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/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 #ifdef CONFIG_IPV6_MIP6 53 #include <net/mip6.h> 54 #endif 55 56 #include <net/rawv6.h> 57 #include <net/xfrm.h> 58 59 #include <linux/proc_fs.h> 60 #include <linux/seq_file.h> 61 62 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE]; 63 DEFINE_RWLOCK(raw_v6_lock); 64 65 static void raw_v6_hash(struct sock *sk) 66 { 67 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num & 68 (RAWV6_HTABLE_SIZE - 1)]; 69 70 write_lock_bh(&raw_v6_lock); 71 sk_add_node(sk, list); 72 sock_prot_inc_use(sk->sk_prot); 73 write_unlock_bh(&raw_v6_lock); 74 } 75 76 static void raw_v6_unhash(struct sock *sk) 77 { 78 write_lock_bh(&raw_v6_lock); 79 if (sk_del_node_init(sk)) 80 sock_prot_dec_use(sk->sk_prot); 81 write_unlock_bh(&raw_v6_lock); 82 } 83 84 85 /* Grumble... icmp and ip_input want to get at this... */ 86 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num, 87 struct in6_addr *loc_addr, struct in6_addr *rmt_addr, 88 int dif) 89 { 90 struct hlist_node *node; 91 int is_multicast = ipv6_addr_is_multicast(loc_addr); 92 93 sk_for_each_from(sk, node) 94 if (inet_sk(sk)->num == num) { 95 struct ipv6_pinfo *np = inet6_sk(sk); 96 97 if (!ipv6_addr_any(&np->daddr) && 98 !ipv6_addr_equal(&np->daddr, rmt_addr)) 99 continue; 100 101 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) 102 continue; 103 104 if (!ipv6_addr_any(&np->rcv_saddr)) { 105 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr)) 106 goto found; 107 if (is_multicast && 108 inet6_mc_check(sk, loc_addr, rmt_addr)) 109 goto found; 110 continue; 111 } 112 goto found; 113 } 114 sk = NULL; 115 found: 116 return sk; 117 } 118 119 /* 120 * 0 - deliver 121 * 1 - block 122 */ 123 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb) 124 { 125 struct icmp6hdr *icmph; 126 struct raw6_sock *rp = raw6_sk(sk); 127 128 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) { 129 __u32 *data = &rp->filter.data[0]; 130 int bit_nr; 131 132 icmph = (struct icmp6hdr *) skb->data; 133 bit_nr = icmph->icmp6_type; 134 135 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0; 136 } 137 return 0; 138 } 139 140 /* 141 * demultiplex raw sockets. 142 * (should consider queueing the skb in the sock receive_queue 143 * without calling rawv6.c) 144 * 145 * Caller owns SKB so we must make clones. 146 */ 147 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 148 { 149 struct in6_addr *saddr; 150 struct in6_addr *daddr; 151 struct sock *sk; 152 int delivered = 0; 153 __u8 hash; 154 155 saddr = &skb->nh.ipv6h->saddr; 156 daddr = saddr + 1; 157 158 hash = nexthdr & (MAX_INET_PROTOS - 1); 159 160 read_lock(&raw_v6_lock); 161 sk = sk_head(&raw_v6_htable[hash]); 162 163 /* 164 * The first socket found will be delivered after 165 * delivery to transport protocols. 166 */ 167 168 if (sk == NULL) 169 goto out; 170 171 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif); 172 173 while (sk) { 174 int filtered; 175 176 delivered = 1; 177 switch (nexthdr) { 178 case IPPROTO_ICMPV6: 179 filtered = icmpv6_filter(sk, skb); 180 break; 181 #ifdef CONFIG_IPV6_MIP6 182 case IPPROTO_MH: 183 /* XXX: To validate MH only once for each packet, 184 * this is placed here. It should be after checking 185 * xfrm policy, however it doesn't. The checking xfrm 186 * policy is placed in rawv6_rcv() because it is 187 * required for each socket. 