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 <asm/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 static struct raw_hashinfo raw_v6_hashinfo = { 69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock), 70 }; 71 72 static 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 struct ipv6_pinfo *np = inet6_sk(sk); 81 82 if (!net_eq(sock_net(sk), net)) 83 continue; 84 85 if (!ipv6_addr_any(&np->daddr) && 86 !ipv6_addr_equal(&np->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(&np->rcv_saddr)) { 93 if (ipv6_addr_equal(&np->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 } 106 107 /* 108 * 0 - deliver 109 * 1 - block 110 */ 111 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb) 112 { 113 struct icmp6hdr _hdr; 114 const struct icmp6hdr *hdr; 115 116 /* We require only the four bytes of the ICMPv6 header, not any 117 * additional bytes of message body in "struct icmp6hdr". 118 */ 119 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 120 ICMPV6_HDRLEN, &_hdr); 121 if (hdr) { 122 const __u32 *data = &raw6_sk(sk)->filter.data[0]; 123 unsigned int type = hdr->icmp6_type; 124 125 return (data[type >> 5] & (1U << (type & 31))) != 0; 126 } 127 return 1; 128 } 129 130 #if IS_ENABLED(CONFIG_IPV6_MIP6) 131 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb); 132 133 static mh_filter_t __rcu *mh_filter __read_mostly; 134 135 int rawv6_mh_filter_register(mh_filter_t filter) 136 { 137 rcu_assign_pointer(mh_filter, filter); 138 return 0; 139 } 140 EXPORT_SYMBOL(rawv6_mh_filter_register); 141 142 int rawv6_mh_filter_unregister(mh_filter_t filter) 143 { 144 RCU_INIT_POINTER(mh_filter, NULL); 145 synchronize_rcu(); 146 return 0; 147 } 148 EXPORT_SYMBOL(rawv6_mh_filter_unregister); 149 150 #endif 151 152 /* 153 * demultiplex raw sockets. 154 * (should consider queueing the skb in the sock receive_queue 155 * without calling rawv6.c) 156 * 157 * Caller owns SKB so we must make clones. 158 */ 159 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 160 { 161 const struct in6_addr *saddr; 162 const struct in6_addr *daddr; 163 struct sock *sk; 164 bool delivered = false; 165 __u8 hash; 166 struct net *net; 167 168 saddr = &ipv6_hdr(skb)->saddr; 169 daddr = saddr + 1; 170 171 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 172 173 read_lock(&raw_v6_hashinfo.lock); 174 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 175 176 if (sk == NULL) 177 goto out; 178 179 net = dev_net(skb->dev); 180 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif); 181 182 while (sk) { 183 int filtered; 184 185 delivered = true; 186 switch (nexthdr) { 187 case IPPROTO_ICMPV6: 188 filtered = icmpv6_filter(sk, skb); 189 break; 190 191 #if IS_ENABLED(CONFIG_IPV6_MIP6) 192 case IPPROTO_MH: 193 { 194 /* XXX: To validate MH only once for each packet, 195 * this is placed here. It should be after checking 196 * xfrm policy, however it doesn't. The checking xfrm 197 * policy is placed in rawv6_rcv() because it is 198 * required for each socket. 199 */ 200 mh_filter_t *filter; 201 202 filter = rcu_dereference(mh_filter); 203 filtered = filter ? (*filter)(sk, skb) : 0; 204 break; 205 } 206 #endif 207 default: 208 filtered = 0; 209 break; 210 } 211 212 if (filtered < 0) 213 break; 214 if (filtered == 0) { 215 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 216 217 /* Not releasing hash table! */ 218 if (clone) { 219 nf_reset(clone); 220 rawv6_rcv(sk, clone); 221 } 222 } 223 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr, 224 IP6CB(skb)->iif); 225 } 226 out: 227 read_unlock(&raw_v6_hashinfo.lock); 228 return delivered; 229 } 230 231 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr) 232 { 233 struct sock *raw_sk; 234 235 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]); 236 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr)) 237 raw_sk = NULL; 238 239 return raw_sk != NULL; 240 } 241 242 /* This cleans up af_inet6 a bit. -DaveM */ 243 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 