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 68 struct raw_hashinfo raw_v6_hashinfo = { 69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock), 70 }; 71 EXPORT_SYMBOL_GPL(raw_v6_hashinfo); 72 73 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk, 74 unsigned short num, const struct in6_addr *loc_addr, 75 const struct in6_addr *rmt_addr, int dif) 76 { 77 bool is_multicast = ipv6_addr_is_multicast(loc_addr); 78 79 sk_for_each_from(sk) 80 if (inet_sk(sk)->inet_num == num) { 81 82 if (!net_eq(sock_net(sk), net)) 83 continue; 84 85 if (!ipv6_addr_any(&sk->sk_v6_daddr) && 86 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) 87 continue; 88 89 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) 90 continue; 91 92 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) { 93 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)) 94 goto found; 95 if (is_multicast && 96 inet6_mc_check(sk, loc_addr, rmt_addr)) 97 goto found; 98 continue; 99 } 100 goto found; 101 } 102 sk = NULL; 103 found: 104 return sk; 105 } 106 EXPORT_SYMBOL_GPL(__raw_v6_lookup); 107 108 /* 109 * 0 - deliver 110 * 1 - block 111 */ 112 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb) 113 { 114 struct icmp6hdr _hdr; 115 const struct icmp6hdr *hdr; 116 117 /* We require only the four bytes of the ICMPv6 header, not any 118 * additional bytes of message body in "struct icmp6hdr". 119 */ 120 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 121 ICMPV6_HDRLEN, &_hdr); 122 if (hdr) { 123 const __u32 *data = &raw6_sk(sk)->filter.data[0]; 124 unsigned int type = hdr->icmp6_type; 125 126 return (data[type >> 5] & (1U << (type & 31))) != 0; 127 } 128 return 1; 129 } 130 131 #if IS_ENABLED(CONFIG_IPV6_MIP6) 132 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb); 133 134 static mh_filter_t __rcu *mh_filter __read_mostly; 135 136 int rawv6_mh_filter_register(mh_filter_t filter) 137 { 138 rcu_assign_pointer(mh_filter, filter); 139 return 0; 140 } 141 EXPORT_SYMBOL(rawv6_mh_filter_register); 142 143 int rawv6_mh_filter_unregister(mh_filter_t filter) 144 { 145 RCU_INIT_POINTER(mh_filter, NULL); 146 synchronize_rcu(); 147 return 0; 148 } 149 EXPORT_SYMBOL(rawv6_mh_filter_unregister); 150 151 #endif 152 153 /* 154 * demultiplex raw sockets. 155 * (should consider queueing the skb in the sock receive_queue 156 * without calling rawv6.c) 157 * 158 * Caller owns SKB so we must make clones. 159 */ 160 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 161 { 162 const struct in6_addr *saddr; 163 const struct in6_addr *daddr; 164 struct sock *sk; 165 bool delivered = false; 166 __u8 hash; 167 struct net *net; 168 169 saddr = &ipv6_hdr(skb)->saddr; 170 daddr = saddr + 1; 171 172 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 173 174 read_lock(&raw_v6_hashinfo.lock); 175 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 176 177 if (!sk) 178 goto out; 179 180 net = dev_net(skb->dev); 181 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, inet6_iif(skb)); 182 183 while (sk) { 184 int filtered; 185 186 delivered = true; 187 switch (nexthdr) { 188 case IPPROTO_ICMPV6: 189 filtered = icmpv6_filter(sk, skb); 190 break; 191 192 #if IS_ENABLED(CONFIG_IPV6_MIP6) 193 case IPPROTO_MH: 194 { 195 /* XXX: To validate MH only once for each packet, 196 * this is placed here. It should be after checking 197 * xfrm policy, however it doesn't. The checking xfrm 198 * policy is placed in rawv6_rcv() because it is 199 * required for each socket. 200 */ 201 mh_filter_t *filter; 202 203 filter = rcu_dereference(mh_filter); 204 filtered = filter ? (*filter)(sk, skb) : 0; 205 break; 206 } 207 #endif 208 default: 209 filtered = 0; 210 break; 211 } 212 213 if (filtered < 0) 214 break; 215 if (filtered == 0) { 216 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 217 218 /* Not releasing hash table! */ 219 if (clone) { 220 nf_reset(clone); 221 rawv6_rcv(sk, clone); 222 } 223 } 224 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr, 225 inet6_iif(skb)); 226 } 227 out: 228 read_unlock(&raw_v6_hashinfo.lock); 229 return delivered; 230 } 231 232 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr) 