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 <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 82 static struct raw_hashinfo raw_v4_hashinfo = { 83 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock), 84 }; 85 86 void raw_hash_sk(struct sock *sk) 87 { 88 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 89 struct hlist_head *head; 90 91 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)]; 92 93 write_lock_bh(&h->lock); 94 sk_add_node(sk, head); 95 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 96 write_unlock_bh(&h->lock); 97 } 98 EXPORT_SYMBOL_GPL(raw_hash_sk); 99 100 void raw_unhash_sk(struct sock *sk) 101 { 102 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 103 104 write_lock_bh(&h->lock); 105 if (sk_del_node_init(sk)) 106 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 107 write_unlock_bh(&h->lock); 108 } 109 EXPORT_SYMBOL_GPL(raw_unhash_sk); 110 111 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk, 112 unsigned short num, __be32 raddr, __be32 laddr, int dif) 113 { 114 sk_for_each_from(sk) { 115 struct inet_sock *inet = inet_sk(sk); 116 117 if (net_eq(sock_net(sk), net) && inet->inet_num == num && 118 !(inet->inet_daddr && inet->inet_daddr != raddr) && 119 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) && 120 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) 121 goto found; /* gotcha */ 122 } 123 sk = NULL; 124 found: 125 return sk; 126 } 127 128 /* 129 * 0 - deliver 130 * 1 - block 131 */ 132 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb) 133 { 134 struct icmphdr _hdr; 135 const struct icmphdr *hdr; 136 137 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 138 sizeof(_hdr), &_hdr); 139 if (!hdr) 140 return 1; 141 142 if (hdr->type < 32) { 143 __u32 data = raw_sk(sk)->filter.data; 144 145 return ((1U << hdr->type) & data) != 0; 146 } 147 148 /* Do not block unknown ICMP types */ 149 return 0; 150 } 151 152 /* IP input processing comes here for RAW socket delivery. 153 * Caller owns SKB, so we must make clones. 154 * 155 * RFC 1122: SHOULD pass TOS value up to the transport layer. 156 * -> It does. And not only TOS, but all IP header. 157 */ 158 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash) 159 { 160 struct sock *sk; 161 struct hlist_head *head; 162 int delivered = 0; 163 struct net *net; 164 165 read_lock(&raw_v4_hashinfo.lock); 166 head = &raw_v4_hashinfo.ht[hash]; 167 if (hlist_empty(head)) 168 goto out; 169 170 net = dev_net(skb->dev); 171 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol, 172 iph->saddr, iph->daddr, 173 skb->dev->ifindex); 174 175 while (sk) { 176 delivered = 1; 177 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) { 178 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 179 180 /* Not releasing hash table! */ 181 if (clone) 182 raw_rcv(sk, clone); 183 } 184 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol, 185 iph->saddr, iph->daddr, 186 skb->dev->ifindex); 187 } 188 out: 189 read_unlock(&raw_v4_hashinfo.lock); 190 return delivered; 191 } 192 193 int raw_local_deliver(struct sk_buff *skb, int protocol) 194 { 195 int hash; 196 struct sock *raw_sk; 197 198 hash = protocol & (RAW_HTABLE_SIZE - 1); 199 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 200 201 /* If there maybe a raw socket we must check - if not we 202 * don't care less 203 */ 204 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash)) 205 raw_sk = NULL; 206 207 return raw_sk != NULL; 208 209 } 210 211 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) 212 { 213 struct inet_sock *inet = inet_sk(sk); 214 const int type = icmp_hdr(skb)->type; 215 const int code = icmp_hdr(skb)->code; 216 int err = 0; 217 int harderr = 0; 218 219 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) 220 ipv4_sk_update_pmtu(skb, sk, info); 221 else if (type == ICMP_REDIRECT) { 222 ipv4_sk_redirect(skb, sk); 223 return; 224 } 225 226 /* Report error on raw socket, if: 227 1. User requested ip_recverr. 