNOTSUPP; 436 437 if (addr_len) 438 *addr_len=sizeof(*sin6); 439 440 if (flags & MSG_ERRQUEUE) 441 return ipv6_recv_error(sk, msg, len); 442 443 skb = skb_recv_datagram(sk, flags, noblock, &err); 444 if (!skb) 445 goto out; 446 447 copied = skb->len; 448 if (copied > len) { 449 copied = len; 450 msg->msg_flags |= MSG_TRUNC; 451 } 452 453 if (skb_csum_unnecessary(skb)) { 454 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 455 } else if (msg->msg_flags&MSG_TRUNC) { 456 if (__skb_checksum_complete(skb)) 457 goto csum_copy_err; 458 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 459 } else { 460 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov); 461 if (err == -EINVAL) 462 goto csum_copy_err; 463 } 464 if (err) 465 goto out_free; 466 467 /* Copy the address. */ 468 if (sin6) { 469 sin6->sin6_family = AF_INET6; 470 sin6->sin6_port = 0; 471 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr); 472 sin6->sin6_flowinfo = 0; 473 sin6->sin6_scope_id = 0; 474 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) 475 sin6->sin6_scope_id = IP6CB(skb)->iif; 476 } 477 478 sock_recv_timestamp(msg, sk, skb); 479 480 if (np->rxopt.all) 481 datagram_recv_ctl(sk, msg, skb); 482 483 err = copied; 484 if (flags & MSG_TRUNC) 485 err = skb->len; 486 487 out_free: 488 skb_free_datagram(sk, skb); 489 out: 490 return err; 491 492 csum_copy_err: 493 skb_kill_datagram(sk, skb, flags); 494 495 /* Error for blocking case is chosen to masquerade 496 as some normal condition. 497 */ 498 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 499 /* FIXME: increment a raw6 drops counter here */ 500 goto out; 501 } 502 503 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl, 504 struct raw6_sock *rp) 505 { 506 struct sk_buff *skb; 507 int err = 0; 508 int offset; 509 int len; 510 int total_len; 511 __wsum tmp_csum; 512 __sum16 csum; 513 514 if (!rp->checksum) 515 goto send; 516 517 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 518 goto out; 519 520 offset = rp->offset; 521 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) - 522 skb->data); 523 if (offset >= total_len - 1) { 524 err = -EINVAL; 525 ip6_flush_pending_frames(sk); 526 goto out; 527 } 528 529 /* should be check HW csum miyazawa */ 530 if (skb_queue_len(&sk->sk_write_queue) == 1) { 531 /* 532 * Only one fragment on the socket. 533 */ 534 tmp_csum = skb->csum; 535 } else { 536 struct sk_buff *csum_skb = NULL; 537 tmp_csum = 0; 538 539 skb_queue_walk(&sk->sk_write_queue, skb) { 540 tmp_csum = csum_add(tmp_csum, skb->csum); 541 542 if (csum_skb) 543 continue; 544 545 len = skb->len - skb_transport_offset(skb); 546 if (offset >= len) { 547 offset -= len; 548 continue; 549 } 550 551 csum_skb = skb; 552 } 553 554 skb = csum_skb; 555 } 556 557 offset += skb_transport_offset(skb); 558 if (skb_copy_bits(skb, offset, &csum, 2)) 559 BUG(); 560 561 /* in case cksum was not initialized */ 562 if (unlikely(csum)) 563 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 564 565 csum = csum_ipv6_magic(&fl->fl6_src, 566 &fl->fl6_dst, 567 total_len, fl->proto, tmp_csum); 568 569 if (csum == 0 && fl->proto == IPPROTO_UDP) 570 csum = CSUM_MANGLED_0; 571 572 if (skb_store_bits(skb, offset, &csum, 2)) 573 BUG(); 574 575 send: 576 err = ip6_push_pending_frames(sk); 577 out: 578 return err; 579 } 580 581 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length, 582 struct flowi *fl, struct rt6_info *rt, 583 unsigned int flags) 584 { 585 struct ipv6_pinfo *np = inet6_sk(sk); 586 struct ipv6hdr *iph; 587 struct sk_buff *skb; 588 unsigned int hh_len; 589 int err; 590 591 if (length > rt->u.dst.dev->mtu) { 592 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu); 593 return -EMSGSIZE; 594 } 595 if (flags&MSG_PROBE) 596 goto out; 597 598 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); 599 600 skb = sock_alloc_send_skb(sk, length+hh_len+15, 601 flags&MSG_DONTWAIT, &err); 602 if (skb == NULL) 603 goto error; 604 skb_reserve(skb, hh_len); 605 606 skb->priority = sk->sk_priority; 607 skb->dst = dst_clone(&rt->u.dst); 608 609 skb_put(skb, length); 610 skb_reset_network_header(skb); 611 iph = ipv6_hdr(skb); 612 613 skb->ip_summed = CHECKSUM_NONE; 614 615 skb->transport_header = skb->network_header; 616 err = memcpy_fromiovecend((void *)iph, from, 0, length); 617 if (err) 618 goto error_fault; 619 620 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS); 621 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev, 622 dst_output); 623 if (err > 0) 624 err = np->recverr ? net_xmit_errno(err) : 0; 625 if (err) 626 goto error; 627 out: 628 return 0; 629 630 error_fault: 631 err = -EFAULT; 632 kfree_skb(skb); 633 error: 634 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 635 return err; 636 } 637 638 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg) 639 { 640 struct iovec *iov; 641 u8 __user *type = NULL; 642 u8 __user *code = NULL; 643 u8 len = 0; 644 int probed = 0; 645 int i; 646 647 if (!msg->msg_iov) 648 return 0; 649 650 for (i = 0; i < msg->msg_iovlen; i++) { 651 iov = &msg->msg_iov[i]; 652 if (!iov) 653 continue; 654 655 switch (fl->proto) { 656 case IPPROTO_ICMPV6: 657 /* check if one-byte field is readable or not. */ 658 if (iov->iov_base && iov->iov_len < 1) 659 break; 660 661 if (!type) { 662 type = iov->iov_base; 663 /* check if code field is readable or not. */ 664 if (iov->iov_len > 1) 665 code = type + 1; 666 } else if (!code) 667 code = iov->iov_base; 668 669 if (type && code) { 670 if (get_user(fl->fl_icmp_type, type) || 671 get_user(fl->fl_icmp_code, code)) 672 return -EFAULT; 673 probed = 1; 674 } 675 break; 676 case IPPROTO_MH: 677 if (iov->iov_base && iov->iov_len < 1) 678 break; 679 /* check if type field is readable or not. */ 680 if (iov->iov_len > 2 - len) { 681 u8 __user *p = iov->iov_base; 682 if (get_user(fl->fl_mh_type, &p[2 - len])) 683 return -EFAULT; 684 probed = 1; 685 } else 686 len += iov->iov_len; 687 688 break; 689 default: 690 probed = 1; 691 break; 692 } 693 if (probed) 694 break; 695 } 696 return 0; 697 } 698 699 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk, 700 struct msghdr *msg, size_t len) 701 { 702 struct ipv6_txoptions opt_space; 703 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name; 704 struct in6_addr *daddr, *final_p = NULL, final; 705 struct inet_sock *inet = inet_sk(sk); 706 struct ipv6_pinfo *np = inet6_sk(sk); 707 struct raw6_sock *rp = raw6_sk(sk); 708 struct ipv6_txoptions *opt = NULL; 709 struct ip6_flowlabel *flowlabel = NULL; 710 struct dst_entry *dst = NULL; 711 struct flowi fl; 712 int addr_len = msg->msg_namelen; 713 int hlimit = -1; 714 int tclass = -1; 715 u16 proto; 716 int err; 717 718 /* Rough check on arithmetic overflow, 719 better check is made in ip6_append_data(). 720 */ 721 if (len > INT_MAX) 722 return -EMSGSIZE; 723 724 /* Mirror BSD error message compatibility */ 725 if (msg->msg_flags & MSG_OOB) 726 return -EOPNOTSUPP; 727 728 /* 729 * Get and verify the address. 