1/* 2 * RAW sockets for IPv6 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 * 8 * Adapted from linux/net/ipv4/raw.c 9 * 10 * Fixes: 11 * Hideaki YOSHIFUJI : sin6_scope_id support 12 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance) 13 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data 14 * 15 * This program is free software; you can redistribute it and/or 16 * modify it under the terms of the GNU General Public License 17 * as published by the Free Software Foundation; either version 18 * 2 of the License, or (at your option) any later version. 19 */ 20 21#include <linux/errno.h> 22#include <linux/types.h> 23#include <linux/socket.h> 24#include <linux/slab.h> 25#include <linux/sockios.h> 26#include <linux/net.h> 27#include <linux/in6.h> 28#include <linux/netdevice.h> 29#include <linux/if_arp.h> 30#include <linux/icmpv6.h> 31#include <linux/netfilter.h> 32#include <linux/netfilter_ipv6.h> 33#include <linux/skbuff.h> 34#include <linux/compat.h> 35#include <linux/uaccess.h> 36#include <asm/ioctls.h> 37 38#include <net/net_namespace.h> 39#include <net/ip.h> 40#include <net/sock.h> 41#include <net/snmp.h> 42 43#include <net/ipv6.h> 44#include <net/ndisc.h> 45#include <net/protocol.h> 46#include <net/ip6_route.h> 47#include <net/ip6_checksum.h> 48#include <net/addrconf.h> 49#include <net/transp_v6.h> 50#include <net/udp.h> 51#include <net/inet_common.h> 52#include <net/tcp_states.h> 53#if IS_ENABLED(CONFIG_IPV6_MIP6) 54#include <net/mip6.h> 55#endif 56#include <linux/mroute6.h> 57 58#include <net/raw.h> 59#include <net/rawv6.h> 60#include <net/xfrm.h> 61 62#include <linux/proc_fs.h> 63#include <linux/seq_file.h> 64#include <linux/export.h> 65 66#define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */ 67 68static struct raw_hashinfo raw_v6_hashinfo = { 69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock), 70}; 71 72static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk, 73 unsigned short num, const struct in6_addr *loc_addr, 74 const struct in6_addr *rmt_addr, int dif) 75{ 76 bool is_multicast = ipv6_addr_is_multicast(loc_addr); 77 78 sk_for_each_from(sk) 79 if (inet_sk(sk)->inet_num == num) { 80 81 if (!net_eq(sock_net(sk), net)) 82 continue; 83 84 if (!ipv6_addr_any(&sk->sk_v6_daddr) && 85 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) 86 continue; 87 88 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) 89 continue; 90 91 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) { 92 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)) 93 goto found; 94 if (is_multicast && 95 inet6_mc_check(sk, loc_addr, rmt_addr)) 96 goto found; 97 continue; 98 } 99 goto found; 100 } 101 sk = NULL; 102found: 103 return sk; 104} 105 106/* 107 * 0 - deliver 108 * 1 - block 109 */ 110static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb) 111{ 112 struct icmp6hdr _hdr; 113 const struct icmp6hdr *hdr; 114 115 /* We require only the four bytes of the ICMPv6 header, not any 116 * additional bytes of message body in "struct icmp6hdr". 117 */ 118 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 119 ICMPV6_HDRLEN, &_hdr); 120 if (hdr) { 121 const __u32 *data = &raw6_sk(sk)->filter.data[0]; 122 unsigned int type = hdr->icmp6_type; 123 124 return (data[type >> 5] & (1U << (type & 31))) != 0; 125 } 126 return 1; 127} 128 129#if IS_ENABLED(CONFIG_IPV6_MIP6) 130typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb); 131 132static mh_filter_t __rcu *mh_filter __read_mostly; 133 134int rawv6_mh_filter_register(mh_filter_t filter) 135{ 136 rcu_assign_pointer(mh_filter, filter); 137 return 0; 138} 139EXPORT_SYMBOL(rawv6_mh_filter_register); 140 141int rawv6_mh_filter_unregister(mh_filter_t filter) 142{ 143 RCU_INIT_POINTER(mh_filter, NULL); 144 synchronize_rcu(); 145 return 0; 146} 147EXPORT_SYMBOL(rawv6_mh_filter_unregister); 148 149#endif 150 151/* 152 * demultiplex raw sockets. 153 * (should consider queueing the skb in the sock receive_queue 154 * without calling rawv6.c) 155 * 156 * Caller owns SKB so we must make clones. 