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 raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif)) 135 goto found; /* gotcha */ 136 } 137 sk = NULL; 138 found: 139 return sk; 140 } 141 EXPORT_SYMBOL_GPL(__raw_v4_lookup); 142 143 /* 144 * 0 - deliver 145 * 1 - block 146 */ 147 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb) 148 { 149 struct icmphdr _hdr; 150 const struct icmphdr *hdr; 151 152 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 153 sizeof(_hdr), &_hdr); 154 if (!hdr) 155 return 1; 156 157 if (hdr->type < 32) { 158 __u32 data = raw_sk(sk)->filter.data; 159 160 return ((1U << hdr->type) & data) != 0; 161 } 162 163 /* Do not block unknown ICMP types */ 164 return 0; 165 } 166 167 /* IP input processing comes here for RAW socket delivery. 168 * Caller owns SKB, so we must make clones. 169 * 170 * RFC 1122: SHOULD pass TOS value up to the transport layer. 171 * -> It does. And not only TOS, but all IP header. 172 */ 173 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash) 174 { 175 int sdif = inet_sdif(skb); 176 struct sock *sk; 177 struct hlist_head *head; 178 int delivered = 0; 179 struct net *net; 180 181 read_lock(&raw_v4_hashinfo.lock); 182 head = &raw_v4_hashinfo.ht[hash]; 183 if (hlist_empty(head)) 184 goto out; 185 186 net = dev_net(skb->dev); 187 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol, 188 iph->saddr, iph->daddr, 189 skb->dev->ifindex, sdif); 190 191 while (sk) { 192 delivered = 1; 193 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) && 194 ip_mc_sf_allow(sk, iph->daddr, iph->saddr, 195 skb->dev->ifindex, sdif)) { 196 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 197 198 /* Not releasing hash table! */ 199 if (clone) 200 raw_rcv(sk, clone); 201 } 202 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol, 203 iph->saddr, iph->daddr, 204 skb->dev->ifindex, sdif); 205 } 206 out: 207 read_unlock(&raw_v4_hashinfo.lock); 208 return delivered; 209 } 210 211 int raw_local_deliver(struct sk_buff *skb, int protocol) 212 { 213 int hash; 214 struct sock *raw_sk; 215 216 hash = protocol & (RAW_HTABLE_SIZE - 1); 217 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 218 219 /* If there maybe a raw socket we must check - if not we 220 * don't care less 221 */ 222 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash)) 223 raw_sk = NULL; 224 225 return raw_sk != NULL; 226 227 } 228 229 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) 230 { 231 struct inet_sock *inet = inet_sk(sk); 232 const int type = icmp_hdr(skb)->type; 233 const int code = icmp_hdr(skb)->code; 234 int err = 0; 235 int harderr = 0; 236 237 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) 238 ipv4_sk_update_pmtu(skb, sk, info); 239 else if (type == ICMP_REDIRECT) { 240 ipv4_sk_redirect(skb, sk); 241 return; 242 } 243 244 /* Report error on raw socket, if: 245 1. User requested ip_recverr. 