1/* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * RAW - implementation of IP "raw" sockets. 7 * 8 * Authors: Ross Biro 9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 10 * 11 * Fixes: 12 * Alan Cox : verify_area() fixed up 13 * Alan Cox : ICMP error handling 14 * Alan Cox : EMSGSIZE if you send too big a packet 15 * Alan Cox : Now uses generic datagrams and shared 16 * skbuff library. No more peek crashes, 17 * no more backlogs 18 * Alan Cox : Checks sk->broadcast. 19 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram 20 * Alan Cox : Raw passes ip options too 21 * Alan Cox : Setsocketopt added 22 * Alan Cox : Fixed error return for broadcasts 23 * Alan Cox : Removed wake_up calls 24 * Alan Cox : Use ttl/tos 25 * Alan Cox : Cleaned up old debugging 26 * Alan Cox : Use new kernel side addresses 27 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets. 28 * Alan Cox : BSD style RAW socket demultiplexing. 29 * Alan Cox : Beginnings of mrouted support. 30 * Alan Cox : Added IP_HDRINCL option. 31 * Alan Cox : Skip broadcast check if BSDism set. 32 * David S. Miller : New socket lookup architecture. 33 * 34 * This program is free software; you can redistribute it and/or 35 * modify it under the terms of the GNU General Public License 36 * as published by the Free Software Foundation; either version 37 * 2 of the License, or (at your option) any later version. 38 */ 39 40#include <linux/types.h> 41#include <linux/atomic.h> 42#include <asm/byteorder.h> 43#include <asm/current.h> 44#include <asm/uaccess.h> 45#include <asm/ioctls.h> 46#include <linux/stddef.h> 47#include <linux/slab.h> 48#include <linux/errno.h> 49#include <linux/kernel.h> 50#include <linux/export.h> 51#include <linux/spinlock.h> 52#include <linux/sockios.h> 53#include <linux/socket.h> 54#include <linux/in.h> 55#include <linux/mroute.h> 56#include <linux/netdevice.h> 57#include <linux/in_route.h> 58#include <linux/route.h> 59#include <linux/skbuff.h> 60#include <linux/igmp.h> 61#include <net/net_namespace.h> 62#include <net/dst.h> 63#include <net/sock.h> 64#include <linux/ip.h> 65#include <linux/net.h> 66#include <net/ip.h> 67#include <net/icmp.h> 68#include <net/udp.h> 69#include <net/raw.h> 70#include <net/snmp.h> 71#include <net/tcp_states.h> 72#include <net/inet_common.h> 73#include <net/checksum.h> 74#include <net/xfrm.h> 75#include <linux/rtnetlink.h> 76#include <linux/proc_fs.h> 77#include <linux/seq_file.h> 78#include <linux/netfilter.h> 79#include <linux/netfilter_ipv4.h> 80#include <linux/compat.h> 81#include <linux/uio.h> 82 83struct raw_frag_vec { 84 struct msghdr *msg; 85 union { 86 struct icmphdr icmph; 87 char c[1]; 88 } hdr; 89 int hlen; 90}; 91 92static struct raw_hashinfo raw_v4_hashinfo = { 93 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock), 94}; 95 96void raw_hash_sk(struct sock *sk) 97{ 98 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 99 struct hlist_head *head; 100 101 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)]; 102 103 write_lock_bh(&h->lock); 104 sk_add_node(sk, head); 105 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 106 write_unlock_bh(&h->lock); 107} 108EXPORT_SYMBOL_GPL(raw_hash_sk); 109 110void raw_unhash_sk(struct sock *sk) 111{ 112 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 113 114 write_lock_bh(&h->lock); 115 if (sk_del_node_init(sk)) 116 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 117 write_unlock_bh(&h->lock); 118} 119EXPORT_SYMBOL_GPL(raw_unhash_sk); 120 121static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk, 122 unsigned short num, __be32 raddr, __be32 laddr, int dif) 123{ 124 sk_for_each_from(sk) { 125 struct inet_sock *inet = inet_sk(sk); 126 127 if (net_eq(sock_net(sk), net) && inet->inet_num == num && 128 !