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 <linux/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 83 struct raw_frag_vec { 84 struct msghdr *msg; 85 union { 86 struct icmphdr icmph; 87 char c[1]; 88 } hdr; 89 int hlen; 90 }; 91 92 struct raw_hashinfo raw_v4_hashinfo = { 93 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock), 94 }; 95 EXPORT_SYMBOL_GPL(raw_v4_hashinfo); 96 97 int raw_hash_sk(struct sock *sk) 98 { 99 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 100 struct hlist_head *head; 101 102 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)]; 103 104 write_lock_bh(&h->lock); 105 sk_add_node(sk, head); 106 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 107 write_unlock_bh(&h->lock); 108 109 return 0; 110 } 111 EXPORT_SYMBOL_GPL(raw_hash_sk); 112 113 void raw_unhash_sk(struct sock *sk) 114 { 115 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 116 117 write_lock_bh(&h->lock); 118 if (sk_del_node_init(sk)) 119 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 120 write_unlock_bh(&h->lock); 121 } 122 EXPORT_SYMBOL_GPL(raw_unhash_sk); 123 124 struct sock *__raw_v4_lookup(struct net *net, struct sock *sk, 125 unsigned short num, __be32 raddr, __be32 laddr, 126 int dif, int sdif) 127 { 128 sk_for_each_from(sk) { 129 struct inet_sock *inet = inet_sk(sk); 130 131 if (net_eq(sock_net(sk), net) && inet->inet_num == num && 132 !(inet->inet_daddr && inet->inet_daddr != raddr) && 133 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) && 134 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif && 135 sk->sk_bound_dev_if != sdif)) 136 goto found; /* gotcha */ 137 } 138 sk = NULL; 139 found: 140 return sk; 141 } 142 EXPORT_SYMBOL_GPL(__raw_v4_lookup); 143 144 /* 145 * 0 - deliver 146 * 1 - block 147 */ 148 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb) 149 { 150 struct icmphdr _hdr; 151 const struct icmphdr *hdr; 152 153 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 154 sizeof(_hdr), &_hdr); 155 if (!hdr) 156 return 1; 157 158 if (hdr->type < 32) { 159 __u32 data = raw_sk(sk)->filter.data; 160 161 return ((1U << hdr->type) & data) != 0; 162 } 163 164 /* Do not block unknown ICMP types */ 165 return 0; 166 } 167 168 /* IP input processing comes here for RAW socket delivery. 169 * Caller owns SKB, so we must make clones. 170 * 171 * RFC 1122: SHOULD pass TOS value up to the transport layer. 172 * -> It does. And not only TOS, but all IP header. 173 */ 174 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash) 175 { 176 int sdif = inet_sdif(skb); 177 struct sock *sk; 178 struct hlist_head *head; 179 int delivered = 0; 180 struct net *net; 181 182 read_lock(&raw_v4_hashinfo.lock); 183 head = &raw_v4_hashinfo.ht[hash]; 184 if (hlist_empty(head)) 185 goto out; 186 187 net = dev_net(skb->dev); 188 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol, 189 iph->saddr, iph->daddr, 190 skb->dev->ifindex, sdif); 191 192 while (sk) { 193 delivered = 1; 194 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) && 195 ip_mc_sf_allow(sk, iph->daddr, iph->saddr, 196 skb->dev->ifindex, sdif)) { 197 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 198 199 /* Not releasing hash table! */ 200 if (clone) 201 raw_rcv(sk, clone); 202 } 203 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol, 204 iph->saddr, iph->daddr, 205 skb->dev->ifindex, sdif); 206 } 207 