157 */ 158static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 159{ 160 const struct in6_addr *saddr; 161 const struct in6_addr *daddr; 162 struct sock *sk; 163 bool delivered = false; 164 __u8 hash; 165 struct net *net; 166 167 saddr = &ipv6_hdr(skb)->saddr; 168 daddr = saddr + 1; 169 170 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 171 172 read_lock(&raw_v6_hashinfo.lock); 173 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 174 175 if (!sk) 176 goto out; 177 178 net = dev_net(skb->dev); 179 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, inet6_iif(skb)); 180 181 while (sk) { 182 int filtered; 183 184 delivered = true; 185 switch (nexthdr) { 186 case IPPROTO_ICMPV6: 187 filtered = icmpv6_filter(sk, skb); 188 break; 189 190#if IS_ENABLED(CONFIG_IPV6_MIP6) 191 case IPPROTO_MH: 192 { 193 /* XXX: To validate MH only once for each packet, 194 * this is placed here. It should be after checking 195 * xfrm policy, however it doesn't. The checking xfrm 196 * policy is placed in rawv6_rcv() because it is 197 * required for each socket. 198 */ 199 mh_filter_t *filter; 200 201 filter = rcu_dereference(mh_filter); 202 filtered = filter ? (*filter)(sk, skb) : 0; 203 break; 204 } 205#endif 206 default: 207 filtered = 0; 208 break; 209 } 210 211 if (filtered < 0) 212 break; 213 if (filtered == 0) { 214 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 215 216 /* Not releasing hash table! */ 217 if (clone) { 218 nf_reset(clone); 219 rawv6_rcv(sk, clone); 220 } 221 } 222 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr, 223 inet6_iif(skb)); 224 } 225out: 226 read_unlock(&raw_v6_hashinfo.lock); 227 return delivered; 228} 229 230bool raw6_local_deliver(struct sk_buff *skb, int nexthdr) 231{ 232 struct sock *raw_sk; 233 234 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]); 235 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr)) 236 raw_sk = NULL; 237 238 return raw_sk != NULL; 239} 240 241/* This cleans up af_inet6 a bit. -DaveM */ 242static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 243{ 244 struct inet_sock *inet = inet_sk(sk); 245 struct ipv6_pinfo *np = inet6_sk(sk); 246 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 247 __be32 v4addr = 0; 248 int addr_type; 249 int err; 250 251 if (addr_len < SIN6_LEN_RFC2133) 252 return -EINVAL; 253 254 if (addr->sin6_family != AF_INET6) 255 return -EINVAL; 256 257 addr_type = ipv6_addr_type(&addr->sin6_addr); 258 259 /* Raw sockets are IPv6 only */ 260 if (addr_type == IPV6_ADDR_MAPPED) 261 return -EADDRNOTAVAIL; 262 263 lock_sock(sk); 264 265 err = -EINVAL; 266 if (sk->sk_state != TCP_CLOSE) 267 goto out; 268 269 rcu_read_lock(); 270 /* Check if the address belongs to the host. */ 271 if (addr_type != IPV6_ADDR_ANY) { 272 struct net_device *dev = NULL; 273 274 if (__ipv6_addr_needs_scope_id(addr_type)) { 275 if (addr_len >= sizeof(struct sockaddr_in6) && 276 addr->sin6_scope_id) { 277 /* Override any existing binding, if another 278 * one is supplied by user. 279 */ 280 sk->sk_bound_dev_if = addr->sin6_scope_id; 281 } 282 283 /* Binding to link-local address requires an interface */ 284 if (!sk->sk_bound_dev_if) 285 goto out_unlock; 286 287 err = -ENODEV; 288 dev = dev_get_by_index_rcu(sock_net(sk), 289 sk->sk_bound_dev_if); 290 if (!dev) 291 goto out_unlock; 292 } 293 294 /* ipv4 addr of the socket is invalid. Only the 295 * unspecified and mapped address have a v4 equivalent. 296 */ 297 v4addr = LOOPBACK4_IPV6; 298 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 299 err = -EADDRNOTAVAIL; 300 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr, 301 dev, 0)) { 302 goto out_unlock; 303 } 304 } 305 } 306 307 inet->inet_rcv_saddr = inet->inet_saddr = v4addr; 308 sk->sk_v6_rcv_saddr = addr->sin6_addr; 309 if (!(addr_type & IPV6_ADDR_MULTICAST)) 310 np->saddr = addr->sin6_addr; 311 err = 0; 312out_unlock: 313 rcu_read_unlock(); 314out: 315 release_sock(sk); 316 return err; 317} 318 319static void rawv6_err(struct sock *sk, struct sk_buff *skb, 320 struct inet6_skb_parm *opt, 321 u8 type, u8 code, int offset, __be32 info) 322{ 323 struct inet_sock *inet = inet_sk(sk); 324 struct ipv6_pinfo *np = inet6_sk(sk); 325 int err; 326 int harderr; 327 328 /* Report error on raw socket, if: 329 1. User requested recverr. 