233 { 234 struct sock *raw_sk; 235 236 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]); 237 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr)) 238 raw_sk = NULL; 239 240 return raw_sk != NULL; 241 } 242 243 /* This cleans up af_inet6 a bit. -DaveM */ 244 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 245 { 246 struct inet_sock *inet = inet_sk(sk); 247 struct ipv6_pinfo *np = inet6_sk(sk); 248 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 249 __be32 v4addr = 0; 250 int addr_type; 251 int err; 252 253 if (addr_len < SIN6_LEN_RFC2133) 254 return -EINVAL; 255 256 if (addr->sin6_family != AF_INET6) 257 return -EINVAL; 258 259 addr_type = ipv6_addr_type(&addr->sin6_addr); 260 261 /* Raw sockets are IPv6 only */ 262 if (addr_type == IPV6_ADDR_MAPPED) 263 return -EADDRNOTAVAIL; 264 265 lock_sock(sk); 266 267 err = -EINVAL; 268 if (sk->sk_state != TCP_CLOSE) 269 goto out; 270 271 rcu_read_lock(); 272 /* Check if the address belongs to the host. */ 273 if (addr_type != IPV6_ADDR_ANY) { 274 struct net_device *dev = NULL; 275 276 if (__ipv6_addr_needs_scope_id(addr_type)) { 277 if (addr_len >= sizeof(struct sockaddr_in6) && 278 addr->sin6_scope_id) { 279 /* Override any existing binding, if another 280 * one is supplied by user. 281 */ 282 sk->sk_bound_dev_if = addr->sin6_scope_id; 283 } 284 285 /* Binding to link-local address requires an interface */ 286 if (!sk->sk_bound_dev_if) 287 goto out_unlock; 288 289 err = -ENODEV; 290 dev = dev_get_by_index_rcu(sock_net(sk), 291 sk->sk_bound_dev_if); 292 if (!dev) 293 goto out_unlock; 294 } 295 296 /* ipv4 addr of the socket is invalid. Only the 297 * unspecified and mapped address have a v4 equivalent. 298 */ 299 v4addr = LOOPBACK4_IPV6; 300 if (!(addr_type & IPV6_ADDR_MULTICAST) && 301 !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) { 302 err = -EADDRNOTAVAIL; 303 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr, 304 dev, 0)) { 305 goto out_unlock; 306 } 307 } 308 } 309 310 inet->inet_rcv_saddr = inet->inet_saddr = v4addr; 311 sk->sk_v6_rcv_saddr = addr->sin6_addr; 312 if (!(addr_type & IPV6_ADDR_MULTICAST)) 313 np->saddr = addr->sin6_addr; 314 err = 0; 315 out_unlock: 316 rcu_read_unlock(); 317 out: 318 release_sock(sk); 319 return err; 320 } 321 322 static void rawv6_err(struct sock *sk, struct sk_buff *skb, 323 struct inet6_skb_parm *opt, 324 u8 type, u8 code, int offset, __be32 info) 325 { 326 struct inet_sock *inet = inet_sk(sk); 327 struct ipv6_pinfo *np = inet6_sk(sk); 328 int err; 329 int harderr; 330 331 /* Report error on raw socket, if: 332 1. User requested recverr. 333 2. Socket is connected (otherwise the error indication 334 is useless without recverr and error is hard. 335 */ 336 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 337 return; 338 339 harderr = icmpv6_err_convert(type, code, &err); 340 if (type == ICMPV6_PKT_TOOBIG) { 341 ip6_sk_update_pmtu(skb, sk, info); 342 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 343 } 344 if (type == NDISC_REDIRECT) { 345 ip6_sk_redirect(skb, sk); 346 return; 347 } 348 if (np->recverr) { 349 u8 *payload = skb->data; 350 if (!inet->hdrincl) 351 payload += offset; 352 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 353 } 354 355 if (np->recverr || harderr) { 356 sk->sk_err = err; 357 sk->sk_error_report(sk); 358 } 359 } 360 361 void raw6_icmp_error(struct sk_buff *skb, int nexthdr, 362 u8 type, u8 code, int inner_offset, __be32 info) 363 { 364 struct sock *sk; 365 int hash; 366 const struct in6_addr *saddr, *daddr; 367 struct net *net; 368 369 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 370 371 read_lock(&raw_v6_hashinfo.lock); 372 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 373 if (sk) { 374 /* Note: ipv6_hdr(skb) != skb->data */ 375 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data; 376 saddr = &ip6h->saddr; 377 daddr = &ip6h->daddr; 378 net = dev_net(skb->dev); 379 380 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr, 381 inet6_iif(skb)))) { 382 rawv6_err(sk, skb, NULL, type, code, 383 inner_offset, info); 384 sk = sk_next(sk); 385 } 386 } 387 read_unlock(&raw_v6_hashinfo.lock); 388 } 389 390 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb) 391 { 392 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) && 393 skb_checksum_complete(skb)) { 394 atomic_inc(&sk->sk_drops); 395 kfree_skb(skb); 396 return NET_RX_DROP; 397 } 398 399 /* Charge it to the socket. */ 400 skb_dst_drop(skb); 401 if (sock_queue_rcv_skb(sk, skb) < 0) { 402 kfree_skb(skb); 403 return NET_RX_DROP; 404 } 405 406 return 0; 407 } 408 409 /* 410 * This is next to useless... 