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  * Version:     $Id: raw.c,v 1.64 2002/02/01 22:01:04 davem Exp $
   9  *
  10  * Authors:     Ross Biro, <bir7@leland.Stanford.Edu>
  11  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  12  *
  13  * Fixes:
  14  *              Alan Cox        :       verify_area() fixed up
  15  *              Alan Cox        :       ICMP error handling
  16  *              Alan Cox        :       EMSGSIZE if you send too big a packet
  17  *              Alan Cox        :       Now uses generic datagrams and shared
  18  *                                      skbuff library. No more peek crashes,
  19  *                                      no more backlogs
  20  *              Alan Cox        :       Checks sk->broadcast.
  21  *              Alan Cox        :       Uses skb_free_datagram/skb_copy_datagram
  22  *              Alan Cox        :       Raw passes ip options too
  23  *              Alan Cox        :       Setsocketopt added
  24  *              Alan Cox        :       Fixed error return for broadcasts
  25  *              Alan Cox        :       Removed wake_up calls
  26  *              Alan Cox        :       Use ttl/tos
  27  *              Alan Cox        :       Cleaned up old debugging
  28  *              Alan Cox        :       Use new kernel side addresses
  29  *      Arnt Gulbrandsen        :       Fixed MSG_DONTROUTE in raw sockets.
  30  *              Alan Cox        :       BSD style RAW socket demultiplexing.
  31  *              Alan Cox        :       Beginnings of mrouted support.
  32  *              Alan Cox        :       Added IP_HDRINCL option.
  33  *              Alan Cox        :       Skip broadcast check if BSDism set.
  34  *              David S. Miller :       New socket lookup architecture.
  35  *
  36  *              This program is free software; you can redistribute it and/or
  37  *              modify it under the terms of the GNU General Public License
  38  *              as published by the Free Software Foundation; either version
  39  *              2 of the License, or (at your option) any later version.
  40  */
  41
  42 #include <linux/config.h>
  43 #include <asm/atomic.h>
  44 #include <asm/byteorder.h>
  45 #include <asm/current.h>
  46 #include <asm/uaccess.h>
  47 #include <asm/ioctls.h>
  48 #include <linux/types.h>
  49 #include <linux/stddef.h>
  50 #include <linux/slab.h>
  51 #include <linux/errno.h>
  52 #include <linux/aio.h>
  53 #include <linux/kernel.h>
  54 #include <linux/spinlock.h>
  55 #include <linux/sockios.h>
  56 #include <linux/socket.h>
  57 #include <linux/in.h>
  58 #include <linux/mroute.h>
  59 #include <linux/netdevice.h>
  60 #include <linux/in_route.h>
  61 #include <linux/route.h>
  62 #include <linux/tcp.h>
  63 #include <linux/skbuff.h>
  64 #include <net/dst.h>
  65 #include <net/sock.h>
  66 #include <linux/gfp.h>
  67 #include <linux/ip.h>
  68 #include <linux/net.h>
  69 #include <net/ip.h>
  70 #include <net/icmp.h>
  71 #include <net/udp.h>
  72 #include <net/raw.h>
  73 #include <net/snmp.h>
  74 #include <net/inet_common.h>
  75 #include <net/checksum.h>
  76 #include <net/xfrm.h>
  77 #include <linux/rtnetlink.h>
  78 #include <linux/proc_fs.h>
  79 #include <linux/seq_file.h>
  80 #include <linux/netfilter.h>
  81 #include <linux/netfilter_ipv4.h>
  82
  83 struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE];
  84 rwlock_t raw_v4_lock = RW_LOCK_UNLOCKED;
  85
  86 static void raw_v4_hash(struct sock *sk)
  87 {
  88         struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num &
  89                                                  (RAWV4_HTABLE_SIZE - 1)];
  90
  91         write_lock_bh(&raw_v4_lock);
  92         sk_add_node(sk, head);
  93         sock_prot_inc_use(sk->sk_prot);
  94         write_unlock_bh(&raw_v4_lock);
  95 }
  96
  97 static void raw_v4_unhash(struct sock *sk)
  98 {
  99         write_lock_bh(&raw_v4_lock);
 100         if (sk_del_node_init(sk))
 101                 sock_prot_dec_use(sk->sk_prot);
 102         write_unlock_bh(&raw_v4_lock);
 103 }
 104
 105 struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
 106                              unsigned long raddr, unsigned long laddr,
 107                              int dif)
 108 {
 109         struct hlist_node *node;
 110