188 */ 189 filtered = mip6_mh_filter(sk, skb); 190 break; 191 #endif 192 default: 193 filtered = 0; 194 break; 195 } 196 197 if (filtered < 0) 198 break; 199 if (filtered == 0) { 200 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 201 202 /* Not releasing hash table! */ 203 if (clone) { 204 nf_reset(clone); 205 rawv6_rcv(sk, clone); 206 } 207 } 208 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr, 209 IP6CB(skb)->iif); 210 } 211 out: 212 read_unlock(&raw_v6_lock); 213 return delivered; 214 } 215 216 /* This cleans up af_inet6 a bit. -DaveM */ 217 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 218 { 219 struct inet_sock *inet = inet_sk(sk); 220 struct ipv6_pinfo *np = inet6_sk(sk); 221 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 222 __be32 v4addr = 0; 223 int addr_type; 224 int err; 225 226 if (addr_len < SIN6_LEN_RFC2133) 227 return -EINVAL; 228 addr_type = ipv6_addr_type(&addr->sin6_addr); 229 230 /* Raw sockets are IPv6 only */ 231 if (addr_type == IPV6_ADDR_MAPPED) 232 return(-EADDRNOTAVAIL); 233 234 lock_sock(sk); 235 236 err = -EINVAL; 237 if (sk->sk_state != TCP_CLOSE) 238 goto out; 239 240 /* Check if the address belongs to the host. */ 241 if (addr_type != IPV6_ADDR_ANY) { 242 struct net_device *dev = NULL; 243 244 if (addr_type & IPV6_ADDR_LINKLOCAL) { 245 if (addr_len >= sizeof(struct sockaddr_in6) && 246 addr->sin6_scope_id) { 247 /* Override any existing binding, if another 248 * one is supplied by user. 249 */ 250 sk->sk_bound_dev_if = addr->sin6_scope_id; 251 } 252 253 /* Binding to link-local address requires an interface */ 254 if (!sk->sk_bound_dev_if) 255 goto out; 256 257 dev = dev_get_by_index(sk->sk_bound_dev_if); 258 if (!dev) { 259 err = -ENODEV; 260 goto out; 261 } 262 } 263 264 /* ipv4 addr of the socket is invalid. Only the 265 * unspecified and mapped address have a v4 equivalent. 266 */ 267 v4addr = LOOPBACK4_IPV6; 268 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 269 err = -EADDRNOTAVAIL; 270 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) { 271 if (dev) 272 dev_put(dev); 273 goto out; 274 } 275 } 276 if (dev) 277 dev_put(dev); 278 } 279 280 inet->rcv_saddr = inet->saddr = v4addr; 281 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr); 282 if (!(addr_type & IPV6_ADDR_MULTICAST)) 283 ipv6_addr_copy(&np->saddr, &addr->sin6_addr); 284 err = 0; 285 out: 286 release_sock(sk); 287 return err; 288 } 289 290 void rawv6_err(struct sock *sk, struct sk_buff *skb, 291 struct inet6_skb_parm *opt, 292 int type, int code, int offset, __be32 info) 293 { 294 struct inet_sock *inet = inet_sk(sk); 295 struct ipv6_pinfo *np = inet6_sk(sk); 296 int err; 297 int harderr; 298 299 /* Report error on raw socket, if: 300 1. User requested recverr. 301 2. Socket is connected (otherwise the error indication 302 is useless without recverr and error is hard. 303 */ 304 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 305 return; 306 307 harderr = icmpv6_err_convert(type, code, &err); 308 if (type == ICMPV6_PKT_TOOBIG) 309 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 310 311 if (np->recverr) { 312 u8 *payload = skb->data; 313 if (!inet->hdrincl) 314 payload += offset; 315 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 316 } 317 318 if (np->recverr || harderr) { 319 sk->sk_err = err; 320 sk->sk_error_report(sk); 321 } 322 } 323 324 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb) 325 { 326 if ((raw6_sk(sk)->checksum || sk->sk_filter) && 327 skb_checksum_complete(skb)) { 328 /* FIXME: increment a raw6 drops counter here */ 329 kfree_skb(skb); 330 return 0; 331 } 332 333 /* Charge it to the socket. */ 334 if (sock_queue_rcv_skb(sk,skb)<0) { 335 /* FIXME: increment a raw6 drops counter here */ 336 kfree_skb(skb); 337 return 0; 338 } 339 340 return 0; 341 } 342 343 /* 344 * This is next to useless... 