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 <linux/skbuff.h> 36 #include <asm/uaccess.h> 37 #include <asm/ioctls.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 skb_kill_datagram(sk, skb, flags); 437 438 /* Error for blocking case is chosen to masquerade 439 as some normal condition. 440 */ 441 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 442 /* FIXME: increment a raw6 drops counter here */ 443 goto out; 444 } 445 446 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl, 447 struct raw6_sock *rp) 448 { 449 struct sk_buff *skb; 450 int err = 0; 451 int offset; 452 int len; 453 int total_len; 454 u32 tmp_csum; 455 u16 csum; 456 457 if (!rp->checksum) 458 goto send; 459 460 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 461 goto out; 462 463 offset = rp->offset; 464 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data); 465 if (offset >= total_len - 1) { 466 err = -EINVAL; 467 ip6_flush_pending_frames(sk); 468 goto out; 469 } 470 471 /* should be check HW csum miyazawa */ 472 if (skb_queue_len(&sk->sk_write_queue) == 1) { 473 /* 474 * Only one fragment on the socket. 475 */ 476 tmp_csum = skb->csum; 477 } else { 478 struct sk_buff *csum_skb = NULL; 479 tmp_csum = 0; 480 481 skb_queue_walk(&sk->sk_write_queue, skb) { 482 tmp_csum = csum_add(tmp_csum, skb->csum); 483 484 if (csum_skb) 485 continue; 486 487 len = skb->len - (skb->h.raw - skb->data); 488 if (offset >= len) { 489 offset -= len; 490 continue; 491 } 492 493 csum_skb = skb; 494 } 495 496 skb = csum_skb; 497 } 498 499 offset += skb->h.raw - skb->data; 500 if (skb_copy_bits(skb, offset, &csum, 2)) 501 BUG(); 502 503 /* in case cksum was not initialized */ 504 if (unlikely(csum)) 505 tmp_csum = csum_sub(tmp_csum, csum); 506 507 tmp_csum = csum_ipv6_magic(&fl->fl6_src, 508 &fl->fl6_dst, 509 total_len, fl->proto, tmp_csum); 510 511 if (tmp_csum == 0) 512 tmp_csum = -1; 513 514 csum = tmp_csum; 515 if (skb_store_bits(skb, offset, &csum, 2)) 516 BUG(); 517 518 send: 519 err = ip6_push_pending_frames(sk); 520 out: 521 return err; 522 } 523 524 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length, 525 struct flowi *fl, struct rt6_info *rt, 526 unsigned int flags) 527 { 528 struct ipv6_pinfo *np = inet6_sk(sk); 529 struct ipv6hdr *iph; 530 struct sk_buff *skb; 531 unsigned int hh_len; 532 int err; 533 534 if (length > rt->u.dst.dev->mtu) { 535 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu); 536 return -EMSGSIZE; 537 } 538 if (flags&MSG_PROBE) 539 goto out; 540 541 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); 542 543 skb = sock_alloc_send_skb(sk, length+hh_len+15, 544 flags&MSG_DONTWAIT, &err); 545 if (skb == NULL) 546 goto error; 547 skb_reserve(skb, hh_len); 548 549 skb->priority = sk->sk_priority; 550 skb->dst = dst_clone(&rt->u.dst); 551 552 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length); 553 554 skb->ip_summed = CHECKSUM_NONE; 555 556 skb->h.raw = skb->nh.raw; 557 err = memcpy_fromiovecend((void *)iph, from, 0, length); 558 if (err) 559 goto error_fault; 560 561 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS); 562 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev, 563 dst_output); 564 if (err > 0) 565 err = np->recverr ? net_xmit_errno(err) : 0; 566 if (err) 567 goto error; 568 out: 569 return 0; 570 571 error_fault: 572 err = -EFAULT; 573 kfree_skb(skb); 574 error: 575 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS); 576 return err; 577 } 578 579 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg) 580 { 581 struct iovec *iov; 582 u8 __user *type = NULL; 583 u8 __user *code = NULL; 584 int probed = 0; 585 int i; 586 587 if (!msg->msg_iov) 588 return; 589 590 for (i = 0; i < msg->msg_iovlen; i++) { 591 iov = &msg->msg_iov[i]; 592 if (!iov) 593 continue; 594 595 switch (fl->proto) { 596 case IPPROTO_ICMPV6: 597 /* check if one-byte field is readable or not. */ 598 if (iov->iov_base && iov->iov_len < 1) 