330 2. Socket is connected (otherwise the error indication 331 is useless without recverr and error is hard. 332 */ 333 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 334 return; 335 336 harderr = icmpv6_err_convert(type, code, &err); 337 if (type == ICMPV6_PKT_TOOBIG) { 338 ip6_sk_update_pmtu(skb, sk, info); 339 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 340 } 341 if (type == NDISC_REDIRECT) { 342 ip6_sk_redirect(skb, sk); 343 return; 344 } 345 if (np->recverr) { 346 u8 *payload = skb->data; 347 if (!inet->hdrincl) 348 payload += offset; 349 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 350 } 351 352 if (np->recverr || harderr) { 353 sk->sk_err = err; 354 sk->sk_error_report(sk); 355 } 356} 357 358void raw6_icmp_error(struct sk_buff *skb, int nexthdr, 359 u8 type, u8 code, int inner_offset, __be32 info) 360{ 361 struct sock *sk; 362 int hash; 363 const struct in6_addr *saddr, *daddr; 364 struct net *net; 365 366 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 367 368 read_lock(&raw_v6_hashinfo.lock); 369 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 370 if (sk) { 371 /* Note: ipv6_hdr(skb) != skb->data */ 372 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data; 373 saddr = &ip6h->saddr; 374 daddr = &ip6h->daddr; 375 net = dev_net(skb->dev); 376 377 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr, 378 inet6_iif(skb)))) { 379 rawv6_err(sk, skb, NULL, type, code, 380 inner_offset, info); 381 sk = sk_next(sk); 382 } 383 } 384 read_unlock(&raw_v6_hashinfo.lock); 385} 386 387static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb) 388{ 389 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) && 390 skb_checksum_complete(skb)) { 391 atomic_inc(&sk->sk_drops); 392 kfree_skb(skb); 393 return NET_RX_DROP; 394 } 395 396 /* Charge it to the socket. */ 397 skb_dst_drop(skb); 398 if (sock_queue_rcv_skb(sk, skb) < 0) { 399 kfree_skb(skb); 400 return NET_RX_DROP; 401 } 402 403 return 0; 404} 405 406/* 407 * This is next to useless... 408 * if we demultiplex in network layer we don't need the extra call 409 * just to queue the skb... 410 * maybe we could have the network decide upon a hint if it 411 * should call raw_rcv for demultiplexing 412 */ 413int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 414{ 415 struct inet_sock *inet = inet_sk(sk); 416 struct raw6_sock *rp = raw6_sk(sk); 417 418 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 419 atomic_inc(&sk->sk_drops); 420 kfree_skb(skb); 421 return NET_RX_DROP; 422 } 423 424 if (!rp->checksum) 425 skb->ip_summed = CHECKSUM_UNNECESSARY; 426 427 if (skb->ip_summed == CHECKSUM_COMPLETE) { 428 skb_postpull_rcsum(skb, skb_network_header(skb), 429 skb_network_header_len(skb)); 430 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 431 &ipv6_hdr(skb)->daddr, 432 skb->len, inet->inet_num, skb->csum)) 433 skb->ip_summed = CHECKSUM_UNNECESSARY; 434 } 435 if (!skb_csum_unnecessary(skb)) 436 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 437 &ipv6_hdr(skb)->daddr, 438 skb->len, 439 inet->inet_num, 0)); 440 441 if (inet->hdrincl) { 442 if (skb_checksum_complete(skb)) { 443 atomic_inc(&sk->sk_drops); 444 kfree_skb(skb); 445 return NET_RX_DROP; 446 } 447 } 448 449 rawv6_rcv_skb(sk, skb); 450 return 0; 451} 452 453 454/* 455 * This should be easy, if there is something there 456 * we return it, otherwise we block. 457 */ 458 459static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 460 int noblock, int flags, int *addr_len) 461{ 462 struct ipv6_pinfo *np = inet6_sk(sk); 463 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 464 struct sk_buff *skb; 465 size_t copied; 466 int err; 467 468 if (flags & MSG_OOB) 469 return -EOPNOTSUPP; 470 471 if (flags & MSG_ERRQUEUE) 472 return ipv6_recv_error(sk, msg, len, addr_len); 473 474 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 475 return ipv6_recv_rxpmtu(sk, msg, len, addr_len); 476 477 skb = skb_recv_datagram(sk, flags, noblock, &err); 478 if (!skb) 479 goto out; 480 481 copied = skb->len; 482 if (copied > len) { 483 copied = len; 484 msg->msg_flags |= MSG_TRUNC; 485 } 486 487 if (skb_csum_unnecessary(skb)) { 488 err = skb_copy_datagram_msg(skb, 0, msg, copied); 489 } else if (msg->msg_flags&MSG_TRUNC) { 490 if (__skb_checksum_complete(skb)) 491 goto csum_copy_err; 492 err = skb_copy_datagram_msg(skb, 0, msg, copied); 