out: 208 read_unlock(&raw_v4_hashinfo.lock); 209 return delivered; 210 } 211 212 int raw_local_deliver(struct sk_buff *skb, int protocol) 213 { 214 int hash; 215 struct sock *raw_sk; 216 217 hash = protocol & (RAW_HTABLE_SIZE - 1); 218 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 219 220 /* If there maybe a raw socket we must check - if not we 221 * don't care less 222 */ 223 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash)) 224 raw_sk = NULL; 225 226 return raw_sk != NULL; 227 228 } 229 230 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) 231 { 232 struct inet_sock *inet = inet_sk(sk); 233 const int type = icmp_hdr(skb)->type; 234 const int code = icmp_hdr(skb)->code; 235 int err = 0; 236 int harderr = 0; 237 238 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) 239 ipv4_sk_update_pmtu(skb, sk, info); 240 else if (type == ICMP_REDIRECT) { 241 ipv4_sk_redirect(skb, sk); 242 return; 243 } 244 245 /* Report error on raw socket, if: 246 1. User requested ip_recverr. 247 2. Socket is connected (otherwise the error indication 248 is useless without ip_recverr and error is hard. 249 */ 250 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) 251 return; 252 253 switch (type) { 254 default: 255 case ICMP_TIME_EXCEEDED: 256 err = EHOSTUNREACH; 257 break; 258 case ICMP_SOURCE_QUENCH: 259 return; 260 case ICMP_PARAMETERPROB: 261 err = EPROTO; 262 harderr = 1; 263 break; 264 case ICMP_DEST_UNREACH: 265 err = EHOSTUNREACH; 266 if (code > NR_ICMP_UNREACH) 267 break; 268 err = icmp_err_convert[code].errno; 269 harderr = icmp_err_convert[code].fatal; 270 if (code == ICMP_FRAG_NEEDED) { 271 harderr = inet->pmtudisc != IP_PMTUDISC_DONT; 272 err = EMSGSIZE; 273 } 274 } 275 276 if (inet->recverr) { 277 const struct iphdr *iph = (const struct iphdr *)skb->data; 278 u8 *payload = skb->data + (iph->ihl << 2); 279 280 if (inet->hdrincl) 281 payload = skb->data; 282 ip_icmp_error(sk, skb, err, 0, info, payload); 283 } 284 285 if (inet->recverr || harderr) { 286 sk->sk_err = err; 287 sk->sk_error_report(sk); 288 } 289 } 290 291 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) 292 { 293 int hash; 294 struct sock *raw_sk; 295 const struct iphdr *iph; 296 struct net *net; 297 298 hash = protocol & (RAW_HTABLE_SIZE - 1); 299 300 read_lock(&raw_v4_hashinfo.lock); 301 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 302 if (raw_sk) { 303 int dif = skb->dev->ifindex; 304 int sdif = inet_sdif(skb); 305 306 iph = (const struct iphdr *)skb->data; 307 net = dev_net(skb->dev); 308 309 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol, 310 iph->daddr, iph->saddr, 311 dif, sdif)) != NULL) { 312 raw_err(raw_sk, skb, info); 313 raw_sk = sk_next(raw_sk); 314 iph = (const struct iphdr *)skb->data; 315 } 316 } 317 read_unlock(&raw_v4_hashinfo.lock); 318 } 319 320 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb) 321 { 322 /* Charge it to the socket. */ 323 324 ipv4_pktinfo_prepare(sk, skb); 325 if (sock_queue_rcv_skb(sk, skb) < 0) { 326 kfree_skb(skb); 327 return NET_RX_DROP; 328 } 329 330 return NET_RX_SUCCESS; 331 } 332 333 int raw_rcv(struct sock *sk, struct sk_buff *skb) 334 { 335 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { 336 atomic_inc(&sk->sk_drops); 337 kfree_skb(skb); 338 return NET_RX_DROP; 339 } 340 nf_reset(skb); 341 342 skb_push(skb, skb->data - skb_network_header(skb)); 343 344 