411 * if we demultiplex in network layer we don't need the extra call 412 * just to queue the skb... 413 * maybe we could have the network decide upon a hint if it 414 * should call raw_rcv for demultiplexing 415 */ 416 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 417 { 418 struct inet_sock *inet = inet_sk(sk); 419 struct raw6_sock *rp = raw6_sk(sk); 420 421 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 422 atomic_inc(&sk->sk_drops); 423 kfree_skb(skb); 424 return NET_RX_DROP; 425 } 426 427 if (!rp->checksum) 428 skb->ip_summed = CHECKSUM_UNNECESSARY; 429 430 if (skb->ip_summed == CHECKSUM_COMPLETE) { 431 skb_postpull_rcsum(skb, skb_network_header(skb), 432 skb_network_header_len(skb)); 433 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 434 &ipv6_hdr(skb)->daddr, 435 skb->len, inet->inet_num, skb->csum)) 436 skb->ip_summed = CHECKSUM_UNNECESSARY; 437 } 438 if (!skb_csum_unnecessary(skb)) 439 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 440 &ipv6_hdr(skb)->daddr, 441 skb->len, 442 inet->inet_num, 0)); 443 444 if (inet->hdrincl) { 445 if (skb_checksum_complete(skb)) { 446 atomic_inc(&sk->sk_drops); 447 kfree_skb(skb); 448 return NET_RX_DROP; 449 } 450 } 451 452 rawv6_rcv_skb(sk, skb); 453 return 0; 454 } 455 456 457 /* 458 * This should be easy, if there is something there 459 * we return it, otherwise we block. 460 */ 461 462 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 463 int noblock, int flags, int *addr_len) 464 { 465 struct ipv6_pinfo *np = inet6_sk(sk); 466 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 467 struct sk_buff *skb; 468 size_t copied; 469 int err; 470 471 if (flags & MSG_OOB) 472 return -EOPNOTSUPP; 473 474 if (flags & MSG_ERRQUEUE) 475 return ipv6_recv_error(sk, msg, len, addr_len); 476 477 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 478 return ipv6_recv_rxpmtu(sk, msg, len, addr_len); 479 480 skb = skb_recv_datagram(sk, flags, noblock, &err); 481 if (!skb) 482 goto out; 483 484 copied = skb->len; 485 if (copied > len) { 486 copied = len; 487 msg->msg_flags |= MSG_TRUNC; 488 } 489 490 if (skb_csum_unnecessary(skb)) { 491 err = skb_copy_datagram_msg(skb, 0, msg, copied); 492 } else if (msg->msg_flags&MSG_TRUNC) { 493 if (__skb_checksum_complete(skb)) 494 goto csum_copy_err; 495 err = skb_copy_datagram_msg(skb, 0, msg, copied); 496 } else { 497 err = skb_copy_and_csum_datagram_msg(skb, 0, msg); 498 if (err == -EINVAL) 499 goto csum_copy_err; 500 } 501 if (err) 502 goto out_free; 503 504 /* Copy the address. */ 505 if (sin6) { 506 sin6->sin6_family = AF_INET6; 507 sin6->sin6_port = 0; 508 sin6->sin6_addr = ipv6_hdr(skb)->saddr; 509 sin6->sin6_flowinfo = 0; 510 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr, 511 inet6_iif(skb)); 512 *addr_len = sizeof(*sin6); 513 } 514 515 sock_recv_ts_and_drops(msg, sk, skb); 516 517 if (np->rxopt.all) 518 ip6_datagram_recv_ctl(sk, msg, skb); 519 520 err = copied; 521 if (flags & MSG_TRUNC) 522 err = skb->len; 523 524 out_free: 525 skb_free_datagram(sk, skb); 526 out: 527 return err; 528 529 csum_copy_err: 530 skb_kill_datagram(sk, skb, flags); 531 532 /* Error for blocking case is chosen to masquerade 533 as some normal condition. 534 */ 535 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 536 goto out; 537 } 538 539 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 540 struct raw6_sock *rp) 541 { 542 struct sk_buff *skb; 543 int err = 0; 544 int offset; 545 int len; 546 int total_len; 547 __wsum tmp_csum; 548 __sum16 csum; 549 550 if (!rp->checksum) 551 goto send; 552 553 skb = skb_peek(&sk->sk_write_queue); 554 if (!skb) 555 goto out; 556 557 offset = rp->offset; 558 total_len = inet_sk(sk)->cork.base.length; 559 if (offset >= total_len - 1) { 560 err = -EINVAL; 561 ip6_flush_pending_frames(sk); 562 goto out; 563 } 564 565 /* should be check HW csum miyazawa */ 566 if (skb_queue_len(&sk->sk_write_queue) == 1) { 567 /* 568 * Only one fragment on the socket. 