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/aio.h> 50 #include <linux/kernel.h> 51 #include <linux/export.h> 52 #include <linux/spinlock.h> 53 #include <linux/sockios.h> 54 #include <linux/socket.h> 55 #include <linux/in.h> 56 #include <linux/mroute.h> 57 #include <linux/netdevice.h> 58 #include <linux/in_route.h> 59 #include <linux/route.h> 60 #include <linux/skbuff.h> 61 #include <linux/igmp.h> 62 #include <net/net_namespace.h> 63 #include <net/dst.h> 64 #include <net/sock.h> 65 #include <linux/ip.h> 66 #include <linux/net.h> 67 #include <net/ip.h> 68 #include <net/icmp.h> 69 #include <net/udp.h> 70 #include <net/raw.h> 71 #include <net/snmp.h> 72 #include <net/tcp_states.h> 73 #include <net/inet_common.h> 74 #include <net/checksum.h> 75 #include <net/xfrm.h> 76 #include <linux/rtnetlink.h> 77 #include <linux/proc_fs.h> 78 #include <linux/seq_file.h> 79 #include <linux/netfilter.h> 80 #include <linux/netfilter_ipv4.h> 81 #include <linux/compat.h> 82 83 static struct raw_hashinfo raw_v4_hashinfo = { 84 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock), 85 }; 86 87 void raw_hash_sk(struct sock *sk) 88 { 89 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 90 struct hlist_head *head; 91 92 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)]; 93 94 write_lock_bh(&h->lock); 95 sk_add_node(sk, head); 96 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 97 write_unlock_bh(&h->lock); 98 } 99 EXPORT_SYMBOL_GPL(raw_hash_sk); 100 101 void raw_unhash_sk(struct sock *sk) 102 { 103 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 104 105 write_lock_bh(&h->lock); 106 if (sk_del_node_init(sk)) 107 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 108 write_unlock_bh(&h->lock); 109 } 110 EXPORT_SYMBOL_GPL(raw_unhash_sk); 111 112 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk, 113 unsigned short num, __be32 raddr, __be32 laddr, int dif) 114 { 115 sk_for_each_from(sk) { 116 struct inet_sock *inet = inet_sk(sk); 117 118 if (net_eq(sock_net(sk), net) && inet->inet_num == num && 119 !(inet->inet_daddr && inet->inet_daddr != raddr) && 120 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) && 121 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) 122 goto found; /* gotcha */ 123 } 124 sk = NULL; 125 found: 126 return sk; 127 } 128 129 /* 130 * 0 - deliver 131 * 1 - block 132 */ 133 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb) 134 { 135 struct icmphdr _hdr; 136 const struct icmphdr *hdr; 137 138 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 139 sizeof(_hdr), &_hdr); 140 if (!hdr) 141 return 1; 142 143 if (hdr->type < 32) { 144 __u32 data = raw_sk(sk)->filter.data; 145 146 return ((1U << hdr->type) & data) != 0; 147 } 148 149 /* Do not block unknown ICMP types */ 150 return 0; 151 } 152 153 /* IP input processing comes here for RAW socket delivery. 154 * Caller owns SKB, so we must make clones. 155 * 156 * RFC 1122: SHOULD pass TOS value up to the transport layer. 157 * -> It does. And not only TOS, but all IP header. 158 */ 159 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash) 160 { 161 struct sock *sk; 162 struct hlist_head *head; 163 int delivered = 0; 164 struct net *net; 165 166 read_lock(&raw_v4_hashinfo.lock); 167 head = &raw_v4_hashinfo.ht[hash]; 168 if (hlist_empty(head)) 169 goto out; 170 171 net = dev_net(skb->dev); 172 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol, 173 iph->saddr, iph->daddr, 174 skb->dev->ifindex); 175 176 while (sk) { 177 delivered = 1; 178 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) && 179 ip_mc_sf_allow(sk, iph->daddr, iph->saddr, 180 skb->dev->ifindex)) { 181 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 182 183 /* Not releasing hash table! */ 