raw_rcv_skb(sk, skb); 345 return 0; 346 } 347 348 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4, 349 struct msghdr *msg, size_t length, 350 struct rtable **rtp, unsigned int flags, 351 const struct sockcm_cookie *sockc) 352 { 353 struct inet_sock *inet = inet_sk(sk); 354 struct net *net = sock_net(sk); 355 struct iphdr *iph; 356 struct sk_buff *skb; 357 unsigned int iphlen; 358 int err; 359 struct rtable *rt = *rtp; 360 int hlen, tlen; 361 362 if (length > rt->dst.dev->mtu) { 363 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, 364 rt->dst.dev->mtu); 365 return -EMSGSIZE; 366 } 367 if (length < sizeof(struct iphdr)) 368 return -EINVAL; 369 370 if (flags&MSG_PROBE) 371 goto out; 372 373 hlen = LL_RESERVED_SPACE(rt->dst.dev); 374 tlen = rt->dst.dev->needed_tailroom; 375 skb = sock_alloc_send_skb(sk, 376 length + hlen + tlen + 15, 377 flags & MSG_DONTWAIT, &err); 378 if (!skb) 379 goto error; 380 skb_reserve(skb, hlen); 381 382 skb->priority = sk->sk_priority; 383 skb->mark = sk->sk_mark; 384 skb_dst_set(skb, &rt->dst); 385 *rtp = NULL; 386 387 skb_reset_network_header(skb); 388 iph = ip_hdr(skb); 389 skb_put(skb, length); 390 391 skb->ip_summed = CHECKSUM_NONE; 392 393 sock_tx_timestamp(sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags); 394 395 if (flags & MSG_CONFIRM) 396 skb_set_dst_pending_confirm(skb, 1); 397 398 skb->transport_header = skb->network_header; 399 err = -EFAULT; 400 if (memcpy_from_msg(iph, msg, length)) 401 goto error_free; 402 403 iphlen = iph->ihl * 4; 404 405 /* 406 * We don't want to modify the ip header, but we do need to 407 * be sure that it won't cause problems later along the network 408 * stack. Specifically we want to make sure that iph->ihl is a 409 * sane value. If ihl points beyond the length of the buffer passed 410 * in, reject the frame as invalid 411 */ 412 err = -EINVAL; 413 if (iphlen > length) 414 goto error_free; 415 416 if (iphlen >= sizeof(*iph)) { 417 if (!iph->saddr) 418 iph->saddr = fl4->saddr; 419 iph->check = 0; 420 iph->tot_len = htons(length); 421 if (!iph->id) 422 ip_select_ident(net, skb, NULL); 423 424 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); 425 skb->transport_header += iphlen; 426 if (iph->protocol == IPPROTO_ICMP && 427 length >= iphlen + sizeof(struct icmphdr)) 428 icmp_out_count(net, ((struct icmphdr *) 429 skb_transport_header(skb))->type); 430 } 431 432 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, 433 net, sk, skb, NULL, rt->dst.dev, 434 dst_output); 435 if (err > 0) 436 err = net_xmit_errno(err); 437 if (err) 438 goto error; 439 out: 440 return 0; 441 442 error_free: 443 kfree_skb(skb); 444 error: 445 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS); 446 if (err == -ENOBUFS && !inet->recverr) 447 err = 0; 448 return err; 449 } 450 451 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4) 452 { 453 int err; 454 455 if (fl4->flowi4_proto != IPPROTO_ICMP) 456 return 0; 457 458 /* We only need the first two bytes. */ 459 rfv->hlen = 2; 460 461 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen); 462 if (err) 463 return err; 464 465 fl4->fl4_icmp_type = rfv->hdr.icmph.type; 466 fl4->fl4_icmp_code = rfv->hdr.icmph.code; 467 468 return 0; 469 } 470 471 static int raw_getfrag(void *from, char *to, int offset, int len, int odd, 472 struct sk_buff *skb) 473 { 474 struct raw_frag_vec *rfv = from; 475 476 