1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * RAW - implementation of IP "raw" sockets. 8 * 9 * Authors: Ross Biro 10 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 11 * 12 * Fixes: 13 * Alan Cox : verify_area() fixed up 14 * Alan Cox : ICMP error handling 15 * Alan Cox : EMSGSIZE if you send too big a packet 16 * Alan Cox : Now uses generic datagrams and shared 17 * skbuff library. No more peek crashes, 18 * no more backlogs 19 * Alan Cox : Checks sk->broadcast. 20 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram 21 * Alan Cox : Raw passes ip options too 22 * Alan Cox : Setsocketopt added 23 * Alan Cox : Fixed error return for broadcasts 24 * Alan Cox : Removed wake_up calls 25 * Alan Cox : Use ttl/tos 26 * Alan Cox : Cleaned up old debugging 27 * Alan Cox : Use new kernel side addresses 28 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets. 29 * Alan Cox : BSD style RAW socket demultiplexing. 30 * Alan Cox : Beginnings of mrouted support. 31 * Alan Cox : Added IP_HDRINCL option. 32 * Alan Cox : Skip broadcast check if BSDism set. 33 * David S. Miller : New socket lookup architecture. 34 */ 35 36 #include <linux/types.h> 37 #include <linux/atomic.h> 38 #include <asm/byteorder.h> 39 #include <asm/current.h> 40 #include <linux/uaccess.h> 41 #include <asm/ioctls.h> 42 #include <linux/stddef.h> 43 #include <linux/slab.h> 44 #include <linux/errno.h> 45 #include <linux/kernel.h> 46 #include <linux/export.h> 47 #include <linux/spinlock.h> 48 #include <linux/sockios.h> 49 #include <linux/socket.h> 50 #include <linux/in.h> 51 #include <linux/mroute.h> 52 #include <linux/netdevice.h> 53 #include <linux/in_route.h> 54 #include <linux/route.h> 55 #include <linux/skbuff.h> 56 #include <linux/igmp.h> 57 #include <net/net_namespace.h> 58 #include <net/dst.h> 59 #include <net/sock.h> 60 #include <linux/ip.h> 61 #include <linux/net.h> 62 #include <net/ip.h> 63 #include <net/icmp.h> 64 #include <net/udp.h> 65 #include <net/raw.h> 66 #include <net/snmp.h> 67 #include <net/tcp_states.h> 68 #include <net/inet_common.h> 69 #include <net/checksum.h> 70 #include <net/xfrm.h> 71 #include <linux/rtnetlink.h> 72 #include <linux/proc_fs.h> 73 #include <linux/seq_file.h> 74 #include <linux/netfilter.h> 75 #include <linux/netfilter_ipv4.h> 76 #include <linux/compat.h> 77 #include <linux/uio.h> 78 79 struct raw_frag_vec { 80 struct msghdr *msg; 81 union { 82 struct icmphdr icmph; 83 char c[1]; 84 } hdr; 85 int hlen; 86 }; 87 88 struct raw_hashinfo raw_v4_hashinfo; 89 EXPORT_SYMBOL_GPL(raw_v4_hashinfo); 90 91 int raw_hash_sk(struct sock *sk) 92 { 93 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 94 struct hlist_head *hlist; 95 96 hlist = &h->ht[raw_hashfunc(sock_net(sk), inet_sk(sk)->inet_num)]; 97 98 spin_lock(&h->lock); 99 sk_add_node_rcu(sk, hlist); 100 sock_set_flag(sk, SOCK_RCU_FREE); 101 spin_unlock(&h->lock); 102 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 103 104 return 0; 105 } 106 EXPORT_SYMBOL_GPL(raw_hash_sk); 107 108 void raw_unhash_sk(struct sock *sk) 109 { 110 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash; 111 112 spin_lock(&h->lock); 113 if (sk_del_node_init_rcu(sk)) 114 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 115 spin_unlock(&h->lock); 116 } 117 EXPORT_SYMBOL_GPL(raw_unhash_sk); 118 119 bool raw_v4_match(struct net *net, const struct sock *sk, unsigned short num, 120 __be32 raddr, __be32 laddr, int dif, int sdif) 121 { 122 const struct inet_sock *inet = inet_sk(sk); 123 124 if (net_eq(sock_net(sk), net) && inet->inet_num == num && 125 !