184 if (clone) 185 raw_rcv(sk, clone); 186 } 187 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol, 188 iph->saddr, iph->daddr, 189 skb->dev->ifindex); 190 } 191 out: 192 read_unlock(&raw_v4_hashinfo.lock); 193 return delivered; 194 } 195 196 int raw_local_deliver(struct sk_buff *skb, int protocol) 197 { 198 int hash; 199 struct sock *raw_sk; 200 201 hash = protocol & (RAW_HTABLE_SIZE - 1); 202 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 203 204 /* If there maybe a raw socket we must check - if not we 205 * don't care less 206 */ 207 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash)) 208 raw_sk = NULL; 209 210 return raw_sk != NULL; 211 212 } 213 214 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) 215 { 216 struct inet_sock *inet = inet_sk(sk); 217 const int type = icmp_hdr(skb)->type; 218 const int code = icmp_hdr(skb)->code; 219 int err = 0; 220 int harderr = 0; 221 222 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) 223 ipv4_sk_update_pmtu(skb, sk, info); 224 else if (type == ICMP_REDIRECT) { 225 ipv4_sk_redirect(skb, sk); 226 return; 227 } 228 229 /* Report error on raw socket, if: 230 1. User requested ip_recverr. 231 2. Socket is connected (otherwise the error indication 232 is useless without ip_recverr and error is hard. 233 */ 234 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) 235 return; 236 237 switch (type) { 238 default: 239 case ICMP_TIME_EXCEEDED: 240 err = EHOSTUNREACH; 241 break; 242 case ICMP_SOURCE_QUENCH: 243 return; 244 case ICMP_PARAMETERPROB: 245 err = EPROTO; 246 harderr = 1; 247 break; 248 case ICMP_DEST_UNREACH: 249 err = EHOSTUNREACH; 250 if (code > NR_ICMP_UNREACH) 251 break; 252 err = icmp_err_convert[code].errno; 253 harderr = icmp_err_convert[code].fatal; 254 if (code == ICMP_FRAG_NEEDED) { 255 harderr = inet->pmtudisc != IP_PMTUDISC_DONT; 256 err = EMSGSIZE; 257 } 258 } 259 260 if (inet->recverr) { 261 const struct iphdr *iph = (const struct iphdr *)skb->data; 262 u8 *payload = skb->data + (iph->ihl << 2); 263 264 if (inet->hdrincl) 265 payload = skb->data; 266 ip_icmp_error(sk, skb, err, 0, info, payload); 267 } 268 269 if (inet->recverr || harderr) { 270 sk->sk_err = err; 271 sk->sk_error_report(sk); 272 } 273 } 274 275 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) 276 { 277 int hash; 278 struct sock *raw_sk; 279 const struct iphdr *iph; 280 struct net *net; 281 282 hash = protocol & (RAW_HTABLE_SIZE - 1); 283 284 read_lock(&raw_v4_hashinfo.lock); 285 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 286 if (raw_sk != NULL) { 287 iph = (const struct iphdr *)skb->data; 288 net = dev_net(skb->dev); 289 290 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol, 291 iph->daddr, iph->saddr, 292 skb->dev->ifindex)) != NULL) { 293 raw_err(raw_sk, skb, info); 294 raw_sk = sk_next(raw_sk); 295 iph = (const struct iphdr *)skb->data; 296 } 297 } 298 read_unlock(&raw_v4_hashinfo.lock); 299 } 300 301 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb) 302 { 303 /* Charge it to the socket. */ 304 305 ipv4_pktinfo_prepare(sk, skb); 306 if (sock_queue_rcv_skb(sk, skb) < 0) { 307 kfree_skb(skb); 308 return NET_RX_DROP; 309 } 310 311 return NET_RX_SUCCESS; 312 } 313 314 int raw_rcv(struct sock *sk, struct sk_buff *skb) 315 { 316 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { 317 atomic_inc(&sk->sk_drops); 318 kfree_skb(skb); 319 return NET_RX_DROP; 320 } 321 nf_reset(skb); 322 323 skb_push(skb, skb->data - skb_network_header(skb)); 324 325 raw_rcv_skb(sk, skb); 326 return 0; 327 } 328 329 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4, 330 void *from, size_t length, 331 struct rtable **rtp, 332 unsigned int flags) 333 { 334 struct inet_sock *inet = inet_sk(sk); 335 