if (offset < rfv->hlen) { 477 int copy = min(rfv->hlen - offset, len); 478 479 if (skb->ip_summed == CHECKSUM_PARTIAL) 480 memcpy(to, rfv->hdr.c + offset, copy); 481 else 482 skb->csum = csum_block_add( 483 skb->csum, 484 csum_partial_copy_nocheck(rfv->hdr.c + offset, 485 to, copy, 0), 486 odd); 487 488 odd = 0; 489 offset += copy; 490 to += copy; 491 len -= copy; 492 493 if (!len) 494 return 0; 495 } 496 497 offset -= rfv->hlen; 498 499 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); 500 } 501 502 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 503 { 504 struct inet_sock *inet = inet_sk(sk); 505 struct net *net = sock_net(sk); 506 struct ipcm_cookie ipc; 507 struct rtable *rt = NULL; 508 struct flowi4 fl4; 509 int free = 0; 510 __be32 daddr; 511 __be32 saddr; 512 u8 tos; 513 int err; 514 struct ip_options_data opt_copy; 515 struct raw_frag_vec rfv; 516 517 err = -EMSGSIZE; 518 if (len > 0xFFFF) 519 goto out; 520 521 /* 522 * Check the flags. 523 */ 524 525 err = -EOPNOTSUPP; 526 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ 527 goto out; /* compatibility */ 528 529 /* 530 * Get and verify the address. 531 */ 532 533 if (msg->msg_namelen) { 534 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 535 err = -EINVAL; 536 if (msg->msg_namelen < sizeof(*usin)) 537 goto out; 538 if (usin->sin_family != AF_INET) { 539 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n", 540 __func__, current->comm); 541 err = -EAFNOSUPPORT; 542 if (usin->sin_family) 543 goto out; 544 } 545 daddr = usin->sin_addr.s_addr; 546 /* ANK: I did not forget to get protocol from port field. 547 * I just do not know, who uses this weirdness. 548 * IP_HDRINCL is much more convenient. 549 */ 550 } else { 551 err = -EDESTADDRREQ; 552 if (sk->sk_state != TCP_ESTABLISHED) 553 goto out; 554 daddr = inet->inet_daddr; 555 } 556 557 ipc.sockc.tsflags = sk->sk_tsflags; 558 ipc.addr = inet->inet_saddr; 559 ipc.opt = NULL; 560 ipc.tx_flags = 0; 561 ipc.ttl = 0; 562 ipc.tos = -1; 563 ipc.oif = sk->sk_bound_dev_if; 564 565 if (msg->msg_controllen) { 566 err = ip_cmsg_send(sk, msg, &ipc, false); 567 if (unlikely(err)) { 568 kfree(ipc.opt); 569 goto out; 570 } 571 if (ipc.opt) 572 free = 1; 573 } 574 575 saddr = ipc.addr; 576 ipc.addr = daddr; 577 578 if (!ipc.opt) { 579 struct ip_options_rcu *inet_opt; 580 581 rcu_read_lock(); 582 inet_opt = rcu_dereference(inet->inet_opt); 583 if (inet_opt) { 584 memcpy(&opt_copy, inet_opt, 585 sizeof(*inet_opt) + inet_opt->opt.optlen); 586 ipc.opt = &opt_copy.opt; 587 } 588 rcu_read_unlock(); 589 } 590 591 if (ipc.opt) { 592 err = -EINVAL; 593 /* Linux does not mangle headers on raw sockets, 594 * so that IP options + IP_HDRINCL is non-sense. 595 */ 596 if (inet->hdrincl) 597 goto done; 598 if (ipc.opt->opt.srr) { 599 if (!daddr) 600 goto done; 601 daddr = ipc.opt->opt.faddr; 602 } 603 } 604 tos = get_rtconn_flags(&ipc, sk); 605 if (msg->msg_flags & MSG_DONTROUTE) 606 tos |= RTO_ONLINK; 607 608 if (ipv4_is_multicast(daddr)) { 609 if (!ipc.oif) 610 ipc.oif = inet->mc_index; 611 if (!saddr) 612 saddr = inet->mc_addr; 613 } else if (!ipc.oif) 614 ipc.oif = inet->uc_index; 615 616 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos, 617 RT_SCOPE_UNIVERSE, 618 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol, 619 inet_sk_flowi_flags(sk) | 620 (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0), 