(inet->inet_daddr && inet->inet_daddr != raddr) && 129 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) && 130 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) 131 goto found; /* gotcha */ 132 } 133 sk = NULL; 134found: 135 return sk; 136} 137 138/* 139 * 0 - deliver 140 * 1 - block 141 */ 142static int icmp_filter(const struct sock *sk, const struct sk_buff *skb) 143{ 144 struct icmphdr _hdr; 145 const struct icmphdr *hdr; 146 147 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 148 sizeof(_hdr), &_hdr); 149 if (!hdr) 150 return 1; 151 152 if (hdr->type < 32) { 153 __u32 data = raw_sk(sk)->filter.data; 154 155 return ((1U << hdr->type) & data) != 0; 156 } 157 158 /* Do not block unknown ICMP types */ 159 return 0; 160} 161 162/* IP input processing comes here for RAW socket delivery. 163 * Caller owns SKB, so we must make clones. 164 * 165 * RFC 1122: SHOULD pass TOS value up to the transport layer. 166 * -> It does. And not only TOS, but all IP header. 167 */ 168static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash) 169{ 170 struct sock *sk; 171 struct hlist_head *head; 172 int delivered = 0; 173 struct net *net; 174 175 read_lock(&raw_v4_hashinfo.lock); 176 head = &raw_v4_hashinfo.ht[hash]; 177 if (hlist_empty(head)) 178 goto out; 179 180 net = dev_net(skb->dev); 181 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol, 182 iph->saddr, iph->daddr, 183 skb->dev->ifindex); 184 185 while (sk) { 186 delivered = 1; 187 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) && 188 ip_mc_sf_allow(sk, iph->daddr, iph->saddr, 189 skb->dev->ifindex)) { 190 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 191 192 /* Not releasing hash table! */ 193 if (clone) 194 raw_rcv(sk, clone); 195 } 196 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol, 197 iph->saddr, iph->daddr, 198 skb->dev->ifindex); 199 } 200out: 201 read_unlock(&raw_v4_hashinfo.lock); 202 return delivered; 203} 204 205int raw_local_deliver(struct sk_buff *skb, int protocol) 206{ 207 int hash; 208 struct sock *raw_sk; 209 210 hash = protocol & (RAW_HTABLE_SIZE - 1); 211 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 212 213 /* If there maybe a raw socket we must check - if not we 214 * don't care less 215 */ 216 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash)) 217 raw_sk = NULL; 218 219 return raw_sk != NULL; 220 221} 222 223static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) 224{ 225 struct inet_sock *inet = inet_sk(sk); 226 const int type = icmp_hdr(skb)->type; 227 const int code = icmp_hdr(skb)->code; 228 int err = 0; 229 int harderr = 0; 230 231 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) 232 ipv4_sk_update_pmtu(skb, sk, info); 233 else if (type == ICMP_REDIRECT) { 234 ipv4_sk_redirect(skb, sk); 235 return; 236 } 237 238 /* Report error on raw socket, if: 239 1. User requested ip_recverr. 