244 { 245 struct inet_sock *inet = inet_sk(sk); 246 struct ipv6_pinfo *np = inet6_sk(sk); 247 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 248 __be32 v4addr = 0; 249 int addr_type; 250 int err; 251 252 if (addr_len < SIN6_LEN_RFC2133) 253 return -EINVAL; 254 addr_type = ipv6_addr_type(&addr->sin6_addr); 255 256 /* Raw sockets are IPv6 only */ 257 if (addr_type == IPV6_ADDR_MAPPED) 258 return -EADDRNOTAVAIL; 259 260 lock_sock(sk); 261 262 err = -EINVAL; 263 if (sk->sk_state != TCP_CLOSE) 264 goto out; 265 266 rcu_read_lock(); 267 /* Check if the address belongs to the host. */ 268 if (addr_type != IPV6_ADDR_ANY) { 269 struct net_device *dev = NULL; 270 271 if (__ipv6_addr_needs_scope_id(addr_type)) { 272 if (addr_len >= sizeof(struct sockaddr_in6) && 273 addr->sin6_scope_id) { 274 /* Override any existing binding, if another 275 * one is supplied by user. 276 */ 277 sk->sk_bound_dev_if = addr->sin6_scope_id; 278 } 279 280 /* Binding to link-local address requires an interface */ 281 if (!sk->sk_bound_dev_if) 282 goto out_unlock; 283 284 err = -ENODEV; 285 dev = dev_get_by_index_rcu(sock_net(sk), 286 sk->sk_bound_dev_if); 287 if (!dev) 288 goto out_unlock; 289 } 290 291 /* ipv4 addr of the socket is invalid. Only the 292 * unspecified and mapped address have a v4 equivalent. 293 */ 294 v4addr = LOOPBACK4_IPV6; 295 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 296 err = -EADDRNOTAVAIL; 297 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr, 298 dev, 0)) { 299 goto out_unlock; 300 } 301 } 302 } 303 304 inet->inet_rcv_saddr = inet->inet_saddr = v4addr; 305 np->rcv_saddr = addr->sin6_addr; 306 if (!(addr_type & IPV6_ADDR_MULTICAST)) 307 np->saddr = addr->sin6_addr; 308 err = 0; 309 out_unlock: 310 rcu_read_unlock(); 311 out: 312 release_sock(sk); 313 return err; 314 } 315 316 static void rawv6_err(struct sock *sk, struct sk_buff *skb, 317 struct inet6_skb_parm *opt, 318 u8 type, u8 code, int offset, __be32 info) 319 { 320 struct inet_sock *inet = inet_sk(sk); 321 struct ipv6_pinfo *np = inet6_sk(sk); 322 int err; 323 int harderr; 324 325 /* Report error on raw socket, if: 326 1. User requested recverr. 327 2. Socket is connected (otherwise the error indication 328 is useless without recverr and error is hard. 329 */ 330 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) 331 return; 332 333 harderr = icmpv6_err_convert(type, code, &err); 334 if (type == ICMPV6_PKT_TOOBIG) { 335 ip6_sk_update_pmtu(skb, sk, info); 336 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); 337 } 338 if (type == NDISC_REDIRECT) 339 ip6_sk_redirect(skb, sk); 340 if (np->recverr) { 341 u8 *payload = skb->data; 342 if (!inet->hdrincl) 343 payload += offset; 344 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); 345 } 346 347 if (np->recverr || harderr) { 348 sk->sk_err = err; 349 sk->sk_error_report(sk); 350 } 351 } 352 353 void raw6_icmp_error(struct sk_buff *skb, int nexthdr, 354 u8 type, u8 code, int inner_offset, __be32 info) 355 { 356 struct sock *sk; 357 int hash; 358 const struct in6_addr *saddr, *daddr; 359 struct net *net; 360 361 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 362 363 read_lock(&raw_v6_hashinfo.lock); 364 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 365 if (sk != NULL) { 366 /* Note: ipv6_hdr(skb) != skb->data */ 367 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data; 368 saddr = &ip6h->saddr; 369 daddr = &ip6h->daddr; 370 net = dev_net(skb->dev); 371 372 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr, 373 IP6CB(skb)->iif))) { 374 rawv6_err(sk, skb, NULL, type, code, 375 inner_offset, info); 376 sk = sk_next(sk); 377 } 378 } 379 read_unlock(&raw_v6_hashinfo.lock); 380 } 381 382 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb) 383 { 384 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) && 385 skb_checksum_complete(skb)) { 386 atomic_inc(&sk->sk_drops); 387 kfree_skb(skb); 388 return NET_RX_DROP; 389 } 390 391 /* Charge it to the socket. */ 392 skb_dst_drop(skb); 393 if (sock_queue_rcv_skb(sk, skb) < 0) { 394 kfree_skb(skb); 395 return NET_RX_DROP; 396 } 397 398 return 0; 399 } 400 401 /* 402 * This is next to useless... 