621 daddr, saddr, 0, 0, sk->sk_uid); 622 623 if (!inet->hdrincl) { 624 rfv.msg = msg; 625 rfv.hlen = 0; 626 627 err = raw_probe_proto_opt(&rfv, &fl4); 628 if (err) 629 goto done; 630 } 631 632 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4)); 633 rt = ip_route_output_flow(net, &fl4, sk); 634 if (IS_ERR(rt)) { 635 err = PTR_ERR(rt); 636 rt = NULL; 637 goto done; 638 } 639 640 err = -EACCES; 641 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) 642 goto done; 643 644 if (msg->msg_flags & MSG_CONFIRM) 645 goto do_confirm; 646 back_from_confirm: 647 648 if (inet->hdrincl) 649 err = raw_send_hdrinc(sk, &fl4, msg, len, 650 &rt, msg->msg_flags, &ipc.sockc); 651 652 else { 653 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags); 654 655 if (!ipc.addr) 656 ipc.addr = fl4.daddr; 657 lock_sock(sk); 658 err = ip_append_data(sk, &fl4, raw_getfrag, 659 &rfv, len, 0, 660 &ipc, &rt, msg->msg_flags); 661 if (err) 662 ip_flush_pending_frames(sk); 663 else if (!(msg->msg_flags & MSG_MORE)) { 664 err = ip_push_pending_frames(sk, &fl4); 665 if (err == -ENOBUFS && !inet->recverr) 666 err = 0; 667 } 668 release_sock(sk); 669 } 670 done: 671 if (free) 672 kfree(ipc.opt); 673 ip_rt_put(rt); 674 675 out: 676 if (err < 0) 677 return err; 678 return len; 679 680 do_confirm: 681 if (msg->msg_flags & MSG_PROBE) 682 dst_confirm_neigh(&rt->dst, &fl4.daddr); 683 if (!(msg->msg_flags & MSG_PROBE) || len) 684 goto back_from_confirm; 685 err = 0; 686 goto done; 687 } 688 689 static void raw_close(struct sock *sk, long timeout) 690 { 691 /* 692 * Raw sockets may have direct kernel references. Kill them. 693 */ 694 rtnl_lock(); 695 ip_ra_control(sk, 0, NULL); 696 rtnl_unlock(); 697 698 sk_common_release(sk); 699 } 700 701 static void raw_destroy(struct sock *sk) 702 { 703 lock_sock(sk); 704 ip_flush_pending_frames(sk); 705 release_sock(sk); 706 } 707 708 /* This gets rid of all the nasties in af_inet. -DaveM */ 709 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 710 { 711 struct inet_sock *inet = inet_sk(sk); 712 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 713 u32 tb_id = RT_TABLE_LOCAL; 714 int ret = -EINVAL; 715 int chk_addr_ret; 716 717 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) 718 goto out; 719 720 if (sk->sk_bound_dev_if) 721 tb_id = l3mdev_fib_table_by_index(sock_net(sk), 722 sk->sk_bound_dev_if) ? : tb_id; 723 724 chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr, 725 tb_id); 726 727 ret = -EADDRNOTAVAIL; 728 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && 729 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) 730 goto out; 731 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 732 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 733 inet->inet_saddr = 0; /* Use device */ 734 sk_dst_reset(sk); 735 ret = 0; 736 out: return ret; 737 } 738 739 /* 740 * This should be easy, if there is something there 741 * we return it, otherwise we block. 