1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * RAW - implementation of IP "raw" sockets. 8 * 9 * Authors: Ross Biro 10 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 11 * 12 * Fixes: 13 * Alan Cox : verify_area() fixed up 14 * Alan Cox : ICMP error handling 15 * Alan Cox : EMSGSIZE if you send too big a packet 16 * Alan Cox : Now uses generic datagrams and shared 17 * skbuff library. No more peek crashes, 18 * no more backlogs 19 * Alan Cox : Checks sk->broadcast. 20 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram 21 * Alan Cox : Raw passes ip options too 22 * Alan Cox : Setsocketopt added 23 * Alan Cox : Fixed error return for broadcasts 24 * Alan Cox : Removed wake_up calls 25 * Alan Cox : Use ttl/tos 26 * Alan Cox : Cleaned up old debugging 27 * Alan Cox : Use new kernel side addresses 28 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets. 29 * Alan Cox : BSD style RAW socket demultiplexing. 30 * Alan Cox : Beginnings of mrouted support. 31 * Alan Cox : Added IP_HDRINCL option. 32 * Alan Cox : Skip broadcast check if BSDism set. 33 * David S. Miller : New socket lookup architecture. 34 */ 35 36 #include <linux/types.h> 37 #include <linux/atomic.h> 38 #include <asm/byteorder.h> 39 #include <asm/current.h> 40 #include <linux/uaccess.h> 41 #include <asm/ioctls.h> 42 #include <linux/stddef.h> 43 #include <linux/slab.h> 44 #include <linux/errno.h> 45 #include <linux/kernel.h> 46 #include <linux/export.h> 47 #include <linux/spinlock.h> 48 #include <linux/sockios.h> 49 #include <linux/socket.h> 50 #include <linux/in.h> 51 #include <linux/mroute.h> 52 #include <linux/netdevice.h> 53 #include <linux/in_route.h> 54 #include <linux/route.h> 55 #include <linux/skbuff.h> 56 #include <linux/igmp.h> 57 #include <net/net_namespace.h> 58 #include <net/dst.h> 59 #include <net/sock.h> 60 #include <linux/ip.h> 61 #include <linux/net.h> 62 #include <net/ip.h> 63 #include <net/icmp.h> 64 #include <net/udp.h> 65 #include <net/raw.h> 66 #include <net/snmp.h> 67 #include <net/tcp_states.h> 68 #include <net/inet_common.h> 69 #include <net/checksum.h> 70 #include <net/xfrm.h> 71 #include <linux/rtnetlink.h> 72 #include <linux/proc_fs.h> 73 #include <linux/seq_file.h> 74 #include <linux/netfilter.h> 75 #include <linux/netfilter_ipv4.h> 76 #include <linux/compat.h> 77 #include <linux/uio.h> 78 79 struct raw_frag_vec { 80 struct msghdr *msg; 81 union { 82 struct icmphdr icmph; 83 char c[1]; 84 } hdr; 85 int hlen; 86 }; 87 88 struct raw_hashinfo raw_v4_hashinfo; 89 EXPORT_SYMBOL_GPL(raw_v4_hashinfo); 90 91 int raw_hash_sk(struct sock *sk) 92 { 93 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 94 struct hlist_head *hlist; 95 96 hlist = &h->ht[raw_hashfunc(sock_net(sk), inet_sk(sk)->inet_num)]; 97 98 spin_lock(&h->lock); 99 sk_add_node_rcu(sk, hlist); 100 sock_set_flag(sk, SOCK_RCU_FREE); 101 spin_unlock(&h->lock); 102 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 103 104 return 0; 105 } 106 EXPORT_SYMBOL_GPL(raw_hash_sk); 107 108 void raw_unhash_sk(struct sock *sk) 109 { 110 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 111 112 spin_lock(&h->lock); 113 if (sk_del_node_init_rcu(sk)) 114 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 115 spin_unlock(&h->lock); 116 } 117 EXPORT_SYMBOL_GPL(raw_unhash_sk); 118 119 bool raw_v4_match(struct net *net, const struct sock *sk, unsigned short num, 120 __be32 raddr, __be32 laddr, int dif, int sdif) 121 { 122 const struct inet_sock *inet = inet_sk(sk); 123 124 if (net_eq(sock_net(sk), net) && inet->inet_num == num && 125 !