246 2. Socket is connected (otherwise the error indication 247 is useless without ip_recverr and error is hard. 248 */ 249 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) 250 return; 251 252 switch (type) { 253 default: 254 case ICMP_TIME_EXCEEDED: 255 err = EHOSTUNREACH; 256 break; 257 case ICMP_SOURCE_QUENCH: 258 return; 259 case ICMP_PARAMETERPROB: 260 err = EPROTO; 261 harderr = 1; 262 break; 263 case ICMP_DEST_UNREACH: 264 err = EHOSTUNREACH; 265 if (code > NR_ICMP_UNREACH) 266 break; 267 err = icmp_err_convert[code].errno; 268 harderr = icmp_err_convert[code].fatal; 269 if (code == ICMP_FRAG_NEEDED) { 270 harderr = inet->pmtudisc != IP_PMTUDISC_DONT; 271 err = EMSGSIZE; 272 } 273 } 274 275 if (inet->recverr) { 276 const struct iphdr *iph = (const struct iphdr *)skb->data; 277 u8 *payload = skb->data + (iph->ihl << 2); 278 279 if (inet->hdrincl) 280 payload = skb->data; 281 ip_icmp_error(sk, skb, err, 0, info, payload); 282 } 283 284 if (inet->recverr || harderr) { 285 sk->sk_err = err; 286 sk->sk_error_report(sk); 287 } 288 } 289 290 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) 291 { 292 int hash; 293 struct sock *raw_sk; 294 const struct iphdr *iph; 295 struct net *net; 296 297 hash = protocol & (RAW_HTABLE_SIZE - 1); 298 299 read_lock(&raw_v4_hashinfo.lock); 300 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); 301 if (raw_sk) { 302 int dif = skb->dev->ifindex; 303 int sdif = inet_sdif(skb); 304 305 iph = (const struct iphdr *)skb->data; 306 net = dev_net(skb->dev); 307 308 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol, 309 iph->daddr, iph->saddr, 310 dif, sdif)) != NULL) { 311 raw_err(raw_sk, skb, info); 312 raw_sk = sk_next(raw_sk); 313 iph = (const struct iphdr *)skb->data; 314 } 315 } 316 read_unlock(&raw_v4_hashinfo.lock); 317 } 318 319 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb) 320 { 321 /* Charge it to the socket. */ 322 323 ipv4_pktinfo_prepare(sk, skb); 324 if (sock_queue_rcv_skb(sk, skb) < 0) { 325 kfree_skb(skb); 326 return NET_RX_DROP; 327 } 328 329 return NET_RX_SUCCESS; 330 } 331 332 int raw_rcv(struct sock *sk, struct sk_buff *skb) 333 { 334 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { 335 atomic_inc(&sk->sk_drops); 336 kfree_skb(skb); 337 return NET_RX_DROP; 338 } 339 nf_reset(skb); 340 341 skb_push(skb, skb->data - skb_network_header(skb)); 342 343 raw_rcv_skb(sk, skb); 344 return 0; 345 } 346 347 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4, 348 struct msghdr *msg, size_t length, 349 struct rtable **rtp, unsigned int flags, 350 const struct sockcm_cookie *sockc) 351 { 352 struct inet_sock *inet = inet_sk(sk); 353 struct net *net = sock_net(sk); 354 struct iphdr *iph; 355 struct sk_buff *skb; 356 unsigned int iphlen; 357 int err; 358 struct rtable *rt = *rtp; 359 int hlen, tlen; 360 361 if (length > rt->dst.dev->mtu) { 362 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport, 363 rt->dst.dev->mtu); 364 return -EMSGSIZE; 365 } 366 if (length < sizeof(struct iphdr)) 367 return -EINVAL; 368 369 if (flags&MSG_PROBE) 370 goto out; 371 372 hlen = LL_RESERVED_SPACE(rt->dst.dev); 373 tlen = rt->dst.dev->needed_tailroom; 374 skb = sock_alloc_send_skb(sk, 375 length + hlen + tlen + 15, 376 flags & MSG_DONTWAIT, &err); 377 if (!skb) 378 goto error; 379 skb_reserve(skb, hlen); 380 381 skb->priority = sk->sk_priority; 382 skb->mark = sk->sk_mark; 383 skb->tstamp = sockc->transmit_time; 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 skb_setup_tx_timestamp(skb, sockc->tsflags); 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 int hdrincl; 517 518 err = -EMSGSIZE; 519 if (len > 0xFFFF) 520 goto out; 521 522 /* hdrincl should be READ_ONCE(inet->hdrincl) 523 * but READ_ONCE() doesn't work with bit fields. 