1// SPDX-License-Identifier: GPL-2.0-or-later 2/* 3 * RAW sockets for IPv6 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * 9 * Adapted from linux/net/ipv4/raw.c 10 * 11 * Fixes: 12 * Hideaki YOSHIFUJI : sin6_scope_id support 13 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance) 14 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data 15 */ 16 17#include <linux/errno.h> 18#include <linux/types.h> 19#include <linux/socket.h> 20#include <linux/slab.h> 21#include <linux/sockios.h> 22#include <linux/net.h> 23#include <linux/in6.h> 24#include <linux/netdevice.h> 25#include <linux/if_arp.h> 26#include <linux/icmpv6.h> 27#include <linux/netfilter.h> 28#include <linux/netfilter_ipv6.h> 29#include <linux/skbuff.h> 30#include <linux/compat.h> 31#include <linux/uaccess.h> 32#include <asm/ioctls.h> 33 34#include <net/net_namespace.h> 35#include <net/ip.h> 36#include <net/sock.h> 37#include <net/snmp.h> 38 39#include <net/ipv6.h> 40#include <net/ndisc.h> 41#include <net/protocol.h> 42#include <net/ip6_route.h> 43#include <net/ip6_checksum.h> 44#include <net/addrconf.h> 45#include <net/transp_v6.h> 46#include <net/udp.h> 47#include <net/inet_common.h> 48#include <net/tcp_states.h> 49#if IS_ENABLED(CONFIG_IPV6_MIP6) 50#include <net/mip6.h> 51#endif 52#include <linux/mroute6.h> 53 54#include <net/raw.h> 55#include <net/rawv6.h> 56#include <net/xfrm.h> 57 58#include <linux/proc_fs.h> 59#include <linux/seq_file.h> 60#include <linux/export.h> 61 62#define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */ 63 64struct raw_hashinfo raw_v6_hashinfo = { 65 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock), 66}; 67EXPORT_SYMBOL_GPL(raw_v6_hashinfo); 68 69struct sock *__raw_v6_lookup(struct net *net, struct sock *sk, 70 unsigned short num, const struct in6_addr *loc_addr, 71 const struct in6_addr *rmt_addr, int dif, int sdif) 72{ 73 bool is_multicast = ipv6_addr_is_multicast(loc_addr); 74 75 sk_for_each_from(sk) 76 if (inet_sk(sk)->inet_num == num) { 77 78 if (!net_eq(sock_net(sk), net)) 79 continue; 80 81 if (!ipv6_addr_any(&sk->sk_v6_daddr) && 82 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) 83 continue; 84 85 if (!raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, 86 dif, sdif)) 87 continue; 88 89 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) { 90 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)) 91 goto found; 92 if (is_multicast && 93 inet6_mc_check(sk, loc_addr, rmt_addr)) 94 goto found; 95 continue; 96 } 97 goto found; 98 } 99 sk = NULL; 100found: 101 return sk; 102} 103EXPORT_SYMBOL_GPL(__raw_v6_lookup); 104 105/* 106 * 0 - deliver 107 * 1 - block 108 */ 109static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb) 110{ 111 struct icmp6hdr _hdr; 112 const struct icmp6hdr *hdr; 113 114 /* We require only the four bytes of the ICMPv6 header, not any 115 * additional bytes of message body in "struct icmp6hdr". 