(inet->inet_daddr && inet->inet_daddr != raddr) && 126 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) && 127 raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif)) 128 return true; 129 return false; 130 } 131 EXPORT_SYMBOL_GPL(raw_v4_match); 132 133 /* 134 * 0 - deliver 135 * 1 - block 136 */ 137 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb) 138 { 139 struct icmphdr _hdr; 140 const struct icmphdr *hdr; 141 142 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 143 sizeof(_hdr), &_hdr); 144 if (!hdr) 145 return 1; 146 147 if (hdr->type < 32) { 148 __u32 data = raw_sk(sk)->filter.data; 149 150 return ((1U << hdr->type) & data) != 0; 151 } 152 153 /* Do not block unknown ICMP types */ 154 return 0; 155 } 156 157 /* IP input processing comes here for RAW socket delivery. 158 * Caller owns SKB, so we must make clones. 159 * 160 * RFC 1122: SHOULD pass TOS value up to the transport layer. 161 * -> It does. And not only TOS, but all IP header. 162 */ 163 static int raw_v4_input(struct net *net, struct sk_buff *skb, 164 const struct iphdr *iph, int hash) 165 { 166 int sdif = inet_sdif(skb); 167 struct hlist_head *hlist; 168 int dif = inet_iif(skb); 169 int delivered = 0; 170 struct sock *sk; 171 172 hlist = &raw_v4_hashinfo.ht[hash]; 173 rcu_read_lock(); 174 sk_for_each_rcu(sk, hlist) { 175 if (!raw_v4_match(net, sk, iph->protocol, 176 iph->saddr, iph->daddr, dif, sdif)) 177 continue; 178 delivered = 1; 179 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) && 180 ip_mc_sf_allow(sk, iph->daddr, iph->saddr, 181 skb->dev->ifindex, sdif)) { 182 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 183 184 /* Not releasing hash table! */ 185 if (clone) 186 raw_rcv(sk, clone); 187 } 188 } 189 rcu_read_unlock(); 190 return delivered; 191 } 192 193 int raw_local_deliver(struct sk_buff *skb, int protocol) 194 { 195 struct net *net = dev_net(skb->dev); 196 197 return raw_v4_input(net, skb, ip_hdr(skb), 198 raw_hashfunc(net, protocol)); 199 } 200 201 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) 202 { 203 struct inet_sock *inet = inet_sk(sk); 204 const int type = icmp_hdr(skb)->type; 205 const int code = icmp_hdr(skb)->code; 206 int harderr = 0; 207 bool recverr; 208 int err = 0; 209 210 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) 211 ipv4_sk_update_pmtu(skb, sk, info); 212 else if (type == ICMP_REDIRECT) { 213 ipv4_sk_redirect(skb, sk); 214 return; 215 } 216 217 /* Report error on raw socket, if: 218 1. User requested ip_recverr. 219 2. Socket is connected (otherwise the error indication 220 is useless without ip_recverr and error is hard. 221 */ 222 recverr = inet_test_bit(RECVERR, sk); 223 if (!recverr && sk->sk_state != TCP_ESTABLISHED) 224 return; 225 226 switch (type) { 227 default: 228 case ICMP_TIME_EXCEEDED: 229 err = EHOSTUNREACH; 230 break; 231 case ICMP_SOURCE_QUENCH: 232 return; 233 case ICMP_PARAMETERPROB: 234 err = EPROTO; 235 harderr = 1; 236 break; 237 case ICMP_DEST_UNREACH: 238 err = EHOSTUNREACH; 239 if (code > NR_ICMP_UNREACH) 240 break; 241 if (code == ICMP_FRAG_NEEDED) { 242 harderr = READ_ONCE(inet->pmtudisc) != IP_PMTUDISC_DONT; 243 err = EMSGSIZE; 244 } else { 245 err = icmp_err_convert[code].errno; 246 harderr = icmp_err_convert[code].fatal; 247 } 248 } 249 250 if (recverr) { 251 const struct iphdr *iph = (const struct iphdr *)skb->data; 252 u8 *payload = skb->data + (iph->ihl << 2); 253 254 if (inet_test_bit(HDRINCL, sk)) 255 payload = skb->data; 256 ip_icmp_error(sk, skb, err, 0, info, payload); 257 } 258 