(inet->inet_daddr && inet->inet_daddr != raddr) && 126 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) && 127 raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif)) 128 return true; 129 return false; 130 } 131 EXPORT_SYMBOL_GPL(raw_v4_match); 132 133 /* 134 * 0 - deliver 135 * 1 - block 136 */ 137 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb) 138 { 139 struct icmphdr _hdr; 140 const struct icmphdr *hdr; 141 142 hdr = skb_header_pointer(skb, skb_transport_offset(skb), 143 sizeof(_hdr), &_hdr); 144 if (!hdr) 145 return 1; 146 147 if (hdr->type < 32) { 148 __u32 data = raw_sk(sk)->filter.data; 149 150 return ((1U << hdr->type) & data) != 0; 151 } 152 153 /* Do not block unknown ICMP types */ 154 return 0; 155 } 156 157 /* IP input processing comes here for RAW socket delivery. 158 * Caller owns SKB, so we must make clones. 159 * 160 * RFC 1122: SHOULD pass TOS value up to the transport layer. 161 * -> It does. And not only TOS, but all IP header. 162 */ 163 static int raw_v4_input(struct net *net, struct sk_buff *skb, 164 const struct iphdr *iph, int hash) 165 { 166 int sdif = inet_sdif(skb); 167 struct hlist_head *hlist; 168 int dif = inet_iif(skb); 169 int delivered = 0; 170 struct sock *sk; 171 172 hlist = &raw_v4_hashinfo.ht[hash]; 173 rcu_read_lock(); 174 sk_for_each_rcu(sk, hlist) { 175 if (!raw_v4_match(net, sk, iph->protocol, 176 iph->saddr, iph->daddr, dif, sdif)) 177 continue; 178 179 if (atomic_read(&sk->sk_rmem_alloc) >= 180 READ_ONCE(sk->sk_rcvbuf)) { 181 atomic_inc(&sk->sk_drops); 182 continue; 183 } 184 185 delivered = 1; 186 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) && 187 ip_mc_sf_allow(sk, iph->daddr, iph->saddr, 188 skb->dev->ifindex, sdif)) { 189 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 190 191 /* Not releasing hash table! */ 192 if (clone) 193 raw_rcv(sk, clone); 194 } 195 } 196 rcu_read_unlock(); 197 return delivered; 198 } 199 200 int raw_local_deliver(struct sk_buff *skb, int protocol) 201 { 202 struct net *net = dev_net(skb->dev); 203 204 return raw_v4_input(net, skb, ip_hdr(skb), 205 raw_hashfunc(net, protocol)); 206 } 207 208 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) 209 { 210 struct inet_sock *inet = inet_sk(sk); 211 const int type = icmp_hdr(skb)->type; 212 const int code = icmp_hdr(skb)->code; 213 int harderr = 0; 214 bool recverr; 215 int err = 0; 216 217 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) 218 ipv4_sk_update_pmtu(skb, sk, info); 219 else if (type == ICMP_REDIRECT) { 220 ipv4_sk_redirect(skb, sk); 221 return; 222 } 223 224 /* Report error on raw socket, if: 225 1. User requested ip_recverr. 226 2. Socket is connected (otherwise the error indication 227 is useless without ip_recverr and error is hard. 228 */ 229 recverr = inet_test_bit(RECVERR, sk); 230 if (!recverr && sk->sk_state != TCP_ESTABLISHED) 231 return; 232 233 switch (type) { 234 default: 235 case ICMP_TIME_EXCEEDED: 236 err = EHOSTUNREACH; 237 break; 238 case ICMP_SOURCE_QUENCH: 239 return; 240 case ICMP_PARAMETERPROB: 241 err = EPROTO; 242 harderr = 1; 243 break; 244 case ICMP_DEST_UNREACH: 245 err = EHOSTUNREACH; 246 if (code > NR_ICMP_UNREACH) 247 break; 248 if (code == ICMP_FRAG_NEEDED) { 249 harderr = READ_ONCE(inet->pmtudisc) != IP_PMTUDISC_DONT; 250 err = EMSGSIZE; 251 } else { 252 err = icmp_err_convert[code].errno; 253 harderr = icmp_err_convert[code].fatal; 254 } 255 } 256 257 if (recverr) { 258 const struct iphdr *iph = (const struct iphdr *)skb->data; 259 u8 *payload = skb->data + (iph->ihl << 2); 260 261 if (inet_test_bit(HDRINCL, sk)) 262 payload = skb->data; 263 ip_icmp_error(sk, skb, err, 0, info, payload); 264 } 265 266 if (recverr || harderr) { 267 sk->sk_err = err; 268 sk_error_report(sk); 269 } 270 } 271 272 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) 273 { 274 struct net *net = dev_net(skb->dev); 275 int dif = skb->dev->ifindex; 276 int sdif = inet_sdif(skb); 277 struct hlist_head *hlist; 278 const struct iphdr *iph; 279 struct sock *sk; 280 int hash; 281 282 hash = raw_hashfunc(net, protocol); 283 hlist = &raw_v4_hashinfo.ht[hash]; 284 285 rcu_read_lock(); 286 sk_for_each_rcu(sk, hlist) { 287 iph = (const struct iphdr *)skb->data; 288 if (!raw_v4_match(net, sk, iph->protocol, 289 iph->daddr, iph->saddr, dif, sdif)) 290 continue; 291 raw_err(sk, skb, info); 292 } 293 rcu_read_unlock(); 294 } 295 296 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb) 297 { 298 enum skb_drop_reason reason; 299 300 /* Charge it to the socket. */ 301 302 ipv4_pktinfo_prepare(sk, skb, true); 303 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) { 304 kfree_skb_reason(skb, reason); 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_reason(skb, SKB_DROP_REASON_XFRM_POLICY); 316 return NET_RX_DROP; 317 } 318 nf_reset_ct(skb); 319 320 skb_push(skb, -skb_network_offset(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 struct msghdr *msg, size_t length, 328 struct rtable **rtp, unsigned int flags, 329 const struct sockcm_cookie *sockc) 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 (length < sizeof(struct iphdr)) 346 return -EINVAL; 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) 357 goto error; 358 skb_reserve(skb, hlen); 359 360 skb->priority = READ_ONCE(sk->sk_priority); 361 skb->mark = sockc->mark; 362 skb->tstamp = sockc->transmit_time; 363 skb_dst_set(skb, &rt->dst); 364 *rtp = NULL; 365 366 skb_reset_network_header(skb); 367 iph = ip_hdr(skb); 368 skb_put(skb, length); 369 370 skb->ip_summed = CHECKSUM_NONE; 371 372 skb_setup_tx_timestamp(skb, sockc->tsflags); 373 374 if (flags & MSG_CONFIRM) 375 skb_set_dst_pending_confirm(skb, 1); 376 377 skb->transport_header = skb->network_header; 378 err = -EFAULT; 379 if (memcpy_from_msg(iph, msg, length)) 380 goto error_free; 381 382 iphlen = iph->ihl * 4; 383 384 /* 385 * We don't want to modify the ip header, but we do need to 386 * be sure that it won't cause problems later along the network 387 * stack. Specifically we want to make sure that iph->ihl is a 388 * sane value. If ihl points beyond the length of the buffer passed 389 * in, reject the frame as invalid 390 */ 391 err = -EINVAL; 392 if (iphlen > length) 393 goto error_free; 394 395 if (iphlen >= sizeof(*iph)) { 396 if (!iph->saddr) 397 iph->saddr = fl4->saddr; 398 iph->check = 0; 399 iph->tot_len = htons(length); 400 if (!iph->id) 401 ip_select_ident(net, skb, NULL); 402 403 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); 404 skb->transport_header += iphlen; 405 if (iph->protocol == IPPROTO_ICMP && 406 length >= iphlen + sizeof(struct icmphdr)) 407 icmp_out_count(net, ((struct icmphdr *) 408 skb_transport_header(skb))->type); 409 } 410 411 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, 412 net, sk, skb, NULL, rt->dst.dev, 413 dst_output); 414 if (err > 0) 415 err = net_xmit_errno(err); 416 if (err) 417 goto error; 418 out: 419 return 0; 420 421 error_free: 422 kfree_skb(skb); 423 error: 424 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS); 425 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk)) 426 err = 0; 427 return err; 428 } 429 430 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4) 431 { 432 int err; 433 434 if (fl4->flowi4_proto != IPPROTO_ICMP) 435 return 0; 436 437 /* We only need the first two bytes. */ 438 rfv->hlen = 2; 439 440 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen); 441 if (err) 442 return err; 443 444 fl4->fl4_icmp_type = rfv->hdr.icmph.type; 445 fl4->fl4_icmp_code = rfv->hdr.icmph.code; 446 447 return 0; 448 } 449 450 static int raw_getfrag(void *from, char *to, int offset, int len, int odd, 451 struct sk_buff *skb) 452 { 453 struct raw_frag_vec *rfv = from; 454 455 if (offset < rfv->hlen) { 456 int copy = min(rfv->hlen - offset, len); 457 458 if (skb->ip_summed == CHECKSUM_PARTIAL) 459 memcpy(to, rfv->hdr.c + offset, copy); 460 else 461 skb->csum = csum_block_add( 462 skb->csum, 463 csum_partial_copy_nocheck(rfv->hdr.c + offset, 464 to, copy), 465 odd); 466 467 odd = 0; 468 offset += copy; 469 to += copy; 470 len -= copy; 471 472 if (!len) 473 return 0; 474 } 475 476 offset -= rfv->hlen; 477 478 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); 479 } 480 481 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 482 { 483 struct inet_sock *inet = inet_sk(sk); 484 struct net *net = sock_net(sk); 485 struct ipcm_cookie ipc; 486 struct rtable *rt = NULL; 487 struct flowi4 fl4; 488 u8 tos, scope; 489 int free = 0; 490 __be32 daddr; 491 __be32 saddr; 492 int uc_index, err; 493 struct ip_options_data opt_copy; 494 struct raw_frag_vec rfv; 495 int hdrincl; 496 497 err = -EMSGSIZE; 498 if (len > 0xFFFF) 499 goto out; 500 501 hdrincl = inet_test_bit(HDRINCL, sk); 502 503 /* 504 * Check the flags. 505 */ 506 507 err = -EOPNOTSUPP; 508 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ 509 goto out; /* compatibility */ 510 511 /* 512 * Get and verify the address. 513 */ 514 515 if (msg->msg_namelen) { 516 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 517 err = -EINVAL; 518 if (msg->msg_namelen < sizeof(*usin)) 519 goto out; 520 if (usin->sin_family != AF_INET) { 521 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n", 522 __func__, current->comm); 523 err = -EAFNOSUPPORT; 524 if (usin->sin_family) 525 goto out; 526 } 527 daddr = usin->sin_addr.s_addr; 528 /* ANK: I did not forget to get protocol from port field. 529 * I just do not know, who uses this weirdness. 530 * IP_HDRINCL is much more convenient. 531 */ 532 } else { 533 err = -EDESTADDRREQ; 534 if (sk->sk_state != TCP_ESTABLISHED) 535 goto out; 536 daddr = inet->inet_daddr; 537 } 538 539 ipcm_init_sk(&ipc, inet); 540 /* Keep backward compat */ 541 if (hdrincl) 542 ipc.protocol = IPPROTO_RAW; 543 544 if (msg->msg_controllen) { 545 err = ip_cmsg_send(sk, msg, &ipc, false); 546 if (unlikely(err)) { 547 kfree(ipc.opt); 548 goto out; 549 } 550 if (ipc.opt) 551 free = 1; 552 } 553 554 saddr = ipc.addr; 555 ipc.addr = daddr; 556 557 if (!ipc.opt) { 558 struct ip_options_rcu *inet_opt; 559 560 rcu_read_lock(); 561 inet_opt = rcu_dereference(inet->inet_opt); 562 if (inet_opt) { 563 memcpy(&opt_copy, inet_opt, 564 sizeof(*inet_opt) + inet_opt->opt.optlen); 565 ipc.opt = &opt_copy.opt; 566 } 567 rcu_read_unlock(); 568 } 569 570 if (ipc.opt) { 571 err = -EINVAL; 572 /* Linux does not mangle headers on raw sockets, 573 * so that IP options + IP_HDRINCL is non-sense. 