569 */ 570 tmp_csum = skb->csum; 571 } else { 572 struct sk_buff *csum_skb = NULL; 573 tmp_csum = 0; 574 575 skb_queue_walk(&sk->sk_write_queue, skb) { 576 tmp_csum = csum_add(tmp_csum, skb->csum); 577 578 if (csum_skb) 579 continue; 580 581 len = skb->len - skb_transport_offset(skb); 582 if (offset >= len) { 583 offset -= len; 584 continue; 585 } 586 587 csum_skb = skb; 588 } 589 590 skb = csum_skb; 591 } 592 593 offset += skb_transport_offset(skb); 594 BUG_ON(skb_copy_bits(skb, offset, &csum, 2)); 595 596 /* in case cksum was not initialized */ 597 if (unlikely(csum)) 598 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 599 600 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 601 total_len, fl6->flowi6_proto, tmp_csum); 602 603 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP) 604 csum = CSUM_MANGLED_0; 605 606 BUG_ON(skb_store_bits(skb, offset, &csum, 2)); 607 608 send: 609 err = ip6_push_pending_frames(sk); 610 out: 611 return err; 612 } 613 614 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length, 615 struct flowi6 *fl6, struct dst_entry **dstp, 616 unsigned int flags) 617 { 618 struct ipv6_pinfo *np = inet6_sk(sk); 619 struct net *net = sock_net(sk); 620 struct ipv6hdr *iph; 621 struct sk_buff *skb; 622 int err; 623 struct rt6_info *rt = (struct rt6_info *)*dstp; 624 int hlen = LL_RESERVED_SPACE(rt->dst.dev); 625 int tlen = rt->dst.dev->needed_tailroom; 626 627 if (length > rt->dst.dev->mtu) { 628 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu); 629 return -EMSGSIZE; 630 } 631 if (flags&MSG_PROBE) 632 goto out; 633 634 skb = sock_alloc_send_skb(sk, 635 length + hlen + tlen + 15, 636 flags & MSG_DONTWAIT, &err); 637 if (!skb) 638 goto error; 639 skb_reserve(skb, hlen); 640 641 skb->protocol = htons(ETH_P_IPV6); 642 skb->priority = sk->sk_priority; 643 skb->mark = sk->sk_mark; 644 skb_dst_set(skb, &rt->dst); 645 *dstp = NULL; 646 647 skb_put(skb, length); 648 skb_reset_network_header(skb); 649 iph = ipv6_hdr(skb); 650 651 skb->ip_summed = CHECKSUM_NONE; 652 653 skb->transport_header = skb->network_header; 654 err = memcpy_from_msg(iph, msg, length); 655 if (err) 656 goto error_fault; 657 658 /* if egress device is enslaved to an L3 master device pass the 659 * skb to its handler for processing 660 */ 661 skb = l3mdev_ip6_out(sk, skb); 662 if (unlikely(!skb)) 663 return 0; 664 665 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); 666 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb, 667 NULL, rt->dst.dev, dst_output); 668 if (err > 0) 669 err = net_xmit_errno(err); 670 if (err) 671 goto error; 672 out: 673 return 0; 674 675 error_fault: 676 err = -EFAULT; 677 kfree_skb(skb); 678 error: 679 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 680 if (err == -ENOBUFS && !np->recverr) 681 err = 0; 682 return err; 683 } 684 685 struct raw6_frag_vec { 686 struct msghdr *msg; 687 int hlen; 688 char c[4]; 689 }; 690 691 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6) 692 { 693 int err = 0; 694 switch (fl6->flowi6_proto) { 695 case IPPROTO_ICMPV6: 696 rfv->hlen = 2; 697 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); 698 if (!err) { 699 fl6->fl6_icmp_type = rfv->c[0]; 700 fl6->fl6_icmp_code = rfv->c[1]; 701 } 702 break; 703 case IPPROTO_MH: 704 rfv->hlen = 4; 705 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); 706 if (!err) 707 fl6->fl6_mh_type = rfv->c[2]; 708 } 709 return err; 710 } 711 712 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd, 713 struct sk_buff *skb) 714 { 715 struct raw6_frag_vec *rfv = from; 716 717 if (offset < rfv->hlen) { 718 int copy = min(rfv->hlen - offset, len); 719 720 if (skb->ip_summed == CHECKSUM_PARTIAL) 721 memcpy(to, rfv->c + offset, copy); 722 else 723 skb->csum = csum_block_add( 724 skb->csum, 725 csum_partial_copy_nocheck(rfv->c + offset, 726 