116 */ 117 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 118 ICMPV6_HDRLEN, &_hdr); 119 if (hdr) { 120 const __u32 *data = &raw6_sk(sk)->filter.data[0]; 121 unsigned int type = hdr->icmp6_type; 122 123 return (data[type >> 5] & (1U << (type & 31))) != 0; 124 } 125 return 1; 126} 127 128#if IS_ENABLED(CONFIG_IPV6_MIP6) 129typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb); 130 131static mh_filter_t __rcu *mh_filter __read_mostly; 132 133int rawv6_mh_filter_register(mh_filter_t filter) 134{ 135 rcu_assign_pointer(mh_filter, filter); 136 return 0; 137} 138EXPORT_SYMBOL(rawv6_mh_filter_register); 139 140int rawv6_mh_filter_unregister(mh_filter_t filter) 141{ 142 RCU_INIT_POINTER(mh_filter, NULL); 143 synchronize_rcu(); 144 return 0; 145} 146EXPORT_SYMBOL(rawv6_mh_filter_unregister); 147 148#endif 149 150/* 151 * demultiplex raw sockets. 152 * (should consider queueing the skb in the sock receive_queue 153 * without calling rawv6.c) 154 * 155 * Caller owns SKB so we must make clones. 156 */ 157static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) 158{ 159 const struct in6_addr *saddr; 160 const struct in6_addr *daddr; 161 struct sock *sk; 162 bool delivered = false; 163 __u8 hash; 164 struct net *net; 165 166 saddr = &ipv6_hdr(skb)->saddr; 167 daddr = saddr + 1; 168 169 hash = nexthdr & (RAW_HTABLE_SIZE - 1); 170 171 read_lock(&raw_v6_hashinfo.lock); 172 sk = sk_head(&raw_v6_hashinfo.ht[hash]); 173 174 if (!sk) 175 goto out; 176 177 net = dev_net(skb->dev); 178 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, 179 inet6_iif(skb), inet6_sdif(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_ct(clone); 219 rawv6_rcv(sk, clone); 220 } 221 } 222 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr, 223 inet6_iif(skb), inet6_sdif(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 } 287 288 if (sk->sk_bound_dev_if) { 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 !ipv6_can_nonlocal_bind(sock_net(sk), inet)) { 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; 315out_unlock: 316 rcu_read_unlock(); 317out: 318 release_sock(sk); 319 return err; 320} 321 322static 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 361void 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), 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 390static 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 */ 416int 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 462static 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 524out_free: 525 skb_free_datagram(sk, skb); 526out: 527 return err; 528 529csum_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 539static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 540 struct raw6_sock *rp) 541{ 542 struct ipv6_txoptions *opt; 543 struct sk_buff *skb; 544 int err = 0; 545 int offset; 546 int len; 547 int total_len; 548 __wsum tmp_csum; 549 __sum16 csum; 550 551 if (!rp->checksum) 552 goto send; 553 554 skb = skb_peek(&sk->sk_write_queue); 555 if (!skb) 556 goto out; 557 558 offset = rp->offset; 559 total_len = inet_sk(sk)->cork.base.length; 560 opt = inet6_sk(sk)->cork.opt; 561 total_len -= opt ? opt->opt_flen : 0; 562 563 if (offset >= total_len - 1) { 564 err = -EINVAL; 565 ip6_flush_pending_frames(sk); 566 goto out; 567 } 568 569 /* should be check HW csum miyazawa */ 570 if (skb_queue_len(&sk->sk_write_queue) == 1) { 571 /* 572 * Only one fragment on the socket. 