403 * if we demultiplex in network layer we don't need the extra call 404 * just to queue the skb... 405 * maybe we could have the network decide upon a hint if it 406 * should call raw_rcv for demultiplexing 407 */ 408 int rawv6_rcv(struct sock *sk, struct sk_buff *skb) 409 { 410 struct inet_sock *inet = inet_sk(sk); 411 struct raw6_sock *rp = raw6_sk(sk); 412 413 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { 414 atomic_inc(&sk->sk_drops); 415 kfree_skb(skb); 416 return NET_RX_DROP; 417 } 418 419 if (!rp->checksum) 420 skb->ip_summed = CHECKSUM_UNNECESSARY; 421 422 if (skb->ip_summed == CHECKSUM_COMPLETE) { 423 skb_postpull_rcsum(skb, skb_network_header(skb), 424 skb_network_header_len(skb)); 425 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 426 &ipv6_hdr(skb)->daddr, 427 skb->len, inet->inet_num, skb->csum)) 428 skb->ip_summed = CHECKSUM_UNNECESSARY; 429 } 430 if (!skb_csum_unnecessary(skb)) 431 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 432 &ipv6_hdr(skb)->daddr, 433 skb->len, 434 inet->inet_num, 0)); 435 436 if (inet->hdrincl) { 437 if (skb_checksum_complete(skb)) { 438 atomic_inc(&sk->sk_drops); 439 kfree_skb(skb); 440 return NET_RX_DROP; 441 } 442 } 443 444 rawv6_rcv_skb(sk, skb); 445 return 0; 446 } 447 448 449 /* 450 * This should be easy, if there is something there 451 * we return it, otherwise we block. 452 */ 453 454 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk, 455 struct msghdr *msg, size_t len, 456 int noblock, int flags, int *addr_len) 457 { 458 struct ipv6_pinfo *np = inet6_sk(sk); 459 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name; 460 struct sk_buff *skb; 461 size_t copied; 462 int err; 463 464 if (flags & MSG_OOB) 465 return -EOPNOTSUPP; 466 467 if (addr_len) 468 *addr_len=sizeof(*sin6); 469 470 if (flags & MSG_ERRQUEUE) 471 return ipv6_recv_error(sk, msg, len); 472 473 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 474 return ipv6_recv_rxpmtu(sk, msg, len); 475 476 skb = skb_recv_datagram(sk, flags, noblock, &err); 477 if (!skb) 478 goto out; 479 480 copied = skb->len; 481 if (copied > len) { 482 copied = len; 483 msg->msg_flags |= MSG_TRUNC; 484 } 485 486 if (skb_csum_unnecessary(skb)) { 487 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 488 } else if (msg->msg_flags&MSG_TRUNC) { 489 if (__skb_checksum_complete(skb)) 490 goto csum_copy_err; 491 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 492 } else { 493 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov); 494 if (err == -EINVAL) 495 goto csum_copy_err; 496 } 497 if (err) 498 goto out_free; 499 500 /* Copy the address. */ 501 if (sin6) { 502 sin6->sin6_family = AF_INET6; 503 sin6->sin6_port = 0; 504 sin6->sin6_addr = ipv6_hdr(skb)->saddr; 505 sin6->sin6_flowinfo = 0; 506 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr, 507 IP6CB(skb)->iif); 508 } 509 510 sock_recv_ts_and_drops(msg, sk, skb); 511 512 if (np->rxopt.all) 513 ip6_datagram_recv_ctl(sk, msg, skb); 514 515 err = copied; 516 if (flags & MSG_TRUNC) 517 err = skb->len; 518 519 out_free: 520 skb_free_datagram(sk, skb); 521 out: 522 return err; 523 524 csum_copy_err: 525 skb_kill_datagram(sk, skb, flags); 526 527 /* Error for blocking case is chosen to masquerade 528 as some normal condition. 529 */ 530 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; 531 goto out; 532 } 533 534 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 535 struct raw6_sock *rp) 536 { 537 struct sk_buff *skb; 538 int err = 0; 539 int offset; 540 int len; 541 int total_len; 542 __wsum tmp_csum; 543 __sum16 csum; 544 545 if (!rp->checksum) 546 goto send; 547 548 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 549 goto out; 550 551 offset = rp->offset; 552 total_len = inet_sk(sk)->cork.base.length; 553 if (offset >= total_len - 1) { 554 err = -EINVAL; 555 ip6_flush_pending_frames(sk); 556 goto out; 557 } 558 559 /* should be check HW csum miyazawa */ 560 if (skb_queue_len(&sk->sk_write_queue) == 1) { 561 /* 562 * Only one fragment on the socket. 