730 */ 731 memset(&fl, 0, sizeof(fl)); 732 733 if (sin6) { 734 if (addr_len < SIN6_LEN_RFC2133) 735 return -EINVAL; 736 737 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 738 return(-EAFNOSUPPORT); 739 740 /* port is the proto value [0..255] carried in nexthdr */ 741 proto = ntohs(sin6->sin6_port); 742 743 if (!proto) 744 proto = inet->num; 745 else if (proto != inet->num) 746 return(-EINVAL); 747 748 if (proto > 255) 749 return(-EINVAL); 750 751 daddr = &sin6->sin6_addr; 752 if (np->sndflow) { 753 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 754 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { 755 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 756 if (flowlabel == NULL) 757 return -EINVAL; 758 daddr = &flowlabel->dst; 759 } 760 } 761 762 /* 763 * Otherwise it will be difficult to maintain 764 * sk->sk_dst_cache. 765 */ 766 if (sk->sk_state == TCP_ESTABLISHED && 767 ipv6_addr_equal(daddr, &np->daddr)) 768 daddr = &np->daddr; 769 770 if (addr_len >= sizeof(struct sockaddr_in6) && 771 sin6->sin6_scope_id && 772 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) 773 fl.oif = sin6->sin6_scope_id; 774 } else { 775 if (sk->sk_state != TCP_ESTABLISHED) 776 return -EDESTADDRREQ; 777 778 proto = inet->num; 779 daddr = &np->daddr; 780 fl.fl6_flowlabel = np->flow_label; 781 } 782 783 if (ipv6_addr_any(daddr)) { 784 /* 785 * unspecified destination address 786 * treated as error... is this correct ? 787 */ 788 fl6_sock_release(flowlabel); 789 return(-EINVAL); 790 } 791 792 if (fl.oif == 0) 793 fl.oif = sk->sk_bound_dev_if; 794 795 if (msg->msg_controllen) { 796 opt = &opt_space; 797 memset(opt, 0, sizeof(struct ipv6_txoptions)); 798 opt->tot_len = sizeof(struct ipv6_txoptions); 799 800 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass); 801 if (err < 0) { 802 fl6_sock_release(flowlabel); 803 return err; 804 } 805 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 806 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 807 if (flowlabel == NULL) 808 return -EINVAL; 809 } 810 if (!(opt->opt_nflen|opt->opt_flen)) 811 opt = NULL; 812 } 813 if (opt == NULL) 814 opt = np->opt; 815 if (flowlabel) 816 opt = fl6_merge_options(&opt_space, flowlabel, opt); 817 opt = ipv6_fixup_options(&opt_space, opt); 818 819 fl.proto = proto; 820 err = rawv6_probe_proto_opt(&fl, msg); 821 if (err) 822 goto out; 823 824 ipv6_addr_copy(&fl.fl6_dst, daddr); 825 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr)) 826 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 827 828 /* merge ip6_build_xmit from ip6_output */ 829 if (opt && opt->srcrt) { 830 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt; 831 ipv6_addr_copy(&final, &fl.fl6_dst); 832 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 833 final_p = &final; 834 } 835 836 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) 837 fl.oif = np->mcast_oif; 838 security_sk_classify_flow(sk, &fl); 839 840 err = ip6_dst_lookup(sk, &dst, &fl); 841 if (err) 842 goto out; 843 if (final_p) 844 ipv6_addr_copy(&fl.fl6_dst, final_p); 845 846 if ((err = __xfrm_lookup(&dst, &fl, sk, 1)) < 0) { 847 if (err == -EREMOTE) 848 err = ip6_dst_blackhole(sk, &dst, &fl); 849 if (err < 0) 850 goto out; 851 } 852 853 if (hlimit < 0) { 854 if (ipv6_addr_is_multicast(&fl.fl6_dst)) 855 hlimit = np->mcast_hops; 856 else 857 hlimit = np->hop_limit; 858 if (hlimit < 0) 859 hlimit = dst_metric(dst, RTAX_HOPLIMIT); 860 if (hlimit < 0) 861 hlimit = ipv6_get_hoplimit(dst->dev); 862 } 863 864 if (tclass < 0) { 865 tclass = np->tclass; 866 if (tclass < 0) 867 tclass = 0; 868 } 869 870 if (msg->msg_flags&MSG_CONFIRM) 871 goto do_confirm; 872 873 back_from_confirm: 874 if (inet->hdrincl) { 875 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags); 876 } else { 877 lock_sock(sk); 878 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, 879 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst, 880 msg->msg_flags); 881 882 if (err) 883 ip6_flush_pending_frames(sk); 884 else if (!