573 */ 574 tmp_csum = skb->csum; 575 } else { 576 struct sk_buff *csum_skb = NULL; 577 tmp_csum = 0; 578 579 skb_queue_walk(&sk->sk_write_queue, skb) { 580 tmp_csum = csum_add(tmp_csum, skb->csum); 581 582 if (csum_skb) 583 continue; 584 585 len = skb->len - skb_transport_offset(skb); 586 if (offset >= len) { 587 offset -= len; 588 continue; 589 } 590 591 csum_skb = skb; 592 } 593 594 skb = csum_skb; 595 } 596 597 offset += skb_transport_offset(skb); 598 err = skb_copy_bits(skb, offset, &csum, 2); 599 if (err < 0) { 600 ip6_flush_pending_frames(sk); 601 goto out; 602 } 603 604 /* in case cksum was not initialized */ 605 if (unlikely(csum)) 606 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); 607 608 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 609 total_len, fl6->flowi6_proto, tmp_csum); 610 611 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP) 612 csum = CSUM_MANGLED_0; 613 614 BUG_ON(skb_store_bits(skb, offset, &csum, 2)); 615 616send: 617 err = ip6_push_pending_frames(sk); 618out: 619 return err; 620} 621 622static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length, 623 struct flowi6 *fl6, struct dst_entry **dstp, 624 unsigned int flags, const struct sockcm_cookie *sockc) 625{ 626 struct ipv6_pinfo *np = inet6_sk(sk); 627 struct net *net = sock_net(sk); 628 struct ipv6hdr *iph; 629 struct sk_buff *skb; 630 int err; 631 struct rt6_info *rt = (struct rt6_info *)*dstp; 632 int hlen = LL_RESERVED_SPACE(rt->dst.dev); 633 int tlen = rt->dst.dev->needed_tailroom; 634 635 if (length > rt->dst.dev->mtu) { 636 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu); 637 return -EMSGSIZE; 638 } 639 if (length < sizeof(struct ipv6hdr)) 640 return -EINVAL; 641 if (flags&MSG_PROBE) 642 goto out; 643 644 skb = sock_alloc_send_skb(sk, 645 length + hlen + tlen + 15, 646 flags & MSG_DONTWAIT, &err); 647 if (!skb) 648 goto error; 649 skb_reserve(skb, hlen); 650 651 skb->protocol = htons(ETH_P_IPV6); 652 skb->priority = sk->sk_priority; 653 skb->mark = sockc->mark; 654 skb->tstamp = sockc->transmit_time; 655 656 skb_put(skb, length); 657 skb_reset_network_header(skb); 658 iph = ipv6_hdr(skb); 659 660 skb->ip_summed = CHECKSUM_NONE; 661 662 skb_setup_tx_timestamp(skb, sockc->tsflags); 663 664 if (flags & MSG_CONFIRM) 665 skb_set_dst_pending_confirm(skb, 1); 666 667 skb->transport_header = skb->network_header; 668 err = memcpy_from_msg(iph, msg, length); 669 if (err) { 670 err = -EFAULT; 671 kfree_skb(skb); 672 goto error; 673 } 674 675 skb_dst_set(skb, &rt->dst); 676 *dstp = NULL; 677 678 /* if egress device is enslaved to an L3 master device pass the 679 * skb to its handler for processing 680 */ 681 skb = l3mdev_ip6_out(sk, skb); 682 if (unlikely(!skb)) 683 return 0; 684 685 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev 686 * in the error path. Since skb has been freed, the dst could 687 * have been queued for deletion. 688 */ 689 rcu_read_lock(); 690 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); 691 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb, 692 NULL, rt->dst.dev, dst_output); 693 if (err > 0) 694 err = net_xmit_errno(err); 695 if (err) { 696 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 697 rcu_read_unlock(); 698 goto error_check; 699 } 700 rcu_read_unlock(); 701out: 702 return 0; 703 704error: 705 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); 706error_check: 707 if (err == -ENOBUFS && !np->recverr) 708 err = 0; 709 return err; 710} 711 712struct raw6_frag_vec { 713 struct msghdr *msg; 714 int hlen; 715 char c[4]; 716}; 717 718static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6) 719{ 720 int err = 0; 721 switch (fl6->flowi6_proto) { 722 case IPPROTO_ICMPV6: 723 rfv->hlen = 2; 724 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); 725 if (!err) { 726 fl6->fl6_icmp_type = rfv->c[0]; 727 fl6->fl6_icmp_code = rfv->c[1]; 728 } 729 break; 730 case IPPROTO_MH: 731 rfv->hlen = 4; 732 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); 733 if (!err) 734 fl6->fl6_mh_type = rfv->c[2]; 735 } 736 return err; 737} 738 739static int raw6_getfrag(void *from, char *to, int offset, int len, int odd, 740 struct sk_buff *skb) 741{ 742 struct raw6_frag_vec *rfv = from; 743 744 if (offset < rfv->hlen) { 745 int copy = min(rfv->hlen - offset, len); 746 747 if (skb->ip_summed == CHECKSUM_PARTIAL) 748 memcpy(to, rfv->c + offset, copy); 749 else 750 skb->csum = csum_block_add( 751 skb->csum, 752 csum_partial_copy_nocheck(rfv->c + offset, 753 to, copy), 754 odd); 755 756 odd = 0; 757 offset += copy; 758 to += copy; 759 len -= copy; 760 761 if (!len) 762 return 0; 763 } 764 765 offset -= rfv->hlen; 766 767 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); 768} 769 770static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 771{ 772 struct ipv6_txoptions *opt_to_free = NULL; 773 struct ipv6_txoptions opt_space; 774 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 775 struct in6_addr *daddr, *final_p, final; 776 struct inet_sock *inet = inet_sk(sk); 777 struct ipv6_pinfo *np = inet6_sk(sk); 778 struct raw6_sock *rp = raw6_sk(sk); 779 struct ipv6_txoptions *opt = NULL; 780 struct ip6_flowlabel *flowlabel = NULL; 781 struct dst_entry *dst = NULL; 782 struct raw6_frag_vec rfv; 783 struct flowi6 fl6; 784 struct ipcm6_cookie ipc6; 785 int addr_len = msg->msg_namelen; 786 int hdrincl; 787 u16 proto; 788 int err; 789 790 /* Rough check on arithmetic overflow, 791 better check is made in ip6_append_data(). 792 */ 793 if (len > INT_MAX) 794 return -EMSGSIZE; 795 796 /* Mirror BSD error message compatibility */ 797 if (msg->msg_flags & MSG_OOB) 798 return -EOPNOTSUPP; 799 800 /* hdrincl should be READ_ONCE(inet->hdrincl) 801 * but READ_ONCE() doesn't work with bit fields. 802 * Doing this indirectly yields the same result. 803 */ 804 hdrincl = inet->hdrincl; 805 hdrincl = READ_ONCE(hdrincl); 806 807 /* 808 * Get and verify the address. 809 */ 810 memset(&fl6, 0, sizeof(fl6)); 811 812 fl6.flowi6_mark = sk->sk_mark; 813 fl6.flowi6_uid = sk->sk_uid; 814 815 ipcm6_init(&ipc6); 816 ipc6.sockc.tsflags = sk->sk_tsflags; 817 ipc6.sockc.mark = sk->sk_mark; 818 819 if (sin6) { 820 if (addr_len < SIN6_LEN_RFC2133) 821 return -EINVAL; 822 823 if (sin6->sin6_family && sin6->sin6_family != AF_INET6) 824 return -EAFNOSUPPORT; 825 826 /* port is the proto value [0..255] carried in nexthdr */ 827 proto = ntohs(sin6->sin6_port); 828 829 if (!proto) 830 proto = inet->inet_num; 831 else if (proto != inet->inet_num && 832 inet->inet_num != IPPROTO_RAW) 833 return -EINVAL; 834 835 if (proto > 255) 836 return -EINVAL; 837 838 daddr = &sin6->sin6_addr; 839 if (np->sndflow) { 840 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 841 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 842 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 843 if (IS_ERR(flowlabel)) 844 return -EINVAL; 845 } 846 } 847 848 /* 849 * Otherwise it will be difficult to maintain 850 * sk->sk_dst_cache. 