 111         sk_for_each_from(sk, node) {
 112                 struct inet_opt *inet = inet_sk(sk);
 113
 114                 if (inet->num == num                                    &&
 115                     !(inet->daddr && inet->daddr != raddr)              &&
 116                     !(inet->rcv_saddr && inet->rcv_saddr != laddr)      &&
 117                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 118                         goto found; /* gotcha */
 119         }
 120         sk = NULL;
 121 found:
 122         return sk;
 123 }
 124
 125 /*
 126  *      0 - deliver
 127  *      1 - block
 128  */
 129 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 130 {
 131         int type;
 132
 133         type = skb->h.icmph->type;
 134         if (type < 32) {
 135                 __u32 data = raw4_sk(sk)->filter.data;
 136
 137                 return ((1 << type) & data) != 0;
 138         }
 139
 140         /* Do not block unknown ICMP types */
 141         return 0;
 142 }
 143
 144 /* IP input processing comes here for RAW socket delivery.
 145  * Caller owns SKB, so we must make clones.
 146  *
 147  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 148  * -> It does. And not only TOS, but all IP header.
 149  */
 150 void raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 151 {
 152         struct sock *sk;
 153         struct hlist_head *head;
 154
 155         read_lock(&raw_v4_lock);
 156         head = &raw_v4_htable[hash];
 157         if (hlist_empty(head))
 158                 goto out;
 159         sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
 160                              iph->saddr, iph->daddr,
 161                              skb->dev->ifindex);
 162
 163         while (sk) {
 164                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 165                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 166
 167                         /* Not releasing hash table! */
 168                         if (clone)
 169                                 raw_rcv(sk, clone);
 170                 }
 171                 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
 172                                      iph->saddr, iph->daddr,
 173                                      skb->dev->ifindex);
 174         }
 175 out:
 176         read_unlock(&raw_v4_lock);
 177 }
 178
 179 void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
 180 {
 181         struct inet_opt *inet = inet_sk(sk);
 182         int type = skb->h.icmph->type;
 183         int code = skb->h.icmph->code;
 184         int err = 0;
 185         int harderr = 0;
 186
 187         /* Report error on raw socket, if:
 188            1. User requested ip_recverr.
 189            2. Socket is connected (otherwise the error indication
 190               is useless without ip_recverr and error is hard.
 191          */
 192         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 193                 return;
 194
 195         switch (type) {
 196         default:
 197         case ICMP_TIME_EXCEEDED:
 198                 err = EHOSTUNREACH;
 199                 break;
 200         case ICMP_SOURCE_QUENCH:
 201                 return;
 202         case ICMP_PARAMETERPROB:
 203                 err = EPROTO;
 204                 harderr = 1;
 205                 break;
 206         case ICMP_DEST_UNREACH:
 207                 err = EHOSTUNREACH;
 208                 if (code > NR_ICMP_UNREACH)
 209                         break;
 210                 err = icmp_err_convert[code].errno;
 211                 harderr = icmp_err_convert[code].fatal;
 212                 if (code == ICMP_FRAG_NEEDED) {
 213                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 214                         err = EMSGSIZE;
 215                 }
 216         }
 217
 218         if (inet->recverr) {
 219                 struct iphdr *iph = (struct iphdr*)skb->data;
 220                 u8 *payload = skb->data + (iph->ihl << 2);
 221
 222                 if (inet->hdrincl)
 223                         payload = skb->data;
 224                 ip_icmp_error(sk, skb, err, 0, info, payload);
 225         }
 226
 227         if (inet->recverr || harderr) {
 228                 sk->sk_err = err;
 229                 sk->sk_error_report(sk);
 230         }
 231 }
 232