742 */ 743 744 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 745 int noblock, int flags, int *addr_len) 746 { 747 struct inet_sock *inet = inet_sk(sk); 748 size_t copied = 0; 749 int err = -EOPNOTSUPP; 750 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 751 struct sk_buff *skb; 752 753 if (flags & MSG_OOB) 754 goto out; 755 756 if (flags & MSG_ERRQUEUE) { 757 err = ip_recv_error(sk, msg, len, addr_len); 758 goto out; 759 } 760 761 skb = skb_recv_datagram(sk, flags, noblock, &err); 762 if (!skb) 763 goto out; 764 765 copied = skb->len; 766 if (len < copied) { 767 msg->msg_flags |= MSG_TRUNC; 768 copied = len; 769 } 770 771 err = skb_copy_datagram_msg(skb, 0, msg, copied); 772 if (err) 773 goto done; 774 775 sock_recv_ts_and_drops(msg, sk, skb); 776 777 /* Copy the address. */ 778 if (sin) { 779 sin->sin_family = AF_INET; 780 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 781 sin->sin_port = 0; 782 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 783 *addr_len = sizeof(*sin); 784 } 785 if (inet->cmsg_flags) 786 ip_cmsg_recv(msg, skb); 787 if (flags & MSG_TRUNC) 788 copied = skb->len; 789 done: 790 skb_free_datagram(sk, skb); 791 out: 792 if (err) 793 return err; 794 return copied; 795 } 796 797 static int raw_init(struct sock *sk) 798 { 799 struct raw_sock *rp = raw_sk(sk); 800 801 if (inet_sk(sk)->inet_num == IPPROTO_ICMP) 802 memset(&rp->filter, 0, sizeof(rp->filter)); 803 return 0; 804 } 805 806 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) 807 { 808 if (optlen > sizeof(struct icmp_filter)) 809 optlen = sizeof(struct icmp_filter); 810 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) 811 return -EFAULT; 812 return 0; 813 } 814 815 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) 816 { 817 int len, ret = -EFAULT; 818 819 if (get_user(len, optlen)) 820 goto out; 821 ret = -EINVAL; 822 if (len < 0) 823 goto out; 824 if (len > sizeof(struct icmp_filter)) 825 len = sizeof(struct icmp_filter); 826 ret = -EFAULT; 827 if (put_user(len, optlen) || 828 copy_to_user(optval, &raw_sk(sk)->filter, len)) 829 goto out; 830 ret = 0; 831 out: return ret; 832 } 833 834 static int do_raw_setsockopt(struct sock *sk, int level, int optname, 835 char __user *optval, unsigned int optlen) 836 { 837 if (optname == ICMP_FILTER) { 838 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 839 return -EOPNOTSUPP; 840 else 841 return raw_seticmpfilter(sk, optval, optlen); 842 } 843 return -ENOPROTOOPT; 844 } 845 846 static int raw_setsockopt(struct sock *sk, int level, int optname, 847 char __user *optval, unsigned int optlen) 848 { 849 if (level != SOL_RAW) 850 return ip_setsockopt(sk, level, optname, optval, optlen); 851 return do_raw_setsockopt(sk, level, optname, optval, optlen); 852 } 853 854 #ifdef CONFIG_COMPAT 855 static int compat_raw_setsockopt(struct sock *sk, int level, int optname, 856 char __user *optval, unsigned int optlen) 857 { 858 if (level != SOL_RAW) 859 return compat_ip_setsockopt(sk, level, optname, optval, optlen); 860 return do_raw_setsockopt(sk, level, optname, optval, optlen); 861 } 862 #endif 863 864 static int do_raw_getsockopt(struct sock *sk, int level, int optname, 865 char __user *optval, int __user *optlen) 866 { 867 if (optname == ICMP_FILTER) { 868 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 869 return -EOPNOTSUPP; 870 else 871 return raw_geticmpfilter(sk, optval, optlen); 872 } 873 return -ENOPROTOOPT; 874 } 875 876 static int raw_getsockopt(struct sock *sk, int level, int optname, 877 char __user *optval, int __user *optlen) 878 { 879 if (level != SOL_RAW) 880 return ip_getsockopt(sk, level, optname, optval, optlen); 881 return do_raw_getsockopt(sk, level, optname, optval, optlen); 882 } 883 884 #ifdef CONFIG_COMPAT 885 static int compat_raw_getsockopt(struct sock *sk, int level, int