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/sched.h> 28 #include <linux/net.h> 29 #include <linux/in6.h> 30 #include <linux/netdevice.h> 31 #include <linux/if_arp.h> 32 #include <linux/icmpv6.h> 33 #include <linux/netfilter.h> 34 #include <linux/netfilter_ipv6.h> 35 #include <asm/uaccess.h> 36 #include <asm/ioctls.h> 37 #include <asm/bug.h> 38 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 54 #include <net/rawv6.h> 55 #include <net/xfrm.h> 56 57 #include <linux/proc_fs.h> 58 #include <linux/seq_file.h> 59 60 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE]; 61 DEFINE_RWLOCK(raw_v6_lock); 62 63 static void raw_v6_hash(struct sock *sk) 64 { 65 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num & 66 (RAWV6_HTABLE_SIZE - 1)]; 67 68 write_lock_bh(&raw_v6_lock); 69 sk_add_node(sk, list); 70 sock_prot_inc_use(sk->sk_prot); 71 write_unlock_bh(&raw_v6_lock); 72 } 73 74 static void raw_v6_unhash(struct sock *sk) 75 { 76 write_lock_bh(&raw_v6_lock); 77 if (sk_del_node_init(sk)) 78 sock_prot_dec_use(sk->sk_prot); 79 write_unlock_bh(&raw_v6_lock); 80 } 81 82 83 /* Grumble... icmp and ip_input want to get at this... */ 84 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num, 85 struct in6_addr *loc_addr, struct in6_addr *rmt_addr, 86 int dif) 87 { 88 struct hlist_node *node; 89 int is_multicast = ipv6_addr_is_multicast(loc_addr); 90 91 sk_for_each_from(sk, node) 92 if (inet_sk(sk)->num == num) { 93 struct ipv6_pinfo *np = inet6_sk(sk); 94 95 if (!ipv6_addr_any(&np->daddr) && 96 !ipv6_addr_equal(&np->daddr, rmt_addr)) 97 continue; 98 99 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) 100 continue; 101 102 if (!ipv6_addr_any(&np->rcv_saddr)) { 103 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr)) 104 goto found; 105 if (is_multicast && 106 inet6_mc_check(sk, loc_addr, rmt_addr)) 107 goto found; 108 continue; 109 } 110 goto found; 111 } 112 sk = NULL; 113 found: 114 return sk; 115 } 116 117 /* 118 * 0 - deliver 119 * 1 - block 120 */ 121 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb) 122 { 123 struct icmp6hdr *icmph; 124 struct raw6_sock *rp = raw6_sk(sk); 125 126 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) { 127 __u32 *data = &rp->filter.data[0]; 128 int bit_nr; 129 130 icmph = (struct icmp6hdr *) skb->data; 131 bit_nr = icmph->icmp6_type; 132 133 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0; 134 } 135 return 0; 136 } 137 138 /* 139 * demultiplex raw sockets. 140 * (should consider queueing the skb in the sock receive_queue 141 * without calling rawv6.c) 142 * 143 * Caller owns SKB so we must make clones. 144 */ 145 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 146 { 147 struct in6_addr *saddr; 148 struct in6_addr *daddr; 149 struct sock *sk; 150 int delivered = 0; 151 __u8 hash; 152 153 saddr = &skb->nh.ipv6h->saddr; 154 daddr = saddr + 1; 155 156 hash = nexthdr & (MAX_INET_PROTOS - 1); 157 158 read_lock(&raw_v6_lock); 159 sk = sk_head(&raw_v6_htable[hash]); 160 161 /* 162 * The first socket found will be delivered after 163 * delivery to transport protocols. 