240 2. Socket is connected (otherwise the error indication 241 is useless without ip_recverr and error is hard. 242 */ 243 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) 244 return; 245 246 switch (type) { 247 default: 248 case ICMP_TIME_EXCEEDED: 249 err = EHOSTUNREACH; 250 break; 251 case ICMP_SOURCE_QUENCH: 252 return; 253 case ICMP_PARAMETERPROB: 254 err = EPROTO; 255 harderr = 1; 256 break; 257 case ICMP_DEST_UNREACH: 258 err = EHOSTUNREACH; 259 if (code > NR_ICMP_UNREACH) 260 break; 261 err = icmp_err_convert[code].errno; 262 harderr = icmp_err_convert[code].fatal; 263 if (code == ICMP_FRAG_NEEDED) { 264 harderr = inet->pmtudisc != IP_PMTUDISC_DONT; 265 err = EMSGSIZE; 266 } 267 } 268 269 if (inet->recverr) { 270 const struct iphdr *iph = (const struct iphdr *)skb->data; 271 u8 *payload = skb->data + (iph->ihl << 2); 272 273 if (inet->hdrincl) 274 payload = skb->data; 275 ip_icmp_error(sk, skb, err, 0, info, payload); 276 } 277 278 if (inet->recverr || harderr) { 279 sk->sk_err = err; 280 sk->sk_error_report(sk); 281 } 282} 283 284void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) 285{ 286 int hash; 287 struct sock *raw_sk; 288 const struct iphdr *iph; 289 struct net *net; 290 291 hash = protocol & (RAW_HTABLE_SIZE - 1); 292 293 read_lock(&raw_v4_hashinfo.lock); 294 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 295 if (raw_sk) { 296 iph = (const struct iphdr *)skb->data; 297 net = dev_net(skb->dev); 298 299 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol, 300 iph->daddr, iph->saddr, 301 skb->dev->ifindex)) != NULL) { 302 raw_err(raw_sk, skb, info); 303 raw_sk = sk_next(raw_sk); 304 iph = (const struct iphdr *)skb->data; 305 } 306 } 307 read_unlock(&raw_v4_hashinfo.lock); 308} 309 310static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb) 311{ 312 /* Charge it to the socket. */ 313 314 ipv4_pktinfo_prepare(sk, skb); 315 if (sock_queue_rcv_skb(sk, skb) < 0) { 316 kfree_skb(skb); 317 return NET_RX_DROP; 318 } 319 320 return NET_RX_SUCCESS; 321} 322 323int raw_rcv(struct sock *sk, struct sk_buff *skb) 324{ 325 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { 326 atomic_inc(&sk->sk_drops); 327 kfree_skb(skb); 328 return NET_RX_DROP; 329 } 330 nf_reset(skb); 331 332 skb_push(skb, skb->data - skb_network_header(skb)); 333 334 raw_rcv_skb(sk, skb); 335 return 0; 336} 337 338static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4, 339 struct msghdr *msg, size_t length, 340 struct rtable **rtp, 341 unsigned int flags) 342{ 343 struct inet_sock *inet = inet_sk(sk); 344 struct net *net = sock_net(sk); 345 struct iphdr *iph; 346 struct sk_buff *skb; 347 unsigned int iphlen; 348 int err; 349 struct rtable *rt = *rtp; 350 int hlen, tlen; 351 352 if (length > rt->dst.dev->mtu) { 353 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, 354 rt->dst.dev->mtu); 355 return -EMSGSIZE; 356 } 357 if (flags&MSG_PROBE) 358 goto out; 359 360 hlen = LL_RESERVED_SPACE(rt->dst.dev); 361 tlen = rt->dst.dev->needed_tailroom; 362 skb = sock_alloc_send_skb(sk, 363 length + hlen + tlen + 15, 364 flags & MSG_DONTWAIT, &err); 365 if (!skb) 366 goto error; 367 skb_reserve(skb, hlen); 368 369 skb->priority = sk->sk_priority; 370 skb->mark = sk->sk_mark; 371 skb_dst_set(skb, &rt->dst); 372 *rtp = NULL; 373 374 skb_reset_network_header(skb); 375 iph = ip_hdr(skb); 376 skb_put(skb, length); 377 378 skb->ip_summed = CHECKSUM_NONE; 379 380 sock_tx_timestamp(sk, &skb_shinfo(skb)->tx_flags); 381 382 skb->transport_header = skb->network_header; 383 err = -EFAULT; 384 if (memcpy_from_msg(iph, msg, length)) 385 goto error_free; 386 387 iphlen = iph->ihl * 4; 388 389 /* 390 * We don't want to modify the ip header, but we do need to 391 * be sure that it won't cause problems later along the network 392 * stack. Specifically we want to make sure that iph->ihl is a 393 * sane value. If ihl points beyond the length of the buffer passed 394 * in, reject the frame as invalid 395 */ 396 err = -EINVAL; 397 if (iphlen > length) 398 goto error_free; 399 400 if (iphlen >= sizeof(*iph)) { 401 if (!iph->saddr) 402 iph->saddr = fl4->saddr; 403 iph->check = 0; 404 iph->tot_len = htons(length); 405 if (!iph->id) 406 ip_select_ident(net, skb, NULL); 407 408 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); 409 skb->transport_header += iphlen; 410 if (iph->protocol == IPPROTO_ICMP && 411 length >= iphlen + sizeof(struct icmphdr)) 412 icmp_out_count(net, ((struct icmphdr *) 413 skb_transport_header(skb))->type); 414 } 415 416 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, 417 net, sk, skb, NULL, rt->dst.dev, 418 dst_output); 419 if (err > 0) 420 err = net_xmit_errno(err); 421 if (err) 422 goto error; 423out: 424 return 0; 425 426error_free: 427 kfree_skb(skb); 428error: 429 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS); 430 if (err == -ENOBUFS && !inet->recverr) 431 err = 0; 432 return err; 433} 434 435static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4) 436{ 437 int err; 438 439 if (fl4->flowi4_proto != IPPROTO_ICMP) 440 return 0; 441 442 /* We only need the first two bytes. */ 443 rfv->hlen = 2; 444 445 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen); 446 if (err) 447 return err; 448 449 fl4->fl4_icmp_type = rfv->hdr.icmph.type; 450 fl4->fl4_icmp_code = rfv->hdr.icmph.code; 451 452 return 0; 453} 454 455static int raw_getfrag(void *from, char *to, int offset, int len, int odd, 456 struct sk_buff *skb) 457{ 458 struct raw_frag_vec *rfv = from; 459 460 if (offset < rfv->hlen) { 461 int copy = min(rfv->hlen - offset, len); 462 463 if (skb->ip_summed == CHECKSUM_PARTIAL) 464 memcpy(to, rfv->hdr.c + offset, copy); 465 else 466 skb->csum = csum_block_add( 467 skb->csum, 468 csum_partial_copy_nocheck(rfv->hdr.c + offset, 469 to, copy, 0), 470 odd); 471 472 odd = 0; 473 offset += copy; 474 to += copy; 475 len -= copy; 476 477 if (!len) 478 return 0; 479 } 480 481 offset -= rfv->hlen; 482 483 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); 484} 485 486static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 487{ 488 struct inet_sock *inet = inet_sk(sk); 489 struct net *net = sock_net(sk); 490 struct ipcm_cookie ipc; 491 struct rtable *rt = NULL; 492 struct flowi4 fl4; 493 int free = 0; 494 __be32 daddr; 495 __be32 saddr; 496 u8 tos; 497 int err; 498 struct ip_options_data opt_copy; 499 struct raw_frag_vec rfv; 500 501 err = -EMSGSIZE; 502 if (len > 0xFFFF) 503 goto out; 504 505 /* 506 * Check the flags. 507 */ 508 509 err = -EOPNOTSUPP; 510 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ 511 goto out; /* compatibility */ 512 513 /* 514 * Get and verify the address. 515 */ 516 517 if (msg->msg_namelen) { 518 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 519 err = -EINVAL; 520 if (msg->msg_namelen < sizeof(*usin)) 521 goto out; 522 if (usin->sin_family != AF_INET) { 523 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n", 524 __func__, current->comm); 525 err = -EAFNOSUPPORT; 526 if (usin->sin_family) 527 goto out; 528 } 529 daddr = usin->sin_addr.s_addr; 530 /* ANK: I did not forget to get protocol from port field. 531 * I just do not know, who uses this weirdness. 532 * IP_HDRINCL is much more convenient. 