optname, 886 char __user *optval, int __user *optlen) 887 { 888 if (level != SOL_RAW) 889 return compat_ip_getsockopt(sk, level, optname, optval, optlen); 890 return do_raw_getsockopt(sk, level, optname, optval, optlen); 891 } 892 #endif 893 894 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) 895 { 896 switch (cmd) { 897 case SIOCOUTQ: { 898 int amount = sk_wmem_alloc_get(sk); 899 900 return put_user(amount, (int __user *)arg); 901 } 902 case SIOCINQ: { 903 struct sk_buff *skb; 904 int amount = 0; 905 906 spin_lock_bh(&sk->sk_receive_queue.lock); 907 skb = skb_peek(&sk->sk_receive_queue); 908 if (skb) 909 amount = skb->len; 910 spin_unlock_bh(&sk->sk_receive_queue.lock); 911 return put_user(amount, (int __user *)arg); 912 } 913 914 default: 915 #ifdef CONFIG_IP_MROUTE 916 return ipmr_ioctl(sk, cmd, (void __user *)arg); 917 #else 918 return -ENOIOCTLCMD; 919 #endif 920 } 921 } 922 923 #ifdef CONFIG_COMPAT 924 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 925 { 926 switch (cmd) { 927 case SIOCOUTQ: 928 case SIOCINQ: 929 return -ENOIOCTLCMD; 930 default: 931 #ifdef CONFIG_IP_MROUTE 932 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg)); 933 #else 934 return -ENOIOCTLCMD; 935 #endif 936 } 937 } 938 #endif 939 940 int raw_abort(struct sock *sk, int err) 941 { 942 lock_sock(sk); 943 944 sk->sk_err = err; 945 sk->sk_error_report(sk); 946 __udp_disconnect(sk, 0); 947 948 release_sock(sk); 949 950 return 0; 951 } 952 EXPORT_SYMBOL_GPL(raw_abort); 953 954 struct proto raw_prot = { 955 .name = "RAW", 956 .owner = THIS_MODULE, 957 .close = raw_close, 958 .destroy = raw_destroy, 959 .connect = ip4_datagram_connect, 960 .disconnect = __udp_disconnect, 961 .ioctl = raw_ioctl, 962 .init = raw_init, 963 .setsockopt = raw_setsockopt, 964 .getsockopt = raw_getsockopt, 965 .sendmsg = raw_sendmsg, 966 .recvmsg = raw_recvmsg, 967 .bind = raw_bind, 968 .backlog_rcv = raw_rcv_skb, 969 .release_cb = ip4_datagram_release_cb, 970 .hash = raw_hash_sk, 971 .unhash = raw_unhash_sk, 972 .obj_size = sizeof(struct raw_sock), 973 .h.raw_hash = &raw_v4_hashinfo, 974 #ifdef CONFIG_COMPAT 975 .compat_setsockopt = compat_raw_setsockopt, 976 .compat_getsockopt = compat_raw_getsockopt, 977 .compat_ioctl = compat_raw_ioctl, 978 #endif 979 .diag_destroy = raw_abort, 980 }; 981 982 #ifdef CONFIG_PROC_FS 983 static struct sock *raw_get_first(struct seq_file *seq) 984 { 985 struct sock *sk; 986 struct raw_iter_state *state = raw_seq_private(seq); 987 988 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE; 989 ++state->bucket) { 990 sk_for_each(sk, &state->h->ht[state->bucket]) 991 if (sock_net(sk) == seq_file_net(seq)) 992 goto found; 993 } 994 sk = NULL; 995 found: 996 return sk; 997 } 998 999 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) 1000 { 1001 struct raw_iter_state *state = raw_seq_private(seq); 1002 1003 do { 1004 sk = sk_next(sk); 1005 try_again: 1006 ; 1007 } while (sk && sock_net(sk) != seq_file_net(seq)); 1008 1009 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) { 1010 sk = sk_head(&state->h->ht[state->bucket]); 1011 goto try_again; 1012 } 1013 return sk; 1014 } 1015 1016 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) 1017 { 1018 struct sock *sk = raw_get_first(seq); 1019 1020 if (sk) 1021 while (pos && (sk = raw_get_next(seq, sk)) != NULL) 1022 --pos; 1023 return pos ? NULL : sk; 1024 } 1025 