563 */ 564 tmp_csum = skb->csum; 565 } else { 566 struct sk_buff *csum_skb = NULL; 567 tmp_csum = 0; 568 569 skb_queue_walk(&sk->sk_write_queue, skb) { 570 tmp_csum = csum_add(tmp_csum, skb->csum); 571 572 if (csum_skb) 573 continue; 574 575 len = skb->len - skb_transport_offset(skb); 576 if (offset >= len) { 577 offset -= len; 578 continue; 579 } 580 581 csum_skb = skb; 582 } 583 584 skb = csum_skb; 585 } 586 587 offset += skb_transport_offset(skb); 588 if (skb_copy_bits(skb, offset, &csum, 2)) 589 BUG(); 590 591 /* in case cksum was not initialized */ 592 if (unlikely(csum)) 593 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 594 595 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 596 total_len, fl6->flowi6_proto, tmp_csum); 597 598 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP) 599 csum = CSUM_MANGLED_0; 600 601 if (skb_store_bits(skb, offset, &csum, 2)) 602 BUG(); 603 604 send: 605 err = ip6_push_pending_frames(sk); 606 out: 607 return err; 608 } 609 610 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length, 611 struct flowi6 *fl6, struct dst_entry **dstp, 612 unsigned int flags) 613 { 614 struct ipv6_pinfo *np = inet6_sk(sk); 615 struct ipv6hdr *iph; 616 struct sk_buff *skb; 617 int err; 618 struct rt6_info *rt = (struct rt6_info *)*dstp; 619 int hlen = LL_RESERVED_SPACE(rt->dst.dev); 620 int tlen = rt->dst.dev->needed_tailroom; 621 622 if (length > rt->dst.dev->mtu) { 623 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu); 624 return -EMSGSIZE; 625 } 626 if (flags&MSG_PROBE) 627 goto out; 628 629 skb = sock_alloc_send_skb(sk, 630 length + hlen + tlen + 15, 631 flags & MSG_DONTWAIT, &err); 632 if (skb == NULL) 633 goto error; 634 skb_reserve(skb, hlen); 635 636 skb->protocol = htons(ETH_P_IPV6); 637 skb->priority = sk->sk_priority; 638 skb->mark = sk->sk_mark; 639 skb_dst_set(skb, &rt->dst); 640 *dstp = NULL; 641 642 skb_put(skb, length); 643 skb_reset_network_header(skb); 644 iph = ipv6_hdr(skb); 645 646 skb->ip_summed = CHECKSUM_NONE; 647 648 skb->transport_header = skb->network_header; 649 err = memcpy_fromiovecend((void *)iph, from, 0, length); 650 if (err) 651 goto error_fault; 652 653 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); 654 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, 655 rt->dst.dev, dst_output); 656 if (err > 0) 657 err = net_xmit_errno(err); 658 if (err) 659 goto error; 660 out: 661 return 0; 662 663 error_fault: 664 err = -EFAULT; 665 kfree_skb(skb); 666 error: 667 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 668 if (err == -ENOBUFS && !np->recverr) 669 err = 0; 670 return err; 671 } 672 673 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg) 674 { 675 struct iovec *iov; 676 u8 __user *type = NULL; 677 u8 __user *code = NULL; 678 u8 len = 0; 679 int probed = 0; 680 int i; 681 682 if (!msg->msg_iov) 683 return 0; 684 685 for (i = 0; i < msg->msg_iovlen; i++) { 686 iov = &msg->msg_iov[i]; 687 if (!iov) 688 continue; 689 690 switch (fl6->flowi6_proto) { 691 case IPPROTO_ICMPV6: 692 /* check if one-byte field is readable or not. */ 693 if (iov->iov_base && iov->iov_len < 1) 694 break; 695 696 if (!type) { 697 type = iov->iov_base; 698 /* check if code field is readable or not. */ 699 if (iov->iov_len > 1) 700 code = type + 1; 701 } else if (!code) 702 code = iov->iov_base; 703 704 if (type && code) { 705 if (get_user(fl6->fl6_icmp_type, type) || 706 get_user(fl6->fl6_icmp_code, code)) 707 return -EFAULT; 708 probed = 1; 709 } 710 break; 711 case IPPROTO_MH: 712 if (iov->iov_base && iov->iov_len < 1) 713 break; 714 /* check if type field is readable or not. */ 715 if (iov->iov_len > 2 - len) { 716 u8 __user *p = iov->iov_base; 717 if (get_user(fl6->fl6_mh_type, &p[2 - len])) 718 return -EFAULT; 719 probed = 1; 720 } else 721 len += iov->iov_len; 722 723 break; 724 default: 725 probed = 1; 726 break; 727 } 728 if (probed) 729 break; 730 } 731 return 0; 732 } 733 734 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk, 735 struct msghdr *msg, size_t len) 736 { 737 struct ipv6_txoptions opt_space; 738 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name; 739 struct in6_addr *daddr, *final_p, final; 740 struct inet_sock *inet = inet_sk(sk); 741 struct ipv6_pinfo *np = inet6_sk(sk); 742 struct raw6_sock *rp = raw6_sk(sk); 743 struct ipv6_txoptions *opt = NULL; 744 struct ip6_flowlabel *flowlabel = NULL; 745 struct dst_entry *dst = NULL; 746 struct flowi6 fl6; 747 int addr_len = msg->msg_namelen; 748 int hlimit = -1; 749 int tclass = -1; 750 int dontfrag = -1; 751 u16 proto; 752 int err; 753 754 /* Rough check on arithmetic overflow, 755 better check is made in ip6_append_data(). 756 */ 757 if (len > INT_MAX) 758 return -EMSGSIZE; 759 760 /* Mirror BSD error message compatibility */ 761 if (msg->msg_flags & MSG_OOB) 762 return -EOPNOTSUPP; 763 764 /* 765 * Get and verify the address. 766 */ 767 memset(&fl6, 0, sizeof(fl6)); 768 769 fl6.flowi6_mark = sk->sk_mark; 770 771 if (sin6) { 772 if (addr_len < SIN6_LEN_RFC2133) 773 return -EINVAL; 774 775 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 776 return -EAFNOSUPPORT; 777 778 /* port is the proto value [0..255] carried in nexthdr */ 779 proto = ntohs(sin6->sin6_port); 780 781 if (!proto) 782 proto = inet->inet_num; 783 else if (proto != inet->inet_num) 784 return -EINVAL; 785 786 if (proto > 255) 787 return -EINVAL; 788 789 daddr = &sin6->sin6_addr; 790 if (np->sndflow) { 791 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 792 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 793 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 794 if (flowlabel == NULL) 795 return -EINVAL; 796 daddr = &flowlabel->dst; 797 } 798 } 799 800 /* 801 * Otherwise it will be difficult to maintain 802 * sk->sk_dst_cache. 803 */ 804 if (sk->sk_state == TCP_ESTABLISHED && 805 ipv6_addr_equal(daddr, &np->daddr)) 806 daddr = &np->daddr; 807 808 if (addr_len >= sizeof(struct sockaddr_in6) && 809 sin6->sin6_scope_id && 810 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) 811 fl6.flowi6_oif = sin6->sin6_scope_id; 812 } else { 813 if (sk->sk_state != TCP_ESTABLISHED) 814 return -EDESTADDRREQ; 815 816 proto = inet->inet_num; 817 daddr = &np->daddr; 818 fl6.flowlabel = np->flow_label; 819 } 820 821 if (fl6.flowi6_oif == 0) 822 fl6.flowi6_oif = sk->sk_bound_dev_if; 823 824 if (msg->msg_controllen) { 825 opt = &opt_space; 826 memset(opt, 0, sizeof(struct ipv6_txoptions)); 827 opt->tot_len = sizeof(struct ipv6_txoptions); 828 829 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt, 830 &hlimit, &tclass, &dontfrag); 831 if (err < 0) { 832 fl6_sock_release(flowlabel); 833 return err; 834 } 835 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 836 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 837 if (flowlabel == NULL) 838 return -EINVAL; 839 } 840 if (!(opt->opt_nflen|opt->opt_flen)) 841 opt = NULL; 842 } 843 if (opt == NULL) 844 opt = np->opt; 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 err = rawv6_probe_proto_opt(&fl6, msg); 851 if (err) 852 goto out; 853 854 if (!ipv6_addr_any(daddr)) 855 fl6.daddr = *daddr; 856 else 857 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 858 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 859 fl6.saddr = np->saddr; 860 861 final_p = fl6_update_dst(&fl6, opt, &final); 862 863 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 864 fl6.flowi6_oif = np->mcast_oif; 865 else if (!fl6.flowi6_oif) 866 fl6.flowi6_oif = np->ucast_oif; 867 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 868 869 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true); 870 if (IS_ERR(dst)) { 871 err = PTR_ERR(dst); 872 goto out; 