851 */ 852 if (sk->sk_state == TCP_ESTABLISHED && 853 ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) 854 daddr = &sk->sk_v6_daddr; 855 856 if (addr_len >= sizeof(struct sockaddr_in6) && 857 sin6->sin6_scope_id && 858 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) 859 fl6.flowi6_oif = sin6->sin6_scope_id; 860 } else { 861 if (sk->sk_state != TCP_ESTABLISHED) 862 return -EDESTADDRREQ; 863 864 proto = inet->inet_num; 865 daddr = &sk->sk_v6_daddr; 866 fl6.flowlabel = np->flow_label; 867 } 868 869 if (fl6.flowi6_oif == 0) 870 fl6.flowi6_oif = sk->sk_bound_dev_if; 871 872 if (msg->msg_controllen) { 873 opt = &opt_space; 874 memset(opt, 0, sizeof(struct ipv6_txoptions)); 875 opt->tot_len = sizeof(struct ipv6_txoptions); 876 ipc6.opt = opt; 877 878 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6); 879 if (err < 0) { 880 fl6_sock_release(flowlabel); 881 return err; 882 } 883 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 884 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 885 if (IS_ERR(flowlabel)) 886 return -EINVAL; 887 } 888 if (!(opt->opt_nflen|opt->opt_flen)) 889 opt = NULL; 890 } 891 if (!opt) { 892 opt = txopt_get(np); 893 opt_to_free = opt; 894 } 895 if (flowlabel) 896 opt = fl6_merge_options(&opt_space, flowlabel, opt); 897 opt = ipv6_fixup_options(&opt_space, opt); 898 899 fl6.flowi6_proto = proto; 900 fl6.flowi6_mark = ipc6.sockc.mark; 901 902 if (!hdrincl) { 903 rfv.msg = msg; 904 rfv.hlen = 0; 905 err = rawv6_probe_proto_opt(&rfv, &fl6); 906 if (err) 907 goto out; 908 } 909 910 if (!ipv6_addr_any(daddr)) 911 fl6.daddr = *daddr; 912 else 913 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 914 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 915 fl6.saddr = np->saddr; 916 917 final_p = fl6_update_dst(&fl6, opt, &final); 918 919 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) 920 fl6.flowi6_oif = np->mcast_oif; 921 else if (!fl6.flowi6_oif) 922 fl6.flowi6_oif = np->ucast_oif; 923 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 924 925 if (hdrincl) 926 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH; 927 928 if (ipc6.tclass < 0) 929 ipc6.tclass = np->tclass; 930 931 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); 932 933 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 934 if (IS_ERR(dst)) { 935 err = PTR_ERR(dst); 936 goto out; 937 } 938 if (ipc6.hlimit < 0) 939 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 940 941 if (ipc6.dontfrag < 0) 942 ipc6.dontfrag = np->dontfrag; 943 944 if (msg->msg_flags&MSG_CONFIRM) 945 goto do_confirm; 946 947back_from_confirm: 948 if (hdrincl) 949 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, 950 msg->msg_flags, &ipc6.sockc); 951 else { 952 ipc6.opt = opt; 953 lock_sock(sk); 954 err = ip6_append_data(sk, raw6_getfrag, &rfv, 955 len, 0, &ipc6, &fl6, (struct rt6_info *)dst, 956 msg->msg_flags); 957 958 if (err) 959 ip6_flush_pending_frames(sk); 960 else if (!