 233 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 234 {
 235         /* Charge it to the socket. */
 236
 237         if (sock_queue_rcv_skb(sk, skb) < 0) {
 238                 /* FIXME: increment a raw drops counter here */
 239                 kfree_skb(skb);
 240                 return NET_RX_DROP;
 241         }
 242
 243         return NET_RX_SUCCESS;
 244 }
 245
 246 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 247 {
 248         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 249                 kfree_skb(skb);
 250                 return NET_RX_DROP;
 251         }
 252
 253         skb_push(skb, skb->data - skb->nh.raw);
 254
 255         raw_rcv_skb(sk, skb);
 256         return 0;
 257 }
 258
 259 static int raw_send_hdrinc(struct sock *sk, void *from, int length,
 260                         struct rtable *rt,
 261                         unsigned int flags)
 262 {
 263         struct inet_opt *inet = inet_sk(sk);
 264         int hh_len;
 265         struct iphdr *iph;
 266         struct sk_buff *skb;
 267         int err;
 268
 269         if (length > rt->u.dst.dev->mtu) {
 270                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
 271                                rt->u.dst.dev->mtu);
 272                 return -EMSGSIZE;
 273         }
 274         if (flags&MSG_PROBE)
 275                 goto out;
 276
 277         hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
 278
 279         skb = sock_alloc_send_skb(sk, length+hh_len+15,
 280                                   flags&MSG_DONTWAIT, &err);
 281         if (skb == NULL)
 282                 goto error;
 283         skb_reserve(skb, hh_len);
 284
 285         skb->priority = sk->sk_priority;
 286         skb->dst = dst_clone(&rt->u.dst);
 287
 288         skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
 289
 290         skb->ip_summed = CHECKSUM_NONE;
 291
 292         skb->h.raw = skb->nh.raw;
 293         err = memcpy_fromiovecend((void *)iph, from, 0, length);
 294         if (err)
 295                 goto error_fault;
 296
 297         /* We don't modify invalid header */
 298         if (length >= sizeof(*iph) && iph->ihl * 4 <= length) {
 299                 if (!iph->saddr)
 300                         iph->saddr = rt->rt_src;
 301                 iph->check   = 0;
 302                 iph->tot_len = htons(length);
 303                 if (!iph->id)
 304                         ip_select_ident(iph, &rt->u.dst, NULL);
 305
 306                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 307         }
 308
 309         err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 310                       dst_output);
 311         if (err > 0)
 312                 err = inet->recverr ? net_xmit_errno(err) : 0;
 313         if (err)
 314                 goto error;
 315 out:
 316         return 0;
 317
 318 error_fault:
 319         err = -EFAULT;
 320         kfree_skb(skb);
 321 error:
 322         IP_INC_STATS(IpOutDiscards);
 323         return err;
 324 }
 325
 326 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 327                        int len)
 328 {
 329         struct inet_opt *inet = inet_sk(sk);
 330         struct ipcm_cookie ipc;
 331         struct rtable *rt = NULL;
 332         int free = 0;
 333         u32 daddr;
 334         u32 saddr;
 335         u8  tos;
 336         int err;
 337
 338         /* This check is ONLY to check for arithmetic overflow
 339            on integer(!) len. Not more! Real check will be made
 340            in ip_build_xmit --ANK
 341
 342            BTW socket.c -> af_*.c -> ... make multiple
 343            invalid conversions size_t -> int. We MUST repair it f.e.
 344            by replacing all of them with size_t and revise all
 345            the places sort of len += sizeof(struct iphdr)
 346            If len was ULONG_MAX-10 it would be cathastrophe  --ANK
 347          */
 348
 349         err = -EMSGSIZE;
 350         if (len < 0 || len > 0xFFFF)
 351                 goto out;
 352
 353         /*
 354          *      Check the flags.
 355          */
 356
 357         err = -EOPNOTSUPP;
 358         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 359                 goto out;               /* compatibility */
 360
 361         /*
 362          *      Get and verify the address.