873 } 874 if (hlimit < 0) { 875 if (ipv6_addr_is_multicast(&fl6.daddr)) 876 hlimit = np->mcast_hops; 877 else 878 hlimit = np->hop_limit; 879 if (hlimit < 0) 880 hlimit = ip6_dst_hoplimit(dst); 881 } 882 883 if (tclass < 0) 884 tclass = np->tclass; 885 886 if (dontfrag < 0) 887 dontfrag = np->dontfrag; 888 889 if (msg->msg_flags&MSG_CONFIRM) 890 goto do_confirm; 891 892 back_from_confirm: 893 if (inet->hdrincl) 894 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags); 895 else { 896 lock_sock(sk); 897 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, 898 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst, 899 msg->msg_flags, dontfrag); 900 901 if (err) 902 ip6_flush_pending_frames(sk); 903 else if (!(msg->msg_flags & MSG_MORE)) 904 err = rawv6_push_pending_frames(sk, &fl6, rp); 905 release_sock(sk); 906 } 907 done: 908 dst_release(dst); 909 out: 910 fl6_sock_release(flowlabel); 911 return err<0?err:len; 912 do_confirm: 913 dst_confirm(dst); 914 if (!(msg->msg_flags & MSG_PROBE) || len) 915 goto back_from_confirm; 916 err = 0; 917 goto done; 918 } 919 920 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 921 char __user *optval, int optlen) 922 { 923 switch (optname) { 924 case ICMPV6_FILTER: 925 if (optlen > sizeof(struct icmp6_filter)) 926 optlen = sizeof(struct icmp6_filter); 927 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) 928 return -EFAULT; 929 return 0; 930 default: 931 return -ENOPROTOOPT; 932 } 933 934 return 0; 935 } 936 937 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 938 char __user *optval, int __user *optlen) 939 { 940 int len; 941 942 switch (optname) { 943 case ICMPV6_FILTER: 944 if (get_user(len, optlen)) 945 return -EFAULT; 946 if (len < 0) 947 return -EINVAL; 948 if (len > sizeof(struct icmp6_filter)) 949 len = sizeof(struct icmp6_filter); 950 if (put_user(len, optlen)) 951 return -EFAULT; 952 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 953 return -EFAULT; 954 return 0; 955 default: 956 return -ENOPROTOOPT; 957 } 958 959 return 0; 960 } 961 962 963 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 964 char __user *optval, unsigned int optlen) 965 { 966 struct raw6_sock *rp = raw6_sk(sk); 967 int val; 968 969 if (get_user(val, (int __user *)optval)) 970 return -EFAULT; 971 972 switch (optname) { 973 case IPV6_CHECKSUM: 974 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && 975 level == IPPROTO_IPV6) { 976 /* 977 * RFC3542 tells that IPV6_CHECKSUM socket 978 * option in the IPPROTO_IPV6 level is not 979 * allowed on ICMPv6 sockets. 980 * If you want to set it, use IPPROTO_RAW 981 * level IPV6_CHECKSUM socket option 982 * (Linux extension). 983 */ 984 return -EINVAL; 985 } 986 987 /* You may get strange result with a positive odd offset; 988 RFC2292bis agrees with me. */ 989 if (val > 0 && (val&1)) 990 return -EINVAL; 991 if (val < 0) { 992 rp->checksum = 0; 993 } else { 994 rp->checksum = 1; 995 rp->offset = val; 996 } 997 998 return 0; 999 1000 default: 1001 return -ENOPROTOOPT; 1002 } 1003 } 1004 1005 static int rawv6_setsockopt(struct sock *sk, int level, int optname, 1006 char __user *optval, unsigned int optlen) 1007 { 1008 switch (level) { 1009 case SOL_RAW: 1010 break; 1011 1012 case SOL_ICMPV6: 1013 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1014 return -EOPNOTSUPP; 1015 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1016 case SOL_IPV6: 1017 if (optname == IPV6_CHECKSUM) 1018 break; 1019 default: 1020 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1021 } 1022 1023 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1024 } 1025 1026 #ifdef CONFIG_COMPAT 1027 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, 1028 char __user *optval, unsigned int optlen) 1029 { 1030 switch (level) { 1031 case SOL_RAW: 1032 break; 1033 case SOL_ICMPV6: 1034 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1035 return -EOPNOTSUPP; 1036 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1037 case