164 */ 165 166 if (sk == NULL) 167 goto out; 168 169 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif); 170 171 while (sk) { 172 delivered = 1; 173 if (nexthdr != IPPROTO_ICMPV6 || !icmpv6_filter(sk, skb)) { 174 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 175 176 /* Not releasing hash table! */ 177 if (clone) { 178 nf_reset(clone); 179 rawv6_rcv(sk, clone); 180 } 181 } 182 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr, 183 IP6CB(skb)->iif); 184 } 185 out: 186 read_unlock(&raw_v6_lock); 187 return delivered; 188 } 189 190 /* This cleans up af_inet6 a bit. -DaveM */ 191 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 192 { 193 struct inet_sock *inet = inet_sk(sk); 194 struct ipv6_pinfo *np = inet6_sk(sk); 195 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 196 __u32 v4addr = 0; 197 int addr_type; 198 int err; 199 200 if (addr_len < SIN6_LEN_RFC2133) 201 return -EINVAL; 202 addr_type = ipv6_addr_type(&addr->sin6_addr); 203 204 /* Raw sockets are IPv6 only */ 205 if (addr_type == IPV6_ADDR_MAPPED) 206 return(-EADDRNOTAVAIL); 207 208 lock_sock(sk); 209 210 err = -EINVAL; 211 if (sk->sk_state != TCP_CLOSE) 212 goto out; 213 214 /* Check if the address belongs to the host. */ 215 if (addr_type != IPV6_ADDR_ANY) { 216 struct net_device *dev = NULL; 217 218 if (addr_type & IPV6_ADDR_LINKLOCAL) { 219 if (addr_len >= sizeof(struct sockaddr_in6) && 220 addr->sin6_scope_id) { 221 /* Override any existing binding, if another 222 * one is supplied by user. 223 */ 224 sk->sk_bound_dev_if = addr->sin6_scope_id; 225 } 226 227 /* Binding to link-local address requires an interface */ 228 if (!sk->sk_bound_dev_if) 229 goto out; 230 231 dev = dev_get_by_index(sk->sk_bound_dev_if); 232 if (!dev) { 233 err = -ENODEV; 234 goto out; 235 } 236 } 237 238 /* ipv4 addr of the socket is invalid. Only the 239 * unspecified and mapped address have a v4 equivalent. 240 */ 241 v4addr = LOOPBACK4_IPV6; 242 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 243 err = -EADDRNOTAVAIL; 244 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) { 245 if (dev) 246 dev_put(dev); 247 goto out; 248 } 249 } 250 if (dev) 251 dev_put(dev); 252 } 253 254 inet->rcv_saddr = inet->saddr = v4addr; 255 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr); 256 if (!(addr_type & IPV6_ADDR_MULTICAST)) 257 ipv6_addr_copy(&np->saddr, &addr->sin6_addr); 258 err = 0; 259 out: 260 release_sock(sk); 261 return err; 262 } 263 264 void rawv6_err(struct sock *sk, struct sk_buff *skb, 265 struct inet6_skb_parm *opt, 266 int type, int code, int offset, u32 info) 267 { 268 struct inet_sock *inet = inet_sk(sk); 269 struct ipv6_pinfo *np = inet6_sk(sk); 270 int err; 271 int harderr; 272 273 /* Report error on raw socket, if: 274 1. User requested recverr. 275 2. Socket is connected (otherwise the error indication 276 is useless without recverr and error is hard. 277 */ 278 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 279 return; 280 281 harderr = icmpv6_err_convert(type, code, &err); 282 if (type == ICMPV6_PKT_TOOBIG) 283 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 284 285 if (np->recverr) { 286 u8 *payload = skb->data; 287 if (!inet->hdrincl) 288 payload += offset; 289 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 290 } 291 292 if (np->recverr || harderr) { 293 sk->sk_err = err; 294 sk->sk_error_report(sk); 295 } 296 } 297 298 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb) 299 { 300 if ((raw6_sk(sk)->checksum || sk->sk_filter) && 301 skb_checksum_complete(skb)) { 302 /* FIXME: increment a raw6 drops counter here */ 303 kfree_skb(skb); 304 return 0; 305 } 306 307 /* Charge it to the socket. */ 308 if (sock_queue_rcv_skb(sk,skb)<0) { 309 /* FIXME: increment a raw6 drops counter here */ 310 kfree_skb(skb); 311 return 0; 312 } 313 314 return 0; 315 } 316 317 /* 318 * This is next to useless... 