struct net *net = sock_net(sk); 336 struct iphdr *iph; 337 struct sk_buff *skb; 338 unsigned int iphlen; 339 int err; 340 struct rtable *rt = *rtp; 341 int hlen, tlen; 342 343 if (length > rt->dst.dev->mtu) { 344 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, 345 rt->dst.dev->mtu); 346 return -EMSGSIZE; 347 } 348 if (flags&MSG_PROBE) 349 goto out; 350 351 hlen = LL_RESERVED_SPACE(rt->dst.dev); 352 tlen = rt->dst.dev->needed_tailroom; 353 skb = sock_alloc_send_skb(sk, 354 length + hlen + tlen + 15, 355 flags & MSG_DONTWAIT, &err); 356 if (skb == NULL) 357 goto error; 358 skb_reserve(skb, hlen); 359 360 skb->priority = sk->sk_priority; 361 skb->mark = sk->sk_mark; 362 skb_dst_set(skb, &rt->dst); 363 *rtp = NULL; 364 365 skb_reset_network_header(skb); 366 iph = ip_hdr(skb); 367 skb_put(skb, length); 368 369 skb->ip_summed = CHECKSUM_NONE; 370 371 sock_tx_timestamp(sk, &skb_shinfo(skb)->tx_flags); 372 373 skb->transport_header = skb->network_header; 374 err = -EFAULT; 375 if (memcpy_fromiovecend((void *)iph, from, 0, length)) 376 goto error_free; 377 378 iphlen = iph->ihl * 4; 379 380 /* 381 * We don't want to modify the ip header, but we do need to 382 * be sure that it won't cause problems later along the network 383 * stack. Specifically we want to make sure that iph->ihl is a 384 * sane value. If ihl points beyond the length of the buffer passed 385 * in, reject the frame as invalid 386 */ 387 err = -EINVAL; 388 if (iphlen > length) 389 goto error_free; 390 391 if (iphlen >= sizeof(*iph)) { 392 if (!iph->saddr) 393 iph->saddr = fl4->saddr; 394 iph->check = 0; 395 iph->tot_len = htons(length); 396 if (!iph->id) 397 ip_select_ident(skb, NULL); 398 399 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); 400 } 401 if (iph->protocol == IPPROTO_ICMP) 402 icmp_out_count(net, ((struct icmphdr *) 403 skb_transport_header(skb))->type); 404 405 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL, 406 rt->dst.dev, 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->recverr) 419 err = 0; 420 return err; 421 } 422 423 static int raw_probe_proto_opt(struct flowi4 *fl4, struct msghdr *msg) 424 { 425 struct iovec *iov; 426 u8 __user *type = NULL; 427 u8 __user *code = NULL; 428 int probed = 0; 429 unsigned int i; 430 431 if (!msg->msg_iov) 432 return 0; 433 434 for (i = 0; i < msg->msg_iovlen; i++) { 435 iov = &msg->msg_iov[i]; 436 if (!iov) 437 continue; 438 439 switch (fl4->flowi4_proto) { 440 case IPPROTO_ICMP: 441 /* check if one-byte field is readable or not. */ 442 if (iov->iov_base && iov->iov_len < 1) 443 break; 444 445 if (!type) { 446 type = iov->iov_base; 447 /* check if code field is readable or not. */ 448 if (iov->iov_len > 1) 449 code = type + 1; 450 } else if (!code) 451 code = iov->iov_base; 452 453 if (type && code) { 454 if (get_user(fl4->fl4_icmp_type, type) || 455 get_user(fl4->fl4_icmp_code, code)) 456 return -EFAULT; 457 probed = 1; 458 } 459 break; 460 default: 461 probed = 1; 462 break; 463 } 464 if (probed) 465 break; 466 } 467 return 0; 468 } 469 470 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 471 size_t len) 472 { 473 struct inet_sock *inet = inet_sk(sk); 474 struct ipcm_cookie ipc; 475 struct rtable *rt = NULL; 476 struct flowi4 fl4; 477 int free = 0; 478 __be32 daddr; 479 __be32 saddr; 480 u8 tos; 481 int err; 482 struct ip_options_data opt_copy; 483 484 err = -EMSGSIZE; 485 if (len > 0xFFFF) 486 goto out; 487 488 /* 489 * Check the flags. 490 */ 491 492 err = -EOPNOTSUPP; 493 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ 494 goto out; /* compatibility */ 495 496 /* 497 * Get and verify the address. 