493 } else { 494 err = skb_copy_and_csum_datagram_msg(skb, 0, msg); 495 if (err == -EINVAL) 496 goto csum_copy_err; 497 } 498 if (err) 499 goto out_free; 500 501 /* Copy the address. */ 502 if (sin6) { 503 sin6->sin6_family = AF_INET6; 504 sin6->sin6_port = 0; 505 sin6->sin6_addr = ipv6_hdr(skb)->saddr; 506 sin6->sin6_flowinfo = 0; 507 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr, 508 inet6_iif(skb)); 509 *addr_len = sizeof(*sin6); 510 } 511 512 sock_recv_ts_and_drops(msg, sk, skb); 513 514 if (np->rxopt.all) 515 ip6_datagram_recv_ctl(sk, msg, skb); 516 517 err = copied; 518 if (flags & MSG_TRUNC) 519 err = skb->len; 520 521out_free: 522 skb_free_datagram(sk, skb); 523out: 524 return err; 525 526csum_copy_err: 527 skb_kill_datagram(sk, skb, flags); 528 529 /* Error for blocking case is chosen to masquerade 530 as some normal condition. 531 */ 532 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 533 goto out; 534} 535 536static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 537 struct raw6_sock *rp) 538{ 539 struct sk_buff *skb; 540 int err = 0; 541 int offset; 542 int len; 543 int total_len; 544 __wsum tmp_csum; 545 __sum16 csum; 546 547 if (!rp->checksum) 548 goto send; 549 550 skb = skb_peek(&sk->sk_write_queue); 551 if (!skb) 552 goto out; 553 554 offset = rp->offset; 555 total_len = inet_sk(sk)->cork.base.length; 556 if (offset >= total_len - 1) { 557 err = -EINVAL; 558 ip6_flush_pending_frames(sk); 559 goto out; 560 } 561 562 /* should be check HW csum miyazawa */ 563 if (skb_queue_len(&sk->sk_write_queue) == 1) { 564 /* 565 * Only one fragment on the socket. 566 */ 567 tmp_csum = skb->csum; 568 } else { 569 struct sk_buff *csum_skb = NULL; 570 tmp_csum = 0; 571 572 skb_queue_walk(&sk->sk_write_queue, skb) { 573 tmp_csum = csum_add(tmp_csum, skb->csum); 574 575 if (csum_skb) 576 continue; 577 578 len = skb->len - skb_transport_offset(skb); 579 if (offset >= len) { 580 offset -= len; 581 continue; 582 } 583 584 csum_skb = skb; 585 } 586 587 skb = csum_skb; 588 } 589 590 offset += skb_transport_offset(skb); 591 BUG_ON(skb_copy_bits(skb, offset, &csum, 2)); 592 593 /* in case cksum was not initialized */ 594 if (unlikely(csum)) 595 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 596 597 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 598 total_len, fl6->flowi6_proto, tmp_csum); 599 600 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP) 601 csum = CSUM_MANGLED_0; 602 603 BUG_ON(skb_store_bits(skb, offset, &csum, 2)); 604 605send: 606 err = ip6_push_pending_frames(sk); 607out: 608 return err; 609} 610 611static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length, 612 struct flowi6 *fl6, struct dst_entry **dstp, 613 unsigned int flags) 614{ 615 struct ipv6_pinfo *np = inet6_sk(sk); 616 struct ipv6hdr *iph; 617 struct sk_buff *skb; 618 int err; 619 struct rt6_info *rt = (struct rt6_info *)*dstp; 620 int hlen = LL_RESERVED_SPACE(rt->dst.dev); 621 int tlen = rt->dst.dev->needed_tailroom; 622 623 if (length > rt->dst.dev->mtu) { 624 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu); 625 return -EMSGSIZE; 626 } 627 if (flags&MSG_PROBE) 628 goto out; 629 630 skb = sock_alloc_send_skb(sk, 631 length + hlen + tlen + 15, 632 flags & MSG_DONTWAIT, &err); 633 if (!skb) 634 goto error; 635 skb_reserve(skb, hlen); 636 637 skb->protocol = htons(ETH_P_IPV6); 638 skb->priority = sk->sk_priority; 639 skb->mark = sk->sk_mark; 640 skb_dst_set(skb, &rt->dst); 641 *dstp = NULL; 642 643 skb_put(skb, length); 644 skb_reset_network_header(skb); 645 iph = ipv6_hdr(skb); 646 647 skb->ip_summed = CHECKSUM_NONE; 648 649 skb->transport_header = skb->network_header; 650 err = memcpy_from_msg(iph, msg, length); 651 if (err) 652 goto error_fault; 653 654 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); 655 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, sk, skb, 656 NULL, rt->dst.dev, dst_output_sk); 657 if (err > 0) 658 err = net_xmit_errno(err); 659 if (err) 660 goto error; 661out: 662 return 0; 663 664error_fault: 665 err = -EFAULT; 666 kfree_skb(skb); 667error: 668 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 669 if (err == -ENOBUFS && !np->recverr) 670 err = 0; 671 return err; 672} 673 674struct