319 * if we demultiplex in network layer we don't need the extra call 320 * just to queue the skb... 321 * maybe we could have the network decide upon a hint if it 322 * should call raw_rcv for demultiplexing 323 */ 324 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 325 { 326 struct inet_sock *inet = inet_sk(sk); 327 struct raw6_sock *rp = raw6_sk(sk); 328 329 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 330 kfree_skb(skb); 331 return NET_RX_DROP; 332 } 333 334 if (!rp->checksum) 335 skb->ip_summed = CHECKSUM_UNNECESSARY; 336 337 if (skb->ip_summed == CHECKSUM_HW) { 338 skb_postpull_rcsum(skb, skb->nh.raw, 339 skb->h.raw - skb->nh.raw); 340 if (!csum_ipv6_magic(&skb->nh.ipv6h->saddr, 341 &skb->nh.ipv6h->daddr, 342 skb->len, inet->num, skb->csum)) 343 skb->ip_summed = CHECKSUM_UNNECESSARY; 344 } 345 if (skb->ip_summed != CHECKSUM_UNNECESSARY) 346 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr, 347 &skb->nh.ipv6h->daddr, 348 skb->len, inet->num, 0); 349 350 if (inet->hdrincl) { 351 if (skb_checksum_complete(skb)) { 352 /* FIXME: increment a raw6 drops counter here */ 353 kfree_skb(skb); 354 return 0; 355 } 356 } 357 358 rawv6_rcv_skb(sk, skb); 359 return 0; 360 } 361 362 363 /* 364 * This should be easy, if there is something there 365 * we return it, otherwise we block. 366 */ 367 368 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk, 369 struct msghdr *msg, size_t len, 370 int noblock, int flags, int *addr_len) 371 { 372 struct ipv6_pinfo *np = inet6_sk(sk); 373 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name; 374 struct sk_buff *skb; 375 size_t copied; 376 int err; 377 378 if (flags & MSG_OOB) 379 return -EOPNOTSUPP; 380 381 if (addr_len) 382 *addr_len=sizeof(*sin6); 383 384 if (flags & MSG_ERRQUEUE) 385 return ipv6_recv_error(sk, msg, len); 386 387 skb = skb_recv_datagram(sk, flags, noblock, &err); 388 if (!skb) 389 goto out; 390 391 copied = skb->len; 392 if (copied > len) { 393 copied = len; 394 msg->msg_flags |= MSG_TRUNC; 395 } 396 397 if (skb->ip_summed==CHECKSUM_UNNECESSARY) { 398 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 399 } else if (msg->msg_flags&MSG_TRUNC) { 400 if (__skb_checksum_complete(skb)) 401 goto csum_copy_err; 402 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 403 } else { 404 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov); 405 if (err == -EINVAL) 406 goto csum_copy_err; 407 } 408 if (err) 409 goto out_free; 410 411 /* Copy the address. */ 412 if (sin6) { 413 sin6->sin6_family = AF_INET6; 414 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr); 415 sin6->sin6_flowinfo = 0; 416 sin6->sin6_scope_id = 0; 417 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) 418 sin6->sin6_scope_id = IP6CB(skb)->iif; 419 } 420 421 sock_recv_timestamp(msg, sk, skb); 422 423 if (np->rxopt.all) 424 datagram_recv_ctl(sk, msg, skb); 425 426 err = copied; 427 if (flags & MSG_TRUNC) 428 err = skb->len; 429 430 out_free: 431 skb_free_datagram(sk, skb); 432 out: 433 return err; 434 435 csum_copy_err: 436 /* Clear queue. */ 437 if (flags&MSG_PEEK) { 438 int clear = 0; 439 spin_lock_bh(&sk->sk_receive_queue.lock); 440 if (skb == skb_peek(&sk->sk_receive_queue)) { 441 __skb_unlink(skb, &sk->sk_receive_queue); 442 clear = 1; 443 } 444 spin_unlock_bh(&sk->sk_receive_queue.lock); 445 if (clear) 446 kfree_skb(skb); 447 } 448 449 /* Error for blocking case is chosen to masquerade 450 as some normal condition. 