228 2. Socket is connected (otherwise the error indication 229 is useless without ip_recverr and error is hard. 230 */ 231 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) 232 return; 233 234 switch (type) { 235 default: 236 case ICMP_TIME_EXCEEDED: 237 err = EHOSTUNREACH; 238 break; 239 case ICMP_SOURCE_QUENCH: 240 return; 241 case ICMP_PARAMETERPROB: 242 err = EPROTO; 243 harderr = 1; 244 break; 245 case ICMP_DEST_UNREACH: 246 err = EHOSTUNREACH; 247 if (code > NR_ICMP_UNREACH) 248 break; 249 err = icmp_err_convert[code].errno; 250 harderr = icmp_err_convert[code].fatal; 251 if (code == ICMP_FRAG_NEEDED) { 252 harderr = inet->pmtudisc != IP_PMTUDISC_DONT; 253 err = EMSGSIZE; 254 } 255 } 256 257 if (inet->recverr) { 258 const struct iphdr *iph = (const struct iphdr *)skb->data; 259 u8 *payload = skb->data + (iph->ihl << 2); 260 261 if (inet->hdrincl) 262 payload = skb->data; 263 ip_icmp_error(sk, skb, err, 0, info, payload); 264 } 265 266 if (inet->recverr || harderr) { 267 sk->sk_err = err; 268 sk->sk_error_report(sk); 269 } 270 } 271 272 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) 273 { 274 int hash; 275 struct sock *raw_sk; 276 const struct iphdr *iph; 277 struct net *net; 278 279 hash = protocol & (RAW_HTABLE_SIZE - 1); 280 281 read_lock(&raw_v4_hashinfo.lock); 282 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 283 if (raw_sk != NULL) { 284 iph = (const struct iphdr *)skb->data; 285 net = dev_net(skb->dev); 286 287 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol, 288 iph->daddr, iph->saddr, 289 skb->dev->ifindex)) != NULL) { 290 raw_err(raw_sk, skb, info); 291 raw_sk = sk_next(raw_sk); 292 iph = (const struct iphdr *)skb->data; 293 } 294 } 295 read_unlock(&raw_v4_hashinfo.lock); 296 } 297 298 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb) 299 { 300 /* Charge it to the socket. */ 301 302 ipv4_pktinfo_prepare(sk, skb); 303 if (sock_queue_rcv_skb(sk, skb) < 0) { 304 kfree_skb(skb); 305 return NET_RX_DROP; 306 } 307 308 return NET_RX_SUCCESS; 309 } 310 311 int raw_rcv(struct sock *sk, struct sk_buff *skb) 312 { 313 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { 314 atomic_inc(&sk->sk_drops); 315 kfree_skb(skb); 316 return NET_RX_DROP; 317 } 318 nf_reset(skb); 319 320 skb_push(skb, skb->data - skb_network_header(skb)); 321 322 raw_rcv_skb(sk, skb); 323 return 0; 324 } 325 326 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4, 327 void *from, size_t length, 328 struct rtable **rtp, 329 unsigned int flags) 330 { 331 struct inet_sock *inet = inet_sk(sk); 332 struct net *net = sock_net(sk); 333 struct iphdr *iph; 334 struct sk_buff *skb; 335 unsigned int iphlen; 336 int err; 337 struct rtable *rt = *rtp; 338 int hlen, tlen; 339 340 if (length > rt->dst.dev->mtu) { 341 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, 342 rt->dst.dev->mtu); 343 return -EMSGSIZE; 344 } 345 if (flags&MSG_PROBE) 346 goto out; 347 348 hlen = LL_RESERVED_SPACE(rt->dst.dev); 349 tlen = rt->dst.dev->needed_tailroom; 350 skb = sock_alloc_send_skb(sk, 351 length + hlen + tlen + 15, 352 flags & MSG_DONTWAIT, &err); 353 if (skb == NULL) 354 goto error; 355 skb_reserve(skb, hlen); 356 357 skb->priority = sk->sk_priority; 358 skb->mark = sk->sk_mark; 359 skb_dst_set(skb, &rt->dst); 