599 break; 600 601 if (!type) { 602 type = iov->iov_base; 603 /* check if code field is readable or not. */ 604 if (iov->iov_len > 1) 605 code = type + 1; 606 } else if (!code) 607 code = iov->iov_base; 608 609 if (type && code) { 610 get_user(fl->fl_icmp_type, type); 611 get_user(fl->fl_icmp_code, code); 612 probed = 1; 613 } 614 break; 615 default: 616 probed = 1; 617 break; 618 } 619 if (probed) 620 break; 621 } 622 } 623 624 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk, 625 struct msghdr *msg, size_t len) 626 { 627 struct ipv6_txoptions opt_space; 628 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name; 629 struct in6_addr *daddr, *final_p = NULL, final; 630 struct inet_sock *inet = inet_sk(sk); 631 struct ipv6_pinfo *np = inet6_sk(sk); 632 struct raw6_sock *rp = raw6_sk(sk); 633 struct ipv6_txoptions *opt = NULL; 634 struct ip6_flowlabel *flowlabel = NULL; 635 struct dst_entry *dst = NULL; 636 struct flowi fl; 637 int addr_len = msg->msg_namelen; 638 int hlimit = -1; 639 int tclass = -1; 640 u16 proto; 641 int err; 642 643 /* Rough check on arithmetic overflow, 644 better check is made in ip6_build_xmit 645 */ 646 if (len < 0) 647 return -EMSGSIZE; 648 649 /* Mirror BSD error message compatibility */ 650 if (msg->msg_flags & MSG_OOB) 651 return -EOPNOTSUPP; 652 653 /* 654 * Get and verify the address. 655 */ 656 memset(&fl, 0, sizeof(fl)); 657 658 if (sin6) { 659 if (addr_len < SIN6_LEN_RFC2133) 660 return -EINVAL; 661 662 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 663 return(-EAFNOSUPPORT); 664 665 /* port is the proto value [0..255] carried in nexthdr */ 666 proto = ntohs(sin6->sin6_port); 667 668 if (!proto) 669 proto = inet->num; 670 else if (proto != inet->num) 671 return(-EINVAL); 672 673 if (proto > 255) 674 return(-EINVAL); 675 676 daddr = &sin6->sin6_addr; 677 if (np->sndflow) { 678 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 679 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { 680 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 681 if (flowlabel == NULL) 682 return -EINVAL; 683 daddr = &flowlabel->dst; 684 } 685 } 686 687 /* 688 * Otherwise it will be difficult to maintain 689 * sk->sk_dst_cache. 690 */ 691 if (sk->sk_state == TCP_ESTABLISHED && 692 ipv6_addr_equal(daddr, &np->daddr)) 693 daddr = &np->daddr; 694 695 if (addr_len >= sizeof(struct sockaddr_in6) && 696 sin6->sin6_scope_id && 697 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) 698 fl.oif = sin6->sin6_scope_id; 699 } else { 700 if (sk->sk_state != TCP_ESTABLISHED) 701 return -EDESTADDRREQ; 702 703 proto = inet->num; 704 daddr = &np->daddr; 705 fl.fl6_flowlabel = np->flow_label; 706 } 707 708 if (ipv6_addr_any(daddr)) { 709 /* 710 * unspecified destination address 711 * treated as error... is this correct ? 712 */ 713 fl6_sock_release(flowlabel); 714 return(-EINVAL); 715 } 716 717 if (fl.oif == 0) 718 fl.oif = sk->sk_bound_dev_if; 719 720 if (msg->msg_controllen) { 721 opt = &opt_space; 722 memset(opt, 0, sizeof(struct ipv6_txoptions)); 723 opt->tot_len = sizeof(struct ipv6_txoptions); 724 725 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass); 726 if (err < 0) { 727 fl6_sock_release(flowlabel); 728 return err; 729 } 730 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 731 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 732 if (flowlabel == NULL) 733 return -EINVAL; 734 } 735 if (!