451 */ 452 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 453 /* FIXME: increment a raw6 drops counter here */ 454 goto out_free; 455 } 456 457 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl, 458 struct raw6_sock *rp) 459 { 460 struct sk_buff *skb; 461 int err = 0; 462 int offset; 463 int len; 464 int total_len; 465 u32 tmp_csum; 466 u16 csum; 467 468 if (!rp->checksum) 469 goto send; 470 471 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 472 goto out; 473 474 offset = rp->offset; 475 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data); 476 if (offset >= total_len - 1) { 477 err = -EINVAL; 478 ip6_flush_pending_frames(sk); 479 goto out; 480 } 481 482 /* should be check HW csum miyazawa */ 483 if (skb_queue_len(&sk->sk_write_queue) == 1) { 484 /* 485 * Only one fragment on the socket. 486 */ 487 tmp_csum = skb->csum; 488 } else { 489 struct sk_buff *csum_skb = NULL; 490 tmp_csum = 0; 491 492 skb_queue_walk(&sk->sk_write_queue, skb) { 493 tmp_csum = csum_add(tmp_csum, skb->csum); 494 495 if (csum_skb) 496 continue; 497 498 len = skb->len - (skb->h.raw - skb->data); 499 if (offset >= len) { 500 offset -= len; 501 continue; 502 } 503 504 csum_skb = skb; 505 } 506 507 skb = csum_skb; 508 } 509 510 offset += skb->h.raw - skb->data; 511 if (skb_copy_bits(skb, offset, &csum, 2)) 512 BUG(); 513 514 /* in case cksum was not initialized */ 515 if (unlikely(csum)) 516 tmp_csum = csum_sub(tmp_csum, csum); 517 518 tmp_csum = csum_ipv6_magic(&fl->fl6_src, 519 &fl->fl6_dst, 520 total_len, fl->proto, tmp_csum); 521 522 if (tmp_csum == 0) 523 tmp_csum = -1; 524 525 csum = tmp_csum; 526 if (skb_store_bits(skb, offset, &csum, 2)) 527 BUG(); 528 529 send: 530 err = ip6_push_pending_frames(sk); 531 out: 532 return err; 533 } 534 535 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length, 536 struct flowi *fl, struct rt6_info *rt, 537 unsigned int flags) 538 { 539 struct ipv6_pinfo *np = inet6_sk(sk); 540 struct ipv6hdr *iph; 541 struct sk_buff *skb; 542 unsigned int hh_len; 543 int err; 544 545 if (length > rt->u.dst.dev->mtu) { 546 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu); 547 return -EMSGSIZE; 548 } 549 if (flags&MSG_PROBE) 550 goto out; 551 552 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); 553 554 skb = sock_alloc_send_skb(sk, length+hh_len+15, 555 flags&MSG_DONTWAIT, &err); 556 if (skb == NULL) 557 goto error; 558 skb_reserve(skb, hh_len); 559 560 skb->priority = sk->sk_priority; 561 skb->dst = dst_clone(&rt->u.dst); 562 563 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length); 564 565 skb->ip_summed = CHECKSUM_NONE; 566 567 skb->h.raw = skb->nh.raw; 568 err = memcpy_fromiovecend((void *)iph, from, 0, length); 569 if (err) 570 goto error_fault; 571 572 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS); 573 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev, 574 dst_output); 575 if (err > 0) 576 err = np->recverr ? net_xmit_errno(err) : 0; 577 if (err) 578 goto error; 579 out: 580 return 0; 581 582 error_fault: 583 err = -EFAULT; 584 kfree_skb(skb); 585 error: 586 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS); 587 return err; 588 } 589 590 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg) 591 { 592 struct iovec *iov; 593 u8 __user *type = NULL; 594 u8 __user *code = NULL; 595 int probed = 0; 596 int i; 597 598 if (!msg->msg_iov) 599 return; 600 601 for (i = 0; i < msg->msg_iovlen; i++) { 602 iov = &msg->msg_iov[i]; 603 if (!iov) 604 continue; 605 606 switch (fl->proto) { 607 case IPPROTO_ICMPV6: 608 /* check if one-byte field is readable or not. */ 609 if (iov->iov_base && iov->iov_len < 1) 610 break; 611 612 if (!type) { 613 type = iov->iov_base; 614 /* check if code