533 */ 534 } else { 535 err = -EDESTADDRREQ; 536 if (sk->sk_state != TCP_ESTABLISHED) 537 goto out; 538 daddr = inet->inet_daddr; 539 } 540 541 ipc.addr = inet->inet_saddr; 542 ipc.opt = NULL; 543 ipc.tx_flags = 0; 544 ipc.ttl = 0; 545 ipc.tos = -1; 546 ipc.oif = sk->sk_bound_dev_if; 547 548 if (msg->msg_controllen) { 549 err = ip_cmsg_send(net, msg, &ipc, false); 550 if (unlikely(err)) { 551 kfree(ipc.opt); 552 goto out; 553 } 554 if (ipc.opt) 555 free = 1; 556 } 557 558 saddr = ipc.addr; 559 ipc.addr = daddr; 560 561 if (!ipc.opt) { 562 struct ip_options_rcu *inet_opt; 563 564 rcu_read_lock(); 565 inet_opt = rcu_dereference(inet->inet_opt); 566 if (inet_opt) { 567 memcpy(&opt_copy, inet_opt, 568 sizeof(*inet_opt) + inet_opt->opt.optlen); 569 ipc.opt = &opt_copy.opt; 570 } 571 rcu_read_unlock(); 572 } 573 574 if (ipc.opt) { 575 err = -EINVAL; 576 /* Linux does not mangle headers on raw sockets, 577 * so that IP options + IP_HDRINCL is non-sense. 578 */ 579 if (inet->hdrincl) 580 goto done; 581 if (ipc.opt->opt.srr) { 582 if (!daddr) 583 goto done; 584 daddr = ipc.opt->opt.faddr; 585 } 586 } 587 tos = get_rtconn_flags(&ipc, sk); 588 if (msg->msg_flags & MSG_DONTROUTE) 589 tos |= RTO_ONLINK; 590 591 if (ipv4_is_multicast(daddr)) { 592 if (!ipc.oif) 593 ipc.oif = inet->mc_index; 594 if (!saddr) 595 saddr = inet->mc_addr; 596 } else if (!ipc.oif) 597 ipc.oif = inet->uc_index; 598 599 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos, 600 RT_SCOPE_UNIVERSE, 601 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol, 602 inet_sk_flowi_flags(sk) | 603 (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0), 604 daddr, saddr, 0, 0); 605 606 if (!saddr && ipc.oif) { 607 err = l3mdev_get_saddr(net, ipc.oif, &fl4); 608 if (err < 0) 609 goto done; 610 } 611 612 if (!inet->hdrincl) { 613 rfv.msg = msg; 614 rfv.hlen = 0; 615 616 err = raw_probe_proto_opt(&rfv, &fl4); 617 if (err) 618 goto done; 619 } 620 621 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4)); 622 rt = ip_route_output_flow(net, &fl4, sk); 623 if (IS_ERR(rt)) { 624 err = PTR_ERR(rt); 625 rt = NULL; 626 goto done; 627 } 628 629 err = -EACCES; 630 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) 631 goto done; 632 633 if (msg->msg_flags & MSG_CONFIRM) 634 goto do_confirm; 635back_from_confirm: 636 637 if (inet->hdrincl) 638 err = raw_send_hdrinc(sk, &fl4, msg, len, 639 &rt, msg->msg_flags); 640 641 else { 642 sock_tx_timestamp(sk, &ipc.tx_flags); 643 644 if (!ipc.addr) 645 ipc.addr = fl4.daddr; 646 lock_sock(sk); 647 err = ip_append_data(sk, &fl4, raw_getfrag, 648 &rfv, len, 0, 649 &ipc, &rt, msg->msg_flags); 650 if (err) 651 ip_flush_pending_frames(sk); 652 else if (!(msg->msg_flags & MSG_MORE)) { 653 err = ip_push_pending_frames(sk, &fl4); 654 if (err == -ENOBUFS && !inet->recverr) 655 err = 0; 656 } 657 release_sock(sk); 658 } 659done: 660 if (free) 661 kfree(ipc.opt); 662 ip_rt_put(rt); 663 664out: 665 if (err < 0) 666 return err; 667 return len; 668 669do_confirm: 670 dst_confirm(&rt->dst); 671 if (!(msg->msg_flags & MSG_PROBE) || len) 672 goto back_from_confirm; 673 err = 0; 674 goto done; 675} 676 677static void raw_close(struct sock *sk, long timeout) 678{ 679 /* 680 * Raw sockets may have direct kernel references. Kill them. 681 */ 682 ip_ra_control(sk, 0, NULL); 683 684 sk_common_release(sk); 685} 686 687static void raw_destroy(struct sock *sk) 688{ 689 lock_sock(sk); 690 ip_flush_pending_frames(sk); 691 release_sock(sk); 692} 693 694/* This gets rid of all the nasties in af_inet. -DaveM */ 695static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 696{ 697 struct inet_sock *inet = inet_sk(sk); 