SOL_IPV6: 1038 if (optname == IPV6_CHECKSUM) 1039 break; 1040 default: 1041 return compat_ipv6_setsockopt(sk, level, optname, 1042 optval, optlen); 1043 } 1044 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1045 } 1046 #endif 1047 1048 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1049 char __user *optval, int __user *optlen) 1050 { 1051 struct raw6_sock *rp = raw6_sk(sk); 1052 int val, len; 1053 1054 if (get_user(len,optlen)) 1055 return -EFAULT; 1056 1057 switch (optname) { 1058 case IPV6_CHECKSUM: 1059 /* 1060 * We allow getsockopt() for IPPROTO_IPV6-level 1061 * IPV6_CHECKSUM socket option on ICMPv6 sockets 1062 * since RFC3542 is silent about it. 1063 */ 1064 if (rp->checksum == 0) 1065 val = -1; 1066 else 1067 val = rp->offset; 1068 break; 1069 1070 default: 1071 return -ENOPROTOOPT; 1072 } 1073 1074 len = min_t(unsigned int, sizeof(int), len); 1075 1076 if (put_user(len, optlen)) 1077 return -EFAULT; 1078 if (copy_to_user(optval,&val,len)) 1079 return -EFAULT; 1080 return 0; 1081 } 1082 1083 static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1084 char __user *optval, int __user *optlen) 1085 { 1086 switch (level) { 1087 case SOL_RAW: 1088 break; 1089 1090 case SOL_ICMPV6: 1091 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1092 return -EOPNOTSUPP; 1093 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1094 case SOL_IPV6: 1095 if (optname == IPV6_CHECKSUM) 1096 break; 1097 default: 1098 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1099 } 1100 1101 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1102 } 1103 1104 #ifdef CONFIG_COMPAT 1105 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, 1106 char __user *optval, int __user *optlen) 1107 { 1108 switch (level) { 1109 case SOL_RAW: 1110 break; 1111 case SOL_ICMPV6: 1112 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1113 return -EOPNOTSUPP; 1114 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1115 case SOL_IPV6: 1116 if (optname == IPV6_CHECKSUM) 1117 break; 1118 default: 1119 return compat_ipv6_getsockopt(sk, level, optname, 1120 optval, optlen); 1121 } 1122 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1123 } 1124 #endif 1125 1126 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1127 { 1128 switch (cmd) { 1129 case SIOCOUTQ: { 1130 int amount = sk_wmem_alloc_get(sk); 1131 1132 return put_user(amount, (int __user *)arg); 1133 } 1134 case SIOCINQ: { 1135 struct sk_buff *skb; 1136 int amount = 0; 1137 1138 spin_lock_bh(&sk->sk_receive_queue.lock); 1139 skb = skb_peek(&sk->sk_receive_queue); 1140 if (skb != NULL) 1141 amount = skb_tail_pointer(skb) - 1142 skb_transport_header(skb); 1143 spin_unlock_bh(&sk->sk_receive_queue.lock); 1144 return put_user(amount, (int __user *)arg); 1145 } 1146 1147 default: 1148 #ifdef CONFIG_IPV6_MROUTE 1149 return ip6mr_ioctl(sk, cmd, (void __user *)arg); 1150 #else 1151 return -ENOIOCTLCMD; 1152 #endif 1153 } 1154 } 1155 1156 #ifdef CONFIG_COMPAT 1157 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 1158 { 1159 switch (cmd) { 1160 case SIOCOUTQ: 1161 case SIOCINQ: 1162 return -ENOIOCTLCMD; 1163 default: 1164 #ifdef CONFIG_IPV6_MROUTE 1165 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg)); 1166 #else 1167 return -ENOIOCTLCMD; 1168 #endif 1169 } 1170 } 1171 #endif 1172 1173 static void rawv6_close(struct sock *sk, long timeout) 1174 { 1175 if (inet_sk(sk)->inet_num == IPPROTO_RAW) 1176 ip6_ra_control(sk, -1); 1177 ip6mr_sk_done(sk); 1178 sk_common_release(sk); 1179 } 1180 1181 static void raw6_destroy(struct sock *sk) 1182 { 1183 lock_sock(sk); 1184 ip6_flush_pending_frames(sk); 1185 release_sock(sk); 1186 1187 inet6_destroy_sock(sk); 1188 } 1189 1190 static int rawv6_init_sk(struct sock *sk) 1191 { 1192 struct raw6_sock *rp = raw6_sk(sk); 1193 1194 switch (inet_sk(sk)->inet_num) { 1195 case IPPROTO_ICMPV6: 1196 rp->checksum = 1; 1197 rp->offset = 2; 1198 break; 1199 case IPPROTO_MH: 1200 rp->checksum = 1; 1201 rp->offset = 4; 1202 