to, copy, 0), 727 odd); 728 729 odd = 0; 730 offset += copy; 731 to += copy; 732 len -= copy; 733 734 if (!len) 735 return 0; 736 } 737 738 offset -= rfv->hlen; 739 740 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); 741 } 742 743 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 744 { 745 struct ipv6_txoptions *opt_to_free = NULL; 746 struct ipv6_txoptions opt_space; 747 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 748 struct in6_addr *daddr, *final_p, final; 749 struct inet_sock *inet = inet_sk(sk); 750 struct ipv6_pinfo *np = inet6_sk(sk); 751 struct raw6_sock *rp = raw6_sk(sk); 752 struct ipv6_txoptions *opt = NULL; 753 struct ip6_flowlabel *flowlabel = NULL; 754 struct dst_entry *dst = NULL; 755 struct raw6_frag_vec rfv; 756 struct flowi6 fl6; 757 struct sockcm_cookie sockc; 758 struct ipcm6_cookie ipc6; 759 int addr_len = msg->msg_namelen; 760 u16 proto; 761 int err; 762 763 /* Rough check on arithmetic overflow, 764 better check is made in ip6_append_data(). 765 */ 766 if (len > INT_MAX) 767 return -EMSGSIZE; 768 769 /* Mirror BSD error message compatibility */ 770 if (msg->msg_flags & MSG_OOB) 771 return -EOPNOTSUPP; 772 773 /* 774 * Get and verify the address. 775 */ 776 memset(&fl6, 0, sizeof(fl6)); 777 778 fl6.flowi6_mark = sk->sk_mark; 779 fl6.flowi6_uid = sk->sk_uid; 780 781 ipc6.hlimit = -1; 782 ipc6.tclass = -1; 783 ipc6.dontfrag = -1; 784 ipc6.opt = NULL; 785 786 if (sin6) { 787 if (addr_len < SIN6_LEN_RFC2133) 788 return -EINVAL; 789 790 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 791 return -EAFNOSUPPORT; 792 793 /* port is the proto value [0..255] carried in nexthdr */ 794 proto = ntohs(sin6->sin6_port); 795 796 if (!proto) 797 proto = inet->inet_num; 798 else if (proto != inet->inet_num) 799 return -EINVAL; 800 801 if (proto > 255) 802 return -EINVAL; 803 804 daddr = &sin6->sin6_addr; 805 if (np->sndflow) { 806 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 807 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 808 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 809 if (!flowlabel) 810 return -EINVAL; 811 } 812 } 813 814 /* 815 * Otherwise it will be difficult to maintain 816 * sk->sk_dst_cache. 817 */ 818 if (sk->sk_state == TCP_ESTABLISHED && 819 ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) 820 daddr = &sk->sk_v6_daddr; 821 822 if (addr_len >= sizeof(struct sockaddr_in6) && 823 sin6->sin6_scope_id && 824 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) 825 fl6.flowi6_oif = sin6->sin6_scope_id; 826 } else { 827 if (sk->sk_state != TCP_ESTABLISHED) 828 return -EDESTADDRREQ; 829 830 proto = inet->inet_num; 831 daddr = &sk->sk_v6_daddr; 832 fl6.flowlabel = np->flow_label; 833 } 834 835 if (fl6.flowi6_oif == 0) 836 fl6.flowi6_oif = sk->sk_bound_dev_if; 837 838 sockc.tsflags = sk->sk_tsflags; 839 if (msg->msg_controllen) { 840 opt = &opt_space; 841 memset(opt, 0, sizeof(struct ipv6_txoptions)); 842 opt->tot_len = sizeof(struct ipv6_txoptions); 843 ipc6.opt = opt; 844 845 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc); 846 if (err < 0) { 847 fl6_sock_release(flowlabel); 848 return err; 849 } 850 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 851 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 852 if (!flowlabel) 853 return -EINVAL; 854 } 855 if (!(opt->opt_nflen|opt->opt_flen)) 856 opt = NULL; 857 } 858 if (!opt) { 859 opt = txopt_get(np); 860 opt_to_free = opt; 861 } 862 if (flowlabel) 863 opt = fl6_merge_options(&opt_space, flowlabel, opt); 864 opt = ipv6_fixup_options(&opt_space, opt); 865 866 fl6.flowi6_proto = proto; 867 rfv.msg = msg; 868 rfv.hlen = 0; 869 err = rawv6_probe_proto_opt(&rfv, &fl6); 870 if (err) 871 goto out; 872 873 if (!ipv6_addr_any(daddr)) 874 fl6.daddr = *daddr; 875 else 876 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 877 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 878 fl6.saddr = np->saddr; 879 880 final_p = fl6_update_dst(&fl6, opt, &final); 881 882 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 883 fl6.flowi6_oif = np->mcast_oif; 884 else if (!fl6.flowi6_oif) 885 fl6.flowi6_oif = np->ucast_oif; 886 