524 * Doing this indirectly yields the same result. 525 */ 526 hdrincl = inet->hdrincl; 527 hdrincl = READ_ONCE(hdrincl); 528 /* 529 * Check the flags. 530 */ 531 532 err = -EOPNOTSUPP; 533 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ 534 goto out; /* compatibility */ 535 536 /* 537 * Get and verify the address. 538 */ 539 540 if (msg->msg_namelen) { 541 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 542 err = -EINVAL; 543 if (msg->msg_namelen < sizeof(*usin)) 544 goto out; 545 if (usin->sin_family != AF_INET) { 546 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n", 547 __func__, current->comm); 548 err = -EAFNOSUPPORT; 549 if (usin->sin_family) 550 goto out; 551 } 552 daddr = usin->sin_addr.s_addr; 553 /* ANK: I did not forget to get protocol from port field. 554 * I just do not know, who uses this weirdness. 555 * IP_HDRINCL is much more convenient. 556 */ 557 } else { 558 err = -EDESTADDRREQ; 559 if (sk->sk_state != TCP_ESTABLISHED) 560 goto out; 561 daddr = inet->inet_daddr; 562 } 563 564 ipcm_init_sk(&ipc, inet); 565 566 if (msg->msg_controllen) { 567 err = ip_cmsg_send(sk, msg, &ipc, false); 568 if (unlikely(err)) { 569 kfree(ipc.opt); 570 goto out; 571 } 572 if (ipc.opt) 573 free = 1; 574 } 575 576 saddr = ipc.addr; 577 ipc.addr = daddr; 578 579 if (!ipc.opt) { 580 struct ip_options_rcu *inet_opt; 581 582 rcu_read_lock(); 583 inet_opt = rcu_dereference(inet->inet_opt); 584 if (inet_opt) { 585 memcpy(&opt_copy, inet_opt, 586 sizeof(*inet_opt) + inet_opt->opt.optlen); 587 ipc.opt = &opt_copy.opt; 588 } 589 rcu_read_unlock(); 590 } 591 592 if (ipc.opt) { 593 err = -EINVAL; 594 /* Linux does not mangle headers on raw sockets, 595 * so that IP options + IP_HDRINCL is non-sense. 596 */ 597 if (hdrincl) 598 goto done; 599 if (ipc.opt->opt.srr) { 600 if (!daddr) 601 goto done; 602 daddr = ipc.opt->opt.faddr; 603 } 604 } 605 tos = get_rtconn_flags(&ipc, sk); 606 if (msg->msg_flags & MSG_DONTROUTE) 607 tos |= RTO_ONLINK; 608 609 if (ipv4_is_multicast(daddr)) { 610 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif)) 611 ipc.oif = inet->mc_index; 612 if (!saddr) 613 saddr = inet->mc_addr; 614 } else if (!ipc.oif) { 615 ipc.oif = inet->uc_index; 616 } else if (ipv4_is_lbcast(daddr) && inet->uc_index) { 617 /* oif is set, packet is to local broadcast and 618 * and uc_index is set. oif is most likely set 619 * by sk_bound_dev_if. If uc_index != oif check if the 620 * oif is an L3 master and uc_index is an L3 slave. 621 * If so, we want to allow the send using the uc_index. 622 */ 623 if (ipc.oif != inet->uc_index && 624 ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk), 625 inet->uc_index)) { 626 ipc.oif = inet->uc_index; 627 } 628 } 629 630 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos, 631 RT_SCOPE_UNIVERSE, 632 hdrincl ? IPPROTO_RAW : sk->sk_protocol, 633 inet_sk_flowi_flags(sk) | 634 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0), 635 daddr, saddr, 0, 0, sk->sk_uid); 636 637 if (!hdrincl) { 638 rfv.msg = msg; 639 rfv.hlen = 0; 640 641 err = raw_probe_proto_opt(&rfv, &fl4); 642 if (err) 643 goto done; 644 } 645 646 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4)); 647 rt = ip_route_output_flow(net, &fl4, sk); 648 if (IS_ERR(rt)) { 649 err = PTR_ERR(rt); 650 rt = NULL; 651 goto done; 652 } 653 654 err = -EACCES; 655 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) 656 goto done; 657 658 if (msg->msg_flags & MSG_CONFIRM) 659 goto do_confirm; 660 back_from_confirm: 661 662 if (hdrincl) 663 err = raw_send_hdrinc(sk, &fl4, msg, len, 664 &rt, msg->msg_flags, &ipc.sockc); 665 666 else { 667 if (!ipc.addr) 668 ipc.addr = fl4.daddr; 669 lock_sock(sk); 670 err = ip_append_data(sk, &fl4, raw_getfrag, 671 &rfv, len, 0, 672 &ipc, &rt, msg->msg_flags); 673 if (err) 674 ip_flush_pending_frames(sk); 675 else if (!(msg->msg_flags & MSG_MORE)) { 676 err = ip_push_pending_frames(sk, &fl4); 677 if (err == -ENOBUFS && !inet->recverr) 678 err = 0; 679 } 680 release_sock(sk); 681 } 682 done: 683 if (free) 684 kfree(ipc.opt); 685 ip_rt_put(rt); 686 687 out: 688 if (err < 0) 689 return err; 690 return len; 691 692 do_confirm: 693 if (msg->msg_flags & MSG_PROBE) 694 dst_confirm_neigh(&rt->dst, &fl4.daddr); 695 if (!(msg->msg_flags & MSG_PROBE) || len) 696 goto back_from_confirm; 697 err = 0; 698 goto done; 699 } 700 701 static void raw_close(struct sock *sk, long timeout) 702 { 703 /* 704 * Raw sockets may have direct kernel references. Kill them. 705 */ 706 ip_ra_control(sk, 0, NULL); 707 708 sk_common_release(sk); 709 } 710 711 static void raw_destroy(struct sock *sk) 712 { 713 lock_sock(sk); 714 ip_flush_pending_frames(sk); 715 release_sock(sk); 716 } 717 718 /* This gets rid of all the nasties in af_inet. -DaveM */ 719 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 720 { 721 struct inet_sock *inet = inet_sk(sk); 722 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 723 u32 tb_id = RT_TABLE_LOCAL; 724 int ret = -EINVAL; 725 int chk_addr_ret; 726 727 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) 728 goto out; 729 730 if (sk->sk_bound_dev_if) 731 tb_id = l3mdev_fib_table_by_index(sock_net(sk), 732 sk->sk_bound_dev_if) ? : tb_id; 733 734 chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr, 735 tb_id); 736 737 ret = -EADDRNOTAVAIL; 738 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && 739 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) 740 goto out; 741 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 742 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 743 inet->inet_saddr = 0; /* Use device */ 744 sk_dst_reset(sk); 745 ret = 0; 746 out: return ret; 747 } 748 749 /* 750 * This should be easy, if there is something there 751 * we return it, otherwise we block. 752 */ 753 754 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 755 int noblock, int flags, int *addr_len) 756 { 757 struct inet_sock *inet = inet_sk(sk); 758 size_t copied = 0; 759 int err = -EOPNOTSUPP; 760 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 761 struct sk_buff *skb; 762 763 if (flags & MSG_OOB) 764 goto out; 765 766 if (flags & MSG_ERRQUEUE) { 767 err = ip_recv_error(sk, msg, len, addr_len); 768 goto out; 769 } 770 771 skb = skb_recv_datagram(sk, flags, noblock, &err); 772 if (!skb) 773 goto out; 774 775 copied = skb->len; 776 if (len < copied) { 777 msg->msg_flags |= MSG_TRUNC; 778 copied = len; 779 } 780 781 err = skb_copy_datagram_msg(skb, 0, msg, copied); 782 if (err) 783 goto done; 784 785 sock_recv_ts_and_drops(msg, sk, skb); 786 787 /* Copy the address. */ 788 if (sin) { 789 sin->sin_family = AF_INET; 790 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 791 sin->sin_port = 0; 792 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 793 *addr_len = sizeof(*sin); 794 } 795 if (inet->cmsg_flags) 796 ip_cmsg_recv(msg, skb); 797 if (flags & MSG_TRUNC) 798 copied = skb->len; 799 done: 800 skb_free_datagram(sk, skb); 801 out: 802 if (err) 803 return err; 804 return copied; 805 } 806 807 static int raw_sk_init(struct sock *sk) 808 { 809 struct raw_sock *rp = raw_sk(sk); 810 811 if (inet_sk(sk)->inet_num == IPPROTO_ICMP) 812 memset(&rp->filter, 0, sizeof(rp->filter)); 813 return 0; 814 } 815 816 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) 817 { 818 if (optlen > sizeof(struct icmp_filter)) 819 optlen = sizeof(struct icmp_filter); 820 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) 821 return -EFAULT; 822 return 0; 823 } 824 825 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) 826 { 827 int len, ret = -EFAULT; 828 829 if (get_user(len, optlen)) 830 goto out; 831 ret = -EINVAL; 832 if (len < 0) 833 goto out; 834 if (len > sizeof(struct icmp_filter)) 835 len = sizeof(struct icmp_filter); 836 ret = -EFAULT; 837 if (put_user(len, optlen) || 838 copy_to_user(optval, &raw_sk(sk)->filter, len)) 839 goto out; 840 ret = 0; 841 out: return ret; 842 } 843 844 static int do_raw_setsockopt(struct sock *sk, int level, int optname, 845 char __user *optval, unsigned int optlen) 846 { 847 if (optname == ICMP_FILTER) { 848 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 849 return -EOPNOTSUPP; 850 else 851 return raw_seticmpfilter(sk, optval, optlen); 852 } 853 return -ENOPROTOOPT; 854 } 855 856 static int raw_setsockopt(struct sock *sk, int level, int optname, 857 char __user *optval, unsigned int optlen) 858 { 859 if (level != SOL_RAW) 860 return ip_setsockopt(sk, level, optname, optval, optlen); 861 return do_raw_setsockopt(sk, level, optname, optval, optlen); 862 } 863 864 #ifdef CONFIG_COMPAT 865 static int compat_raw_setsockopt(struct sock *sk, int level, int optname, 866 char __user *optval, unsigned int optlen) 867 { 868 if (level != SOL_RAW) 869 return compat_ip_setsockopt(sk, level, optname, optval, optlen); 870 return do_raw_setsockopt(sk, level, optname, optval, optlen); 871 } 872 #endif 873 874 static int do_raw_getsockopt(struct sock *sk, int level, int optname, 875 char __user *optval, int __user *optlen) 876 { 877 if (optname == ICMP_FILTER) { 878 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 879 return -EOPNOTSUPP; 880 else 881 return raw_geticmpfilter(sk, optval, optlen); 882 } 883 return -ENOPROTOOPT; 884 } 885 886 static int raw_getsockopt(struct sock *sk, int level, int optname, 887 char __user *optval, int __user *optlen) 888 { 889 if (level != SOL_RAW) 890 return ip_getsockopt(sk, level, optname, optval, optlen); 891 return do_raw_getsockopt(sk, level, optname, optval, optlen); 892 } 893 894 #ifdef CONFIG_COMPAT 895 static int compat_raw_getsockopt(struct sock *sk, int level, int optname, 896 char __user *optval, int __user *optlen) 897 { 898 if (level != SOL_RAW) 899 return compat_ip_getsockopt(sk, level, optname, optval, optlen); 900 return do_raw_getsockopt(sk, level, optname, optval, optlen); 901 } 902 #endif 903 904 