(msg->msg_flags & MSG_MORE)) 961 err = rawv6_push_pending_frames(sk, &fl6, rp); 962 release_sock(sk); 963 } 964done: 965 dst_release(dst); 966out: 967 fl6_sock_release(flowlabel); 968 txopt_put(opt_to_free); 969 return err < 0 ? err : len; 970do_confirm: 971 if (msg->msg_flags & MSG_PROBE) 972 dst_confirm_neigh(dst, &fl6.daddr); 973 if (!(msg->msg_flags & MSG_PROBE) || len) 974 goto back_from_confirm; 975 err = 0; 976 goto done; 977} 978 979static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, 980 sockptr_t optval, int optlen) 981{ 982 switch (optname) { 983 case ICMPV6_FILTER: 984 if (optlen > sizeof(struct icmp6_filter)) 985 optlen = sizeof(struct icmp6_filter); 986 if (copy_from_sockptr(&raw6_sk(sk)->filter, optval, optlen)) 987 return -EFAULT; 988 return 0; 989 default: 990 return -ENOPROTOOPT; 991 } 992 993 return 0; 994} 995 996static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, 997 char __user *optval, int __user *optlen) 998{ 999 int len; 1000 1001 switch (optname) { 1002 case ICMPV6_FILTER: 1003 if (get_user(len, optlen)) 1004 return -EFAULT; 1005 if (len < 0) 1006 return -EINVAL; 1007 if (len > sizeof(struct icmp6_filter)) 1008 len = sizeof(struct icmp6_filter); 1009 if (put_user(len, optlen)) 1010 return -EFAULT; 1011 if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) 1012 return -EFAULT; 1013 return 0; 1014 default: 1015 return -ENOPROTOOPT; 1016 } 1017 1018 return 0; 1019} 1020 1021 1022static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, 1023 sockptr_t optval, unsigned int optlen) 1024{ 1025 struct raw6_sock *rp = raw6_sk(sk); 1026 int val; 1027 1028 if (optlen < sizeof(val)) 1029 return -EINVAL; 1030 1031 if (copy_from_sockptr(&val, optval, sizeof(val))) 1032 return -EFAULT; 1033 1034 switch (optname) { 1035 case IPV6_HDRINCL: 1036 if (sk->sk_type != SOCK_RAW) 1037 return -EINVAL; 1038 inet_sk(sk)->hdrincl = !!val; 1039 return 0; 1040 case IPV6_CHECKSUM: 1041 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && 1042 level == IPPROTO_IPV6) { 1043 /* 1044 * RFC3542 tells that IPV6_CHECKSUM socket 1045 * option in the IPPROTO_IPV6 level is not 1046 * allowed on ICMPv6 sockets. 1047 * If you want to set it, use IPPROTO_RAW 1048 * level IPV6_CHECKSUM socket option 1049 * (Linux extension). 1050 */ 1051 return -EINVAL; 1052 } 1053 1054 /* You may get strange result with a positive odd offset; 1055 RFC2292bis agrees with me. */ 1056 if (val > 0 && (val&1)) 1057 return -EINVAL; 1058 if (val < 0) { 1059 rp->checksum = 0; 1060 } else { 1061 rp->checksum = 1; 1062 rp->offset = val; 1063 } 1064 1065 return 0; 1066 1067 default: 1068 return -ENOPROTOOPT; 1069 } 1070} 1071 1072static int rawv6_setsockopt(struct sock *sk, int level, int optname, 1073 sockptr_t optval, unsigned int optlen) 1074{ 1075 switch (level) { 1076 case SOL_RAW: 1077 break; 1078 1079 case SOL_ICMPV6: 1080 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1081 return -EOPNOTSUPP; 1082 return rawv6_seticmpfilter(sk, level, optname, optval, optlen); 1083 case SOL_IPV6: 1084 if (optname == IPV6_CHECKSUM || 1085 optname == IPV6_HDRINCL) 1086 break; 1087 fallthrough; 1088 default: 1089 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1090 } 1091 1092 return do_rawv6_setsockopt(sk, level, optname, optval, optlen); 1093} 1094 1095static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, 1096 char __user *optval, int __user *optlen) 1097{ 1098 struct raw6_sock *rp = raw6_sk(sk); 