698 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 699 int ret = -EINVAL; 700 int chk_addr_ret; 701 702 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) 703 goto out; 704 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr); 705 ret = -EADDRNOTAVAIL; 706 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && 707 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) 708 goto out; 709 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 710 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 711 inet->inet_saddr = 0; /* Use device */ 712 sk_dst_reset(sk); 713 ret = 0; 714out: return ret; 715} 716 717/* 718 * This should be easy, if there is something there 719 * we return it, otherwise we block. 720 */ 721 722static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 723 int noblock, int flags, int *addr_len) 724{ 725 struct inet_sock *inet = inet_sk(sk); 726 size_t copied = 0; 727 int err = -EOPNOTSUPP; 728 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 729 struct sk_buff *skb; 730 731 if (flags & MSG_OOB) 732 goto out; 733 734 if (flags & MSG_ERRQUEUE) { 735 err = ip_recv_error(sk, msg, len, addr_len); 736 goto out; 737 } 738 739 skb = skb_recv_datagram(sk, flags, noblock, &err); 740 if (!skb) 741 goto out; 742 743 copied = skb->len; 744 if (len < copied) { 745 msg->msg_flags |= MSG_TRUNC; 746 copied = len; 747 } 748 749 err = skb_copy_datagram_msg(skb, 0, msg, copied); 750 if (err) 751 goto done; 752 753 sock_recv_ts_and_drops(msg, sk, skb); 754 755 /* Copy the address. */ 756 if (sin) { 757 sin->sin_family = AF_INET; 758 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 759 sin->sin_port = 0; 760 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 761 *addr_len = sizeof(*sin); 762 } 763 if (inet->cmsg_flags) 764 ip_cmsg_recv(msg, skb); 765 if (flags & MSG_TRUNC) 766 copied = skb->len; 767done: 768 skb_free_datagram(sk, skb); 769out: 770 if (err) 771 return err; 772 return copied; 773} 774 775static int raw_init(struct sock *sk) 776{ 777 struct raw_sock *rp = raw_sk(sk); 778 779 if (inet_sk(sk)->inet_num == IPPROTO_ICMP) 780 memset(&rp->filter, 0, sizeof(rp->filter)); 781 return 0; 782} 783 784static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) 785{ 786 if (optlen > sizeof(struct icmp_filter)) 787 optlen = sizeof(struct icmp_filter); 788 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) 789 return -EFAULT; 790 return 0; 791} 792 793static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) 794{ 795 int len, ret = -EFAULT; 796 797 if (get_user(len, optlen)) 798 goto out; 799 ret = -EINVAL; 800 if (len < 0) 801 goto out; 802 if (len > sizeof(struct icmp_filter)) 803 len = sizeof(struct icmp_filter); 804 ret = -EFAULT; 805 if (put_user(len, optlen) || 806 copy_to_user(optval, &raw_sk(sk)->filter, len)) 807 goto out; 808 ret = 0; 809out: return ret; 810} 811 812static int do_raw_setsockopt(struct sock *sk, int level, int optname, 813 char __user *optval, unsigned int optlen) 814{ 815 if (optname == ICMP_FILTER) { 816 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 817 return -EOPNOTSUPP; 818 else 819 return raw_seticmpfilter(sk, optval, optlen); 820 } 821 return -ENOPROTOOPT; 822} 823 824static int raw_setsockopt(struct sock *sk, int level, int optname, 825 char __user *optval, unsigned int optlen) 826{ 827 if (level != SOL_RAW) 828 return ip_setsockopt(sk, level, optname, optval, optlen); 829 return do_raw_setsockopt(sk, level, optname, optval, optlen); 830} 831 832#ifdef