259 if (recverr || harderr) { 260 sk->sk_err = err; 261 sk_error_report(sk); 262 } 263 } 264 265 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) 266 { 267 struct net *net = dev_net(skb->dev); 268 int dif = skb->dev->ifindex; 269 int sdif = inet_sdif(skb); 270 struct hlist_head *hlist; 271 const struct iphdr *iph; 272 struct sock *sk; 273 int hash; 274 275 hash = raw_hashfunc(net, protocol); 276 hlist = &raw_v4_hashinfo.ht[hash]; 277 278 rcu_read_lock(); 279 sk_for_each_rcu(sk, hlist) { 280 iph = (const struct iphdr *)skb->data; 281 if (!raw_v4_match(net, sk, iph->protocol, 282 iph->daddr, iph->saddr, dif, sdif)) 283 continue; 284 raw_err(sk, skb, info); 285 } 286 rcu_read_unlock(); 287 } 288 289 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb) 290 { 291 enum skb_drop_reason reason; 292 293 /* Charge it to the socket. */ 294 295 ipv4_pktinfo_prepare(sk, skb, true); 296 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) { 297 kfree_skb_reason(skb, reason); 298 return NET_RX_DROP; 299 } 300 301 return NET_RX_SUCCESS; 302 } 303 304 int raw_rcv(struct sock *sk, struct sk_buff *skb) 305 { 306 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { 307 atomic_inc(&sk->sk_drops); 308 kfree_skb_reason(skb, SKB_DROP_REASON_XFRM_POLICY); 309 return NET_RX_DROP; 310 } 311 nf_reset_ct(skb); 312 313 skb_push(skb, skb->data - skb_network_header(skb)); 314 315 raw_rcv_skb(sk, skb); 316 return 0; 317 } 318 319 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4, 320 struct msghdr *msg, size_t length, 321 struct rtable **rtp, unsigned int flags, 322 const struct sockcm_cookie *sockc) 323 { 324 struct inet_sock *inet = inet_sk(sk); 325 struct net *net = sock_net(sk); 326 struct iphdr *iph; 327 struct sk_buff *skb; 328 unsigned int iphlen; 329 int err; 330 struct rtable *rt = *rtp; 331 int hlen, tlen; 332 333 if (length > rt->dst.dev->mtu) { 334 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, 335 rt->dst.dev->mtu); 336 return -EMSGSIZE; 337 } 338 if (length < sizeof(struct iphdr)) 339 return -EINVAL; 340 341 if (flags&MSG_PROBE) 342 goto out; 343 344 hlen = LL_RESERVED_SPACE(rt->dst.dev); 345 tlen = rt->dst.dev->needed_tailroom; 346 skb = sock_alloc_send_skb(sk, 347 length + hlen + tlen + 15, 348 flags & MSG_DONTWAIT, &err); 349 if (!skb) 350 goto error; 351 skb_reserve(skb, hlen); 352 353 skb->priority = READ_ONCE(sk->sk_priority); 354 skb->mark = sockc->mark; 355 skb->tstamp = sockc->transmit_time; 356 skb_dst_set(skb, &rt->dst); 357 *rtp = NULL; 358 359 skb_reset_network_header(skb); 360 iph = ip_hdr(skb); 361 skb_put(skb, length); 362 363 skb->ip_summed = CHECKSUM_NONE; 364 365 skb_setup_tx_timestamp(skb, sockc->tsflags); 366 367 if (flags & MSG_CONFIRM) 368 skb_set_dst_pending_confirm(skb, 1); 369 370 skb->transport_header = skb->network_header; 371 err = -EFAULT; 372 if (memcpy_from_msg(iph, msg, length)) 373 goto error_free; 374 375 iphlen = iph->ihl * 4; 376 377 /* 378 * We don't want to modify the ip header, but we do need to 379 * be sure that it won't cause problems later along the network 380 * stack. Specifically we want to make sure that iph->ihl is a 381 * sane value. If ihl points beyond the length of the buffer passed 382 * in, reject the frame as invalid 383 */ 384 err = -EINVAL; 385 if (iphlen > length) 386 goto error_free; 387 388 if (iphlen >= sizeof(*iph)) { 389 if (!iph->saddr) 390 iph->saddr = fl4->saddr; 391 iph->check = 0; 392 iph->tot_len = htons(length); 393 if (!iph->id) 394 ip_select_ident(net, skb, NULL); 395 396 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); 