498 */ 499 500 if (msg->msg_namelen) { 501 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 502 err = -EINVAL; 503 if (msg->msg_namelen < sizeof(*usin)) 504 goto out; 505 if (usin->sin_family != AF_INET) { 506 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n", 507 __func__, current->comm); 508 err = -EAFNOSUPPORT; 509 if (usin->sin_family) 510 goto out; 511 } 512 daddr = usin->sin_addr.s_addr; 513 /* ANK: I did not forget to get protocol from port field. 514 * I just do not know, who uses this weirdness. 515 * IP_HDRINCL is much more convenient. 516 */ 517 } else { 518 err = -EDESTADDRREQ; 519 if (sk->sk_state != TCP_ESTABLISHED) 520 goto out; 521 daddr = inet->inet_daddr; 522 } 523 524 ipc.addr = inet->inet_saddr; 525 ipc.opt = NULL; 526 ipc.tx_flags = 0; 527 ipc.ttl = 0; 528 ipc.tos = -1; 529 ipc.oif = sk->sk_bound_dev_if; 530 531 if (msg->msg_controllen) { 532 err = ip_cmsg_send(sock_net(sk), msg, &ipc, false); 533 if (err) 534 goto out; 535 if (ipc.opt) 536 free = 1; 537 } 538 539 saddr = ipc.addr; 540 ipc.addr = daddr; 541 542 if (!ipc.opt) { 543 struct ip_options_rcu *inet_opt; 544 545 rcu_read_lock(); 546 inet_opt = rcu_dereference(inet->inet_opt); 547 if (inet_opt) { 548 memcpy(&opt_copy, inet_opt, 549 sizeof(*inet_opt) + inet_opt->opt.optlen); 550 ipc.opt = &opt_copy.opt; 551 } 552 rcu_read_unlock(); 553 } 554 555 if (ipc.opt) { 556 err = -EINVAL; 557 /* Linux does not mangle headers on raw sockets, 558 * so that IP options + IP_HDRINCL is non-sense. 559 */ 560 if (inet->hdrincl) 561 goto done; 562 if (ipc.opt->opt.srr) { 563 if (!daddr) 564 goto done; 565 daddr = ipc.opt->opt.faddr; 566 } 567 } 568 tos = get_rtconn_flags(&ipc, sk); 569 if (msg->msg_flags & MSG_DONTROUTE) 570 tos |= RTO_ONLINK; 571 572 if (ipv4_is_multicast(daddr)) { 573 if (!ipc.oif) 574 ipc.oif = inet->mc_index; 575 if (!saddr) 576 saddr = inet->mc_addr; 577 } else if (!ipc.oif) 578 ipc.oif = inet->uc_index; 579 580 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos, 581 RT_SCOPE_UNIVERSE, 582 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol, 583 inet_sk_flowi_flags(sk) | 584 (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0), 585 daddr, saddr, 0, 0); 586 587 if (!inet->hdrincl) { 588 err = raw_probe_proto_opt(&fl4, msg); 589 if (err) 590 goto done; 591 } 592 593 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4)); 594 rt = ip_route_output_flow(sock_net(sk), &fl4, sk); 595 if (IS_ERR(rt)) { 596 err = PTR_ERR(rt); 597 rt = NULL; 598 goto done; 599 } 600 601 err = -EACCES; 602 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) 603 goto done; 604 605 if (msg->msg_flags & MSG_CONFIRM) 606 goto do_confirm; 607 back_from_confirm: 608 609 if (inet->hdrincl) 610 err = raw_send_hdrinc(sk, &fl4, msg->msg_iov, len, 611 &rt, msg->msg_flags); 612 613 else { 614 sock_tx_timestamp(sk, &ipc.tx_flags); 615 616 if (!ipc.addr) 617 ipc.addr = fl4.daddr; 618 lock_sock(sk); 619 err = ip_append_data(sk, &fl4, ip_generic_getfrag, 620 msg->msg_iov, len, 0, 621 &ipc, &rt, msg->msg_flags); 622 if (err) 623 ip_flush_pending_frames(sk); 624 else if (!(msg->msg_flags & MSG_MORE)) { 625 err = ip_push_pending_frames(sk, &fl4); 626 if (err == -ENOBUFS && !inet->recverr) 627 err = 0; 628 } 629 release_sock(sk); 630 } 631 done: 632 if (free) 633 kfree(ipc.opt); 634 ip_rt_put(rt); 635 636 out: 637 if (err < 0) 638 return err; 639 return len; 640 641 do_confirm: 642 dst_confirm(&rt->dst); 643 if (!