break; 1203 default: 1204 break; 1205 } 1206 return 0; 1207 } 1208 1209 struct proto rawv6_prot = { 1210 .name = "RAWv6", 1211 .owner = THIS_MODULE, 1212 .close = rawv6_close, 1213 .destroy = raw6_destroy, 1214 .connect = ip6_datagram_connect, 1215 .disconnect = udp_disconnect, 1216 .ioctl = rawv6_ioctl, 1217 .init = rawv6_init_sk, 1218 .setsockopt = rawv6_setsockopt, 1219 .getsockopt = rawv6_getsockopt, 1220 .sendmsg = rawv6_sendmsg, 1221 .recvmsg = rawv6_recvmsg, 1222 .bind = rawv6_bind, 1223 .backlog_rcv = rawv6_rcv_skb, 1224 .hash = raw_hash_sk, 1225 .unhash = raw_unhash_sk, 1226 .obj_size = sizeof(struct raw6_sock), 1227 .h.raw_hash = &raw_v6_hashinfo, 1228 #ifdef CONFIG_COMPAT 1229 .compat_setsockopt = compat_rawv6_setsockopt, 1230 .compat_getsockopt = compat_rawv6_getsockopt, 1231 .compat_ioctl = compat_rawv6_ioctl, 1232 #endif 1233 }; 1234 1235 #ifdef CONFIG_PROC_FS 1236 static int raw6_seq_show(struct seq_file *seq, void *v) 1237 { 1238 if (v == SEQ_START_TOKEN) { 1239 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); 1240 } else { 1241 struct sock *sp = v; 1242 __u16 srcp = inet_sk(sp)->inet_num; 1243 ip6_dgram_sock_seq_show(seq, v, srcp, 0, 1244 raw_seq_private(seq)->bucket); 1245 } 1246 return 0; 1247 } 1248 1249 static const struct seq_operations raw6_seq_ops = { 1250 .start = raw_seq_start, 1251 .next = raw_seq_next, 1252 .stop = raw_seq_stop, 1253 .show = raw6_seq_show, 1254 }; 1255 1256 static int raw6_seq_open(struct inode *inode, struct file *file) 1257 { 1258 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops); 1259 } 1260 1261 static const struct file_operations raw6_seq_fops = { 1262 .owner = THIS_MODULE, 1263 .open = raw6_seq_open, 1264 .read = seq_read, 1265 .llseek = seq_lseek, 1266 .release = seq_release_net, 1267 }; 1268 1269 static int __net_init raw6_init_net(struct net *net) 1270 { 1271 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops)) 1272 return -ENOMEM; 1273 1274 return 0; 1275 } 1276 1277 static void __net_exit raw6_exit_net(struct net *net) 1278 { 1279 remove_proc_entry("raw6", net->proc_net); 1280 } 1281 1282 static struct pernet_operations raw6_net_ops = { 1283 .init = raw6_init_net, 1284 .exit = raw6_exit_net, 1285 }; 1286 1287 int __init raw6_proc_init(void) 1288 { 1289 return register_pernet_subsys(&raw6_net_ops); 1290 } 1291 1292 void raw6_proc_exit(void) 1293 { 1294 unregister_pernet_subsys(&raw6_net_ops); 1295 } 1296 #endif /* CONFIG_PROC_FS */ 1297 1298 /* Same as inet6_dgram_ops, sans udp_poll. */ 1299 static const struct proto_ops inet6_sockraw_ops = { 1300 .family = PF_INET6, 1301 .owner = THIS_MODULE, 1302 .release = inet6_release, 1303 .bind = inet6_bind, 1304 .connect = inet_dgram_connect, /* ok */ 1305 .socketpair = sock_no_socketpair, /* a do nothing */ 1306 .accept = sock_no_accept, /* a do nothing */ 1307 .getname = inet6_getname, 1308 .poll = datagram_poll, /* ok */ 1309 .ioctl = inet6_ioctl, /* must change */ 1310 .listen = sock_no_listen, /* ok */ 1311 .shutdown = inet_shutdown, /* ok */ 1312 .setsockopt = sock_common_setsockopt, /* ok */ 1313 .getsockopt = sock_common_getsockopt, /* ok */ 1314 .sendmsg = inet_sendmsg, /* ok */ 1315 .recvmsg = sock_common_recvmsg, /* ok */ 1316 .mmap = sock_no_mmap, 1317 .sendpage = sock_no_sendpage, 1318 #ifdef CONFIG_COMPAT 1319 .compat_setsockopt = compat_sock_common_setsockopt, 1320 .compat_getsockopt = compat_sock_common_getsockopt, 1321 #endif 1322 }; 1323 1324 static struct inet_protosw rawv6_protosw = { 1325 .type = SOCK_RAW, 1326 .protocol = IPPROTO_IP, /* wild card */ 1327 .prot = &rawv6_prot, 1328 .ops = &inet6_sockraw_ops, 1329 .no_check = UDP_CSUM_DEFAULT, 1330 .flags = INET_PROTOSW_REUSE, 1331 }; 1332 1333 int __init rawv6_init(void) 1334 { 1335 int ret; 1336 1337 ret = inet6_register_protosw(&rawv6_protosw); 1338 if (ret) 1339 goto out; 1340 out: 1341 return ret; 1342 } 1343 1344 void rawv6_exit(void) 1345 { 1346 inet6_unregister_protosw(&rawv6_protosw); 1347 } 1348