345 * if we demultiplex in network layer we don't need the extra call 346 * just to queue the skb... 347 * maybe we could have the network decide upon a hint if it 348 * should call raw_rcv for demultiplexing 349 */ 350 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 351 { 352 struct inet_sock *inet = inet_sk(sk); 353 struct raw6_sock *rp = raw6_sk(sk); 354 355 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 356 kfree_skb(skb); 357 return NET_RX_DROP; 358 } 359 360 if (!rp->checksum) 361 skb->ip_summed = CHECKSUM_UNNECESSARY; 362 363 if (skb->ip_summed == CHECKSUM_COMPLETE) { 364 skb_postpull_rcsum(skb, skb->nh.raw, 365 skb->h.raw - skb->nh.raw); 366 if (!csum_ipv6_magic(&skb->nh.ipv6h->saddr, 367 &skb->nh.ipv6h->daddr, 368 skb->len, inet->num, skb->csum)) 369 skb->ip_summed = CHECKSUM_UNNECESSARY; 370 } 371 if (skb->ip_summed != CHECKSUM_UNNECESSARY) 372 skb->csum = ~csum_unfold(csum_ipv6_magic(&skb->nh.ipv6h->saddr, 373 &skb->nh.ipv6h->daddr, 374 skb->len, inet->num, 0)); 375 376 if (inet->hdrincl) { 377 if (skb_checksum_complete(skb)) { 378 /* FIXME: increment a raw6 drops counter here */ 379 kfree_skb(skb); 380 return 0; 381 } 382 } 383 384 rawv6_rcv_skb(sk, skb); 385 return 0; 386 } 387 388 389 /* 390 * This should be easy, if there is something there 391 * we return it, otherwise we block. 392 */ 393 394 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk, 395 struct msghdr *msg, size_t len, 396 int noblock, int flags, int *addr_len) 397 { 398 struct ipv6_pinfo *np = inet6_sk(sk); 399 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name; 400 struct sk_buff *skb; 401 size_t copied; 402 int err; 403 404 if (flags & MSG_OOB) 405 return -EOPNOTSUPP; 406 407 if (addr_len) 408 *addr_len=sizeof(*sin6); 409 410 if (flags & MSG_ERRQUEUE) 411 return ipv6_recv_error(sk, msg, len); 412 413 skb = skb_recv_datagram(sk, flags, noblock, &err); 414 if (!skb) 415 goto out; 416 417 copied = skb->len; 418 if (copied > len) { 419 copied = len; 420 msg->msg_flags |= MSG_TRUNC; 421 } 422 423 if (skb->ip_summed==CHECKSUM_UNNECESSARY) { 424 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 425 } else if (msg->msg_flags&MSG_TRUNC) { 426 if (__skb_checksum_complete(skb)) 427 goto csum_copy_err; 428 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 429 } else { 430 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov); 431 if (err == -EINVAL) 432 goto csum_copy_err; 433 } 434 if (err) 435 goto out_free; 436 437 /* Copy the address. */ 438 if (sin6) { 439 sin6->sin6_family = AF_INET6; 440 sin6->sin6_port = 0; 441 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr); 442 sin6->sin6_flowinfo = 0; 443 sin6->sin6_scope_id = 0; 444 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) 445 sin6->sin6_scope_id = IP6CB(skb)->iif; 446 } 447 448 sock_recv_timestamp(msg, sk, skb); 449 450 if (np->rxopt.all) 451 datagram_recv_ctl(sk, msg, skb); 452 453 err = copied; 454 if (flags & MSG_TRUNC) 455 err = skb->len; 456 457 out_free: 458 skb_free_datagram(sk, skb); 459 out: 460 return err; 461 462 csum_copy_err: 463 skb_kill_datagram(sk, skb, flags); 464 465 /* Error for blocking case is chosen to masquerade 466 as some normal condition. 467 */ 468 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 469 /* FIXME: increment a raw6 drops counter here */ 470 goto out; 471 } 472 473 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl, 474 struct raw6_sock *rp) 475 { 476 struct sk_buff *skb; 477 int err = 0; 478 int offset; 479 int len; 480 int total_len; 481 __wsum tmp_csum; 482 __sum16 csum; 483 484 if (!rp->checksum) 485 goto send; 486 487 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 488 goto out; 489 490 offset = rp->offset; 491 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data); 492 if (offset >= total_len - 1) { 493 err = -EINVAL; 494 ip6_flush_pending_frames(sk); 495 goto out; 496 } 497 498 /* should be check HW csum miyazawa */ 499 if (skb_queue_len(&sk->sk_write_queue) == 1) { 500 /* 501 * Only one fragment on the socket. 