 363          */
 364
 365         if (msg->msg_namelen) {
 366                 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
 367                 err = -EINVAL;
 368                 if (msg->msg_namelen < sizeof(*usin))
 369                         goto out;
 370                 if (usin->sin_family != AF_INET) {
 371                         static int complained;
 372                         if (!complained++)
 373                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 374                                                  "raw sendmsg. Fix it!\n",
 375                                                  current->comm);
 376                         err = -EINVAL;
 377                         if (usin->sin_family)
 378                                 goto out;
 379                 }
 380                 daddr = usin->sin_addr.s_addr;
 381                 /* ANK: I did not forget to get protocol from port field.
 382                  * I just do not know, who uses this weirdness.
 383                  * IP_HDRINCL is much more convenient.
 384                  */
 385         } else {
 386                 err = -EDESTADDRREQ;
 387                 if (sk->sk_state != TCP_ESTABLISHED)
 388                         goto out;
 389                 daddr = inet->daddr;
 390         }
 391
 392         ipc.addr = inet->saddr;
 393         ipc.opt = NULL;
 394         ipc.oif = sk->sk_bound_dev_if;
 395
 396         if (msg->msg_controllen) {
 397                 err = ip_cmsg_send(msg, &ipc);
 398                 if (err)
 399                         goto out;
 400                 if (ipc.opt)
 401                         free = 1;
 402         }
 403
 404         saddr = ipc.addr;
 405         ipc.addr = daddr;
 406
 407         if (!ipc.opt)
 408                 ipc.opt = inet->opt;
 409
 410         if (ipc.opt) {
 411                 err = -EINVAL;
 412                 /* Linux does not mangle headers on raw sockets,
 413                  * so that IP options + IP_HDRINCL is non-sense.
 414                  */
 415                 if (inet->hdrincl)
 416                         goto done;
 417                 if (ipc.opt->srr) {
 418                         if (!daddr)
 419                                 goto done;
 420                         daddr = ipc.opt->faddr;
 421                 }
 422         }
 423         tos = RT_TOS(inet->tos) | sk->sk_localroute;
 424         if (msg->msg_flags & MSG_DONTROUTE)
 425                 tos |= RTO_ONLINK;
 426
 427         if (MULTICAST(daddr)) {
 428                 if (!ipc.oif)
 429                         ipc.oif = inet->mc_index;
 430                 if (!saddr)
 431                         saddr = inet->mc_addr;
 432         }
 433
 434         {
 435                 struct flowi fl = { .oif = ipc.oif,
 436                                     .nl_u = { .ip4_u =
 437                                               { .daddr = daddr,
 438                                                 .saddr = saddr,
 439                                                 .tos = tos } },
 440                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 441                                                              sk->sk_protocol,
 442                                   };
 443                 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
 444         }
 445         if (err)
 446                 goto done;
 447
 448         err = -EACCES;
 449         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 450                 goto done;
 451
 452         if (msg->msg_flags & MSG_CONFIRM)
 453                 goto do_confirm;
 454 back_from_confirm:
 455
 456         if (inet->hdrincl)
 457                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 458                                         rt, msg->msg_flags);
 459
 460          else {
 461                 if (!ipc.addr)
 462                         ipc.addr = rt->rt_dst;
 463                 lock_sock(sk);
 464                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 465                                         &ipc, rt, msg->msg_flags);
 466                 if (err)
 467                         ip_flush_pending_frames(sk);
 468                 else if (!(msg->msg_flags & MSG_MORE))
 469                         err = ip_push_pending_frames(sk);
 470                 release_sock(sk);
 471         }
 472 done:
 473         if (free)
 474                 kfree(ipc.opt);
 475         ip_rt_put(rt);
 476
 477 out:    return err < 0 ? err : len;
 478
 479 do_confirm:
 480         dst_confirm(&rt->u.dst);
 481         if (!(msg->msg_flags & MSG_PROBE) || len)
 482                 goto back_from_confirm;
 483         err = 0;
 484         goto done;
 485 }
 486
 487 static void raw_close(struct sock *sk, long timeout)
 488 {
 489         /*
 490          * Raw sockets may have direct kernel refereneces. Kill them.