CONFIG_COMPAT 833static int compat_raw_setsockopt(struct sock *sk, int level, int optname, 834 char __user *optval, unsigned int optlen) 835{ 836 if (level != SOL_RAW) 837 return compat_ip_setsockopt(sk, level, optname, optval, optlen); 838 return do_raw_setsockopt(sk, level, optname, optval, optlen); 839} 840#endif 841 842static int do_raw_getsockopt(struct sock *sk, int level, int optname, 843 char __user *optval, int __user *optlen) 844{ 845 if (optname == ICMP_FILTER) { 846 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 847 return -EOPNOTSUPP; 848 else 849 return raw_geticmpfilter(sk, optval, optlen); 850 } 851 return -ENOPROTOOPT; 852} 853 854static int raw_getsockopt(struct sock *sk, int level, int optname, 855 char __user *optval, int __user *optlen) 856{ 857 if (level != SOL_RAW) 858 return ip_getsockopt(sk, level, optname, optval, optlen); 859 return do_raw_getsockopt(sk, level, optname, optval, optlen); 860} 861 862#ifdef CONFIG_COMPAT 863static int compat_raw_getsockopt(struct sock *sk, int level, int optname, 864 char __user *optval, int __user *optlen) 865{ 866 if (level != SOL_RAW) 867 return compat_ip_getsockopt(sk, level, optname, optval, optlen); 868 return do_raw_getsockopt(sk, level, optname, optval, optlen); 869} 870#endif 871 872static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) 873{ 874 switch (cmd) { 875 case SIOCOUTQ: { 876 int amount = sk_wmem_alloc_get(sk); 877 878 return put_user(amount, (int __user *)arg); 879 } 880 case SIOCINQ: { 881 struct sk_buff *skb; 882 int amount = 0; 883 884 spin_lock_bh(&sk->sk_receive_queue.lock); 885 skb = skb_peek(&sk->sk_receive_queue); 886 if (skb) 887 amount = skb->len; 888 spin_unlock_bh(&sk->sk_receive_queue.lock); 889 return put_user(amount, (int __user *)arg); 890 } 891 892 default: 893#ifdef CONFIG_IP_MROUTE 894 return ipmr_ioctl(sk, cmd, (void __user *)arg); 895#else 896 return -ENOIOCTLCMD; 897#endif 898 } 899} 900 901#ifdef CONFIG_COMPAT 902static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 903{ 904 switch (cmd) { 905 case SIOCOUTQ: 906 case SIOCINQ: 907 return -ENOIOCTLCMD; 908 default: 909#ifdef CONFIG_IP_MROUTE 910 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg)); 911#else 912 return -ENOIOCTLCMD; 913#endif 914 } 915} 916#endif 917 918struct 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_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 .h.raw_hash = &raw_v4_hashinfo, 938#ifdef CONFIG_COMPAT 939 .compat_setsockopt = compat_raw_setsockopt, 940 .compat_getsockopt = compat_raw_getsockopt, 941 .compat_ioctl = compat_raw_ioctl, 942#endif 943}; 944 945#ifdef CONFIG_PROC_FS 946static struct sock *raw_get_first(struct seq_file *seq) 947{ 948 struct sock *sk; 949 struct raw_iter_state *state = raw_seq_private(seq); 950 951 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE; 952 ++state->bucket) { 953 sk_for_each(sk, &state->h->ht[state->bucket]) 954 if (sock_net(sk) == seq_file_net(seq)) 955 goto found; 956 } 957 sk = NULL; 958found: 959 return sk; 960} 961 962static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) 963{ 964 struct raw_iter_state *state = raw_seq_private(seq); 965 966 do { 967 sk = sk_next(sk); 968try_again: 969 ; 970 } while (sk && sock_net(sk) != seq_file_net(seq)); 971 972 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) { 973 sk = sk_head(&state->h->ht[state->bucket]); 974 goto try_again; 975 } 976 return sk; 977} 978 979static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) 980{ 