(opt->opt_nflen|opt->opt_flen)) 736 opt = NULL; 737 } 738 if (opt == NULL) 739 opt = np->opt; 740 if (flowlabel) 741 opt = fl6_merge_options(&opt_space, flowlabel, opt); 742 opt = ipv6_fixup_options(&opt_space, opt); 743 744 fl.proto = proto; 745 rawv6_probe_proto_opt(&fl, msg); 746 747 ipv6_addr_copy(&fl.fl6_dst, daddr); 748 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr)) 749 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 750 751 /* merge ip6_build_xmit from ip6_output */ 752 if (opt && opt->srcrt) { 753 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt; 754 ipv6_addr_copy(&final, &fl.fl6_dst); 755 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 756 final_p = &final; 757 } 758 759 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) 760 fl.oif = np->mcast_oif; 761 762 err = ip6_dst_lookup(sk, &dst, &fl); 763 if (err) 764 goto out; 765 if (final_p) 766 ipv6_addr_copy(&fl.fl6_dst, final_p); 767 768 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) 769 goto out; 770 771 if (hlimit < 0) { 772 if (ipv6_addr_is_multicast(&fl.fl6_dst)) 773 hlimit = np->mcast_hops; 774 else 775 hlimit = np->hop_limit; 776 if (hlimit < 0) 777 hlimit = dst_metric(dst, RTAX_HOPLIMIT); 778 if (hlimit < 0) 779 hlimit = ipv6_get_hoplimit(dst->dev); 780 } 781 782 if (tclass < 0) { 783 tclass = np->cork.tclass; 784 if (tclass < 0) 785 tclass = 0; 786 } 787 788 if (msg->msg_flags&MSG_CONFIRM) 789 goto do_confirm; 790 791 back_from_confirm: 792 if (inet->hdrincl) { 793 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags); 794 } else { 795 lock_sock(sk); 796 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, 797 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst, 798 msg->msg_flags); 799 800 if (err) 801 ip6_flush_pending_frames(sk); 802 else if (!(msg->msg_flags & MSG_MORE)) 803 err = rawv6_push_pending_frames(sk, &fl, rp); 804 } 805 done: 806 dst_release(dst); 807 release_sock(sk); 808 out: 809 fl6_sock_release(flowlabel); 810 return err<0?err:len; 811 do_confirm: 812 dst_confirm(dst); 813 if (!(msg->msg_flags & MSG_PROBE) || len) 814 goto back_from_confirm; 815 err = 0; 816 goto done; 817 } 818 819 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 820 char __user *optval, int optlen) 821 { 822 switch (optname) { 823 case ICMPV6_FILTER: 824 if (optlen > sizeof(struct icmp6_filter)) 825 optlen = sizeof(struct icmp6_filter); 826 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 827 return -EFAULT; 828 return 0; 829 default: 830 return -ENOPROTOOPT; 831 }; 832 833 return 0; 834 } 835 836 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 837 char __user *optval, int __user *optlen) 838 { 839 int len; 840 841 switch (optname) { 842 case ICMPV6_FILTER: 843 if (get_user(len, optlen)) 844 return -EFAULT; 845 if (len < 0) 846 return -EINVAL; 847 if (len > sizeof(struct icmp6_filter)) 848 len = sizeof(struct icmp6_filter); 849 if (put_user(len, optlen)) 850 return -EFAULT; 851 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 852 return -EFAULT; 853 return 0; 854 default: 855 return -ENOPROTOOPT; 856 }; 857 858 return 0; 859 } 860 861 862 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 863 char __user *optval, int optlen) 864 { 865 struct raw6_sock *rp = raw6_sk(sk); 866 int val; 867 868 if (get_user(val, (int __user *)optval)) 869 return -EFAULT; 870 871 switch (optname) { 872 case IPV6_CHECKSUM: 873 /* You may get strange result with a positive odd offset; 874 RFC2292bis agrees with me. */ 875 if (val > 0 && (val&1)) 876 return(-EINVAL); 877 if (val < 0) { 878 rp->checksum = 0; 879 } else { 880 rp->checksum = 1; 881 rp->offset = val; 882 } 883 884 return 0; 885 break; 886 887 default: 888 return(-ENOPROTOOPT); 889 } 890 } 891 892 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 893 char __user *optval, int optlen) 894 { 895 switch(level) { 896 case SOL_RAW: 897 break; 898 899 case SOL_ICMPV6: 900 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 901 return -EOPNOTSUPP; 902 return rawv6_seticmpfilter(sk, level, optname, optval, 903 optlen); 904 case SOL_IPV6: 905 if (optname == IPV6_CHECKSUM) 906 break; 907 default: 908 return ipv6_setsockopt(sk, level, optname, optval, 909 optlen); 910 }; 911 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 912 } 913 914 #ifdef CONFIG_COMPAT 915 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 916 char __user *optval, int optlen) 917 { 918 switch (level) { 919 case SOL_RAW: 920 break; 921 case SOL_ICMPV6: 