397 skb->transport_header += iphlen; 398 if (iph->protocol == IPPROTO_ICMP && 399 length >= iphlen + sizeof(struct icmphdr)) 400 icmp_out_count(net, ((struct icmphdr *) 401 skb_transport_header(skb))->type); 402 } 403 404 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, 405 net, sk, skb, NULL, rt->dst.dev, 406 dst_output); 407 if (err > 0) 408 err = net_xmit_errno(err); 409 if (err) 410 goto error; 411 out: 412 return 0; 413 414 error_free: 415 kfree_skb(skb); 416 error: 417 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS); 418 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk)) 419 err = 0; 420 return err; 421 } 422 423 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4) 424 { 425 int err; 426 427 if (fl4->flowi4_proto != IPPROTO_ICMP) 428 return 0; 429 430 /* We only need the first two bytes. */ 431 rfv->hlen = 2; 432 433 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen); 434 if (err) 435 return err; 436 437 fl4->fl4_icmp_type = rfv->hdr.icmph.type; 438 fl4->fl4_icmp_code = rfv->hdr.icmph.code; 439 440 return 0; 441 } 442 443 static int raw_getfrag(void *from, char *to, int offset, int len, int odd, 444 struct sk_buff *skb) 445 { 446 struct raw_frag_vec *rfv = from; 447 448 if (offset < rfv->hlen) { 449 int copy = min(rfv->hlen - offset, len); 450 451 if (skb->ip_summed == CHECKSUM_PARTIAL) 452 memcpy(to, rfv->hdr.c + offset, copy); 453 else 454 skb->csum = csum_block_add( 455 skb->csum, 456 csum_partial_copy_nocheck(rfv->hdr.c + offset, 457 to, copy), 458 odd); 459 460 odd = 0; 461 offset += copy; 462 to += copy; 463 len -= copy; 464 465 if (!len) 466 return 0; 467 } 468 469 offset -= rfv->hlen; 470 471 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); 472 } 473 474 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 475 { 476 struct inet_sock *inet = inet_sk(sk); 477 struct net *net = sock_net(sk); 478 struct ipcm_cookie ipc; 479 struct rtable *rt = NULL; 480 struct flowi4 fl4; 481 u8 tos, scope; 482 int free = 0; 483 __be32 daddr; 484 __be32 saddr; 485 int uc_index, err; 486 struct ip_options_data opt_copy; 487 struct raw_frag_vec rfv; 488 int hdrincl; 489 490 err = -EMSGSIZE; 491 if (len > 0xFFFF) 492 goto out; 493 494 hdrincl = inet_test_bit(HDRINCL, sk); 495 496 /* 497 * Check the flags. 498 */ 499 500 err = -EOPNOTSUPP; 501 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ 502 goto out; /* compatibility */ 503 504 /* 505 * Get and verify the address. 506 */ 507 508 if (msg->msg_namelen) { 509 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 510 err = -EINVAL; 511 if (msg->msg_namelen < sizeof(*usin)) 512 goto out; 513 if (usin->sin_family != AF_INET) { 514 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n", 515 __func__, current->comm); 516 err = -EAFNOSUPPORT; 517 if (usin->sin_family) 518 goto out; 519 } 520 daddr = usin->sin_addr.s_addr; 521 /* ANK: I did not forget to get protocol from port field. 522 * I just do not know, who uses this weirdness. 523 * IP_HDRINCL is much more convenient. 524 */ 525 } else { 526 err = -EDESTADDRREQ; 527 if (sk->sk_state != TCP_ESTABLISHED) 528 goto out; 529 daddr = inet->inet_daddr; 530 } 531 532 ipcm_init_sk(&ipc, inet); 533 /* Keep backward compat */ 534 if (hdrincl) 535 ipc.protocol = IPPROTO_RAW; 536 537 if (msg->msg_controllen) { 538 err = ip_cmsg_send(sk, msg, &ipc, false); 539 if (unlikely(err)) { 540 kfree(ipc.opt); 541 goto out; 542 } 543 if (ipc.opt) 544 free = 1; 545 } 546 547 saddr = ipc.addr; 548 ipc.addr = daddr; 549 550 if (!ipc.opt) { 551 struct ip_options_rcu *inet_opt; 552 553 rcu_read_lock(); 554 inet_opt = rcu_dereference(inet->inet_opt); 555 if (inet_opt) { 556 memcpy(&opt_copy, inet_opt, 557 sizeof(*inet_opt) + inet_opt->opt.optlen); 558 ipc.opt = &opt_copy.opt; 559 } 560 rcu_read_unlock(); 561 } 562 563 if (ipc.opt) { 564 err = -EINVAL; 565 /* Linux does not mangle headers on raw sockets, 566 * so that IP options + IP_HDRINCL is non-sense. 