(msg->msg_flags & MSG_PROBE) || len) 644 goto back_from_confirm; 645 err = 0; 646 goto done; 647 } 648 649 static void raw_close(struct sock *sk, long timeout) 650 { 651 /* 652 * Raw sockets may have direct kernel references. Kill them. 653 */ 654 ip_ra_control(sk, 0, NULL); 655 656 sk_common_release(sk); 657 } 658 659 static void raw_destroy(struct sock *sk) 660 { 661 lock_sock(sk); 662 ip_flush_pending_frames(sk); 663 release_sock(sk); 664 } 665 666 /* This gets rid of all the nasties in af_inet. -DaveM */ 667 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 668 { 669 struct inet_sock *inet = inet_sk(sk); 670 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 671 int ret = -EINVAL; 672 int chk_addr_ret; 673 674 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) 675 goto out; 676 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr); 677 ret = -EADDRNOTAVAIL; 678 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && 679 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) 680 goto out; 681 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 682 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 683 inet->inet_saddr = 0; /* Use device */ 684 sk_dst_reset(sk); 685 ret = 0; 686 out: return ret; 687 } 688 689 /* 690 * This should be easy, if there is something there 691 * we return it, otherwise we block. 692 */ 693 694 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 695 size_t len, int noblock, int flags, int *addr_len) 696 { 697 struct inet_sock *inet = inet_sk(sk); 698 size_t copied = 0; 699 int err = -EOPNOTSUPP; 700 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 701 struct sk_buff *skb; 702 703 if (flags & MSG_OOB) 704 goto out; 705 706 if (flags & MSG_ERRQUEUE) { 707 err = ip_recv_error(sk, msg, len, addr_len); 708 goto out; 709 } 710 711 skb = skb_recv_datagram(sk, flags, noblock, &err); 712 if (!skb) 713 goto out; 714 715 copied = skb->len; 716 if (len < copied) { 717 msg->msg_flags |= MSG_TRUNC; 718 copied = len; 719 } 720 721 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 722 if (err) 723 goto done; 724 725 sock_recv_ts_and_drops(msg, sk, skb); 726 727 /* Copy the address. */ 728 if (sin) { 729 sin->sin_family = AF_INET; 730 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 731 sin->sin_port = 0; 732 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 733 *addr_len = sizeof(*sin); 734 } 735 if (inet->cmsg_flags) 736 ip_cmsg_recv(msg, skb); 737 if (flags & MSG_TRUNC) 738 copied = skb->len; 739 done: 740 skb_free_datagram(sk, skb); 741 out: 742 if (err) 743 return err; 744 return copied; 745 } 746 747 static int raw_init(struct sock *sk) 748 { 749 struct raw_sock *rp = raw_sk(sk); 750 751 if (inet_sk(sk)->inet_num == IPPROTO_ICMP) 752 memset(&rp->filter, 0, sizeof(rp->filter)); 753 return 0; 754 } 755 756 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) 757 { 758 if (optlen > sizeof(struct icmp_filter)) 759 optlen = sizeof(struct icmp_filter); 760 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) 761 return -EFAULT; 762 return 0; 763 } 764 765 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) 766 { 767 int len, ret = -EFAULT; 768 769 if (get_user(len, optlen)) 770 goto out; 771 ret = -EINVAL; 772 if (len < 0) 773 goto out; 774 if (len > sizeof(struct icmp_filter)) 775 len = sizeof(struct icmp_filter); 776 ret = -EFAULT; 777 if (put_user(len, optlen) || 778 copy_to_user(optval, &raw_sk(sk)->filter, len)) 779 goto out; 780 ret = 0; 781 out: return ret; 782 } 783 784 static int do_raw_setsockopt(struct sock *sk, int level, int optname, 785 char __user *optval, unsigned int optlen) 786 { 787 if (optname == ICMP_FILTER) { 788 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 789 return -EOPNOTSUPP; 790 else 791 return raw_seticmpfilter(sk, optval, optlen); 792 } 793 return -ENOPROTOOPT; 794 } 795 796 static int raw_setsockopt(struct sock *sk, int level, int optname, 797 char __user *optval, unsigned int optlen) 798 { 799 if (level != SOL_RAW) 800 return ip_setsockopt(sk, level, optname, optval, optlen); 801 return do_raw_setsockopt(sk, level, optname, optval, optlen); 802 } 803 804 #ifdef