net/ipv4/raw.c

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