922 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 923 return -EOPNOTSUPP; 924 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 925 case SOL_IPV6: 926 if (optname == IPV6_CHECKSUM) 927 break; 928 default: 929 return compat_ipv6_setsockopt(sk, level, optname, 930 optval, optlen); 931 }; 932 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 933 } 934 #endif 935 936 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 937 char __user *optval, int __user *optlen) 938 { 939 struct raw6_sock *rp = raw6_sk(sk); 940 int val, len; 941 942 if (get_user(len,optlen)) 943 return -EFAULT; 944 945 switch (optname) { 946 case IPV6_CHECKSUM: 947 if (rp->checksum == 0) 948 val = -1; 949 else 950 val = rp->offset; 951 break; 952 953 default: 954 return -ENOPROTOOPT; 955 } 956 957 len = min_t(unsigned int, sizeof(int), len); 958 959 if (put_user(len, optlen)) 960 return -EFAULT; 961 if (copy_to_user(optval,&val,len)) 962 return -EFAULT; 963 return 0; 964 } 965 966 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 967 char __user *optval, int __user *optlen) 968 { 969 switch(level) { 970 case SOL_RAW: 971 break; 972 973 case SOL_ICMPV6: 974 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 975 return -EOPNOTSUPP; 976 return rawv6_geticmpfilter(sk, level, optname, optval, 977 optlen); 978 case SOL_IPV6: 979 if (optname == IPV6_CHECKSUM) 980 break; 981 default: 982 return ipv6_getsockopt(sk, level, optname, optval, 983 optlen); 984 }; 985 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 986 } 987 988 #ifdef CONFIG_COMPAT 989 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 990 char __user *optval, int __user *optlen) 991 { 992 switch (level) { 993 case SOL_RAW: 994 break; 995 case SOL_ICMPV6: 996 if (inet_sk(sk)->num != IPPROTO_ICMPV6) 997 return -EOPNOTSUPP; 998 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 999 case SOL_IPV6: 1000 if (optname == IPV6_CHECKSUM) 1001 break; 1002 default: 1003 return compat_ipv6_getsockopt(sk, level, optname, 1004 optval, optlen); 1005 }; 1006 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1007 } 1008 #endif 1009 1010 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1011 { 1012 switch(cmd) { 1013 case SIOCOUTQ: 1014 { 1015 int amount = atomic_read(&sk->sk_wmem_alloc); 1016 return put_user(amount, (int __user *)arg); 1017 } 1018 case SIOCINQ: 1019 { 1020 struct sk_buff *skb; 1021 int amount = 0; 1022 1023 spin_lock_bh(&sk->sk_receive_queue.lock); 1024 skb = skb_peek(&sk->sk_receive_queue); 1025 if (skb != NULL) 1026 amount = skb->tail - skb->h.raw; 1027 spin_unlock_bh(&sk->sk_receive_queue.lock); 1028 return put_user(amount, (int __user *)arg); 1029 } 1030 1031 default: 1032 return -ENOIOCTLCMD; 1033 } 1034 } 1035 1036 static void rawv6_close(struct sock *sk, long timeout) 1037 { 1038 if (inet_sk(sk)->num == IPPROTO_RAW) 1039 ip6_ra_control(sk, -1, NULL); 1040 1041 sk_common_release(sk); 1042 } 1043 1044 static int rawv6_init_sk(struct sock *sk) 1045 { 1046 if (inet_sk(sk)->num == IPPROTO_ICMPV6) { 1047 struct raw6_sock *rp = raw6_sk(sk); 1048 rp->checksum = 1; 1049 rp->offset = 2; 1050 } 1051 return(0); 1052 } 1053 1054 struct proto rawv6_prot = { 1055 .name = "RAWv6", 1056 .owner = THIS_MODULE, 1057 .close = rawv6_close, 1058 .connect = ip6_datagram_connect, 1059 .disconnect = udp_disconnect, 1060 .ioctl = rawv6_ioctl, 1061 .init = rawv6_init_sk, 1062 .destroy = inet6_destroy_sock, 1063 .setsockopt = rawv6_setsockopt, 1064 .getsockopt = rawv6_getsockopt, 1065 .sendmsg = rawv6_sendmsg, 1066 .recvmsg = rawv6_recvmsg, 1067 .bind = rawv6_bind, 1068 .backlog_rcv = rawv6_rcv_skb, 1069 .hash = raw_v6_hash, 1070 .unhash = raw_v6_unhash, 1071 .obj_size = sizeof(struct raw6_sock), 1072 #ifdef CONFIG_COMPAT 1073 .compat_setsockopt = compat_rawv6_setsockopt, 1074 .compat_getsockopt = compat_rawv6_getsockopt, 1075 #endif 1076 }; 1077 1078 #ifdef CONFIG_PROC_FS 1079 struct raw6_iter_state { 1080 int bucket; 