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) 905 { 906 switch (cmd) { 907 case SIOCOUTQ: { 908 int amount = sk_wmem_alloc_get(sk); 909 910 return put_user(amount, (int __user *)arg); 911 } 912 case SIOCINQ: { 913 struct sk_buff *skb; 914 int amount = 0; 915 916 spin_lock_bh(&sk->sk_receive_queue.lock); 917 skb = skb_peek(&sk->sk_receive_queue); 918 if (skb) 919 amount = skb->len; 920 spin_unlock_bh(&sk->sk_receive_queue.lock); 921 return put_user(amount, (int __user *)arg); 922 } 923 924 default: 925 #ifdef CONFIG_IP_MROUTE 926 return ipmr_ioctl(sk, cmd, (void __user *)arg); 927 #else 928 return -ENOIOCTLCMD; 929 #endif 930 } 931 } 932 933 #ifdef CONFIG_COMPAT 934 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 935 { 936 switch (cmd) { 937 case SIOCOUTQ: 938 case SIOCINQ: 939 return -ENOIOCTLCMD; 940 default: 941 #ifdef CONFIG_IP_MROUTE 942 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg)); 943 #else 944 return -ENOIOCTLCMD; 945 #endif 946 } 947 } 948 #endif 949 950 int raw_abort(struct sock *sk, int err) 951 { 952 lock_sock(sk); 953 954 sk->sk_err = err; 955 sk->sk_error_report(sk); 956 __udp_disconnect(sk, 0); 957 958 release_sock(sk); 959 960 return 0; 961 } 962 EXPORT_SYMBOL_GPL(raw_abort); 963 964 struct proto raw_prot = { 965 .name = "RAW", 966 .owner = THIS_MODULE, 967 .close = raw_close, 968 .destroy = raw_destroy, 969 .connect = ip4_datagram_connect, 970 .disconnect = __udp_disconnect, 971 .ioctl = raw_ioctl, 972 .init = raw_sk_init, 973 .setsockopt = raw_setsockopt, 974 .getsockopt = raw_getsockopt, 975 .sendmsg = raw_sendmsg, 976 .recvmsg = raw_recvmsg, 977 .bind = raw_bind, 978 .backlog_rcv = raw_rcv_skb, 979 .release_cb = ip4_datagram_release_cb, 980 .hash = raw_hash_sk, 981 .unhash = raw_unhash_sk, 982 .obj_size = sizeof(struct raw_sock), 983 .useroffset = offsetof(struct raw_sock, filter), 984 .usersize = sizeof_field(struct raw_sock, filter), 985 .h.raw_hash = &raw_v4_hashinfo, 986 #ifdef CONFIG_COMPAT 987 .compat_setsockopt = compat_raw_setsockopt, 988 .compat_getsockopt = compat_raw_getsockopt, 989 .compat_ioctl = compat_raw_ioctl, 990 #endif 991 .diag_destroy = raw_abort, 992 }; 993 994 #ifdef CONFIG_PROC_FS 995 static struct sock *raw_get_first(struct seq_file *seq) 996 { 997 struct sock *sk; 998 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file)); 999 struct raw_iter_state *state = raw_seq_private(seq); 1000 1001 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE; 1002 ++state->bucket) { 1003 sk_for_each(sk, &h->ht[state->bucket]) 1004 if (sock_net(sk) == seq_file_net(seq)) 1005 goto found; 1006 } 1007 sk = NULL; 1008 found: 1009 return sk; 1010 } 1011 1012 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) 1013 { 1014 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file)); 1015 struct raw_iter_state *state = raw_seq_private(seq); 1016 1017 do { 1018 sk = sk_next(sk); 1019 try_again: 1020 ; 1021 } while (sk && sock_net(sk) != seq_file_net(seq)); 1022 1023 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) { 1024 sk = sk_head(&h->ht[state->bucket]); 1025 goto try_again; 1026 } 1027 return sk; 1028 } 1029 1030 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) 1031 { 1032 struct sock *sk = raw_get_first(seq); 1033 1034 if (sk) 1035 while (pos && (sk = raw_get_next(seq, sk)) != NULL) 