567 */ 568 if (hdrincl) 569 goto done; 570 if (ipc.opt->opt.srr) { 571 if (!daddr) 572 goto done; 573 daddr = ipc.opt->opt.faddr; 574 } 575 } 576 tos = get_rttos(&ipc, inet); 577 scope = ip_sendmsg_scope(inet, &ipc, msg); 578 579 uc_index = READ_ONCE(inet->uc_index); 580 if (ipv4_is_multicast(daddr)) { 581 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif)) 582 ipc.oif = READ_ONCE(inet->mc_index); 583 if (!saddr) 584 saddr = READ_ONCE(inet->mc_addr); 585 } else if (!ipc.oif) { 586 ipc.oif = uc_index; 587 } else if (ipv4_is_lbcast(daddr) && uc_index) { 588 /* oif is set, packet is to local broadcast 589 * and uc_index is set. oif is most likely set 590 * by sk_bound_dev_if. If uc_index != oif check if the 591 * oif is an L3 master and uc_index is an L3 slave. 592 * If so, we want to allow the send using the uc_index. 593 */ 594 if (ipc.oif != uc_index && 595 ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk), 596 uc_index)) { 597 ipc.oif = uc_index; 598 } 599 } 600 601 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos, scope, 602 hdrincl ? ipc.protocol : sk->sk_protocol, 603 inet_sk_flowi_flags(sk) | 604 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0), 605 daddr, saddr, 0, 0, sk->sk_uid); 606 607 if (!hdrincl) { 608 rfv.msg = msg; 609 rfv.hlen = 0; 610 611 err = raw_probe_proto_opt(&rfv, &fl4); 612 if (err) 613 goto done; 614 } 615 616 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4)); 617 rt = ip_route_output_flow(net, &fl4, sk); 618 if (IS_ERR(rt)) { 619 err = PTR_ERR(rt); 620 rt = NULL; 621 goto done; 622 } 623 624 err = -EACCES; 625 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) 626 goto done; 627 628 if (msg->msg_flags & MSG_CONFIRM) 629 goto do_confirm; 630 back_from_confirm: 631 632 if (hdrincl) 633 err = raw_send_hdrinc(sk, &fl4, msg, len, 634 &rt, msg->msg_flags, &ipc.sockc); 635 636 else { 637 if (!ipc.addr) 638 ipc.addr = fl4.daddr; 639 lock_sock(sk); 640 err = ip_append_data(sk, &fl4, raw_getfrag, 641 &rfv, len, 0, 642 &ipc, &rt, msg->msg_flags); 643 if (err) 644 ip_flush_pending_frames(sk); 645 else if (!(msg->msg_flags & MSG_MORE)) { 646 err = ip_push_pending_frames(sk, &fl4); 647 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk)) 648 err = 0; 649 } 650 release_sock(sk); 651 } 652 done: 653 if (free) 654 kfree(ipc.opt); 655 ip_rt_put(rt); 656 657 out: 658 if (err < 0) 659 return err; 660 return len; 661 662 do_confirm: 663 if (msg->msg_flags & MSG_PROBE) 664 dst_confirm_neigh(&rt->dst, &fl4.daddr); 665 if (!(msg->msg_flags & MSG_PROBE) || len) 666 goto back_from_confirm; 667 err = 0; 668 goto done; 669 } 670 671 static void raw_close(struct sock *sk, long timeout) 672 { 673 /* 674 * Raw sockets may have direct kernel references. Kill them. 675 */ 676 ip_ra_control(sk, 0, NULL); 677 678 sk_common_release(sk); 679 } 680 681 static void raw_destroy(struct sock *sk) 682 { 683 lock_sock(sk); 684 ip_flush_pending_frames(sk); 685 release_sock(sk); 686 } 687 688 /* This gets rid of all the nasties in af_inet. -DaveM */ 689 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 690 { 691 struct inet_sock *inet = inet_sk(sk); 