field is readable or not. */ 615 if (iov->iov_len > 1) 616 code = type + 1; 617 } else if (!code) 618 code = iov->iov_base; 619 620 if (type && code) { 621 get_user(fl->fl_icmp_type, type); 622 get_user(fl->fl_icmp_code, code); 623 probed = 1; 624 } 625 break; 626 default: 627 probed = 1; 628 break; 629 } 630 if (probed) 631 break; 632 } 633 } 634 635 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk, 636 struct msghdr *msg, size_t len) 637 { 638 struct ipv6_txoptions opt_space; 639 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name; 640 struct in6_addr *daddr, *final_p = NULL, final; 641 struct inet_sock *inet = inet_sk(sk); 642 struct ipv6_pinfo *np = inet6_sk(sk); 643 struct raw6_sock *rp = raw6_sk(sk); 644 struct ipv6_txoptions *opt = NULL; 645 struct ip6_flowlabel *flowlabel = NULL; 646 struct dst_entry *dst = NULL; 647 struct flowi fl; 648 int addr_len = msg->msg_namelen; 649 int hlimit = -1; 650 int tclass = -1; 651 u16 proto; 652 int err; 653 654 /* Rough check on arithmetic overflow, 655 better check is made in ip6_build_xmit 656 */ 657 if (len < 0) 658 return -EMSGSIZE; 659 660 /* Mirror BSD error message compatibility */ 661 if (msg->msg_flags & MSG_OOB) 662 return -EOPNOTSUPP; 663 664 /* 665 * Get and verify the address. 666 */ 667 memset(&fl, 0, sizeof(fl)); 668 669 if (sin6) { 670 if (addr_len < SIN6_LEN_RFC2133) 671 return -EINVAL; 672 673 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 674 return(-EAFNOSUPPORT); 675 676 /* port is the proto value [0..255] carried in nexthdr */ 677 proto = ntohs(sin6->sin6_port); 678 679 if (!proto) 680 proto = inet->num; 681 else if (proto != inet->num) 682 return(-EINVAL); 683 684 if (proto > 255) 685 return(-EINVAL); 686 687 daddr = &sin6->sin6_addr; 688 if (np->sndflow) { 689 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 690 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { 691 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 692 if (flowlabel == NULL) 693 return -EINVAL; 694 daddr = &flowlabel->dst; 695 } 696 } 697 698 /* 699 * Otherwise it will be difficult to maintain 700 * sk->sk_dst_cache. 701 */ 702 if (sk->sk_state == TCP_ESTABLISHED && 703 ipv6_addr_equal(daddr, &np->daddr)) 704 daddr = &np->daddr; 705 706 if (addr_len >= sizeof(struct sockaddr_in6) && 707 sin6->sin6_scope_id && 708 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) 709 fl.oif = sin6->sin6_scope_id; 710 } else { 711 if (sk->sk_state != TCP_ESTABLISHED) 712 return -EDESTADDRREQ; 713 714 proto = inet->num; 715 daddr = &np->daddr; 716 fl.fl6_flowlabel = np->flow_label; 717 } 718 719 if (ipv6_addr_any(daddr)) { 720 /* 721 * unspecified destination address 722 * treated as error... is this correct ? 723 */ 724 fl6_sock_release(flowlabel); 725 return(-EINVAL); 726 } 727 728 if (fl.oif == 0) 729 fl.oif = sk->sk_bound_dev_if; 730 731 if (msg->msg_controllen) { 732 opt = &opt_space; 733 memset(opt, 0, sizeof(struct ipv6_txoptions)); 734 opt->tot_len = sizeof(struct ipv6_txoptions); 735 736 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass); 737 if (err < 0) { 738 fl6_sock_release(flowlabel); 739 return err; 740 } 741 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 742 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 743 if (flowlabel == NULL) 744 return -EINVAL; 745 } 746 if (!