raw6_frag_vec { 675 struct msghdr *msg; 676 int hlen; 677 char c[4]; 678}; 679 680static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6) 681{ 682 int err = 0; 683 switch (fl6->flowi6_proto) { 684 case IPPROTO_ICMPV6: 685 rfv->hlen = 2; 686 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); 687 if (!err) { 688 fl6->fl6_icmp_type = rfv->c[0]; 689 fl6->fl6_icmp_code = rfv->c[1]; 690 } 691 break; 692 case IPPROTO_MH: 693 rfv->hlen = 4; 694 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); 695 if (!err) 696 fl6->fl6_mh_type = rfv->c[2]; 697 } 698 return err; 699} 700 701static int raw6_getfrag(void *from, char *to, int offset, int len, int odd, 702 struct sk_buff *skb) 703{ 704 struct raw6_frag_vec *rfv = from; 705 706 if (offset < rfv->hlen) { 707 int copy = min(rfv->hlen - offset, len); 708 709 if (skb->ip_summed == CHECKSUM_PARTIAL) 710 memcpy(to, rfv->c + offset, copy); 711 else 712 skb->csum = csum_block_add( 713 skb->csum, 714 csum_partial_copy_nocheck(rfv->c + offset, 715 to, copy, 0), 716 odd); 717 718 odd = 0; 719 offset += copy; 720 to += copy; 721 len -= copy; 722 723 if (!len) 724 return 0; 725 } 726 727 offset -= rfv->hlen; 728 729 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); 730} 731 732static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 733{ 734 struct ipv6_txoptions *opt_to_free = NULL; 735 struct ipv6_txoptions opt_space; 736 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 737 struct in6_addr *daddr, *final_p, final; 738 struct inet_sock *inet = inet_sk(sk); 739 struct ipv6_pinfo *np = inet6_sk(sk); 740 struct raw6_sock *rp = raw6_sk(sk); 741 struct ipv6_txoptions *opt = NULL; 742 struct ip6_flowlabel *flowlabel = NULL; 743 struct dst_entry *dst = NULL; 744 struct raw6_frag_vec rfv; 745 struct flowi6 fl6; 746 int addr_len = msg->msg_namelen; 747 int hlimit = -1; 748 int tclass = -1; 749 int dontfrag = -1; 750 u16 proto; 751 int err; 752 753 /* Rough check on arithmetic overflow, 754 better check is made in ip6_append_data(). 755 */ 756 if (len > INT_MAX) 757 return -EMSGSIZE; 758 759 /* Mirror BSD error message compatibility */ 760 if (msg->msg_flags & MSG_OOB) 761 return -EOPNOTSUPP; 762 763 /* 764 * Get and verify the address. 765 */ 766 memset(&fl6, 0, sizeof(fl6)); 767 768 fl6.flowi6_mark = sk->sk_mark; 769 770 if (sin6) { 771 if (addr_len < SIN6_LEN_RFC2133) 772 return -EINVAL; 773 774 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 775 return -EAFNOSUPPORT; 776 777 /* port is the proto value [0..255] carried in nexthdr */ 778 proto = ntohs(sin6->sin6_port); 779 780 if (!proto) 781 proto = inet->inet_num; 782 else if (proto != inet->inet_num) 783 return -EINVAL; 784 785 if (proto > 255) 786 return -EINVAL; 787 788 daddr = &sin6->sin6_addr; 789 if (np->sndflow) { 790 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 791 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 792 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 793 if (!flowlabel) 794 return -EINVAL; 795 } 796 } 797 798 /* 799 * Otherwise it will be difficult to maintain 800 * sk->sk_dst_cache. 801 */ 802 if (sk->sk_state == TCP_ESTABLISHED && 803 ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) 804 daddr = &sk->sk_v6_daddr; 805 806 if (addr_len >= sizeof(struct sockaddr_in6) && 807 sin6->sin6_scope_id && 808 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) 809 fl6.flowi6_oif = sin6->sin6_scope_id; 810 } else { 811 if (sk->sk_state != TCP_ESTABLISHED) 812 return -EDESTADDRREQ; 813 814 proto = inet->inet_num; 815 daddr = &sk->sk_v6_daddr; 816 fl6.flowlabel = np->flow_label; 817 } 818 819 if (fl6.flowi6_oif == 0) 820 fl6.flowi6_oif = sk->sk_bound_dev_if; 821 822 if (msg->msg_controllen) { 823 opt = &opt_space; 824 memset(opt, 0, sizeof(struct ipv6_txoptions)); 825 opt->tot_len = sizeof(struct ipv6_txoptions); 826 827 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt, 828 &hlimit, &tclass, &dontfrag); 829 if (err < 0) { 830 fl6_sock_release(flowlabel); 831 return err; 832 } 833 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 834 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 835 if (!flowlabel) 836 return -EINVAL; 837 } 838 if (!