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 887 888 if (inet->hdrincl) 889 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH; 890 891 if (ipc6.tclass < 0) 892 ipc6.tclass = np->tclass; 893 894 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); 895 896 dst = ip6_dst_lookup_flow(sk, &fl6, final_p); 897 if (IS_ERR(dst)) { 898 err = PTR_ERR(dst); 899 goto out; 900 } 901 if (ipc6.hlimit < 0) 902 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 903 904 if (ipc6.dontfrag < 0) 905 ipc6.dontfrag = np->dontfrag; 906 907 if (msg->msg_flags&MSG_CONFIRM) 908 goto do_confirm; 909 910 back_from_confirm: 911 if (inet->hdrincl) 912 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, msg->msg_flags); 913 else { 914 ipc6.opt = opt; 915 lock_sock(sk); 916 err = ip6_append_data(sk, raw6_getfrag, &rfv, 917 len, 0, &ipc6, &fl6, (struct rt6_info *)dst, 918 msg->msg_flags, &sockc); 919 920 if (err) 921 ip6_flush_pending_frames(sk); 922 else if (!(msg->msg_flags & MSG_MORE)) 923 err = rawv6_push_pending_frames(sk, &fl6, rp); 924 release_sock(sk); 925 } 926 done: 927 dst_release(dst); 928 out: 929 fl6_sock_release(flowlabel); 930 txopt_put(opt_to_free); 931 return err < 0 ? err : len; 932 do_confirm: 933 dst_confirm(dst); 934 if (!(msg->msg_flags & MSG_PROBE) || len) 935 goto back_from_confirm; 936 err = 0; 937 goto done; 938 } 939 940 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 941 char __user *optval, int optlen) 942 { 943 switch (optname) { 944 case ICMPV6_FILTER: 945 if (optlen > sizeof(struct icmp6_filter)) 946 optlen = sizeof(struct icmp6_filter); 947 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 948 return -EFAULT; 949 return 0; 950 default: 951 return -ENOPROTOOPT; 952 } 953 954 return 0; 955 } 956 957 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 958 char __user *optval, int __user *optlen) 959 { 960 int len; 961 962 switch (optname) { 963 case ICMPV6_FILTER: 964 if (get_user(len, optlen)) 965 return -EFAULT; 966 if (len < 0) 967 return -EINVAL; 968 if (len > sizeof(struct icmp6_filter)) 969 len = sizeof(struct icmp6_filter); 970 if (put_user(len, optlen)) 971 return -EFAULT; 972 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 973 return -EFAULT; 974 return 0; 975 default: 976 return -ENOPROTOOPT; 977 } 978 979 return 0; 980 } 981 982 983 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 984 char __user *optval, unsigned int optlen) 985 { 986 struct raw6_sock *rp = raw6_sk(sk); 987 int val; 988 989 if (get_user(val, (int __user *)optval)) 990 return -EFAULT; 991 992 switch (optname) { 993 case IPV6_HDRINCL: 994 if (sk->sk_type != SOCK_RAW) 995 return -EINVAL; 996 inet_sk(sk)->hdrincl = !!val; 997 return 0; 998 case IPV6_CHECKSUM: 999 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && 1000 level == IPPROTO_IPV6) { 1001 /* 1002 * RFC3542 tells that IPV6_CHECKSUM socket 1003 * option in the IPPROTO_IPV6 level is not 1004 * allowed on ICMPv6 sockets. 1005 * If you want to set it, use IPPROTO_RAW 1006 * level IPV6_CHECKSUM socket option 1007 * (Linux extension). 1008 */ 1009 return -EINVAL; 1010 } 1011 1012 /* You may get strange result with a positive odd offset; 1013 RFC2292bis agrees with me. */ 1014 if (val > 0 && (val&1)) 1015 return -EINVAL; 1016 if (val < 0) { 1017 rp->checksum = 0; 1018 } else { 1019 rp->checksum = 1; 1020 rp->offset = val; 1021 } 1022 1023 return 0; 1024 1025 default: 1026 return -ENOPROTOOPT; 1027 } 1028 } 1029 1030 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 1031 char __user *optval, unsigned int optlen) 1032 { 1033 switch (level) { 1034 case SOL_RAW: 1035 break; 1036 1037 case SOL_ICMPV6: 1038 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1039 return -EOPNOTSUPP; 1040 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1041 case SOL_IPV6: 1042 if (optname == IPV6_CHECKSUM || 1043 optname == IPV6_HDRINCL) 1044 break; 1045 default: 1046 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1047 } 1048 1049 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1050 } 1051 1052 #ifdef CONFIG_COMPAT 1053 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 1054 char __user *optval, unsigned int optlen) 1055 { 1056 switch (level) { 1057 case SOL_RAW: 1058 break; 1059 case SOL_ICMPV6: 1060 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1061 return -EOPNOTSUPP; 1062 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1063 case SOL_IPV6: 1064 if (optname == IPV6_CHECKSUM || 1065 optname == IPV6_HDRINCL) 1066 break; 1067 default: 1068 return compat_ipv6_setsockopt(sk, level, optname, 1069 optval, optlen); 1070 } 1071 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1072 } 1073 #endif 1074 1075 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1076 char __user *optval, int __user *optlen) 1077 { 1078 struct raw6_sock *rp = raw6_sk(sk); 1079 int val, len; 1080 1081 if (get_user(len, optlen)) 1082 return -EFAULT; 1083 1084 switch (optname) { 1085 case IPV6_HDRINCL: 1086 val = inet_sk(sk)->hdrincl; 1087 break; 1088 case IPV6_CHECKSUM: 1089 /* 1090 * We allow getsockopt() for IPPROTO_IPV6-level 1091 * IPV6_CHECKSUM socket option on ICMPv6 sockets 1092 * since RFC3542 is silent about it. 1093 */ 1094 if (rp->checksum == 0) 1095 val = -1; 1096 else 1097 val = rp->offset; 1098 break; 1099 1100 default: 1101 return -ENOPROTOOPT; 1102 } 1103 1104 len = min_t(unsigned int, sizeof(int), len); 1105 1106 if (put_user(len, optlen)) 1107 return -EFAULT; 1108 if (copy_to_user(optval, &val, len)) 1109 return -EFAULT; 1110 return 0; 1111 } 1112 1113 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1114 char __user *optval, int __user *optlen) 1115 { 1116 switch (level) { 1117 case SOL_RAW: 1118 break; 1119 1120 case SOL_ICMPV6: 1121 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1122 return -EOPNOTSUPP; 1123 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1124 case SOL_IPV6: 1125 if (optname == IPV6_CHECKSUM || 1126 optname == IPV6_HDRINCL) 1127 break; 1128 default: 1129 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1130 } 1131 1132 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1133 } 1134 1135 #ifdef CONFIG_COMPAT 1136 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1137 char __user *optval, int __user *optlen) 1138 { 1139 switch (level) { 1140 case SOL_RAW: 1141 break; 1142 case SOL_ICMPV6: 1143 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1144 return -EOPNOTSUPP; 1145 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1146 case SOL_IPV6: 1147 if (optname == IPV6_CHECKSUM || 1148 optname == IPV6_HDRINCL) 1149 break; 1150 default: 1151 return compat_ipv6_getsockopt(sk, level, optname, 1152 optval, optlen); 1153 } 1154 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1155 } 1156 #endif 1157 1158 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1159 { 1160 switch (cmd) { 1161 case SIOCOUTQ: { 1162 int amount = sk_wmem_alloc_get(sk); 1163 1164 return put_user(amount, (int __user *)arg); 1165 } 1166 case SIOCINQ: { 1167 struct sk_buff *skb; 1168 int amount = 0; 1169 1170 spin_lock_bh(&sk->sk_receive_queue.lock); 1171 skb = skb_peek(&sk->sk_receive_queue); 1172 if (skb) 1173 amount = skb_tail_pointer(skb) - 1174 skb_transport_header(skb); 1175 spin_unlock_bh(&sk->sk_receive_queue.lock); 1176 return put_user(amount, (int __user *)arg); 1177 } 1178 1179 default: 1180 #ifdef CONFIG_IPV6_MROUTE 1181 return ip6mr_ioctl(sk, cmd, (void __user *)arg); 1182 #else 1183 return -ENOIOCTLCMD; 1184 #endif 1185 } 1186 } 1187 1188 #ifdef CONFIG_COMPAT 1189 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 1190 { 1191 switch (cmd) { 1192 case SIOCOUTQ: 1193 case SIOCINQ: 1194 return -ENOIOCTLCMD; 1195 default: 1196 #ifdef CONFIG_IPV6_MROUTE 1197 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg)); 1198 #else 1199 return -ENOIOCTLCMD; 1200 #endif 1201 } 1202 } 1203 #endif 1204 1205 static void rawv6_close(struct sock *sk, long timeout) 1206 { 1207 if (inet_sk(sk)->inet_num == IPPROTO_RAW) 1208 ip6_ra_control(sk, -1); 1209 ip6mr_sk_done(sk); 1210 sk_common_release(sk); 1211 } 1212 1213 static void raw6_destroy(struct sock *sk) 1214 { 1215 lock_sock(sk); 1216 ip6_flush_pending_frames(sk); 1217 release_sock(sk); 1218 1219 inet6_destroy_sock(sk); 1220 } 1221 1222 static int rawv6_init_sk(struct sock *sk) 1223 { 1224 struct raw6_sock *rp = raw6_sk(sk); 1225 1226 switch (inet_sk(sk)->inet_num) { 1227 case IPPROTO_ICMPV6: 1228 rp->checksum = 1; 1229 rp->offset = 2; 1230 break; 1231 case IPPROTO_MH: 1232 rp->checksum = 1; 1233 rp->offset = 4; 1234 break; 1235 default: 1236 break; 1237 } 1238 return 0; 1239 } 1240 1241 struct proto rawv6_prot = { 1242 .name = "RAWv6", 1243 .owner = THIS_MODULE, 1244 .close = rawv6_close, 1245 .destroy = raw6_destroy, 1246 .connect = ip6_datagram_connect_v6_only, 1247 .disconnect = __udp_disconnect, 1248 .ioctl = rawv6_ioctl, 1249 .init = rawv6_init_sk, 1250 .setsockopt = rawv6_setsockopt, 1251 .getsockopt = rawv6_getsockopt, 1252 .sendmsg = rawv6_sendmsg, 1253 .recvmsg = rawv6_recvmsg, 1254 .bind = rawv6_bind, 1255 .backlog_rcv = rawv6_rcv_skb, 1256 .hash = raw_hash_sk, 1257 .unhash = raw_unhash_sk, 1258 .obj_size = sizeof(struct raw6_sock), 1259 .h.raw_hash = &raw_v6_hashinfo, 1260 #ifdef CONFIG_COMPAT 1261 .compat_setsockopt = compat_rawv6_setsockopt, 1262 .compat_getsockopt = compat_rawv6_getsockopt, 1263 .compat_ioctl = compat_rawv6_ioctl, 1264 #endif 1265 .diag_destroy = raw_abort, 1266 }; 1267 1268 #ifdef CONFIG_PROC_FS 1269 static int raw6_seq_show(struct seq_file *seq, void *v) 1270 { 1271 if (v == SEQ_START_TOKEN) { 1272 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); 1273 } else { 1274 struct sock *sp = v; 1275 __u16 srcp = inet_sk(sp)->inet_num; 1276 ip6_dgram_sock_seq_show(seq, v, srcp, 0, 1277 raw_seq_private(seq)->bucket); 1278 } 1279 return 0; 1280 } 1281 1282 static const struct seq_operations raw6_seq_ops = { 1283 .start = raw_seq_start, 1284 .next = raw_seq_next, 1285 .stop = raw_seq_stop, 1286 .show = raw6_seq_show, 1287 }; 1288 1289 static int raw6_seq_open(struct inode *inode, struct file *file) 1290 { 1291 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops); 1292 } 1293 1294 static const struct file_operations raw6_seq_fops = { 1295 .owner = THIS_MODULE, 1296 .open = raw6_seq_open, 1297 .read = seq_read, 1298 .llseek = seq_lseek, 1299 .release = seq_release_net, 1300 }; 1301 1302 static int __net_init raw6_init_net(struct net *net) 1303 { 1304 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops)) 1305 return -ENOMEM; 1306 1307 return 0; 1308 } 1309 1310 static void __net_exit raw6_exit_net(struct net *net) 1311 { 1312 remove_proc_entry("raw6", net->proc_net); 1313 } 1314 1315 static struct pernet_operations raw6_net_ops = { 1316 .init = raw6_init_net, 1317 .exit = raw6_exit_net, 1318 }; 1319 1320 int __init raw6_proc_init(void) 1321 { 1322 return register_pernet_subsys(&raw6_net_ops); 1323 } 1324 1325 void raw6_proc_exit(void) 1326 { 1327 unregister_pernet_subsys(&raw6_net_ops); 1328 } 1329 #endif /* CONFIG_PROC_FS */ 1330 1331 /* Same as inet6_dgram_ops, sans udp_poll. */ 1332 static const struct proto_ops inet6_sockraw_ops = { 1333 .family = PF_INET6, 1334 .owner = THIS_MODULE, 1335 .release = inet6_release, 1336 .bind = inet6_bind, 1337 .connect = inet_dgram_connect, /* ok */ 1338 .socketpair = sock_no_socketpair, /* a do nothing */ 1339 .accept = sock_no_accept, /* a do nothing */ 1340 .getname = inet6_getname, 1341 .poll = datagram_poll, /* ok */ 1342 .ioctl = inet6_ioctl, /* must change */ 1343 .listen = sock_no_listen, /* ok */ 1344 .shutdown = inet_shutdown, /* ok */ 1345 .setsockopt = sock_common_setsockopt, /* ok */ 1346 .getsockopt = sock_common_getsockopt, /* ok */ 1347 .sendmsg = inet_sendmsg, /* ok */ 1348 .recvmsg = sock_common_recvmsg, /* ok */ 1349 .mmap = sock_no_mmap, 1350 .sendpage = sock_no_sendpage, 1351 #ifdef CONFIG_COMPAT 1352 .compat_setsockopt = compat_sock_common_setsockopt, 1353 .compat_getsockopt = compat_sock_common_getsockopt, 1354 #endif 1355 }; 1356 1357 static struct inet_protosw rawv6_protosw = { 1358 .type = SOCK_RAW, 1359 .protocol = IPPROTO_IP, /* wild card */ 1360 .prot = &rawv6_prot, 1361 .ops = &inet6_sockraw_ops, 1362 .flags = INET_PROTOSW_REUSE, 1363 }; 1364 1365 int __init rawv6_init(void) 1366 { 1367 return inet6_register_protosw(&rawv6_protosw); 1368 } 1369 1370 void rawv6_exit(void) 1371 { 1372 inet6_unregister_protosw(&rawv6_protosw); 1373 } 1374