1081 }; 1082 1083 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private) 1084 1085 static struct sock *raw6_get_first(struct seq_file *seq) 1086 { 1087 struct sock *sk; 1088 struct hlist_node *node; 1089 struct raw6_iter_state* state = raw6_seq_private(seq); 1090 1091 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket) 1092 sk_for_each(sk, node, &raw_v6_htable[state->bucket]) 1093 if (sk->sk_family == PF_INET6) 1094 goto out; 1095 sk = NULL; 1096 out: 1097 return sk; 1098 } 1099 1100 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk) 1101 { 1102 struct raw6_iter_state* state = raw6_seq_private(seq); 1103 1104 do { 1105 sk = sk_next(sk); 1106 try_again: 1107 ; 1108 } while (sk && sk->sk_family != PF_INET6); 1109 1110 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) { 1111 sk = sk_head(&raw_v6_htable[state->bucket]); 1112 goto try_again; 1113 } 1114 return sk; 1115 } 1116 1117 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos) 1118 { 1119 struct sock *sk = raw6_get_first(seq); 1120 if (sk) 1121 while (pos && (sk = raw6_get_next(seq, sk)) != NULL) 1122 --pos; 1123 return pos ? NULL : sk; 1124 } 1125 1126 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos) 1127 { 1128 read_lock(&raw_v6_lock); 1129 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 1130 } 1131 1132 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1133 { 1134 struct sock *sk; 1135 1136 if (v == SEQ_START_TOKEN) 1137 sk = raw6_get_first(seq); 1138 else 1139 sk = raw6_get_next(seq, v); 1140 ++*pos; 1141 return sk; 1142 } 1143 1144 static void raw6_seq_stop(struct seq_file *seq, void *v) 1145 { 1146 read_unlock(&raw_v6_lock); 1147 } 1148 1149 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1150 { 1151 struct ipv6_pinfo *np = inet6_sk(sp); 1152 struct in6_addr *dest, *src; 1153 __u16 destp, srcp; 1154 1155 dest = &np->daddr; 1156 src = &np->rcv_saddr; 1157 destp = 0; 1158 srcp = inet_sk(sp)->num; 1159 seq_printf(seq, 1160 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1161 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n", 1162 i, 1163 src->s6_addr32[0], src->s6_addr32[1], 1164 src->s6_addr32[2], src->s6_addr32[3], srcp, 1165 dest->s6_addr32[0], dest->s6_addr32[1], 1166 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1167 sp->sk_state, 1168 atomic_read(&sp->sk_wmem_alloc), 1169 atomic_read(&sp->sk_rmem_alloc), 1170 0, 0L, 0, 1171 sock_i_uid(sp), 0, 1172 sock_i_ino(sp), 1173 atomic_read(&sp->sk_refcnt), sp); 1174 } 1175 1176 static int raw6_seq_show(struct seq_file *seq, void *v) 1177 { 1178 if (v == SEQ_START_TOKEN) 1179 seq_printf(seq, 1180 " sl " 1181 "local_address " 1182 "remote_address " 1183 "st tx_queue rx_queue tr tm->when retrnsmt" 1184 " uid timeout inode\n"); 1185 else 1186 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket); 1187 return 0; 1188 } 1189 1190 static struct seq_operations raw6_seq_ops = { 1191 .start = raw6_seq_start, 1192 .next = raw6_seq_next, 1193 .stop = raw6_seq_stop, 1194 .show = raw6_seq_show, 1195 }; 1196 1197 static int raw6_seq_open(struct inode *inode, struct file *file) 1198 { 1199 struct seq_file *seq; 1200 int rc = -ENOMEM; 1201 struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL); 1202 if (!s) 1203 goto out; 1204 rc = seq_open(file, &raw6_seq_ops); 1205 if (rc) 1206 goto out_kfree; 1207 seq = file->private_data; 1208 seq->private = s; 1209 out: 1210 return rc; 1211 out_kfree: 1212 kfree(s); 1213 goto out; 1214 } 1215 1216 static struct file_operations raw6_seq_fops = { 1217 .owner = THIS_MODULE, 1218 .open = raw6_seq_open, 1219 .read = seq_read, 1220 .llseek = seq_lseek, 1221 .release = seq_release_private, 1222 }; 1223 1224 int __init raw6_proc_init(void) 1225 { 1226 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops)) 1227 return -ENOMEM; 1228 return 0; 1229 } 1230 1231 void raw6_proc_exit(void) 1232 { 1233 proc_net_remove("raw6"); 1234 } 1235 #endif /* CONFIG_PROC_FS */ 1236