502 */ 503 tmp_csum = skb->csum; 504 } else { 505 struct sk_buff *csum_skb = NULL; 506 tmp_csum = 0; 507 508 skb_queue_walk(&sk->sk_write_queue, skb) { 509 tmp_csum = csum_add(tmp_csum, skb->csum); 510 511 if (csum_skb) 512 continue; 513 514 len = skb->len - (skb->h.raw - skb->data); 515 if (offset >= len) { 516 offset -= len; 517 continue; 518 } 519 520 csum_skb = skb; 521 } 522 523 skb = csum_skb; 524 } 525 526 offset += skb->h.raw - skb->data; 527 if (skb_copy_bits(skb, offset, &csum, 2)) 528 BUG(); 529 530 /* in case cksum was not initialized */ 531 if (unlikely(csum)) 532 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 533 534 csum = csum_ipv6_magic(&fl->fl6_src, 535 &fl->fl6_dst, 536 total_len, fl->proto, tmp_csum); 537 538 if (csum == 0 && fl->proto == IPPROTO_UDP) 539 csum = CSUM_MANGLED_0; 540 541 if (skb_store_bits(skb, offset, &csum, 2)) 542 BUG(); 543 544 send: 545 err = ip6_push_pending_frames(sk); 546 out: 547 return err; 548 } 549 550 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length, 551 struct flowi *fl, struct rt6_info *rt, 552 unsigned int flags) 553 { 554 struct ipv6_pinfo *np = inet6_sk(sk); 555 struct ipv6hdr *iph; 556 struct sk_buff *skb; 557 unsigned int hh_len; 558 int err; 559 560 if (length > rt->u.dst.dev->mtu) { 561 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu); 562 return -EMSGSIZE; 563 } 564 if (flags&MSG_PROBE) 565 goto out; 566 567 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); 568 569 skb = sock_alloc_send_skb(sk, length+hh_len+15, 570 flags&MSG_DONTWAIT, &err); 571 if (skb == NULL) 572 goto error; 573 skb_reserve(skb, hh_len); 574 575 skb->priority = sk->sk_priority; 576 skb->dst = dst_clone(&rt->u.dst); 577 578 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length); 579 580 skb->ip_summed = CHECKSUM_NONE; 581 582 skb->h.raw = skb->nh.raw; 583 err = memcpy_fromiovecend((void *)iph, from, 0, length); 584 if (err) 585 goto error_fault; 586 587 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS); 588 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev, 589 dst_output); 590 if (err > 0) 591 err = np->recverr ? net_xmit_errno(err) : 0; 592 if (err) 593 goto error; 594 out: 595 return 0; 596 597 error_fault: 598 err = -EFAULT; 599 kfree_skb(skb); 600 error: 601 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 602 return err; 603 } 604 605 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg) 606 { 607 struct iovec *iov; 608 u8 __user *type = NULL; 609 u8 __user *code = NULL; 610 #ifdef CONFIG_IPV6_MIP6 611 u8 len = 0; 612 #endif 613 int probed = 0; 614 int i; 615 616 if (!msg->msg_iov) 617 return 0; 618 619 for (i = 0; i < msg->msg_iovlen; i++) { 620 iov = &msg->msg_iov[i]; 621 if (!iov) 622 continue; 623 624 switch (fl->proto) { 625 case IPPROTO_ICMPV6: 626 /* check if one-byte field is readable or not. */ 627 if (iov->iov_base && iov->iov_len < 1) 628 break; 629 630 if (!type) { 631 type = iov->iov_base; 632 /* check if code field is readable or not. */ 633 if (iov->iov_len > 1) 634 code = type + 1; 635 } else if (!code) 636 code = iov->iov_base; 637 638 if (type && code) { 639 if (get_user(fl->fl_icmp_type, type) || 640 get_user(fl->fl_icmp_code, code)) 641 return -EFAULT; 642 probed = 1; 643 } 644 break; 645 #ifdef CONFIG_IPV6_MIP6 646 case IPPROTO_MH: 647 if (iov->iov_base && iov->iov_len < 1) 648 break; 649 /* check if type field is readable or not. */ 650 if (iov->iov_len > 2 - len) { 651 u8 __user *p = iov->iov_base; 652 if (get_user(fl->fl_mh_type, &p[2 - len])) 653 return -EFAULT; 654 probed = 1; 655 } else 656 len += iov->iov_len; 657 658 break; 659 #endif 660 default: 661 probed = 1; 662 break; 663 } 664 if (probed) 665 break; 666 } 667 return 0; 668 } 669 670 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk, 671 struct msghdr *msg, size_t len) 672 { 673 struct ipv6_txoptions opt_space; 674 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name; 675 struct in6_addr *daddr, *final_p = NULL, final; 676 struct inet_sock *inet = inet_sk(sk); 677 struct ipv6_pinfo *np = inet6_sk(sk); 678 struct raw6_sock *rp = raw6_sk(sk); 679 struct ipv6_txoptions *opt = NULL; 680 struct ip6_flowlabel *flowlabel = NULL; 681 struct dst_entry *dst = NULL; 682 struct flowi fl; 683 int addr_len = msg->msg_namelen; 684 int hlimit = -1; 685 int tclass = -1; 686 u16 proto; 687 int err; 688 689 /* Rough check on arithmetic overflow, 690 better check is made in ip6_build_xmit 691 */ 692 if (len < 0) 693 return -EMSGSIZE; 694 695 /* Mirror BSD error message compatibility */ 696 if (msg->msg_flags & MSG_OOB) 697 return -EOPNOTSUPP; 698 699 /* 700 * Get and verify the address. 701 */ 702 memset(&fl, 0, sizeof(fl)); 703 704 if (sin6) { 705 if (addr_len < SIN6_LEN_RFC2133) 706 return -EINVAL; 707 708 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 709 return(-EAFNOSUPPORT); 710 711 /* port is the proto value [0..255] carried in nexthdr */ 712 proto = ntohs(sin6->sin6_port); 713 714 if (!proto) 715 proto = inet->num; 716 else if (proto != inet->num) 717 return(-EINVAL); 718 719 if (proto > 255) 720 return(-EINVAL); 721 722 daddr = &sin6->sin6_addr; 723 if (np->sndflow) { 724 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 725 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { 726 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 727 if (flowlabel == NULL) 728 return -EINVAL; 729 daddr = &flowlabel->dst; 730 } 731 } 732 733 /* 734 * Otherwise it will be difficult to maintain 735 * sk->sk_dst_cache. 736 */ 737 if (sk->sk_state == TCP_ESTABLISHED && 738 ipv6_addr_equal(daddr, &np->daddr)) 739 daddr = &np->daddr; 740 741 if (addr_len >= sizeof(struct sockaddr_in6) && 742 sin6->sin6_scope_id && 743 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) 744 fl.oif = sin6->sin6_scope_id; 745 } else { 746 if (sk->sk_state != TCP_ESTABLISHED) 747 return -EDESTADDRREQ; 748 749 proto = inet->num; 750 daddr = &np->daddr; 751 fl.fl6_flowlabel = np->flow_label; 752 } 753 754 if (ipv6_addr_any(daddr)) { 755 /* 756 * unspecified destination address 757 * treated as error... is this correct ? 758 */ 759 fl6_sock_release(flowlabel); 760 return(-EINVAL); 761 } 762 763 if (fl.oif == 0) 764 fl.oif = sk->sk_bound_dev_if; 765 766 if (msg->msg_controllen) { 767 opt = &opt_space; 768 memset(opt, 0, sizeof(struct ipv6_txoptions)); 769 opt->tot_len = sizeof(struct ipv6_txoptions); 770 771 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass); 772 if (err < 0) { 773 fl6_sock_release(flowlabel); 774 return err; 775 } 776 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 777 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 778 if (flowlabel == NULL) 779 return -EINVAL; 780 } 781 if (!(opt->opt_nflen|opt->opt_flen)) 782 opt = NULL; 783 } 784 if (opt == NULL) 785 opt = np->opt; 786 if (flowlabel) 787 opt = fl6_merge_options(&opt_space, flowlabel, opt); 788 opt = ipv6_fixup_options(&opt_space, opt); 789 790 fl.proto = proto; 791 err = rawv6_probe_proto_opt(&fl, msg); 792 if (err) 793 goto out; 794 795 ipv6_addr_copy(&fl.fl6_dst, daddr); 796 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr)) 797 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 798 799 /* merge ip6_build_xmit from ip6_output */ 800 if (opt && opt->srcrt) { 801 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt; 802 ipv6_addr_copy(&final, &fl.fl6_dst); 803 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 804 final_p = &final; 805 } 806 807 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) 808 fl.oif = np->mcast_oif; 809 security_sk_classify_flow(sk, &fl); 810 811 err = ip6_dst_lookup(sk, &dst, &fl); 812 if (err) 813 goto out; 814 if (final_p) 815 ipv6_addr_copy(&fl.fl6_dst, final_p); 816 817 if ((err = xfrm_lookup(&dst, &fl, sk, 1)) < 0) 818 goto out; 819 820 if (hlimit < 0) { 821 if (ipv6_addr_is_multicast(&fl.fl6_dst)) 