1099 int val, len; 1100 1101 if (get_user(len, optlen)) 1102 return -EFAULT; 1103 1104 switch (optname) { 1105 case IPV6_HDRINCL: 1106 val = inet_sk(sk)->hdrincl; 1107 break; 1108 case IPV6_CHECKSUM: 1109 /* 1110 * We allow getsockopt() for IPPROTO_IPV6-level 1111 * IPV6_CHECKSUM socket option on ICMPv6 sockets 1112 * since RFC3542 is silent about it. 1113 */ 1114 if (rp->checksum == 0) 1115 val = -1; 1116 else 1117 val = rp->offset; 1118 break; 1119 1120 default: 1121 return -ENOPROTOOPT; 1122 } 1123 1124 len = min_t(unsigned int, sizeof(int), len); 1125 1126 if (put_user(len, optlen)) 1127 return -EFAULT; 1128 if (copy_to_user(optval, &val, len)) 1129 return -EFAULT; 1130 return 0; 1131} 1132 1133static int rawv6_getsockopt(struct sock *sk, int level, int optname, 1134 char __user *optval, int __user *optlen) 1135{ 1136 switch (level) { 1137 case SOL_RAW: 1138 break; 1139 1140 case SOL_ICMPV6: 1141 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) 1142 return -EOPNOTSUPP; 1143 return rawv6_geticmpfilter(sk, level, optname, optval, optlen); 1144 case SOL_IPV6: 1145 if (optname == IPV6_CHECKSUM || 1146 optname == IPV6_HDRINCL) 1147 break; 1148 fallthrough; 1149 default: 1150 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1151 } 1152 1153 return do_rawv6_getsockopt(sk, level, optname, optval, optlen); 1154} 1155 1156static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) 1157{ 1158 switch (cmd) { 1159 case SIOCOUTQ: { 1160 int amount = sk_wmem_alloc_get(sk); 1161 1162 return put_user(amount, (int __user *)arg); 1163 } 1164 case SIOCINQ: { 1165 struct sk_buff *skb; 1166 int amount = 0; 1167 1168 spin_lock_bh(&sk->sk_receive_queue.lock); 1169 skb = skb_peek(&sk->sk_receive_queue); 1170 if (skb) 1171 amount = skb->len; 1172 spin_unlock_bh(&sk->sk_receive_queue.lock); 1173 return put_user(amount, (int __user *)arg); 1174 } 1175 1176 default: 1177#ifdef CONFIG_IPV6_MROUTE 1178 return ip6mr_ioctl(sk, cmd, (void __user *)arg); 1179#else 1180 return -ENOIOCTLCMD; 1181#endif 1182 } 1183} 1184 1185#ifdef CONFIG_COMPAT 1186static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 1187{ 1188 switch (cmd) { 1189 case SIOCOUTQ: 1190 case SIOCINQ: 1191 return -ENOIOCTLCMD; 1192 default: 1193#ifdef CONFIG_IPV6_MROUTE 1194 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg)); 1195#else 1196 return -ENOIOCTLCMD; 1197#endif 1198 } 1199} 1200#endif 1201 1202static void rawv6_close(struct sock *sk, long timeout) 1203{ 1204 if (inet_sk(sk)->inet_num == IPPROTO_RAW) 1205 ip6_ra_control(sk, -1); 1206 ip6mr_sk_done(sk); 1207 sk_common_release(sk); 1208} 1209 1210static void raw6_destroy(struct sock *sk) 1211{ 1212 lock_sock(sk); 1213 ip6_flush_pending_frames(sk); 1214 release_sock(sk); 1215} 1216 1217static int rawv6_init_sk(struct sock *sk) 1218{ 1219 struct raw6_sock *rp = raw6_sk(sk); 1220 1221 switch (inet_sk(sk)->inet_num) { 1222 case IPPROTO_ICMPV6: 1223 rp->checksum = 1; 1224 rp->offset = 2; 1225 break; 1226 case IPPROTO_MH: 1227 rp->checksum = 1; 1228 rp->offset = 4; 1229 break; 1230 default: 1231 break; 1232 } 1233 return 0; 1234} 1235 1236struct proto rawv6_prot = { 1237 .name = "RAWv6", 1238 .owner = THIS_MODULE, 1239 .close = rawv6_close, 1240 .destroy = raw6_destroy, 1241 .connect = ip6_datagram_connect_v6_only, 