692 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 693 struct net *net = sock_net(sk); 694 u32 tb_id = RT_TABLE_LOCAL; 695 int ret = -EINVAL; 696 int chk_addr_ret; 697 698 lock_sock(sk); 699 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) 700 goto out; 701 702 if (sk->sk_bound_dev_if) 703 tb_id = l3mdev_fib_table_by_index(net, 704 sk->sk_bound_dev_if) ? : tb_id; 705 706 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); 707 708 ret = -EADDRNOTAVAIL; 709 if (!inet_addr_valid_or_nonlocal(net, inet, addr->sin_addr.s_addr, 710 chk_addr_ret)) 711 goto out; 712 713 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 714 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 715 inet->inet_saddr = 0; /* Use device */ 716 sk_dst_reset(sk); 717 ret = 0; 718 out: 719 release_sock(sk); 720 return ret; 721 } 722 723 /* 724 * This should be easy, if there is something there 725 * we return it, otherwise we block. 726 */ 727 728 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 729 int flags, int *addr_len) 730 { 731 struct inet_sock *inet = inet_sk(sk); 732 size_t copied = 0; 733 int err = -EOPNOTSUPP; 734 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 735 struct sk_buff *skb; 736 737 if (flags & MSG_OOB) 738 goto out; 739 740 if (flags & MSG_ERRQUEUE) { 741 err = ip_recv_error(sk, msg, len, addr_len); 742 goto out; 743 } 744 745 skb = skb_recv_datagram(sk, flags, &err); 746 if (!skb) 747 goto out; 748 749 copied = skb->len; 750 if (len < copied) { 751 msg->msg_flags |= MSG_TRUNC; 752 copied = len; 753 } 754 755 err = skb_copy_datagram_msg(skb, 0, msg, copied); 756 if (err) 757 goto done; 758 759 sock_recv_cmsgs(msg, sk, skb); 760 761 /* Copy the address. */ 762 if (sin) { 763 sin->sin_family = AF_INET; 764 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 765 sin->sin_port = 0; 766 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 767 *addr_len = sizeof(*sin); 768 } 769 if (inet_cmsg_flags(inet)) 770 ip_cmsg_recv(msg, skb); 771 if (flags & MSG_TRUNC) 772 copied = skb->len; 773 done: 774 skb_free_datagram(sk, skb); 775 out: 776 if (err) 777 return err; 778 return copied; 779 } 780 781 static int raw_sk_init(struct sock *sk) 782 { 783 struct raw_sock *rp = raw_sk(sk); 784 785 if (inet_sk(sk)->inet_num == IPPROTO_ICMP) 786 memset(&rp->filter, 0, sizeof(rp->filter)); 787 return 0; 788 } 789 790 static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen) 791 { 792 if (optlen > sizeof(struct icmp_filter)) 793 optlen = sizeof(struct icmp_filter); 794 if (copy_from_sockptr(&raw_sk(sk)->filter, optval, optlen)) 795 return -EFAULT; 796 return 0; 797 } 798 799 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) 800 { 801 int len, ret = -EFAULT; 802 803 if (get_user(len, optlen)) 804 goto out; 805 ret = -EINVAL; 806 if (len < 0) 807 goto out; 808 if (len > sizeof(struct icmp_filter)) 809 len = sizeof(struct icmp_filter); 810 ret = -EFAULT; 811 if (put_user(len, optlen) || 812 copy_to_user(optval, &raw_sk(sk)->filter, len)) 813 goto out; 814 ret = 0; 815 out: return ret; 816 } 817 818 static int do_raw_setsockopt(struct sock *sk, int level, int optname, 819 sockptr_t optval, unsigned int optlen) 820 { 821 if (optname == ICMP_FILTER) { 822 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 823 return -EOPNOTSUPP; 824 else 825 return raw_seticmpfilter(sk, optval, optlen); 826 } 827 return -ENOPROTOOPT; 828 } 829 830 static int raw_setsockopt(struct sock *sk, int level, int optname, 831 sockptr_t optval, unsigned int optlen) 832 { 833 if (level != SOL_RAW) 834 