574 */ 575 if (hdrincl) 576 goto done; 577 if (ipc.opt->opt.srr) { 578 if (!daddr) 579 goto done; 580 daddr = ipc.opt->opt.faddr; 581 } 582 } 583 tos = get_rttos(&ipc, inet); 584 scope = ip_sendmsg_scope(inet, &ipc, msg); 585 586 uc_index = READ_ONCE(inet->uc_index); 587 if (ipv4_is_multicast(daddr)) { 588 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif)) 589 ipc.oif = READ_ONCE(inet->mc_index); 590 if (!saddr) 591 saddr = READ_ONCE(inet->mc_addr); 592 } else if (!ipc.oif) { 593 ipc.oif = uc_index; 594 } else if (ipv4_is_lbcast(daddr) && uc_index) { 595 /* oif is set, packet is to local broadcast 596 * and uc_index is set. oif is most likely set 597 * by sk_bound_dev_if. If uc_index != oif check if the 598 * oif is an L3 master and uc_index is an L3 slave. 599 * If so, we want to allow the send using the uc_index. 600 */ 601 if (ipc.oif != uc_index && 602 ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk), 603 uc_index)) { 604 ipc.oif = uc_index; 605 } 606 } 607 608 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos, scope, 609 hdrincl ? ipc.protocol : sk->sk_protocol, 610 inet_sk_flowi_flags(sk) | 611 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0), 612 daddr, saddr, 0, 0, sk->sk_uid); 613 614 if (!hdrincl) { 615 rfv.msg = msg; 616 rfv.hlen = 0; 617 618 err = raw_probe_proto_opt(&rfv, &fl4); 619 if (err) 620 goto done; 621 } 622 623 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4)); 624 rt = ip_route_output_flow(net, &fl4, sk); 625 if (IS_ERR(rt)) { 626 err = PTR_ERR(rt); 627 rt = NULL; 628 goto done; 629 } 630 631 err = -EACCES; 632 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) 633 goto done; 634 635 if (msg->msg_flags & MSG_CONFIRM) 636 goto do_confirm; 637 back_from_confirm: 638 639 if (hdrincl) 640 err = raw_send_hdrinc(sk, &fl4, msg, len, 641 &rt, msg->msg_flags, &ipc.sockc); 642 643 else { 644 if (!ipc.addr) 645 ipc.addr = fl4.daddr; 646 lock_sock(sk); 647 err = ip_append_data(sk, &fl4, raw_getfrag, 648 &rfv, len, 0, 649 &ipc, &rt, msg->msg_flags); 650 if (err) 651 ip_flush_pending_frames(sk); 652 else if (!(msg->msg_flags & MSG_MORE)) { 653 err = ip_push_pending_frames(sk, &fl4); 654 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk)) 655 err = 0; 656 } 657 release_sock(sk); 658 } 659 done: 660 if (free) 661 kfree(ipc.opt); 662 ip_rt_put(rt); 663 664 out: 665 if (err < 0) 666 return err; 667 return len; 668 669 do_confirm: 670 if (msg->msg_flags & MSG_PROBE) 671 dst_confirm_neigh(&rt->dst, &fl4.daddr); 672 if (!(msg->msg_flags & MSG_PROBE) || len) 673 goto back_from_confirm; 674 err = 0; 675 goto done; 676 } 677 678 static void raw_close(struct sock *sk, long timeout) 679 { 680 /* 681 * Raw sockets may have direct kernel references. Kill them. 682 */ 683 ip_ra_control(sk, 0, NULL); 684 685 sk_common_release(sk); 686 } 687 688 static void raw_destroy(struct sock *sk) 689 { 690 lock_sock(sk); 691 ip_flush_pending_frames(sk); 692 release_sock(sk); 693 } 694 695 /* This gets rid of all the nasties in af_inet. -DaveM */ 696 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 697 { 698 struct inet_sock *inet = inet_sk(sk); 699 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 700 struct net *net = sock_net(sk); 701 u32 tb_id = RT_TABLE_LOCAL; 702 int ret = -EINVAL; 703 int chk_addr_ret; 704 705 lock_sock(sk); 706 