(msg->msg_flags & MSG_MORE)) 885 err = rawv6_push_pending_frames(sk, &fl, rp); 886 release_sock(sk); 887 } 888 done: 889 dst_release(dst); 890 out: 891 fl6_sock_release(flowlabel); 892 return err<0?err:len; 893 do_confirm: 894 dst_confirm(dst); 895 if (!(msg->msg_flags & MSG_PROBE) || len) 896 goto back_from_confirm; 897 err = 0; 898 goto done; 899 } 900 901 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 902 char __user *optval, int optlen) 903 { 904 switch (optname) { 905 case ICMPV6_FILTER: 906 if (optlen > sizeof(struct icmp6_filter)) 907 optlen = sizeof(struct icmp6_filter); 908 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 909 return -EFAULT; 910 return 0; 911 default: 912 return -ENOPROTOOPT; 913 } 914 915 return 0; 916 } 917 918 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 919 char __user *optval, int __user *optlen) 920 { 921 int len; 922 923 switch (optname) { 924 case ICMPV6_FILTER: 925 if (get_user(len, optlen)) 926 return -EFAULT; 927 if (len < 0) 928 return -EINVAL; 929 if (len > sizeof(struct icmp6_filter)) 930 len = sizeof(struct icmp6_filter); 931 if (put_user(len, optlen)) 932 return -EFAULT; 933 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 934 return -EFAULT; 935 return 0; 936 default: 937 return -ENOPROTOOPT; 938 } 939 940 return 0; 941 } 942 943 944 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 945 char __user *optval, int optlen) 946 { 947 struct raw6_sock *rp = raw6_sk(sk); 948 int val; 949 950 if (get_user(val, (int __user *)optval)) 951 return -EFAULT; 952 953 switch (optname) { 954 case IPV6_CHECKSUM: 955 /* You may get strange result with a positive odd offset; 956 RFC2292bis agrees with me. */ 957 if (val > 0 && (val&1)) 958 return(-EINVAL); 959 if (val < 0) { 960 rp->checksum = 0; 961 } else { 962 rp->checksum = 1; 963 rp->offset = val; 964 } 965 966 return 0; 967 break; 968 969 default: 970 return(-ENOPROTOOPT); 971 } 972 } 973 974 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 975 char __user *optval, int optlen) 976 { 977 switch(level) { 978 case SOL_RAW: 979 break; 980 981 case SOL_ICMPV6: 982 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 983 return -EOPNOTSUPP; 984 return rawv6_seticmpfilter(sk, level, optname, optval, 985 optlen); 986 case SOL_IPV6: 987 if (optname == IPV6_CHECKSUM) 988 break; 989 default: 990 return ipv6_setsockopt(sk, level, optname, optval, 991 optlen); 992 } 993 994 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 995 } 996 997 #ifdef CONFIG_COMPAT 998 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 999 char __user *optval, int optlen) 1000 { 1001 switch (level) { 1002 case SOL_RAW: 1003 break; 1004 case SOL_ICMPV6: 1005 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 1006 return -EOPNOTSUPP; 1007 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1008 case SOL_IPV6: 1009 if (optname == IPV6_CHECKSUM) 1010 break; 1011 default: 1012 return compat_ipv6_setsockopt(sk, level, optname, 1013 optval, optlen); 1014 } 1015 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1016 } 1017 #endif 1018 1019 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1020 char __user *optval, int __user *optlen) 1021 { 1022 struct raw6_sock *rp = raw6_sk(sk); 1023 int val, len; 1024 1025 if (get_user(len,optlen)) 1026 return -EFAULT; 1027 1028 switch (optname) { 1029 case IPV6_CHECKSUM: 1030 if (rp->checksum == 0) 1031 