822 hlimit = np->mcast_hops; 823 else 824 hlimit = np->hop_limit; 825 if (hlimit < 0) 826 hlimit = dst_metric(dst, RTAX_HOPLIMIT); 827 if (hlimit < 0) 828 hlimit = ipv6_get_hoplimit(dst->dev); 829 } 830 831 if (tclass < 0) { 832 tclass = np->tclass; 833 if (tclass < 0) 834 tclass = 0; 835 } 836 837 if (msg->msg_flags&MSG_CONFIRM) 838 goto do_confirm; 839 840 back_from_confirm: 841 if (inet->hdrincl) { 842 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags); 843 } else { 844 lock_sock(sk); 845 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, 846 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst, 847 msg->msg_flags); 848 849 if (err) 850 ip6_flush_pending_frames(sk); 851 else if (!(msg->msg_flags & MSG_MORE)) 852 err = rawv6_push_pending_frames(sk, &fl, rp); 853 } 854 done: 855 dst_release(dst); 856 if (!inet->hdrincl) 857 release_sock(sk); 858 out: 859 fl6_sock_release(flowlabel); 860 return err<0?err:len; 861 do_confirm: 862 dst_confirm(dst); 863 if (!(msg->msg_flags & MSG_PROBE) || len) 864 goto back_from_confirm; 865 err = 0; 866 goto done; 867 } 868 869 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 870 char __user *optval, int optlen) 871 { 872 switch (optname) { 873 case ICMPV6_FILTER: 874 if (optlen > sizeof(struct icmp6_filter)) 875 optlen = sizeof(struct icmp6_filter); 876 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 877 return -EFAULT; 878 return 0; 879 default: 880 return -ENOPROTOOPT; 881 }; 882 883 return 0; 884 } 885 886 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 887 char __user *optval, int __user *optlen) 888 { 889 int len; 890 891 switch (optname) { 892 case ICMPV6_FILTER: 893 if (get_user(len, optlen)) 894 return -EFAULT; 895 if (len < 0) 896 return -EINVAL; 897 if (len > sizeof(struct icmp6_filter)) 898 len = sizeof(struct icmp6_filter); 899 if (put_user(len, optlen)) 900 return -EFAULT; 901 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 902 return -EFAULT; 903 return 0; 904 default: 905 return -ENOPROTOOPT; 906 }; 907 908 return 0; 909 } 910 911 912 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 913 char __user *optval, int optlen) 914 { 915 struct raw6_sock *rp = raw6_sk(sk); 916 int val; 917 918 if (get_user(val, (int __user *)optval)) 919 return -EFAULT; 920 921 switch (optname) { 922 case IPV6_CHECKSUM: 923 /* You may get strange result with a positive odd offset; 924 RFC2292bis agrees with me. */ 925 if (val > 0 && (val&1)) 926 return(-EINVAL); 927 if (val < 0) { 928 rp->checksum = 0; 929 } else { 930 rp->checksum = 1; 931 rp->offset = val; 932 } 933 934 return 0; 935 break; 936 937 default: 938 return(-ENOPROTOOPT); 939 } 940 } 941 942 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 943 char __user *optval, int optlen) 944 { 945 switch(level) { 946 case SOL_RAW: 947 break; 948 949 case SOL_ICMPV6: 950 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 951 return -EOPNOTSUPP; 952 return rawv6_seticmpfilter(sk, level, optname, optval, 953 optlen); 954 case SOL_IPV6: 955 if (optname == IPV6_CHECKSUM) 956 break; 957 default: 958 return ipv6_setsockopt(sk, level, optname, optval, 959 optlen); 960 }; 961 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 962 } 963 964 #ifdef CONFIG_COMPAT 965 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 966 char __user *optval, int optlen) 967 { 968 switch (level) { 969 case SOL_RAW: 970 break; 971 case SOL_ICMPV6: 972 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 973 return -EOPNOTSUPP; 974 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 975 case SOL_IPV6: 976 if (optname == IPV6_CHECKSUM) 977 break; 978 default: 979 return compat_ipv6_setsockopt(sk, level, optname, 980 optval, optlen); 981 }; 982 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 983 } 984 #endif 985 986 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 