360 *rtp = NULL; 361 362 skb_reset_network_header(skb); 363 iph = ip_hdr(skb); 364 skb_put(skb, length); 365 366 skb->ip_summed = CHECKSUM_NONE; 367 368 skb->transport_header = skb->network_header; 369 err = -EFAULT; 370 if (memcpy_fromiovecend((void *)iph, from, 0, length)) 371 goto error_free; 372 373 iphlen = iph->ihl * 4; 374 375 /* 376 * We don't want to modify the ip header, but we do need to 377 * be sure that it won't cause problems later along the network 378 * stack. Specifically we want to make sure that iph->ihl is a 379 * sane value. If ihl points beyond the length of the buffer passed 380 * in, reject the frame as invalid 381 */ 382 err = -EINVAL; 383 if (iphlen > length) 384 goto error_free; 385 386 if (iphlen >= sizeof(*iph)) { 387 if (!iph->saddr) 388 iph->saddr = fl4->saddr; 389 iph->check = 0; 390 iph->tot_len = htons(length); 391 if (!iph->id) 392 ip_select_ident(skb, &rt->dst, NULL); 393 394 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); 395 } 396 if (iph->protocol == IPPROTO_ICMP) 397 icmp_out_count(net, ((struct icmphdr *) 398 skb_transport_header(skb))->type); 399 400 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL, 401 rt->dst.dev, dst_output); 402 if (err > 0) 403 err = net_xmit_errno(err); 404 if (err) 405 goto error; 406 out: 407 return 0; 408 409 error_free: 410 kfree_skb(skb); 411 error: 412 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS); 413 if (err == -ENOBUFS && !inet->recverr) 414 err = 0; 415 return err; 416 } 417 418 static int raw_probe_proto_opt(struct flowi4 *fl4, struct msghdr *msg) 419 { 420 struct iovec *iov; 421 u8 __user *type = NULL; 422 u8 __user *code = NULL; 423 int probed = 0; 424 unsigned int i; 425 426 if (!msg->msg_iov) 427 return 0; 428 429 for (i = 0; i < msg->msg_iovlen; i++) { 430 iov = &msg->msg_iov[i]; 431 if (!iov) 432 continue; 433 434 switch (fl4->flowi4_proto) { 435 case IPPROTO_ICMP: 436 /* check if one-byte field is readable or not. */ 437 if (iov->iov_base && iov->iov_len < 1) 438 break; 439 440 if (!type) { 441 type = iov->iov_base; 442 /* check if code field is readable or not. */ 443 if (iov->iov_len > 1) 444 code = type + 1; 445 } else if (!code) 446 code = iov->iov_base; 447 448 if (type && code) { 449 if (get_user(fl4->fl4_icmp_type, type) || 450 get_user(fl4->fl4_icmp_code, code)) 451 return -EFAULT; 452 probed = 1; 453 } 454 break; 455 default: 456 probed = 1; 457 break; 458 } 459 if (probed) 460 break; 461 } 462 return 0; 463 } 464 465 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 466 size_t len) 467 { 468 struct inet_sock *inet = inet_sk(sk); 469 struct ipcm_cookie ipc; 470 struct rtable *rt = NULL; 471 struct flowi4 fl4; 472 int free = 0; 473 __be32 daddr; 474 __be32 saddr; 475 u8 tos; 476 int err; 477 struct ip_options_data opt_copy; 478 479 err = -EMSGSIZE; 480 if (len > 0xFFFF) 481 goto out; 482 483 /* 484 * Check the flags. 485 */ 486 487 err = -EOPNOTSUPP; 488 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ 489 goto out; /* compatibility */ 490 491 /* 492 * Get and verify the address. 493 */ 494 495 if (msg->msg_namelen) { 496 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name; 497 err = -EINVAL; 498 if (msg->msg_namelen < sizeof(*usin)) 499 goto out; 500 if (usin->sin_family != AF_INET) { 501 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n", 502 __func__, current->comm); 503 err = -EAFNOSUPPORT; 504 if (usin->sin_family) 505 goto out; 506 } 507 daddr = usin->sin_addr.s_addr; 508 /* ANK: I did not forget to get protocol from port field. 509 * I just do not know, who uses this weirdness. 510 * IP_HDRINCL is much more convenient. 