val = -1; 1032 else 1033 val = rp->offset; 1034 break; 1035 1036 default: 1037 return -ENOPROTOOPT; 1038 } 1039 1040 len = min_t(unsigned int, sizeof(int), len); 1041 1042 if (put_user(len, optlen)) 1043 return -EFAULT; 1044 if (copy_to_user(optval,&val,len)) 1045 return -EFAULT; 1046 return 0; 1047 } 1048 1049 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1050 char __user *optval, int __user *optlen) 1051 { 1052 switch(level) { 1053 case SOL_RAW: 1054 break; 1055 1056 case SOL_ICMPV6: 1057 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 1058 return -EOPNOTSUPP; 1059 return rawv6_geticmpfilter(sk, level, optname, optval, 1060 optlen); 1061 case SOL_IPV6: 1062 if (optname == IPV6_CHECKSUM) 1063 break; 1064 default: 1065 return ipv6_getsockopt(sk, level, optname, optval, 1066 optlen); 1067 } 1068 1069 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1070 } 1071 1072 #ifdef CONFIG_COMPAT 1073 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1074 char __user *optval, int __user *optlen) 1075 { 1076 switch (level) { 1077 case SOL_RAW: 1078 break; 1079 case SOL_ICMPV6: 1080 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 1081 return -EOPNOTSUPP; 1082 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1083 case SOL_IPV6: 1084 if (optname == IPV6_CHECKSUM) 1085 break; 1086 default: 1087 return compat_ipv6_getsockopt(sk, level, optname, 1088 optval, optlen); 1089 } 1090 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1091 } 1092 #endif 1093 1094 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1095 { 1096 switch(cmd) { 1097 case SIOCOUTQ: 1098 { 1099 int amount = atomic_read(&sk->sk_wmem_alloc); 1100 return put_user(amount, (int __user *)arg); 1101 } 1102 case SIOCINQ: 1103 { 1104 struct sk_buff *skb; 1105 int amount = 0; 1106 1107 spin_lock_bh(&sk->sk_receive_queue.lock); 1108 skb = skb_peek(&sk->sk_receive_queue); 1109 if (skb != NULL) 1110 amount = skb->tail - skb->transport_header; 1111 spin_unlock_bh(&sk->sk_receive_queue.lock); 1112 return put_user(amount, (int __user *)arg); 1113 } 1114 1115 default: 1116 return -ENOIOCTLCMD; 1117 } 1118 } 1119 1120 static void rawv6_close(struct sock *sk, long timeout) 1121 { 1122 if (inet_sk(sk)->num == IPPROTO_RAW) 1123 ip6_ra_control(sk, -1, NULL); 1124 1125 sk_common_release(sk); 1126 } 1127 1128 static int rawv6_init_sk(struct sock *sk) 1129 { 1130 struct raw6_sock *rp = raw6_sk(sk); 1131 1132 switch (inet_sk(sk)->num) { 1133 case IPPROTO_ICMPV6: 1134 rp->checksum = 1; 1135 rp->offset = 2; 1136 break; 1137 case IPPROTO_MH: 1138 rp->checksum = 1; 1139 rp->offset = 4; 1140 break; 1141 default: 1142 break; 1143 } 1144 return(0); 1145 } 1146 1147 struct proto rawv6_prot = { 1148 .name = "RAWv6", 1149 .owner = THIS_MODULE, 1150 .close = rawv6_close, 1151 .connect = ip6_datagram_connect, 1152 .disconnect = udp_disconnect, 1153 .ioctl = rawv6_ioctl, 1154 .init = rawv6_init_sk, 1155 .destroy = inet6_destroy_sock, 1156 .setsockopt = rawv6_setsockopt, 1157 .getsockopt = rawv6_getsockopt, 1158 .sendmsg = rawv6_sendmsg, 1159 .recvmsg = rawv6_recvmsg, 1160 .bind = rawv6_bind, 1161 .backlog_rcv = rawv6_rcv_skb, 1162 .hash = raw_v6_hash, 1163 .unhash = raw_v6_unhash, 1164 .obj_size = sizeof(struct raw6_sock), 1165 #ifdef CONFIG_COMPAT 1166 .compat_setsockopt = compat_rawv6_setsockopt, 1167 .compat_getsockopt = compat_rawv6_getsockopt, 1168 #endif 1169 }; 1170 1171 #ifdef CONFIG_PROC_FS 1172 struct raw6_iter_state { 1173 int bucket; 1174 }; 1175 1176 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private) 1177 1178 static struct sock *raw6_get_first(struct