CONFIG_COMPAT 805 static int compat_raw_setsockopt(struct sock *sk, int level, int optname, 806 char __user *optval, unsigned int optlen) 807 { 808 if (level != SOL_RAW) 809 return compat_ip_setsockopt(sk, level, optname, optval, optlen); 810 return do_raw_setsockopt(sk, level, optname, optval, optlen); 811 } 812 #endif 813 814 static int do_raw_getsockopt(struct sock *sk, int level, int optname, 815 char __user *optval, int __user *optlen) 816 { 817 if (optname == ICMP_FILTER) { 818 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 819 return -EOPNOTSUPP; 820 else 821 return raw_geticmpfilter(sk, optval, optlen); 822 } 823 return -ENOPROTOOPT; 824 } 825 826 static int raw_getsockopt(struct sock *sk, int level, int optname, 827 char __user *optval, int __user *optlen) 828 { 829 if (level != SOL_RAW) 830 return ip_getsockopt(sk, level, optname, optval, optlen); 831 return do_raw_getsockopt(sk, level, optname, optval, optlen); 832 } 833 834 #ifdef CONFIG_COMPAT 835 static int compat_raw_getsockopt(struct sock *sk, int level, int optname, 836 char __user *optval, int __user *optlen) 837 { 838 if (level != SOL_RAW) 839 return compat_ip_getsockopt(sk, level, optname, optval, optlen); 840 return do_raw_getsockopt(sk, level, optname, optval, optlen); 841 } 842 #endif 843 844 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) 845 { 846 switch (cmd) { 847 case SIOCOUTQ: { 848 int amount = sk_wmem_alloc_get(sk); 849 850 return put_user(amount, (int __user *)arg); 851 } 852 case SIOCINQ: { 853 struct sk_buff *skb; 854 int amount = 0; 855 856 spin_lock_bh(&sk->sk_receive_queue.lock); 857 skb = skb_peek(&sk->sk_receive_queue); 858 if (skb != NULL) 859 amount = skb->len; 860 spin_unlock_bh(&sk->sk_receive_queue.lock); 861 return put_user(amount, (int __user *)arg); 862 } 863 864 default: 865 #ifdef CONFIG_IP_MROUTE 866 return ipmr_ioctl(sk, cmd, (void __user *)arg); 867 #else 868 return -ENOIOCTLCMD; 869 #endif 870 } 871 } 872 873 #ifdef CONFIG_COMPAT 874 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 875 { 876 switch (cmd) { 877 case SIOCOUTQ: 878 case SIOCINQ: 879 return -ENOIOCTLCMD; 880 default: 881 #ifdef CONFIG_IP_MROUTE 882 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg)); 883 #else 884 return -ENOIOCTLCMD; 885 #endif 886 } 887 } 888 #endif 889 890 struct proto raw_prot = { 891 .name = "RAW", 892 .owner = THIS_MODULE, 893 .close = raw_close, 894 .destroy = raw_destroy, 895 .connect = ip4_datagram_connect, 896 .disconnect = udp_disconnect, 897 .ioctl = raw_ioctl, 898 .init = raw_init, 899 .setsockopt = raw_setsockopt, 900 .getsockopt = raw_getsockopt, 901 .sendmsg = raw_sendmsg, 902 .recvmsg = raw_recvmsg, 903 .bind = raw_bind, 904 .backlog_rcv = raw_rcv_skb, 905 .release_cb = ip4_datagram_release_cb, 906 .hash = raw_hash_sk, 907 .unhash = raw_unhash_sk, 908 .obj_size = sizeof(struct raw_sock), 909 .h.raw_hash = &raw_v4_hashinfo, 910 #ifdef CONFIG_COMPAT 911 .compat_setsockopt = compat_raw_setsockopt, 912 .compat_getsockopt = compat_raw_getsockopt, 913 .compat_ioctl = compat_raw_ioctl, 914 #endif 915 }; 916 917 #ifdef CONFIG_PROC_FS 918 static struct sock *raw_get_first(struct seq_file *seq) 919 { 920 struct sock *sk; 921 struct raw_iter_state *state = raw_seq_private(seq); 922 923 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE; 924 ++state->bucket) { 925 sk_for_each(sk, &state->h->ht[state->bucket]) 926 if (sock_net(sk) == seq_file_net(seq)) 927 goto found; 928 } 929 sk = NULL; 930 found: 931 return sk; 932 } 933 934 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) 935 { 936 struct raw_iter_state *state = raw_seq_private(seq); 937 938 do { 939 sk = sk_next(sk); 940 try_again: 941 ; 942 } while (sk && sock_net(sk) != seq_file_net(seq)); 