511 */ 512 } else { 513 err = -EDESTADDRREQ; 514 if (sk->sk_state != TCP_ESTABLISHED) 515 goto out; 516 daddr = inet->inet_daddr; 517 } 518 519 ipc.addr = inet->inet_saddr; 520 ipc.opt = NULL; 521 ipc.tx_flags = 0; 522 ipc.ttl = 0; 523 ipc.tos = -1; 524 ipc.oif = sk->sk_bound_dev_if; 525 526 if (msg->msg_controllen) { 527 err = ip_cmsg_send(sock_net(sk), msg, &ipc); 528 if (err) 529 goto out; 530 if (ipc.opt) 531 free = 1; 532 } 533 534 saddr = ipc.addr; 535 ipc.addr = daddr; 536 537 if (!ipc.opt) { 538 struct ip_options_rcu *inet_opt; 539 540 rcu_read_lock(); 541 inet_opt = rcu_dereference(inet->inet_opt); 542 if (inet_opt) { 543 memcpy(&opt_copy, inet_opt, 544 sizeof(*inet_opt) + inet_opt->opt.optlen); 545 ipc.opt = &opt_copy.opt; 546 } 547 rcu_read_unlock(); 548 } 549 550 if (ipc.opt) { 551 err = -EINVAL; 552 /* Linux does not mangle headers on raw sockets, 553 * so that IP options + IP_HDRINCL is non-sense. 554 */ 555 if (inet->hdrincl) 556 goto done; 557 if (ipc.opt->opt.srr) { 558 if (!daddr) 559 goto done; 560 daddr = ipc.opt->opt.faddr; 561 } 562 } 563 tos = get_rtconn_flags(&ipc, sk); 564 if (msg->msg_flags & MSG_DONTROUTE) 565 tos |= RTO_ONLINK; 566 567 if (ipv4_is_multicast(daddr)) { 568 if (!ipc.oif) 569 ipc.oif = inet->mc_index; 570 if (!saddr) 571 saddr = inet->mc_addr; 572 } else if (!ipc.oif) 573 ipc.oif = inet->uc_index; 574 575 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos, 576 RT_SCOPE_UNIVERSE, 577 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol, 578 inet_sk_flowi_flags(sk) | FLOWI_FLAG_CAN_SLEEP | 579 (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0), 580 daddr, saddr, 0, 0); 581 582 if (!inet->hdrincl) { 583 err = raw_probe_proto_opt(&fl4, msg); 584 if (err) 585 goto done; 586 } 587 588 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4)); 589 rt = ip_route_output_flow(sock_net(sk), &fl4, sk); 590 if (IS_ERR(rt)) { 591 err = PTR_ERR(rt); 592 rt = NULL; 593 goto done; 594 } 595 596 err = -EACCES; 597 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) 598 goto done; 599 600 if (msg->msg_flags & MSG_CONFIRM) 601 goto do_confirm; 602 back_from_confirm: 603 604 if (inet->hdrincl) 605 err = raw_send_hdrinc(sk, &fl4, msg->msg_iov, len, 606 &rt, msg->msg_flags); 607 608 else { 609 if (!ipc.addr) 610 ipc.addr = fl4.daddr; 611 lock_sock(sk); 612 err = ip_append_data(sk, &fl4, ip_generic_getfrag, 613 msg->msg_iov, len, 0, 614 &ipc, &rt, msg->msg_flags); 615 if (err) 616 ip_flush_pending_frames(sk); 617 else if (!(msg->msg_flags & MSG_MORE)) { 618 err = ip_push_pending_frames(sk, &fl4); 619 if (err == -ENOBUFS && !inet->recverr) 620 err = 0; 621 } 622 release_sock(sk); 623 } 624 done: 625 if (free) 626 kfree(ipc.opt); 627 ip_rt_put(rt); 628 629 out: 630 if (err < 0) 631 return err; 632 return len; 633 634 do_confirm: 635 dst_confirm(&rt->dst); 636 if (!(msg->msg_flags & MSG_PROBE) || len) 637 goto back_from_confirm; 638 err = 0; 639 goto done; 640 } 641 642 static void raw_close(struct sock *sk, long timeout) 643 { 644 /* 645 * Raw sockets may have direct kernel references. Kill them. 646 */ 647 ip_ra_control(sk, 0, NULL); 648 649 sk_common_release(sk); 650 } 651 652 static void raw_destroy(struct sock *sk) 653 { 654 lock_sock(sk); 655 ip_flush_pending_frames(sk); 656 release_sock(sk); 657 } 658 659 /* This gets rid of all the nasties in af_inet. -DaveM */ 660 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 661 { 662 struct inet_sock *inet = inet_sk(sk); 663 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 664 int ret = -EINVAL; 665 int chk_addr_ret; 666 667 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) 