987 char __user *optval, int __user *optlen) 988 { 989 struct raw6_sock *rp = raw6_sk(sk); 990 int val, len; 991 992 if (get_user(len,optlen)) 993 return -EFAULT; 994 995 switch (optname) { 996 case IPV6_CHECKSUM: 997 if (rp->checksum == 0) 998 val = -1; 999 else 1000 val = rp->offset; 1001 break; 1002 1003 default: 1004 return -ENOPROTOOPT; 1005 } 1006 1007 len = min_t(unsigned int, sizeof(int), len); 1008 1009 if (put_user(len, optlen)) 1010 return -EFAULT; 1011 if (copy_to_user(optval,&val,len)) 1012 return -EFAULT; 1013 return 0; 1014 } 1015 1016 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1017 char __user *optval, int __user *optlen) 1018 { 1019 switch(level) { 1020 case SOL_RAW: 1021 break; 1022 1023 case SOL_ICMPV6: 1024 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 1025 return -EOPNOTSUPP; 1026 return rawv6_geticmpfilter(sk, level, optname, optval, 1027 optlen); 1028 case SOL_IPV6: 1029 if (optname == IPV6_CHECKSUM) 1030 break; 1031 default: 1032 return ipv6_getsockopt(sk, level, optname, optval, 1033 optlen); 1034 }; 1035 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1036 } 1037 1038 #ifdef CONFIG_COMPAT 1039 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1040 char __user *optval, int __user *optlen) 1041 { 1042 switch (level) { 1043 case SOL_RAW: 1044 break; 1045 case SOL_ICMPV6: 1046 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 1047 return -EOPNOTSUPP; 1048 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1049 case SOL_IPV6: 1050 if (optname == IPV6_CHECKSUM) 1051 break; 1052 default: 1053 return compat_ipv6_getsockopt(sk, level, optname, 1054 optval, optlen); 1055 }; 1056 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1057 } 1058 #endif 1059 1060 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1061 { 1062 switch(cmd) { 1063 case SIOCOUTQ: 1064 { 1065 int amount = atomic_read(&sk->sk_wmem_alloc); 1066 return put_user(amount, (int __user *)arg); 1067 } 1068 case SIOCINQ: 1069 { 1070 struct sk_buff *skb; 1071 int amount = 0; 1072 1073 spin_lock_bh(&sk->sk_receive_queue.lock); 1074 skb = skb_peek(&sk->sk_receive_queue); 1075 if (skb != NULL) 1076 amount = skb->tail - skb->h.raw; 1077 spin_unlock_bh(&sk->sk_receive_queue.lock); 1078 return put_user(amount, (int __user *)arg); 1079 } 1080 1081 default: 1082 return -ENOIOCTLCMD; 1083 } 1084 } 1085 1086 static void rawv6_close(struct sock *sk, long timeout) 1087 { 1088 if (inet_sk(sk)->num == IPPROTO_RAW) 1089 ip6_ra_control(sk, -1, NULL); 1090 1091 sk_common_release(sk); 1092 } 1093 1094 static int rawv6_init_sk(struct sock *sk) 1095 { 1096 struct raw6_sock *rp = raw6_sk(sk); 1097 1098 switch (inet_sk(sk)->num) { 1099 case IPPROTO_ICMPV6: 1100 rp->checksum = 1; 1101 rp->offset = 2; 1102 break; 1103 case IPPROTO_MH: 1104 rp->checksum = 1; 1105 rp->offset = 4; 1106 break; 1107 default: 1108 break; 1109 } 1110 return(0); 1111 } 1112 1113 struct proto rawv6_prot = { 1114 .name = "RAWv6", 1115 .owner = THIS_MODULE, 1116 .close = rawv6_close, 1117 .connect = ip6_datagram_connect, 1118 .disconnect = udp_disconnect, 1119 .ioctl = rawv6_ioctl, 1120 .init = rawv6_init_sk, 1121 .destroy = inet6_destroy_sock, 1122 .setsockopt = rawv6_setsockopt, 1123 .getsockopt = rawv6_getsockopt, 1124 .sendmsg = rawv6_sendmsg, 1125 .recvmsg = rawv6_recvmsg, 1126 .bind = rawv6_bind, 1127 .backlog_rcv = rawv6_rcv_skb, 1128 .hash = raw_v6_hash, 1129 .unhash = raw_v6_unhash, 1130 .obj_size = sizeof(struct raw6_sock), 1131 #ifdef CONFIG_COMPAT 1132 .compat_setsockopt = compat_rawv6_setsockopt, 1133 .compat_getsockopt = compat_rawv6_getsockopt, 1134 #endif 1135 }; 1136 1137 #ifdef CONFIG_PROC_FS 1138 struct raw6_iter_state { 1139 int bucket; 1140 }; 1141 1142 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private) 1143 1144 static struct sock *raw6_get_first(struct seq_file *seq) 1145 { 1146 struct sock *sk; 1147 struct hlist_node *node; 1148 struct raw6_iter_state* state = raw6_seq_private(seq); 1149 1150 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket) 1151 sk_for_each(sk, node, &raw_v6_htable[state->bucket]) 1152 if (sk->sk_family == PF_INET6) 1153 goto out; 1154 sk = NULL; 1155 out: 1156 return sk; 1157 } 1158 1159 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk) 1160 { 1161 struct raw6_iter_state* state = raw6_seq_private(seq); 1162 1163 do { 1164 sk = sk_next(sk); 1165 try_again: 1166 ; 1167 } while (sk && sk->sk_family != PF_INET6); 1168 1169 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) { 1170 sk = sk_head(&raw_v6_htable[state->bucket]); 1171 goto try_again; 1172 } 1173 return sk; 1174 } 1175 1176 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos) 1177 { 1178 struct sock *sk = raw6_get_first(seq); 1179 if (sk) 1180 while (pos && (sk = raw6_get_next(seq, sk)) != NULL) 1181 --pos; 1182 return pos ? NULL : sk; 1183 } 1184 1185 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos) 1186 { 1187 read_lock(&raw_v6_lock); 1188 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 1189 } 1190 1191 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1192 { 1193 struct sock *sk; 1194 1195 if (v == SEQ_START_TOKEN) 1196 sk = raw6_get_first(seq); 1197 else 1198 sk = raw6_get_next(seq, v); 1199 ++*pos; 1200 return sk; 1201 } 1202 1203 static void raw6_seq_stop(struct seq_file *seq, void *v) 1204 { 1205 read_unlock(&raw_v6_lock); 1206 } 1207 1208 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1209 { 1210 struct ipv6_pinfo *np = inet6_sk(sp); 1211 struct in6_addr *dest, *src; 1212 __u16 destp, srcp; 1213 1214 dest = &np->daddr; 1215 src = &np->rcv_saddr; 1216 destp = 0; 1217 srcp = inet_sk(sp)->num; 1218 seq_printf(seq, 1219 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1220 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n", 1221 i, 1222 src->s6_addr32[0], src->s6_addr32[1], 1223 src->s6_addr32[2], src->s6_addr32[3], srcp, 1224 dest->s6_addr32[0], dest->s6_addr32[1], 1225 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1226 sp->sk_state, 1227 atomic_read(&sp->sk_wmem_alloc), 1228 atomic_read(&sp->sk_rmem_alloc), 1229 0, 0L, 0, 1230 sock_i_uid(sp), 0, 1231 sock_i_ino(sp), 1232 atomic_read(&sp->sk_refcnt), sp); 1233 } 1234 1235 static int raw6_seq_show(struct seq_file *seq, void *v) 1236 { 1237 if (v == SEQ_START_TOKEN) 1238 seq_printf(seq, 1239 " sl " 1240 "local_address " 1241 "remote_address " 1242 "st tx_queue rx_queue tr tm->when retrnsmt" 1243 " uid timeout inode\n"); 1244 else 1245 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket); 1246 return 0; 1247 } 1248 1249 static struct seq_operations raw6_seq_ops = { 1250 .start = raw6_seq_start, 1251 .next = raw6_seq_next, 1252 .stop = raw6_seq_stop, 1253 .show = raw6_seq_show, 1254 }; 1255 1256 static int raw6_seq_open(struct inode *inode, struct file *file) 1257 { 1258 struct seq_file *seq; 1259 int rc = -ENOMEM; 1260 struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL); 1261 if (!s) 1262 goto out; 1263 rc = seq_open(file, &raw6_seq_ops); 1264 if (rc) 1265 goto out_kfree; 1266 seq = file->private_data; 1267 seq->private = s; 1268 out: 1269 return rc; 1270 out_kfree: 1271 kfree(s); 1272 goto out; 1273 } 1274 1275 static const struct file_operations raw6_seq_fops = { 1276 .owner = THIS_MODULE, 1277 .open = raw6_seq_open, 1278 .read = seq_read, 1279 .llseek = seq_lseek, 1280 .release = seq_release_private, 1281 }; 1282 1283 int __init raw6_proc_init(void) 1284 { 1285 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops)) 1286 return -ENOMEM; 1287 return 0; 1288 } 1289 1290 void raw6_proc_exit(void) 1291 { 1292 proc_net_remove("raw6"); 1293 } 1294 #endif /* CONFIG_PROC_FS */ 1295