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) 707 goto out; 708 709 if (sk->sk_bound_dev_if) 710 tb_id = l3mdev_fib_table_by_index(net, 711 sk->sk_bound_dev_if) ? : tb_id; 712 713 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); 714 715 ret = -EADDRNOTAVAIL; 716 if (!inet_addr_valid_or_nonlocal(net, inet, addr->sin_addr.s_addr, 717 chk_addr_ret)) 718 goto out; 719 720 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 721 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 722 inet->inet_saddr = 0; /* Use device */ 723 sk_dst_reset(sk); 724 ret = 0; 725 out: 726 release_sock(sk); 727 return ret; 728 } 729 730 /* 731 * This should be easy, if there is something there 732 * we return it, otherwise we block. 733 */ 734 735 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 736 int flags, int *addr_len) 737 { 738 struct inet_sock *inet = inet_sk(sk); 739 size_t copied = 0; 740 int err = -EOPNOTSUPP; 741 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 742 struct sk_buff *skb; 743 744 if (flags & MSG_OOB) 745 goto out; 746 747 if (flags & MSG_ERRQUEUE) { 748 err = ip_recv_error(sk, msg, len, addr_len); 749 goto out; 750 } 751 752 skb = skb_recv_datagram(sk, flags, &err); 753 if (!skb) 754 goto out; 755 756 copied = skb->len; 757 if (len < copied) { 758 msg->msg_flags |= MSG_TRUNC; 759 copied = len; 760 } 761 762 err = skb_copy_datagram_msg(skb, 0, msg, copied); 763 if (err) 764 goto done; 765 766 sock_recv_cmsgs(msg, sk, skb); 767 768 /* Copy the address. */ 769 if (sin) { 770 sin->sin_family = AF_INET; 771 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 772 sin->sin_port = 0; 773 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 774 *addr_len = sizeof(*sin); 775 } 776 if (inet_cmsg_flags(inet)) 777 ip_cmsg_recv(msg, skb); 778 if (flags & MSG_TRUNC) 779 copied = skb->len; 780 done: 781 skb_free_datagram(sk, skb); 782 out: 783 if (err) 784 return err; 785 return copied; 786 } 787 788 static int raw_sk_init(struct sock *sk) 789 { 790 struct raw_sock *rp = raw_sk(sk); 791 792 if (inet_sk(sk)->inet_num == IPPROTO_ICMP) 793 memset(&rp->filter, 0, sizeof(rp->filter)); 794 return 0; 795 } 796 797 static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen) 798 { 799 if (optlen > sizeof(struct icmp_filter)) 800 optlen = sizeof(struct icmp_filter); 801 if (copy_from_sockptr(&raw_sk(sk)->filter, optval, optlen)) 802 return -EFAULT; 803 return 0; 804 } 805 806 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) 807 { 808 int len, ret = -EFAULT; 809 810 if (get_user(len, optlen)) 811 goto out; 812 ret = -EINVAL; 813 if (len < 0) 814 goto out; 815 if (len > sizeof(struct icmp_filter)) 816 len = sizeof(struct icmp_filter); 817 ret = -EFAULT; 818 if (put_user(len, optlen) || 819 copy_to_user(optval, &raw_sk(sk)->filter, len)) 820 goto out; 821 ret = 0; 822 out: return ret; 823 } 824 825 static int do_raw_setsockopt(struct sock *sk, int optname, 826 sockptr_t optval, unsigned int optlen) 827 { 828 if (optname == ICMP_FILTER) { 829 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 830 return -EOPNOTSUPP; 831 else 832 return raw_seticmpfilter(sk, optval, optlen); 833 } 834 return -ENOPROTOOPT; 835 } 836 837 static int raw_setsockopt(struct sock *sk, int level, int optname, 838 sockptr_t optval, unsigned int optlen) 839 { 840 if (level != SOL_RAW) 841 return ip_setsockopt(sk, level, optname, optval, optlen); 842 return do_raw_setsockopt(sk, optname, optval, optlen); 843 } 844 845 static int