668 goto out; 669 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr); 670 ret = -EADDRNOTAVAIL; 671 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && 672 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) 673 goto out; 674 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 675 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 676 inet->inet_saddr = 0; /* Use device */ 677 sk_dst_reset(sk); 678 ret = 0; 679 out: return ret; 680 } 681 682 /* 683 * This should be easy, if there is something there 684 * we return it, otherwise we block. 685 */ 686 687 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 688 size_t len, int noblock, int flags, int *addr_len) 689 { 690 struct inet_sock *inet = inet_sk(sk); 691 size_t copied = 0; 692 int err = -EOPNOTSUPP; 693 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; 694 struct sk_buff *skb; 695 696 if (flags & MSG_OOB) 697 goto out; 698 699 if (addr_len) 700 *addr_len = sizeof(*sin); 701 702 if (flags & MSG_ERRQUEUE) { 703 err = ip_recv_error(sk, msg, len); 704 goto out; 705 } 706 707 skb = skb_recv_datagram(sk, flags, noblock, &err); 708 if (!skb) 709 goto out; 710 711 copied = skb->len; 712 if (len < copied) { 713 msg->msg_flags |= MSG_TRUNC; 714 copied = len; 715 } 716 717 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 718 if (err) 719 goto done; 720 721 sock_recv_ts_and_drops(msg, sk, skb); 722 723 /* Copy the address. */ 724 if (sin) { 725 sin->sin_family = AF_INET; 726 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 727 sin->sin_port = 0; 728 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 729 } 730 if (inet->cmsg_flags) 731 ip_cmsg_recv(msg, skb); 732 if (flags & MSG_TRUNC) 733 copied = skb->len; 734 done: 735 skb_free_datagram(sk, skb); 736 out: 737 if (err) 738 return err; 739 return copied; 740 } 741 742 static int raw_init(struct sock *sk) 743 { 744 struct raw_sock *rp = raw_sk(sk); 745 746 if (inet_sk(sk)->inet_num == IPPROTO_ICMP) 747 memset(&rp->filter, 0, sizeof(rp->filter)); 748 return 0; 749 } 750 751 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) 752 { 753 if (optlen > sizeof(struct icmp_filter)) 754 optlen = sizeof(struct icmp_filter); 755 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) 756 return -EFAULT; 757 return 0; 758 } 759 760 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) 761 { 762 int len, ret = -EFAULT; 763 764 if (get_user(len, optlen)) 765 goto out; 766 ret = -EINVAL; 767 if (len < 0) 768 goto out; 769 if (len > sizeof(struct icmp_filter)) 770 len = sizeof(struct icmp_filter); 771 ret = -EFAULT; 772 if (put_user(len, optlen) || 773 copy_to_user(optval, &raw_sk(sk)->filter, len)) 774 goto out; 775 ret = 0; 776 out: return ret; 777 } 778 779 static int do_raw_setsockopt(struct sock *sk, int level, int optname, 780 char __user *optval, unsigned int optlen) 781 { 782 if (optname == ICMP_FILTER) { 783 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 784 return -EOPNOTSUPP; 785 else 786 return raw_seticmpfilter(sk, optval, optlen); 787 } 788 return -ENOPROTOOPT; 789 } 790 791 static int raw_setsockopt(struct sock *sk, int level, int optname, 792 char __user *optval, unsigned int optlen) 793 { 794 if (level != SOL_RAW) 795 return ip_setsockopt(sk, level, optname, optval, optlen); 796 return do_raw_setsockopt(sk, level, optname, optval, optlen); 797 } 798 799 #ifdef CONFIG_COMPAT 800 static int compat_raw_setsockopt(struct sock *sk, int level, int optname, 801 char __user *optval, unsigned int optlen) 802 { 803 if (level != SOL_RAW) 804 return compat_ip_setsockopt(sk, level, optname, optval, optlen); 