943 944 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) { 945 sk = sk_head(&state->h->ht[state->bucket]); 946 goto try_again; 947 } 948 return sk; 949 } 950 951 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) 952 { 953 struct sock *sk = raw_get_first(seq); 954 955 if (sk) 956 while (pos && (sk = raw_get_next(seq, sk)) != NULL) 957 --pos; 958 return pos ? NULL : sk; 959 } 960 961 void *raw_seq_start(struct seq_file *seq, loff_t *pos) 962 { 963 struct raw_iter_state *state = raw_seq_private(seq); 964 965 read_lock(&state->h->lock); 966 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 967 } 968 EXPORT_SYMBOL_GPL(raw_seq_start); 969 970 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) 971 { 972 struct sock *sk; 973 974 if (v == SEQ_START_TOKEN) 975 sk = raw_get_first(seq); 976 else 977 sk = raw_get_next(seq, v); 978 ++*pos; 979 return sk; 980 } 981 EXPORT_SYMBOL_GPL(raw_seq_next); 982 983 void raw_seq_stop(struct seq_file *seq, void *v) 984 { 985 struct raw_iter_state *state = raw_seq_private(seq); 986 987 read_unlock(&state->h->lock); 988 } 989 EXPORT_SYMBOL_GPL(raw_seq_stop); 990 991 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 992 { 993 struct inet_sock *inet = inet_sk(sp); 994 __be32 dest = inet->inet_daddr, 995 src = inet->inet_rcv_saddr; 996 __u16 destp = 0, 997 srcp = inet->inet_num; 998 999 seq_printf(seq, "%4d: %08X:%04X %08X:%04X" 1000 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n", 1001 i, src, srcp, dest, destp, sp->sk_state, 1002 sk_wmem_alloc_get(sp), 1003 sk_rmem_alloc_get(sp), 1004 0, 0L, 0, 1005 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1006 0, sock_i_ino(sp), 1007 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); 1008 } 1009 1010 static int raw_seq_show(struct seq_file *seq, void *v) 1011 { 1012 if (v == SEQ_START_TOKEN) 1013 seq_printf(seq, " sl local_address rem_address st tx_queue " 1014 "rx_queue tr tm->when retrnsmt uid timeout " 1015 "inode ref pointer drops\n"); 1016 else 1017 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket); 1018 return 0; 1019 } 1020 1021 static const struct seq_operations raw_seq_ops = { 1022 .start = raw_seq_start, 1023 .next = raw_seq_next, 1024 .stop = raw_seq_stop, 1025 .show = raw_seq_show, 1026 }; 1027 1028 int raw_seq_open(struct inode *ino, struct file *file, 1029 struct raw_hashinfo *h, const struct seq_operations *ops) 1030 { 1031 int err; 1032 struct raw_iter_state *i; 1033 1034 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state)); 1035 if (err < 0) 1036 return err; 1037 1038 i = raw_seq_private((struct seq_file *)file->private_data); 1039 i->h = h; 1040 return 0; 1041 } 1042 EXPORT_SYMBOL_GPL(raw_seq_open); 1043 1044 static int raw_v4_seq_open(struct inode *inode, struct file *file) 1045 { 1046 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops); 1047 } 1048 1049 static const struct file_operations raw_seq_fops = { 1050 .owner = THIS_MODULE, 1051 .open = raw_v4_seq_open, 1052 .read = seq_read, 1053 .llseek = seq_lseek, 1054 .release = seq_release_net, 1055 }; 1056 1057 static __net_init int raw_init_net(struct net *net) 1058 { 1059 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops)) 1060 return -ENOMEM; 1061 1062 return 0; 1063 } 1064 1065 static __net_exit void raw_exit_net(struct net *net) 1066 { 1067 remove_proc_entry("raw", net->proc_net); 1068 } 1069 1070 static __net_initdata struct pernet_operations raw_net_ops = { 1071 .init = raw_init_net, 1072 .exit = raw_exit_net, 1073 }; 1074 1075 int __init raw_proc_init(void) 1076 { 1077 return register_pernet_subsys(&raw_net_ops); 1078 } 1079 1080 void __init raw_proc_exit(void) 1081 { 1082 unregister_pernet_subsys(&raw_net_ops); 1083 } 1084 #endif /* CONFIG_PROC_FS */ 1085