(opt->opt_nflen|opt->opt_flen)) 839 opt = NULL; 840 } 841 if (!opt) { 842 opt = txopt_get(np); 843 opt_to_free = opt; 844 } 845 if (flowlabel) 846 opt = fl6_merge_options(&opt_space, flowlabel, opt); 847 opt = ipv6_fixup_options(&opt_space, opt); 848 849 fl6.flowi6_proto = proto; 850 rfv.msg = msg; 851 rfv.hlen = 0; 852 err = rawv6_probe_proto_opt(&rfv, &fl6); 853 if (err) 854 goto out; 855 856 if (!ipv6_addr_any(daddr)) 857 fl6.daddr = *daddr; 858 else 859 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 860 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 861 fl6.saddr = np->saddr; 862 863 final_p = fl6_update_dst(&fl6, opt, &final); 864 865 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 866 fl6.flowi6_oif = np->mcast_oif; 867 else if (!fl6.flowi6_oif) 868 fl6.flowi6_oif = np->ucast_oif; 869 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 870 871 dst = ip6_dst_lookup_flow(sk, &fl6, final_p); 872 if (IS_ERR(dst)) { 873 err = PTR_ERR(dst); 874 goto out; 875 } 876 if (hlimit < 0) 877 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 878 879 if (tclass < 0) 880 tclass = np->tclass; 881 882 if (dontfrag < 0) 883 dontfrag = np->dontfrag; 884 885 if (msg->msg_flags&MSG_CONFIRM) 886 goto do_confirm; 887 888back_from_confirm: 889 if (inet->hdrincl) 890 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, msg->msg_flags); 891 else { 892 lock_sock(sk); 893 err = ip6_append_data(sk, raw6_getfrag, &rfv, 894 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info *)dst, 895 msg->msg_flags, dontfrag); 896 897 if (err) 898 ip6_flush_pending_frames(sk); 899 else if (!(msg->msg_flags & MSG_MORE)) 900 err = rawv6_push_pending_frames(sk, &fl6, rp); 901 release_sock(sk); 902 } 903done: 904 dst_release(dst); 905out: 906 fl6_sock_release(flowlabel); 907 txopt_put(opt_to_free); 908 return err < 0 ? err : len; 909do_confirm: 910 dst_confirm(dst); 911 if (!(msg->msg_flags & MSG_PROBE) || len) 912 goto back_from_confirm; 913 err = 0; 914 goto done; 915} 916 917static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 918 char __user *optval, int optlen) 919{ 920 switch (optname) { 921 case ICMPV6_FILTER: 922 if (optlen > sizeof(struct icmp6_filter)) 923 optlen = sizeof(struct icmp6_filter); 924 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 925 return -EFAULT; 926 return 0; 927 default: 928 return -ENOPROTOOPT; 929 } 930 931 return 0; 932} 933 934static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 935 char __user *optval, int __user *optlen) 936{ 937 int len; 938 939 switch (optname) { 940 case ICMPV6_FILTER: 941 if (get_user(len, optlen)) 942 return -EFAULT; 943 if (len < 0) 944 return -EINVAL; 945 if (len > sizeof(struct icmp6_filter)) 946 len = sizeof(struct icmp6_filter); 947 if (put_user(len, optlen)) 948 return -EFAULT; 949 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 950 return -EFAULT; 951 return 0; 952 default: 953 return -ENOPROTOOPT; 954 } 955 956 return 0; 957} 958 959 960static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 961 char __user *optval, unsigned int optlen) 962{ 963 struct raw6_sock *rp = raw6_sk(sk); 964 int val; 965 966 if (get_user(val, (int __user *)optval)) 967 return -EFAULT; 968 969 switch (optname) { 970 case IPV6_CHECKSUM: 971 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && 972 level == IPPROTO_IPV6) { 973 /* 974 * RFC3542 tells that IPV6_CHECKSUM socket 975 * option in the IPPROTO_IPV6 level is not 976 * allowed on ICMPv6 sockets. 977 * If you want to set it, use IPPROTO_RAW 978 * level IPV6_CHECKSUM socket option 979 * (Linux extension). 980 */ 981 return -EINVAL; 982 } 983 984 /* You may get strange result with a positive odd offset; 985 RFC2292bis agrees with me. */ 986 if (val > 0 && (val&1)) 987 return -EINVAL; 988 if (val < 0) { 989 rp->checksum = 0; 990 } else { 991 rp->checksum = 1; 992 rp->offset = val; 993 } 994 995 return 0; 996 997 default: 998 return -ENOPROTOOPT; 999 } 1000} 1001 1002static int rawv6_setsockopt(struct sock *sk, int level, int optname, 1003 char __user *optval, unsigned int optlen) 1004{ 1005 switch (level) { 1006 case SOL_RAW: 1007 break; 1008 1009 case SOL_ICMPV6: 1010 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1011 return -EOPNOTSUPP; 1012 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1013 