805 return do_raw_setsockopt(sk, level, optname, optval, optlen); 806 } 807 #endif 808 809 static int do_raw_getsockopt(struct sock *sk, int level, int optname, 810 char __user *optval, int __user *optlen) 811 { 812 if (optname == ICMP_FILTER) { 813 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 814 return -EOPNOTSUPP; 815 else 816 return raw_geticmpfilter(sk, optval, optlen); 817 } 818 return -ENOPROTOOPT; 819 } 820 821 static int raw_getsockopt(struct sock *sk, int level, int optname, 822 char __user *optval, int __user *optlen) 823 { 824 if (level != SOL_RAW) 825 return ip_getsockopt(sk, level, optname, optval, optlen); 826 return do_raw_getsockopt(sk, level, optname, optval, optlen); 827 } 828 829 #ifdef CONFIG_COMPAT 830 static int compat_raw_getsockopt(struct sock *sk, int level, int optname, 831 char __user *optval, int __user *optlen) 832 { 833 if (level != SOL_RAW) 834 return compat_ip_getsockopt(sk, level, optname, optval, optlen); 835 return do_raw_getsockopt(sk, level, optname, optval, optlen); 836 } 837 #endif 838 839 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) 840 { 841 switch (cmd) { 842 case SIOCOUTQ: { 843 int amount = sk_wmem_alloc_get(sk); 844 845 return put_user(amount, (int __user *)arg); 846 } 847 case SIOCINQ: { 848 struct sk_buff *skb; 849 int amount = 0; 850 851 spin_lock_bh(&sk->sk_receive_queue.lock); 852 skb = skb_peek(&sk->sk_receive_queue); 853 if (skb != NULL) 854 amount = skb->len; 855 spin_unlock_bh(&sk->sk_receive_queue.lock); 856 return put_user(amount, (int __user *)arg); 857 } 858 859 default: 860 #ifdef CONFIG_IP_MROUTE 861 return ipmr_ioctl(sk, cmd, (void __user *)arg); 862 #else 863 return -ENOIOCTLCMD; 864 #endif 865 } 866 } 867 868 #ifdef CONFIG_COMPAT 869 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 870 { 871 switch (cmd) { 872 case SIOCOUTQ: 873 case SIOCINQ: 874 return -ENOIOCTLCMD; 875 default: 876 #ifdef CONFIG_IP_MROUTE 877 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg)); 878 #else 879 return -ENOIOCTLCMD; 880 #endif 881 } 882 } 883 #endif 884 885 struct proto raw_prot = { 886 .name = "RAW", 887 .owner = THIS_MODULE, 888 .close = raw_close, 889 .destroy = raw_destroy, 890 .connect = ip4_datagram_connect, 891 .disconnect = udp_disconnect, 892 .ioctl = raw_ioctl, 893 .init = raw_init, 894 .setsockopt = raw_setsockopt, 895 .getsockopt = raw_getsockopt, 896 .sendmsg = raw_sendmsg, 897 .recvmsg = raw_recvmsg, 898 .bind = raw_bind, 899 .backlog_rcv = raw_rcv_skb, 900 .release_cb = ip4_datagram_release_cb, 901 .hash = raw_hash_sk, 902 .unhash = raw_unhash_sk, 903 .obj_size = sizeof(struct raw_sock), 904 .h.raw_hash = &raw_v4_hashinfo, 905 #ifdef CONFIG_COMPAT 906 .compat_setsockopt = compat_raw_setsockopt, 907 .compat_getsockopt = compat_raw_getsockopt, 908 .compat_ioctl = compat_raw_ioctl, 909 #endif 910 }; 911 912 #ifdef CONFIG_PROC_FS 913 static struct sock *raw_get_first(struct seq_file *seq) 914 { 915 struct sock *sk; 916 struct raw_iter_state *state = raw_seq_private(seq); 917 918 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE; 919 ++state->bucket) { 920 sk_for_each(sk, &state->h->ht[state->bucket]) 921 if (sock_net(sk) == seq_file_net(seq)) 922 goto found; 923 } 924 sk = NULL; 925 found: 926 return sk; 927 } 928 929 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) 930 { 931 struct raw_iter_state *state = raw_seq_private(seq); 932 933 do { 934 sk = sk_next(sk); 935 try_again: 936 ; 937 } while (sk && sock_net(sk) != seq_file_net(seq)); 938 939 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) { 940 sk = sk_head(&state->h->ht[state->bucket]); 