1242 .disconnect = __udp_disconnect, 1243 .ioctl = rawv6_ioctl, 1244 .init = rawv6_init_sk, 1245 .setsockopt = rawv6_setsockopt, 1246 .getsockopt = rawv6_getsockopt, 1247 .sendmsg = rawv6_sendmsg, 1248 .recvmsg = rawv6_recvmsg, 1249 .bind = rawv6_bind, 1250 .backlog_rcv = rawv6_rcv_skb, 1251 .hash = raw_hash_sk, 1252 .unhash = raw_unhash_sk, 1253 .obj_size = sizeof(struct raw6_sock), 1254 .useroffset = offsetof(struct raw6_sock, filter), 1255 .usersize = sizeof_field(struct raw6_sock, filter), 1256 .h.raw_hash = &raw_v6_hashinfo, 1257#ifdef CONFIG_COMPAT 1258 .compat_ioctl = compat_rawv6_ioctl, 1259#endif 1260 .diag_destroy = raw_abort, 1261}; 1262 1263#ifdef CONFIG_PROC_FS 1264static int raw6_seq_show(struct seq_file *seq, void *v) 1265{ 1266 if (v == SEQ_START_TOKEN) { 1267 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); 1268 } else { 1269 struct sock *sp = v; 1270 __u16 srcp = inet_sk(sp)->inet_num; 1271 ip6_dgram_sock_seq_show(seq, v, srcp, 0, 1272 raw_seq_private(seq)->bucket); 1273 } 1274 return 0; 1275} 1276 1277static const struct seq_operations raw6_seq_ops = { 1278 .start = raw_seq_start, 1279 .next = raw_seq_next, 1280 .stop = raw_seq_stop, 1281 .show = raw6_seq_show, 1282}; 1283 1284static int __net_init raw6_init_net(struct net *net) 1285{ 1286 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops, 1287 sizeof(struct raw_iter_state), &raw_v6_hashinfo)) 1288 return -ENOMEM; 1289 1290 return 0; 1291} 1292 1293static void __net_exit raw6_exit_net(struct net *net) 1294{ 1295 remove_proc_entry("raw6", net->proc_net); 1296} 1297 1298static struct pernet_operations raw6_net_ops = { 1299 .init = raw6_init_net, 1300 .exit = raw6_exit_net, 1301}; 1302 1303int __init raw6_proc_init(void) 1304{ 1305 return register_pernet_subsys(&raw6_net_ops); 1306} 1307 1308void raw6_proc_exit(void) 1309{ 1310 unregister_pernet_subsys(&raw6_net_ops); 1311} 1312#endif /* CONFIG_PROC_FS */ 1313 1314/* Same as inet6_dgram_ops, sans udp_poll. */ 1315const struct proto_ops inet6_sockraw_ops = { 1316 .family = PF_INET6, 1317 .owner = THIS_MODULE, 1318 .release = inet6_release, 1319 .bind = inet6_bind, 1320 .connect = inet_dgram_connect, /* ok */ 1321 .socketpair = sock_no_socketpair, /* a do nothing */ 1322 .accept = sock_no_accept, /* a do nothing */ 1323 .getname = inet6_getname, 1324 .poll = datagram_poll, /* ok */ 1325 .ioctl = inet6_ioctl, /* must change */ 1326 .gettstamp = sock_gettstamp, 1327 .listen = sock_no_listen, /* ok */ 1328 .shutdown = inet_shutdown, /* ok */ 1329 .setsockopt = sock_common_setsockopt, /* ok */ 1330 .getsockopt = sock_common_getsockopt, /* ok */ 1331 .sendmsg = inet_sendmsg, /* ok */ 1332 .recvmsg = sock_common_recvmsg, /* ok */ 1333 .mmap = sock_no_mmap, 1334 .sendpage = sock_no_sendpage, 1335#ifdef CONFIG_COMPAT 1336 .compat_ioctl = inet6_compat_ioctl, 1337#endif 1338}; 1339 1340static struct inet_protosw rawv6_protosw = { 1341 .type = SOCK_RAW, 1342 .protocol = IPPROTO_IP, /* wild card */ 1343 .prot = &rawv6_prot, 1344 .ops = &inet6_sockraw_ops, 1345 .flags = INET_PROTOSW_REUSE, 1346}; 1347 1348int __init rawv6_init(void) 1349{ 1350 return inet6_register_protosw(&rawv6_protosw); 1351} 1352 1353void rawv6_exit(void) 1354{ 1355 inet6_unregister_protosw(&rawv6_protosw); 1356} 1357