(opt->opt_nflen|opt->opt_flen)) 747 opt = NULL; 748 } 749 if (opt == NULL) 750 opt = np->opt; 751 if (flowlabel) 752 opt = fl6_merge_options(&opt_space, flowlabel, opt); 753 opt = ipv6_fixup_options(&opt_space, opt); 754 755 fl.proto = proto; 756 rawv6_probe_proto_opt(&fl, msg); 757 758 ipv6_addr_copy(&fl.fl6_dst, daddr); 759 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr)) 760 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 761 762 /* merge ip6_build_xmit from ip6_output */ 763 if (opt && opt->srcrt) { 764 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt; 765 ipv6_addr_copy(&final, &fl.fl6_dst); 766 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 767 final_p = &final; 768 } 769 770 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) 771 fl.oif = np->mcast_oif; 772 773 err = ip6_dst_lookup(sk, &dst, &fl); 774 if (err) 775 goto out; 776 if (final_p) 777 ipv6_addr_copy(&fl.fl6_dst, final_p); 778 779 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) 780 goto out; 781 782 if (hlimit < 0) { 783 if (ipv6_addr_is_multicast(&fl.fl6_dst)) 784 hlimit = np->mcast_hops; 785 else 786 hlimit = np->hop_limit; 787 if (hlimit < 0) 788 hlimit = dst_metric(dst, RTAX_HOPLIMIT); 789 if (hlimit < 0) 790 hlimit = ipv6_get_hoplimit(dst->dev); 791 } 792 793 if (tclass < 0) { 794 tclass = np->cork.tclass; 795 if (tclass < 0) 796 tclass = 0; 797 } 798 799 if (msg->msg_flags&MSG_CONFIRM) 800 goto do_confirm; 801 802 back_from_confirm: 803 if (inet->hdrincl) { 804 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags); 805 } else { 806 lock_sock(sk); 807 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, 808 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst, 809 msg->msg_flags); 810 811 if (err) 812 ip6_flush_pending_frames(sk); 813 else if (!(msg->msg_flags & MSG_MORE)) 814 err = rawv6_push_pending_frames(sk, &fl, rp); 815 } 816 done: 817 ip6_dst_store(sk, dst, 818 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ? 819 &np->daddr : NULL); 820 821 release_sock(sk); 822 out: 823 fl6_sock_release(flowlabel); 824 return err<0?err:len; 825 do_confirm: 826 dst_confirm(dst); 827 if (!(msg->msg_flags & MSG_PROBE) || len) 828 goto back_from_confirm; 829 err = 0; 830 goto done; 831 } 832 833 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 834 char __user *optval, int optlen) 835 { 836 switch (optname) { 837 case ICMPV6_FILTER: 838 if (optlen > sizeof(struct icmp6_filter)) 839 optlen = sizeof(struct icmp6_filter); 840 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 841 return -EFAULT; 842 return 0; 843 default: 844 return -ENOPROTOOPT; 845 }; 846 847 return 0; 848 } 849 850 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 851 char __user *optval, int __user *optlen) 852 { 853 int len; 854 855 switch (optname) { 856 case ICMPV6_FILTER: 857 if (get_user(len, optlen)) 858 return -EFAULT; 859 if (len < 0) 860 return -EINVAL; 861 if (len > sizeof(struct icmp6_filter)) 862 len = sizeof(struct icmp6_filter); 863 if (put_user(len, optlen)) 864 return -EFAULT; 865 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 866 return -EFAULT; 867 return 0; 868 default: 869 return -ENOPROTOOPT; 870 }; 871 872 return 0; 873 } 874 875 876 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 877 char __user *optval, int optlen) 878 { 879 struct raw6_sock *rp = raw6_sk(sk); 880 int val; 881 882 switch(level) { 883 case SOL_RAW: 884 break; 885 886 case SOL_ICMPV6: 887 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 888 return -EOPNOTSUPP; 889 return rawv6_seticmpfilter(sk, level, optname, optval, 890 optlen); 891 case SOL_IPV6: 892 if (optname == IPV6_CHECKSUM) 893 break; 894 default: 895 return ipv6_setsockopt(sk, level, optname, optval, 896 optlen); 897 }; 898 899 if (get_user(val, (int __user *)optval)) 900 return -EFAULT; 901 902 switch (optname) { 903 case IPV6_CHECKSUM: 904 /* You may get strange result with a positive odd offset; 905 RFC2292bis agrees with me. */ 906 if (val > 0 && (val&1)) 907 return(-EINVAL); 908 if (val < 0) { 909 rp->checksum = 0; 910 } else { 911 rp->checksum = 1; 912 rp->offset = val; 913 } 914 915 return 0; 916 break; 917 918 default: 919 return(-ENOPROTOOPT); 920 } 921 } 922 923 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 924 char __user *optval, int __user *optlen) 925 { 926 struct raw6_sock *rp = raw6_sk(sk); 927 int val, len; 928 929 switch(level) { 930 case SOL_RAW: 931 break; 932 933 case SOL_ICMPV6: 934 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 935 return -EOPNOTSUPP; 936 return rawv6_geticmpfilter(sk, level, optname, optval, 937 optlen); 938 case SOL_IPV6: 939 if (optname == IPV6_CHECKSUM) 940 break; 941 default: 942 return ipv6_getsockopt(sk, level, optname, optval, 943 optlen); 944 }; 945 946 if (get_user(len,optlen)) 947 return -EFAULT; 948 949 switch (optname) { 950 case IPV6_CHECKSUM: 951 if (rp->checksum == 0) 952 val = -1; 953 else 954 val = rp->offset; 955 break; 956 957 default: 958 return -ENOPROTOOPT; 959 } 960 961 len = min_t(unsigned int, sizeof(int), len); 962 963 if (put_user(len, optlen)) 964 return -EFAULT; 965 if (copy_to_user(optval,&val,len)) 966 return -EFAULT; 967 return 0; 968 } 969 970 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 971 { 972 switch(cmd) { 973 case SIOCOUTQ: 974 { 975 int amount = atomic_read(&sk->sk_wmem_alloc); 976 return put_user(amount, (int __user *)arg); 977 } 978 case SIOCINQ: 979 { 980 struct sk_buff *skb; 981 int amount = 0; 982 983 spin_lock_bh(&sk->sk_receive_queue.lock); 984 skb = skb_peek(&sk->sk_receive_queue); 985 if (skb != NULL) 986 amount = skb->tail - skb->h.raw; 987 spin_unlock_bh(&sk->sk_receive_queue.lock); 988 return put_user(amount, (int __user *)arg); 989 } 990 991 default: 992 return -ENOIOCTLCMD; 993 } 994 } 995 996 static void rawv6_close(struct sock *sk, long timeout) 997 { 998 if (inet_sk(sk)->num == IPPROTO_RAW) 999 ip6_ra_control(sk, -1, NULL); 1000 1001 sk_common_release(sk); 1002 } 1003 1004 static int rawv6_init_sk(struct sock *sk) 1005 { 1006 if (inet_sk(sk)->num == IPPROTO_ICMPV6) { 1007 struct raw6_sock *rp = raw6_sk(sk); 1008 rp->checksum = 1; 1009 rp->offset = 2; 1010 } 1011 return(0); 1012 } 1013 1014 struct proto rawv6_prot = { 1015 .name = "RAWv6", 1016 .owner = THIS_MODULE, 1017 .close = rawv6_close, 1018 .connect = ip6_datagram_connect, 1019 .disconnect = udp_disconnect, 1020 .ioctl = rawv6_ioctl, 1021 .init = rawv6_init_sk, 1022 .destroy = inet6_destroy_sock, 1023 .setsockopt = rawv6_setsockopt, 1024 .getsockopt = rawv6_getsockopt, 1025 .sendmsg = rawv6_sendmsg, 1026 .recvmsg = rawv6_recvmsg, 1027 .bind = rawv6_bind, 1028 .backlog_rcv = rawv6_rcv_skb, 1029 .hash = raw_v6_hash, 1030 .unhash = raw_v6_unhash, 1031 .obj_size = sizeof(struct raw6_sock), 1032 }; 1033 1034 #ifdef CONFIG_PROC_FS 1035 struct raw6_iter_state { 1036 int bucket; 1037 }; 1038 1039 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private) 1040 1041 static struct sock *raw6_get_first(struct seq_file *seq) 1042 { 1043 struct sock *sk; 1044 struct hlist_node *node; 1045 struct raw6_iter_state* state = raw6_seq_private(seq); 1046 1047 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket) 1048 sk_for_each(sk, node, &raw_v6_htable[state->bucket]) 1049 if (sk->sk_family == PF_INET6) 1050 