1026 void *raw_seq_start(struct seq_file *seq, loff_t *pos) 1027 { 1028 struct raw_iter_state *state = raw_seq_private(seq); 1029 1030 read_lock(&state->h->lock); 1031 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 1032 } 1033 EXPORT_SYMBOL_GPL(raw_seq_start); 1034 1035 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1036 { 1037 struct sock *sk; 1038 1039 if (v == SEQ_START_TOKEN) 1040 sk = raw_get_first(seq); 1041 else 1042 sk = raw_get_next(seq, v); 1043 ++*pos; 1044 return sk; 1045 } 1046 EXPORT_SYMBOL_GPL(raw_seq_next); 1047 1048 void raw_seq_stop(struct seq_file *seq, void *v) 1049 { 1050 struct raw_iter_state *state = raw_seq_private(seq); 1051 1052 read_unlock(&state->h->lock); 1053 } 1054 EXPORT_SYMBOL_GPL(raw_seq_stop); 1055 1056 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1057 { 1058 struct inet_sock *inet = inet_sk(sp); 1059 __be32 dest = inet->inet_daddr, 1060 src = inet->inet_rcv_saddr; 1061 __u16 destp = 0, 1062 srcp = inet->inet_num; 1063 1064 seq_printf(seq, "%4d: %08X:%04X %08X:%04X" 1065 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n", 1066 i, src, srcp, dest, destp, sp->sk_state, 1067 sk_wmem_alloc_get(sp), 1068 sk_rmem_alloc_get(sp), 1069 0, 0L, 0, 1070 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1071 0, sock_i_ino(sp), 1072 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); 1073 } 1074 1075 static int raw_seq_show(struct seq_file *seq, void *v) 1076 { 1077 if (v == SEQ_START_TOKEN) 1078 seq_printf(seq, " sl local_address rem_address st tx_queue " 1079 "rx_queue tr tm->when retrnsmt uid timeout " 1080 "inode ref pointer drops\n"); 1081 else 1082 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket); 1083 return 0; 1084 } 1085 1086 static const struct seq_operations raw_seq_ops = { 1087 .start = raw_seq_start, 1088 .next = raw_seq_next, 1089 .stop = raw_seq_stop, 1090 .show = raw_seq_show, 1091 }; 1092 1093 int raw_seq_open(struct inode *ino, struct file *file, 1094 struct raw_hashinfo *h, const struct seq_operations *ops) 1095 { 1096 int err; 1097 struct raw_iter_state *i; 1098 1099 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state)); 1100 if (err < 0) 1101 return err; 1102 1103 i = raw_seq_private((struct seq_file *)file->private_data); 1104 i->h = h; 1105 return 0; 1106 } 1107 EXPORT_SYMBOL_GPL(raw_seq_open); 1108 1109 static int raw_v4_seq_open(struct inode *inode, struct file *file) 1110 { 1111 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops); 1112 } 1113 1114 static const struct file_operations raw_seq_fops = { 1115 .owner = THIS_MODULE, 1116 .open = raw_v4_seq_open, 1117 .read = seq_read, 1118 .llseek = seq_lseek, 1119 .release = seq_release_net, 1120 }; 1121 1122 static __net_init int raw_init_net(struct net *net) 1123 { 1124 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops)) 1125 return -ENOMEM; 1126 1127 return 0; 1128 } 1129 1130 static __net_exit void raw_exit_net(struct net *net) 1131 { 1132 remove_proc_entry("raw", net->proc_net); 1133 } 1134 1135 static __net_initdata struct pernet_operations raw_net_ops = { 1136 .init = raw_init_net, 1137 .exit = raw_exit_net, 1138 }; 1139 1140 int __init raw_proc_init(void) 1141 { 1142 return register_pernet_subsys(&raw_net_ops); 1143 } 1144 1145 void __init raw_proc_exit(void) 1146 { 1147 unregister_pernet_subsys(&raw_net_ops); 1148 } 1149 #endif /* CONFIG_PROC_FS */ 1150