seq_file *seq) 1179 { 1180 struct sock *sk; 1181 struct hlist_node *node; 1182 struct raw6_iter_state* state = raw6_seq_private(seq); 1183 1184 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket) 1185 sk_for_each(sk, node, &raw_v6_htable[state->bucket]) 1186 if (sk->sk_family == PF_INET6) 1187 goto out; 1188 sk = NULL; 1189 out: 1190 return sk; 1191 } 1192 1193 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk) 1194 { 1195 struct raw6_iter_state* state = raw6_seq_private(seq); 1196 1197 do { 1198 sk = sk_next(sk); 1199 try_again: 1200 ; 1201 } while (sk && sk->sk_family != PF_INET6); 1202 1203 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) { 1204 sk = sk_head(&raw_v6_htable[state->bucket]); 1205 goto try_again; 1206 } 1207 return sk; 1208 } 1209 1210 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos) 1211 { 1212 struct sock *sk = raw6_get_first(seq); 1213 if (sk) 1214 while (pos && (sk = raw6_get_next(seq, sk)) != NULL) 1215 --pos; 1216 return pos ? NULL : sk; 1217 } 1218 1219 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos) 1220 { 1221 read_lock(&raw_v6_lock); 1222 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 1223 } 1224 1225 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1226 { 1227 struct sock *sk; 1228 1229 if (v == SEQ_START_TOKEN) 1230 sk = raw6_get_first(seq); 1231 else 1232 sk = raw6_get_next(seq, v); 1233 ++*pos; 1234 return sk; 1235 } 1236 1237 static void raw6_seq_stop(struct seq_file *seq, void *v) 1238 { 1239 read_unlock(&raw_v6_lock); 1240 } 1241 1242 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1243 { 1244 struct ipv6_pinfo *np = inet6_sk(sp); 1245 struct in6_addr *dest, *src; 1246 __u16 destp, srcp; 1247 1248 dest = &np->daddr; 1249 src = &np->rcv_saddr; 1250 destp = 0; 1251 srcp = inet_sk(sp)->num; 1252 seq_printf(seq, 1253 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1254 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n", 1255 i, 1256 src->s6_addr32[0], src->s6_addr32[1], 1257 src->s6_addr32[2], src->s6_addr32[3], srcp, 1258 dest->s6_addr32[0], dest->s6_addr32[1], 1259 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1260 sp->sk_state, 1261 atomic_read(&sp->sk_wmem_alloc), 1262 atomic_read(&sp->sk_rmem_alloc), 1263 0, 0L, 0, 1264 sock_i_uid(sp), 0, 1265 sock_i_ino(sp), 1266 atomic_read(&sp->sk_refcnt), sp); 1267 } 1268 1269 static int raw6_seq_show(struct seq_file *seq, void *v) 1270 { 1271 if (v == SEQ_START_TOKEN) 1272 seq_printf(seq, 1273 " sl " 1274 "local_address " 1275 "remote_address " 1276 "st tx_queue rx_queue tr tm->when retrnsmt" 1277 " uid timeout inode\n"); 1278 else 1279 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket); 1280 return 0; 1281 } 1282 1283 static const struct seq_operations raw6_seq_ops = { 1284 .start = raw6_seq_start, 1285 .next = raw6_seq_next, 1286 .stop = raw6_seq_stop, 1287 .show = raw6_seq_show, 1288 }; 1289 1290 static int raw6_seq_open(struct inode *inode, struct file *file) 1291 { 1292 return seq_open_private(file, &raw6_seq_ops, 1293 sizeof(struct raw6_iter_state)); 1294 } 1295 1296 static const struct file_operations raw6_seq_fops = { 1297 .owner = THIS_MODULE, 1298 .open = raw6_seq_open, 1299 .read = seq_read, 1300 .llseek = seq_lseek, 1301 .release = seq_release_private, 1302 }; 1303 1304 int __init raw6_proc_init(void) 1305 { 1306 if (!proc_net_fops_create(&init_net, "raw6", S_IRUGO, &raw6_seq_fops)) 1307 return -ENOMEM; 1308 return 0; 1309 } 1310 1311 void raw6_proc_exit(void) 1312 { 1313 proc_net_remove(&init_net, "raw6"); 1314 } 1315 #endif /* CONFIG_PROC_FS */ 1316