case SOL_IPV6: 1014 if (optname == IPV6_CHECKSUM) 1015 break; 1016 default: 1017 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1018 } 1019 1020 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1021} 1022 1023#ifdef CONFIG_COMPAT 1024static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 1025 char __user *optval, unsigned int optlen) 1026{ 1027 switch (level) { 1028 case SOL_RAW: 1029 break; 1030 case SOL_ICMPV6: 1031 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1032 return -EOPNOTSUPP; 1033 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1034 case SOL_IPV6: 1035 if (optname == IPV6_CHECKSUM) 1036 break; 1037 default: 1038 return compat_ipv6_setsockopt(sk, level, optname, 1039 optval, optlen); 1040 } 1041 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1042} 1043#endif 1044 1045static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1046 char __user *optval, int __user *optlen) 1047{ 1048 struct raw6_sock *rp = raw6_sk(sk); 1049 int val, len; 1050 1051 if (get_user(len, optlen)) 1052 return -EFAULT; 1053 1054 switch (optname) { 1055 case IPV6_CHECKSUM: 1056 /* 1057 * We allow getsockopt() for IPPROTO_IPV6-level 1058 * IPV6_CHECKSUM socket option on ICMPv6 sockets 1059 * since RFC3542 is silent about it. 1060 */ 1061 if (rp->checksum == 0) 1062 val = -1; 1063 else 1064 val = rp->offset; 1065 break; 1066 1067 default: 1068 return -ENOPROTOOPT; 1069 } 1070 1071 len = min_t(unsigned int, sizeof(int), len); 1072 1073 if (put_user(len, optlen)) 1074 return -EFAULT; 1075 if (copy_to_user(optval, &val, len)) 1076 return -EFAULT; 1077 return 0; 1078} 1079 1080static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1081 char __user *optval, int __user *optlen) 1082{ 1083 switch (level) { 1084 case SOL_RAW: 1085 break; 1086 1087 case SOL_ICMPV6: 1088 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1089 return -EOPNOTSUPP; 1090 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1091 case SOL_IPV6: 1092 if (optname == IPV6_CHECKSUM) 1093 break; 1094 default: 1095 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1096 } 1097 1098 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1099} 1100 1101#ifdef CONFIG_COMPAT 1102static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1103 char __user *optval, int __user *optlen) 1104{ 1105 switch (level) { 1106 case SOL_RAW: 1107 break; 1108 case SOL_ICMPV6: 1109 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1110 return -EOPNOTSUPP; 1111 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1112 case SOL_IPV6: 1113 if (optname == IPV6_CHECKSUM) 1114 break; 1115 default: 1116 return compat_ipv6_getsockopt(sk, level, optname, 1117 optval, optlen); 1118 } 1119 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1120} 1121#endif 1122 1123static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1124{ 1125 switch (cmd) { 1126 case SIOCOUTQ: { 1127 int amount = sk_wmem_alloc_get(sk); 1128 1129 return put_user(amount, (int __user *)arg); 1130 } 1131 case SIOCINQ: { 1132 struct sk_buff *skb; 1133 int amount = 0; 1134 1135 spin_lock_bh(&sk->sk_receive_queue.lock); 1136 skb = skb_peek(&sk->sk_receive_queue); 1137 if (skb) 1138 amount = skb_tail_pointer(skb) - 1139 skb_transport_header(skb); 1140 spin_unlock_bh(&sk->sk_receive_queue.lock); 1141 return put_user(amount, (int __user *)arg); 1142 } 1143 1144 default: 1145#ifdef CONFIG_IPV6_MROUTE 1146 return ip6mr_ioctl(sk, cmd, (void __user *)arg); 1147#else 1148 return -ENOIOCTLCMD; 1149#endif 1150 } 1151} 1152 1153#ifdef CONFIG_COMPAT 1154static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 1155{ 1156 switch (cmd) { 1157 case SIOCOUTQ: 1158 case SIOCINQ: 1159 return -ENOIOCTLCMD; 1160 default: 1161#ifdef CONFIG_IPV6_MROUTE 1162 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg)); 1163#else 1164 return -ENOIOCTLCMD; 1165#endif 1166 } 1167} 1168#endif 1169 1170static void rawv6_close(struct sock *sk, long timeout) 1171{ 1172 if (inet_sk(sk)->inet_num == IPPROTO_RAW) 1173 ip6_ra_control(sk, -1); 1174 ip6mr_sk_done(sk); 1175 sk_common_release(sk); 1176} 1177 1178static void raw6_destroy(struct sock *sk) 1179{ 1180 lock_sock(sk); 1181 ip6_flush_pending_frames(sk); 1182 release_sock(sk); 