do_raw_getsockopt(struct sock *sk, int optname, 846 char __user *optval, int __user *optlen) 847 { 848 if (optname == ICMP_FILTER) { 849 if (inet_sk(sk)->inet_num != IPPROTO_ICMP) 850 return -EOPNOTSUPP; 851 else 852 return raw_geticmpfilter(sk, optval, optlen); 853 } 854 return -ENOPROTOOPT; 855 } 856 857 static int raw_getsockopt(struct sock *sk, int level, int optname, 858 char __user *optval, int __user *optlen) 859 { 860 if (level != SOL_RAW) 861 return ip_getsockopt(sk, level, optname, optval, optlen); 862 return do_raw_getsockopt(sk, optname, optval, optlen); 863 } 864 865 static int raw_ioctl(struct sock *sk, int cmd, int *karg) 866 { 867 switch (cmd) { 868 case SIOCOUTQ: { 869 *karg = sk_wmem_alloc_get(sk); 870 return 0; 871 } 872 case SIOCINQ: { 873 struct sk_buff *skb; 874 875 spin_lock_bh(&sk->sk_receive_queue.lock); 876 skb = skb_peek(&sk->sk_receive_queue); 877 if (skb) 878 *karg = skb->len; 879 else 880 *karg = 0; 881 spin_unlock_bh(&sk->sk_receive_queue.lock); 882 return 0; 883 } 884 885 default: 886 #ifdef CONFIG_IP_MROUTE 887 return ipmr_ioctl(sk, cmd, karg); 888 #else 889 return -ENOIOCTLCMD; 890 #endif 891 } 892 } 893 894 #ifdef CONFIG_COMPAT 895 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) 896 { 897 switch (cmd) { 898 case SIOCOUTQ: 899 case SIOCINQ: 900 return -ENOIOCTLCMD; 901 default: 902 #ifdef CONFIG_IP_MROUTE 903 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg)); 904 #else 905 return -ENOIOCTLCMD; 906 #endif 907 } 908 } 909 #endif 910 911 int raw_abort(struct sock *sk, int err) 912 { 913 lock_sock(sk); 914 915 sk->sk_err = err; 916 sk_error_report(sk); 917 __udp_disconnect(sk, 0); 918 919 release_sock(sk); 920 921 return 0; 922 } 923 EXPORT_SYMBOL_GPL(raw_abort); 924 925 struct proto raw_prot = { 926 .name = "RAW", 927 .owner = THIS_MODULE, 928 .close = raw_close, 929 .destroy = raw_destroy, 930 .connect = ip4_datagram_connect, 931 .disconnect = __udp_disconnect, 932 .ioctl = raw_ioctl, 933 .init = raw_sk_init, 934 .setsockopt = raw_setsockopt, 935 .getsockopt = raw_getsockopt, 936 .sendmsg = raw_sendmsg, 937 .recvmsg = raw_recvmsg, 938 .bind = raw_bind, 939 .backlog_rcv = raw_rcv_skb, 940 .release_cb = ip4_datagram_release_cb, 941 .hash = raw_hash_sk, 942 .unhash = raw_unhash_sk, 943 .obj_size = sizeof(struct raw_sock), 944 .useroffset = offsetof(struct raw_sock, filter), 945 .usersize = sizeof_field(struct raw_sock, filter), 946 .h.raw_hash = &raw_v4_hashinfo, 947 #ifdef CONFIG_COMPAT 948 .compat_ioctl = compat_raw_ioctl, 949 #endif 950 .diag_destroy = raw_abort, 951 }; 952 953 #ifdef CONFIG_PROC_FS 954 static struct sock *raw_get_first(struct seq_file *seq, int bucket) 955 { 956 struct raw_hashinfo *h = pde_data(file_inode(seq->file)); 957 struct raw_iter_state *state = raw_seq_private(seq); 958 struct hlist_head *hlist; 959 struct sock *sk; 960 961 for (state->bucket = bucket; state->bucket < RAW_HTABLE_SIZE; 962 ++state->bucket) { 963 hlist = &h->ht[state->bucket]; 964 sk_for_each(sk, hlist) { 965 if (sock_net(sk) == seq_file_net(seq)) 966 return sk; 967 } 968 } 969 return NULL; 970 } 971 972 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) 973 { 974 struct raw_iter_state *state = raw_seq_private(seq); 975 976 do { 977 sk = sk_next(sk); 978 } while (sk && sock_net(sk) != seq_file_net(seq)); 979 980 if (!sk) 981 return raw_get_first(seq, state->bucket + 1); 982 return sk; 983 } 984 985 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) 986 { 987 struct sock *sk = raw_get_first(seq, 0); 988 989 if (sk) 990 while (pos && (sk = raw_get_next(seq, sk)) != NULL) 991 --pos; 992 return pos ? NULL : sk; 993 } 994 995 void *raw_seq_start(struct seq_file *seq, loff_t *pos) 996 __acquires(&h->lock) 997 { 998 struct raw_hashinfo *h = pde_data(file_inode(seq->file)); 999 1000 spin_lock(&h->lock); 1001 1002 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 1003 } 1004 EXPORT_SYMBOL_GPL(raw_seq_start); 1005 1006 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1007 { 1008 struct sock *sk; 1009 1010 if (v == SEQ_START_TOKEN) 1011 sk = raw_get_first(seq, 0); 1012 else 1013 sk = raw_get_next(seq, v); 1014 ++*pos; 1015 return sk; 1016 } 1017 EXPORT_SYMBOL_GPL(raw_seq_next); 1018 1019 void raw_seq_stop(struct seq_file *seq, void *v) 1020 __releases(&h->lock) 1021 { 1022 struct raw_hashinfo *h = pde_data(file_inode(seq->file)); 1023 1024 spin_unlock(&h->lock); 1025 } 1026 EXPORT_SYMBOL_GPL(raw_seq_stop); 1027 1028 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i) 1029 { 1030 struct inet_sock *inet = inet_sk(sp); 1031 __be32 dest = inet->inet_daddr, 1032 src = inet->inet_rcv_saddr; 1033 __u16 destp = 0, 1034 srcp = inet->inet_num; 1035 1036 seq_printf(seq, "%4d: %08X:%04X %08X:%04X" 1037 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n", 1038 i, src, srcp, dest, destp, sp->sk_state, 1039 sk_wmem_alloc_get(sp), 1040 sk_rmem_alloc_get(sp), 1041 0, 0L, 0, 1042 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1043 0, sock_i_ino(sp), 1044 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); 1045 } 1046 1047 static int raw_seq_show(struct seq_file *seq, void *v) 1048 { 1049 if (v == SEQ_START_TOKEN) 1050 seq_printf(seq, " sl local_address rem_address st tx_queue " 1051 "rx_queue tr tm->when retrnsmt uid timeout " 1052 "inode ref pointer drops\n"); 1053 else 1054 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket); 1055 return 0; 1056 } 1057 1058 static const struct seq_operations raw_seq_ops = { 1059 .start = raw_seq_start, 1060 .next = raw_seq_next, 1061 .stop = raw_seq_stop, 1062 .show = raw_seq_show, 1063 }; 1064 1065 static __net_init int raw_init_net(struct net *net) 1066 { 1067 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops, 1068 sizeof(struct raw_iter_state), &raw_v4_hashinfo)) 1069 return -ENOMEM; 1070 1071 return 0; 1072 } 1073 1074 static __net_exit void raw_exit_net(struct net *net) 1075 { 1076 remove_proc_entry("raw", net->proc_net); 1077 } 1078 1079 static __net_initdata struct pernet_operations raw_net_ops = { 1080 .init = raw_init_net, 1081 .exit = raw_exit_net, 1082 }; 1083 1084 int __init raw_proc_init(void) 1085 { 1086 1087 return register_pernet_subsys(&raw_net_ops); 1088 } 1089 1090 void __init raw_proc_exit(void) 1091 { 1092 unregister_pernet_subsys(&raw_net_ops); 1093 } 1094 #endif /* CONFIG_PROC_FS */ 1095 1096 static void raw_sysctl_init_net(struct net *net) 1097 { 1098 #ifdef CONFIG_NET_L3_MASTER_DEV 1099 net->ipv4.sysctl_raw_l3mdev_accept = 1; 1100 #endif 1101 } 1102 1103 static int __net_init raw_sysctl_init(struct net *net) 1104 { 1105 raw_sysctl_init_net(net); 1106 return 0; 1107 } 1108 1109 static struct pernet_operations __net_initdata raw_sysctl_ops = { 1110 .init = raw_sysctl_init, 1111 }; 1112 1113 void __init raw_init(void) 1114 { 1115 raw_sysctl_init_net(&init_net); 1116 if (register_pernet_subsys(&raw_sysctl_ops)) 1117 panic("RAW: failed to init sysctl parameters.\n"); 1118 } 1119