return ip_setsockopt(sk, level, optname, optval, optlen); 835 return do_raw_setsockopt(sk, level, optname, optval, optlen); 836 } 837 838 static int do_raw_getsockopt(struct sock *sk, int level, int optname, 839 char __user *optval, int __user *optlen) 840 { 841 if (optname == ICMP_FILTER) { 842 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 843 return -EOPNOTSUPP; 844 else 845 return raw_geticmpfilter(sk, optval, optlen); 846 } 847 return -ENOPROTOOPT; 848 } 849 850 static int raw_getsockopt(struct sock *sk, int level, int optname, 851 char __user *optval, int __user *optlen) 852 { 853 if (level != SOL_RAW) 854 return ip_getsockopt(sk, level, optname, optval, optlen); 855 return do_raw_getsockopt(sk, level, optname, optval, optlen); 856 } 857 858 static int raw_ioctl(struct sock *sk, int cmd, int *karg) 859 { 860 switch (cmd) { 861 case SIOCOUTQ: { 862 *karg = sk_wmem_alloc_get(sk); 863 return 0; 864 } 865 case SIOCINQ: { 866 struct sk_buff *skb; 867 868 spin_lock_bh(&sk->sk_receive_queue.lock); 869 skb = skb_peek(&sk->sk_receive_queue); 870 if (skb) 871 *karg = skb->len; 872 else 873 *karg = 0; 874 spin_unlock_bh(&sk->sk_receive_queue.lock); 875 return 0; 876 } 877 878 default: 879 #ifdef CONFIG_IP_MROUTE 880 return ipmr_ioctl(sk, cmd, karg); 881 #else 882 return -ENOIOCTLCMD; 883 #endif 884 } 885 } 886 887 #ifdef CONFIG_COMPAT 888 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 889 { 890 switch (cmd) { 891 case SIOCOUTQ: 892 case SIOCINQ: 893 return -ENOIOCTLCMD; 894 default: 895 #ifdef CONFIG_IP_MROUTE 896 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg)); 897 #else 898 return -ENOIOCTLCMD; 899 #endif 900 } 901 } 902 #endif 903 904 int raw_abort(struct sock *sk, int err) 905 { 906 lock_sock(sk); 907 908 sk->sk_err = err; 909 sk_error_report(sk); 910 __udp_disconnect(sk, 0); 911 912 release_sock(sk); 913 914 return 0; 915 } 916 EXPORT_SYMBOL_GPL(raw_abort); 917 918 struct proto raw_prot = { 919 .name = "RAW", 920 .owner = THIS_MODULE, 921 .close = raw_close, 922 .destroy = raw_destroy, 923 .connect = ip4_datagram_connect, 924 .disconnect = __udp_disconnect, 925 .ioctl = raw_ioctl, 926 .init = raw_sk_init, 927 .setsockopt = raw_setsockopt, 928 .getsockopt = raw_getsockopt, 929 .sendmsg = raw_sendmsg, 930 .recvmsg = raw_recvmsg, 931 .bind = raw_bind, 932 .backlog_rcv = raw_rcv_skb, 933 .release_cb = ip4_datagram_release_cb, 934 .hash = raw_hash_sk, 935 .unhash = raw_unhash_sk, 936 .obj_size = sizeof(struct raw_sock), 937 .useroffset = offsetof(struct raw_sock, filter), 938 .usersize = sizeof_field(struct raw_sock, filter), 939 .h.raw_hash = &raw_v4_hashinfo, 940 #ifdef CONFIG_COMPAT 941 .compat_ioctl = compat_raw_ioctl, 942 #endif 943 .diag_destroy = raw_abort, 944 }; 945 946 #ifdef CONFIG_PROC_FS 947 static struct sock *raw_get_first(struct seq_file *seq, int bucket) 948 { 949 struct raw_hashinfo *h = pde_data(file_inode(seq->file)); 950 struct raw_iter_state *state = raw_seq_private(seq); 951 struct hlist_head *hlist; 952 struct sock *sk; 953 954 for (state->bucket = bucket; state->bucket < RAW_HTABLE_SIZE; 955 ++state->bucket) { 956 hlist = &h->ht[state->bucket]; 957 sk_for_each(sk, hlist) { 958 if (sock_net(sk) == seq_file_net(seq)) 959 return sk; 960 } 961 } 962 return NULL; 963 } 964 965 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) 966 { 967 struct raw_iter_state *state = raw_seq_private(seq); 968 969 do { 970 sk = sk_next(sk); 971 } while (sk && sock_net(sk) != seq_file_net(seq)); 972 973 if (!sk) 974 return raw_get_first(seq, state->bucket + 1); 975 return sk; 976 } 977 978 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) 979 { 980 struct sock *sk = raw_get_first(seq, 0); 981 982 if (sk) 983 while (pos && (sk = raw_get_next(seq, sk)) != NULL) 984 --pos; 985 return pos ? NULL : sk; 986 } 987 988 void *raw_seq_start(struct seq_file *seq, loff_t *pos) 989 __acquires(&h->lock) 990 { 991 struct raw_hashinfo *h = pde_data(file_inode(seq->file)); 992 993 spin_lock(&h->lock); 994 995 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 996 } 997 EXPORT_SYMBOL_GPL(raw_seq_start); 998 999 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1000 { 1001 struct sock *sk; 1002 1003 if (v == SEQ_START_TOKEN) 1004 sk = raw_get_first(seq, 0); 1005 else 1006 sk = raw_get_next(seq, v); 1007 ++*pos; 1008 return sk; 1009 } 1010 EXPORT_SYMBOL_GPL(raw_seq_next); 1011 1012 void raw_seq_stop(struct seq_file *seq, void *v) 1013 __releases(&h->lock) 1014 { 1015 struct raw_hashinfo *h = pde_data(file_inode(seq->file)); 1016 1017 spin_unlock(&h->lock); 1018 } 1019 EXPORT_SYMBOL_GPL(raw_seq_stop); 1020 1021 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1022 { 1023 struct inet_sock *inet = inet_sk(sp); 1024 __be32 dest = inet->inet_daddr, 1025 src = inet->inet_rcv_saddr; 1026 __u16 destp = 0, 1027 srcp = inet->inet_num; 1028 1029 seq_printf(seq, "%4d: %08X:%04X %08X:%04X" 1030 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n", 1031 i, src, srcp, dest, destp, sp->sk_state, 1032 sk_wmem_alloc_get(sp), 1033 sk_rmem_alloc_get(sp), 1034 0, 0L, 0, 1035 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1036 0, sock_i_ino(sp), 1037 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); 1038 } 1039 1040 static int raw_seq_show(struct seq_file *seq, void *v) 1041 { 1042 if (v == SEQ_START_TOKEN) 1043 seq_printf(seq, " sl local_address rem_address st tx_queue " 1044 "rx_queue tr tm->when retrnsmt uid timeout " 1045 "inode ref pointer drops\n"); 1046 else 1047 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket); 1048 return 0; 1049 } 1050 1051 static const struct seq_operations raw_seq_ops = { 1052 .start = raw_seq_start, 1053 .next = raw_seq_next, 1054 .stop = raw_seq_stop, 1055 .show = raw_seq_show, 1056 }; 1057 1058 static __net_init int raw_init_net(struct net *net) 1059 { 1060 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops, 1061 sizeof(struct raw_iter_state), &raw_v4_hashinfo)) 1062 return -ENOMEM; 1063 1064 return 0; 1065 } 1066 1067 static __net_exit void raw_exit_net(struct net *net) 1068 { 1069 remove_proc_entry("raw", net->proc_net); 1070 } 1071 1072 static __net_initdata struct pernet_operations raw_net_ops = { 1073 .init = raw_init_net, 1074 .exit = raw_exit_net, 1075 }; 1076 1077 int __init raw_proc_init(void) 1078 { 1079 1080 return register_pernet_subsys(&raw_net_ops); 1081 } 1082 1083 void __init raw_proc_exit(void) 1084 { 1085 unregister_pernet_subsys(&raw_net_ops); 1086 } 1087 #endif /* CONFIG_PROC_FS */ 1088 1089 static void raw_sysctl_init_net(struct net *net) 1090 { 1091 #ifdef CONFIG_NET_L3_MASTER_DEV 1092 net->ipv4.sysctl_raw_l3mdev_accept = 1; 1093 #endif 1094 } 1095 1096 static int __net_init raw_sysctl_init(struct net *net) 1097 { 1098 raw_sysctl_init_net(net); 1099 return 0; 1100 } 1101 1102 static struct pernet_operations __net_initdata raw_sysctl_ops = { 1103 .init = raw_sysctl_init, 1104 }; 1105 1106 void __init raw_init(void) 1107 { 1108 raw_sysctl_init_net(&init_net); 1109 if (register_pernet_subsys(&raw_sysctl_ops)) 1110 panic("RAW: failed to init sysctl parameters.\n"); 1111 } 1112