941 goto try_again; 942 } 943 return sk; 944 } 945 946 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) 947 { 948 struct sock *sk = raw_get_first(seq); 949 950 if (sk) 951 while (pos && (sk = raw_get_next(seq, sk)) != NULL) 952 --pos; 953 return pos ? NULL : sk; 954 } 955 956 void *raw_seq_start(struct seq_file *seq, loff_t *pos) 957 { 958 struct raw_iter_state *state = raw_seq_private(seq); 959 960 read_lock(&state->h->lock); 961 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 962 } 963 EXPORT_SYMBOL_GPL(raw_seq_start); 964 965 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) 966 { 967 struct sock *sk; 968 969 if (v == SEQ_START_TOKEN) 970 sk = raw_get_first(seq); 971 else 972 sk = raw_get_next(seq, v); 973 ++*pos; 974 return sk; 975 } 976 EXPORT_SYMBOL_GPL(raw_seq_next); 977 978 void raw_seq_stop(struct seq_file *seq, void *v) 979 { 980 struct raw_iter_state *state = raw_seq_private(seq); 981 982 read_unlock(&state->h->lock); 983 } 984 EXPORT_SYMBOL_GPL(raw_seq_stop); 985 986 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 987 { 988 struct inet_sock *inet = inet_sk(sp); 989 __be32 dest = inet->inet_daddr, 990 src = inet->inet_rcv_saddr; 991 __u16 destp = 0, 992 srcp = inet->inet_num; 993 994 seq_printf(seq, "%4d: %08X:%04X %08X:%04X" 995 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n", 996 i, src, srcp, dest, destp, sp->sk_state, 997 sk_wmem_alloc_get(sp), 998 sk_rmem_alloc_get(sp), 999 0, 0L, 0, 1000 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1001 0, sock_i_ino(sp), 1002 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); 1003 } 1004 1005 static int raw_seq_show(struct seq_file *seq, void *v) 1006 { 1007 if (v == SEQ_START_TOKEN) 1008 seq_printf(seq, " sl local_address rem_address st tx_queue " 1009 "rx_queue tr tm->when retrnsmt uid timeout " 1010 "inode ref pointer drops\n"); 1011 else 1012 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket); 1013 return 0; 1014 } 1015 1016 static const struct seq_operations raw_seq_ops = { 1017 .start = raw_seq_start, 1018 .next = raw_seq_next, 1019 .stop = raw_seq_stop, 1020 .show = raw_seq_show, 1021 }; 1022 1023 int raw_seq_open(struct inode *ino, struct file *file, 1024 struct raw_hashinfo *h, const struct seq_operations *ops) 1025 { 1026 int err; 1027 struct raw_iter_state *i; 1028 1029 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state)); 1030 if (err < 0) 1031 return err; 1032 1033 i = raw_seq_private((struct seq_file *)file->private_data); 1034 i->h = h; 1035 return 0; 1036 } 1037 EXPORT_SYMBOL_GPL(raw_seq_open); 1038 1039 static int raw_v4_seq_open(struct inode *inode, struct file *file) 1040 { 1041 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops); 1042 } 1043 1044 static const struct file_operations raw_seq_fops = { 1045 .owner = THIS_MODULE, 1046 .open = raw_v4_seq_open, 1047 .read = seq_read, 1048 .llseek = seq_lseek, 1049 .release = seq_release_net, 1050 }; 1051 1052 static __net_init int raw_init_net(struct net *net) 1053 { 1054 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops)) 1055 return -ENOMEM; 1056 1057 return 0; 1058 } 1059 1060 static __net_exit void raw_exit_net(struct net *net) 1061 { 1062 remove_proc_entry("raw", net->proc_net); 1063 } 1064 1065 static __net_initdata struct pernet_operations raw_net_ops = { 1066 .init = raw_init_net, 1067 .exit = raw_exit_net, 1068 }; 1069 1070 int __init raw_proc_init(void) 1071 { 1072 return register_pernet_subsys(&raw_net_ops); 1073 } 1074 1075 void __init raw_proc_exit(void) 1076 { 1077 unregister_pernet_subsys(&raw_net_ops); 1078 } 1079 #endif /* CONFIG_PROC_FS */ 1080