goto out; 1051 sk = NULL; 1052 out: 1053 return sk; 1054 } 1055 1056 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk) 1057 { 1058 struct raw6_iter_state* state = raw6_seq_private(seq); 1059 1060 do { 1061 sk = sk_next(sk); 1062 try_again: 1063 ; 1064 } while (sk && sk->sk_family != PF_INET6); 1065 1066 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) { 1067 sk = sk_head(&raw_v6_htable[state->bucket]); 1068 goto try_again; 1069 } 1070 return sk; 1071 } 1072 1073 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos) 1074 { 1075 struct sock *sk = raw6_get_first(seq); 1076 if (sk) 1077 while (pos && (sk = raw6_get_next(seq, sk)) != NULL) 1078 --pos; 1079 return pos ? NULL : sk; 1080 } 1081 1082 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos) 1083 { 1084 read_lock(&raw_v6_lock); 1085 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 1086 } 1087 1088 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1089 { 1090 struct sock *sk; 1091 1092 if (v == SEQ_START_TOKEN) 1093 sk = raw6_get_first(seq); 1094 else 1095 sk = raw6_get_next(seq, v); 1096 ++*pos; 1097 return sk; 1098 } 1099 1100 static void raw6_seq_stop(struct seq_file *seq, void *v) 1101 { 1102 read_unlock(&raw_v6_lock); 1103 } 1104 1105 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1106 { 1107 struct ipv6_pinfo *np = inet6_sk(sp); 1108 struct in6_addr *dest, *src; 1109 __u16 destp, srcp; 1110 1111 dest = &np->daddr; 1112 src = &np->rcv_saddr; 1113 destp = 0; 1114 srcp = inet_sk(sp)->num; 1115 seq_printf(seq, 1116 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1117 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n", 1118 i, 1119 src->s6_addr32[0], src->s6_addr32[1], 1120 src->s6_addr32[2], src->s6_addr32[3], srcp, 1121 dest->s6_addr32[0], dest->s6_addr32[1], 1122 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1123 sp->sk_state, 1124 atomic_read(&sp->sk_wmem_alloc), 1125 atomic_read(&sp->sk_rmem_alloc), 1126 0, 0L, 0, 1127 sock_i_uid(sp), 0, 1128 sock_i_ino(sp), 1129 atomic_read(&sp->sk_refcnt), sp); 1130 } 1131 1132 static int raw6_seq_show(struct seq_file *seq, void *v) 1133 { 1134 if (v == SEQ_START_TOKEN) 1135 seq_printf(seq, 1136 " sl " 1137 "local_address " 1138 "remote_address " 1139 "st tx_queue rx_queue tr tm->when retrnsmt" 1140 " uid timeout inode\n"); 1141 else 1142 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket); 1143 return 0; 1144 } 1145 1146 static struct seq_operations raw6_seq_ops = { 1147 .start = raw6_seq_start, 1148 .next = raw6_seq_next, 1149 .stop = raw6_seq_stop, 1150 .show = raw6_seq_show, 1151 }; 1152 1153 static int raw6_seq_open(struct inode *inode, struct file *file) 1154 { 1155 struct seq_file *seq; 1156 int rc = -ENOMEM; 1157 struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL); 1158 if (!s) 1159 goto out; 1160 rc = seq_open(file, &raw6_seq_ops); 1161 if (rc) 1162 goto out_kfree; 1163 seq = file->private_data; 1164 seq->private = s; 1165 memset(s, 0, sizeof(*s)); 1166 out: 1167 return rc; 1168 out_kfree: 1169 kfree(s); 1170 goto out; 1171 } 1172 1173 static struct file_operations raw6_seq_fops = { 1174 .owner = THIS_MODULE, 1175 .open = raw6_seq_open, 1176 .read = seq_read, 1177 .llseek = seq_lseek, 1178 .release = seq_release_private, 1179 }; 1180 1181 int __init raw6_proc_init(void) 1182 { 1183 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops)) 1184 return -ENOMEM; 1185 return 0; 1186 } 1187 1188 void raw6_proc_exit(void) 1189 { 1190 proc_net_remove("raw6"); 1191 } 1192 #endif /* CONFIG_PROC_FS */ 1193