1183 1184 inet6_destroy_sock(sk); 1185} 1186 1187static int rawv6_init_sk(struct sock *sk) 1188{ 1189 struct raw6_sock *rp = raw6_sk(sk); 1190 1191 switch (inet_sk(sk)->inet_num) { 1192 case IPPROTO_ICMPV6: 1193 rp->checksum = 1; 1194 rp->offset = 2; 1195 break; 1196 case IPPROTO_MH: 1197 rp->checksum = 1; 1198 rp->offset = 4; 1199 break; 1200 default: 1201 break; 1202 } 1203 return 0; 1204} 1205 1206struct proto rawv6_prot = { 1207 .name = "RAWv6", 1208 .owner = THIS_MODULE, 1209 .close = rawv6_close, 1210 .destroy = raw6_destroy, 1211 .connect = ip6_datagram_connect_v6_only, 1212 .disconnect = udp_disconnect, 1213 .ioctl = rawv6_ioctl, 1214 .init = rawv6_init_sk, 1215 .setsockopt = rawv6_setsockopt, 1216 .getsockopt = rawv6_getsockopt, 1217 .sendmsg = rawv6_sendmsg, 1218 .recvmsg = rawv6_recvmsg, 1219 .bind = rawv6_bind, 1220 .backlog_rcv = rawv6_rcv_skb, 1221 .hash = raw_hash_sk, 1222 .unhash = raw_unhash_sk, 1223 .obj_size = sizeof(struct raw6_sock), 1224 .h.raw_hash = &raw_v6_hashinfo, 1225#ifdef CONFIG_COMPAT 1226 .compat_setsockopt = compat_rawv6_setsockopt, 1227 .compat_getsockopt = compat_rawv6_getsockopt, 1228 .compat_ioctl = compat_rawv6_ioctl, 1229#endif 1230}; 1231 1232#ifdef CONFIG_PROC_FS 1233static int raw6_seq_show(struct seq_file *seq, void *v) 1234{ 1235 if (v == SEQ_START_TOKEN) { 1236 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); 1237 } else { 1238 struct sock *sp = v; 1239 __u16 srcp = inet_sk(sp)->inet_num; 1240 ip6_dgram_sock_seq_show(seq, v, srcp, 0, 1241 raw_seq_private(seq)->bucket); 1242 } 1243 return 0; 1244} 1245 1246static const struct seq_operations raw6_seq_ops = { 1247 .start = raw_seq_start, 1248 .next = raw_seq_next, 1249 .stop = raw_seq_stop, 1250 .show = raw6_seq_show, 1251}; 1252 1253static int raw6_seq_open(struct inode *inode, struct file *file) 1254{ 1255 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops); 1256} 1257 1258static const struct file_operations raw6_seq_fops = { 1259 .owner = THIS_MODULE, 1260 .open = raw6_seq_open, 1261 .read = seq_read, 1262 .llseek = seq_lseek, 1263 .release = seq_release_net, 1264}; 1265 1266static int __net_init raw6_init_net(struct net *net) 1267{ 1268 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops)) 1269 return -ENOMEM; 1270 1271 return 0; 1272} 1273 1274static void __net_exit raw6_exit_net(struct net *net) 1275{ 1276 remove_proc_entry("raw6", net->proc_net); 1277} 1278 1279static struct pernet_operations raw6_net_ops = { 1280 .init = raw6_init_net, 1281 .exit = raw6_exit_net, 1282}; 1283 1284int __init raw6_proc_init(void) 1285{ 1286 return register_pernet_subsys(&raw6_net_ops); 1287} 1288 1289void raw6_proc_exit(void) 1290{ 1291 unregister_pernet_subsys(&raw6_net_ops); 1292} 1293#endif /* CONFIG_PROC_FS */ 1294 1295/* Same as inet6_dgram_ops, sans udp_poll. */ 1296static const struct proto_ops inet6_sockraw_ops = { 1297 .family = PF_INET6, 1298 .owner = THIS_MODULE, 1299 .release = inet6_release, 1300 .bind = inet6_bind, 1301 .connect = inet_dgram_connect, /* ok */ 1302 .socketpair = sock_no_socketpair, /* a do nothing */ 1303 .accept = sock_no_accept, /* a do nothing */ 1304 .getname = inet6_getname, 1305 .poll = datagram_poll, /* ok */ 1306 .ioctl = inet6_ioctl, /* must change */ 1307 .listen = sock_no_listen, /* ok */ 1308 .shutdown = inet_shutdown, /* ok */ 1309 .setsockopt = sock_common_setsockopt, /* ok */ 1310 .getsockopt = sock_common_getsockopt, /* ok */ 1311 .sendmsg = inet_sendmsg, /* ok */ 1312 .recvmsg = sock_common_recvmsg, /* ok */ 1313 .mmap = sock_no_mmap, 1314 .sendpage = sock_no_sendpage, 1315#ifdef CONFIG_COMPAT 1316 .compat_setsockopt = compat_sock_common_setsockopt, 1317 .compat_getsockopt = compat_sock_common_getsockopt, 1318#endif 1319}; 1320 1321static struct inet_protosw rawv6_protosw = { 1322 .type = SOCK_RAW, 1323 .protocol = IPPROTO_IP, /* wild card */ 1324 .prot = &rawv6_prot, 1325 .ops = &inet6_sockraw_ops, 1326 .flags = INET_PROTOSW_REUSE, 1327}; 1328 1329int __init rawv6_init(void) 1330{ 1331 int ret; 1332 1333 ret = inet6_register_protosw(&rawv6_protosw); 1334 if (ret) 1335 goto out; 1336out: 1337 return ret; 1338} 1339 1340void rawv6_exit(void) 1341{ 1342 inet6_unregister_protosw(&rawv6_protosw); 1343} 1344