1036 --pos; 1037 return pos ? NULL : sk; 1038 } 1039 1040 void *raw_seq_start(struct seq_file *seq, loff_t *pos) 1041 { 1042 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file)); 1043 1044 read_lock(&h->lock); 1045 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 1046 } 1047 EXPORT_SYMBOL_GPL(raw_seq_start); 1048 1049 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1050 { 1051 struct sock *sk; 1052 1053 if (v == SEQ_START_TOKEN) 1054 sk = raw_get_first(seq); 1055 else 1056 sk = raw_get_next(seq, v); 1057 ++*pos; 1058 return sk; 1059 } 1060 EXPORT_SYMBOL_GPL(raw_seq_next); 1061 1062 void raw_seq_stop(struct seq_file *seq, void *v) 1063 { 1064 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file)); 1065 1066 read_unlock(&h->lock); 1067 } 1068 EXPORT_SYMBOL_GPL(raw_seq_stop); 1069 1070 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1071 { 1072 struct inet_sock *inet = inet_sk(sp); 1073 __be32 dest = inet->inet_daddr, 1074 src = inet->inet_rcv_saddr; 1075 __u16 destp = 0, 1076 srcp = inet->inet_num; 1077 1078 seq_printf(seq, "%4d: %08X:%04X %08X:%04X" 1079 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n", 1080 i, src, srcp, dest, destp, sp->sk_state, 1081 sk_wmem_alloc_get(sp), 1082 sk_rmem_alloc_get(sp), 1083 0, 0L, 0, 1084 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1085 0, sock_i_ino(sp), 1086 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); 1087 } 1088 1089 static int raw_seq_show(struct seq_file *seq, void *v) 1090 { 1091 if (v == SEQ_START_TOKEN) 1092 seq_printf(seq, " sl local_address rem_address st tx_queue " 1093 "rx_queue tr tm->when retrnsmt uid timeout " 1094 "inode ref pointer drops\n"); 1095 else 1096 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket); 1097 return 0; 1098 } 1099 1100 static const struct seq_operations raw_seq_ops = { 1101 .start = raw_seq_start, 1102 .next = raw_seq_next, 1103 .stop = raw_seq_stop, 1104 .show = raw_seq_show, 1105 }; 1106 1107 static __net_init int raw_init_net(struct net *net) 1108 { 1109 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops, 1110 sizeof(struct raw_iter_state), &raw_v4_hashinfo)) 1111 return -ENOMEM; 1112 1113 return 0; 1114 } 1115 1116 static __net_exit void raw_exit_net(struct net *net) 1117 { 1118 remove_proc_entry("raw", net->proc_net); 1119 } 1120 1121 static __net_initdata struct pernet_operations raw_net_ops = { 1122 .init = raw_init_net, 1123 .exit = raw_exit_net, 1124 }; 1125 1126 int __init raw_proc_init(void) 1127 { 1128 return register_pernet_subsys(&raw_net_ops); 1129 } 1130 1131 void __init raw_proc_exit(void) 1132 { 1133 unregister_pernet_subsys(&raw_net_ops); 1134 } 1135 #endif /* CONFIG_PROC_FS */ 1136 1137 static void raw_sysctl_init_net(struct net *net) 1138 { 1139 #ifdef CONFIG_NET_L3_MASTER_DEV 1140 net->ipv4.sysctl_raw_l3mdev_accept = 1; 1141 #endif 1142 } 1143 1144 static int __net_init raw_sysctl_init(struct net *net) 1145 { 1146 raw_sysctl_init_net(net); 1147 return 0; 1148 } 1149 1150 static struct pernet_operations __net_initdata raw_sysctl_ops = { 1151 .init = raw_sysctl_init, 1152 }; 1153 1154 void __init raw_init(void) 1155 { 1156 raw_sysctl_init_net(&init_net); 1157 if (register_pernet_subsys(&raw_sysctl_ops)) 1158 panic("RAW: failed to init sysctl parameters.\n"); 1159 } 1160