981 struct sock *sk = raw_get_first(seq); 982 983 if (sk) 984 while (pos && (sk = raw_get_next(seq, sk)) != NULL) 985 --pos; 986 return pos ? NULL : sk; 987} 988 989void *raw_seq_start(struct seq_file *seq, loff_t *pos) 990{ 991 struct raw_iter_state *state = raw_seq_private(seq); 992 993 read_lock(&state->h->lock); 994 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 995} 996EXPORT_SYMBOL_GPL(raw_seq_start); 997 998void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) 999{ 1000 struct sock *sk; 1001 1002 if (v == SEQ_START_TOKEN) 1003 sk = raw_get_first(seq); 1004 else 1005 sk = raw_get_next(seq, v); 1006 ++*pos; 1007 return sk; 1008} 1009EXPORT_SYMBOL_GPL(raw_seq_next); 1010 1011void raw_seq_stop(struct seq_file *seq, void *v) 1012{ 1013 struct raw_iter_state *state = raw_seq_private(seq); 1014 1015 read_unlock(&state->h->lock); 1016} 1017EXPORT_SYMBOL_GPL(raw_seq_stop); 1018 1019static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1020{ 1021 struct inet_sock *inet = inet_sk(sp); 1022 __be32 dest = inet->inet_daddr, 1023 src = inet->inet_rcv_saddr; 1024 __u16 destp = 0, 1025 srcp = inet->inet_num; 1026 1027 seq_printf(seq, "%4d: %08X:%04X %08X:%04X" 1028 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n", 1029 i, src, srcp, dest, destp, sp->sk_state, 1030 sk_wmem_alloc_get(sp), 1031 sk_rmem_alloc_get(sp), 1032 0, 0L, 0, 1033 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1034 0, sock_i_ino(sp), 1035 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); 1036} 1037 1038static int raw_seq_show(struct seq_file *seq, void *v) 1039{ 1040 if (v == SEQ_START_TOKEN) 1041 seq_printf(seq, " sl local_address rem_address st tx_queue " 1042 "rx_queue tr tm->when retrnsmt uid timeout " 1043 "inode ref pointer drops\n"); 1044 else 1045 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket); 1046 return 0; 1047} 1048 1049static const struct seq_operations raw_seq_ops = { 1050 .start = raw_seq_start, 1051 .next = raw_seq_next, 1052 .stop = raw_seq_stop, 1053 .show = raw_seq_show, 1054}; 1055 1056int raw_seq_open(struct inode *ino, struct file *file, 1057 struct raw_hashinfo *h, const struct seq_operations *ops) 1058{ 1059 int err; 1060 struct raw_iter_state *i; 1061 1062 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state)); 1063 if (err < 0) 1064 return err; 1065 1066 i = raw_seq_private((struct seq_file *)file->private_data); 1067 i->h = h; 1068 return 0; 1069} 1070EXPORT_SYMBOL_GPL(raw_seq_open); 1071 1072static int raw_v4_seq_open(struct inode *inode, struct file *file) 1073{ 1074 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops); 1075} 1076 1077static const struct file_operations raw_seq_fops = { 1078 .owner = THIS_MODULE, 1079 .open = raw_v4_seq_open, 1080 .read = seq_read, 1081 .llseek = seq_lseek, 1082 .release = seq_release_net, 1083}; 1084 1085static __net_init int raw_init_net(struct net *net) 1086{ 1087 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops)) 1088 return -ENOMEM; 1089 1090 return 0; 1091} 1092 1093static __net_exit void raw_exit_net(struct net *net) 1094{ 1095 remove_proc_entry("raw", net->proc_net); 1096} 1097 1098static __net_initdata struct pernet_operations raw_net_ops = { 1099 .init = raw_init_net, 1100 .exit = raw_exit_net, 1101}; 1102 1103int __init raw_proc_init(void) 1104{ 1105 return register_pernet_subsys(&raw_net_ops); 1106} 1107 1108void __init raw_proc_exit(void) 1109{ 1110 unregister_pernet_subsys(&raw_net_ops); 1111} 1112#endif /* CONFIG_PROC_FS */ 1113