 491          */
 492         ip_ra_control(sk, 0, NULL);
 493
 494         inet_sock_release(sk);
 495 }
 496
 497 /* This gets rid of all the nasties in af_inet. -DaveM */
 498 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 499 {
 500         struct inet_opt *inet = inet_sk(sk);
 501         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 502         int ret = -EINVAL;
 503         int chk_addr_ret;
 504
 505         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 506                 goto out;
 507         chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
 508         ret = -EADDRNOTAVAIL;
 509         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 510             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 511                 goto out;
 512         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 513         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 514                 inet->saddr = 0;  /* Use device */
 515         sk_dst_reset(sk);
 516         ret = 0;
 517 out:    return ret;
 518 }
 519
 520 /*
 521  *      This should be easy, if there is something there
 522  *      we return it, otherwise we block.
 523  */
 524
 525 int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 526                 int len, int noblock, int flags, int *addr_len)
 527 {
 528         struct inet_opt *inet = inet_sk(sk);
 529         int copied = 0;
 530         int err = -EOPNOTSUPP;
 531         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 532         struct sk_buff *skb;
 533
 534         if (flags & MSG_OOB)
 535                 goto out;
 536
 537         if (addr_len)
 538                 *addr_len = sizeof(*sin);
 539
 540         if (flags & MSG_ERRQUEUE) {
 541                 err = ip_recv_error(sk, msg, len);
 542                 goto out;
 543         }
 544
 545         skb = skb_recv_datagram(sk, flags, noblock, &err);
 546         if (!skb)
 547                 goto out;
 548
 549         copied = skb->len;
 550         if (len < copied) {
 551                 msg->msg_flags |= MSG_TRUNC;
 552                 copied = len;
 553         }
 554
 555         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 556         if (err)
 557                 goto done;
 558
 559         sock_recv_timestamp(msg, sk, skb);
 560
 561         /* Copy the address. */
 562         if (sin) {
 563                 sin->sin_family = AF_INET;
 564                 sin->sin_addr.s_addr = skb->nh.iph->saddr;
 565                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 566         }
 567         if (inet->cmsg_flags)
 568                 ip_cmsg_recv(msg, skb);
 569 done:
 570         skb_free_datagram(sk, skb);
 571 out:    return err ? : copied;
 572 }
 573
 574 static int raw_init(struct sock *sk)
 575 {
 576         struct raw_opt *tp = raw4_sk(sk);
 577         if (inet_sk(sk)->num == IPPROTO_ICMP)
 578                 memset(&tp->filter, 0, sizeof(tp->filter));
 579         return 0;
 580 }
 581
 582 static int raw_seticmpfilter(struct sock *sk, char *optval, int optlen)
 583 {
 584         if (optlen > sizeof(struct icmp_filter))
 585                 optlen = sizeof(struct icmp_filter);
 586         if (copy_from_user(&raw4_sk(sk)->filter, optval, optlen))
 587                 return -EFAULT;
 588         return 0;
 589 }
 590
 591 static int raw_geticmpfilter(struct sock *sk, char *optval, int *optlen)
 592 {
 593         int len, ret = -EFAULT;
 594
 595         if (get_user(len, optlen))
 596                 goto out;
 597         ret = -EINVAL;
 598         if (len < 0)
 599                 goto out;
 600         if (len > sizeof(struct icmp_filter))
 601                 len = sizeof(struct icmp_filter);
 602         ret = -EFAULT;
 603         if (put_user(len, optlen) ||
 604             copy_to_user(optval, &raw4_sk(sk)->filter, len))
 605                 goto out;
 606         ret = 0;
 607 out:    return ret;
 608 }
 609
 610 static int raw_setsockopt(struct sock *sk, int level, int optname,
 611                           char *optval, int optlen)
 612 {
 613         if (level != SOL_RAW)
 614                 return ip_setsockopt(sk, level, optname, optval, optlen);
 615
 616         if (optname == ICMP_FILTER) {
 617                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 618                         return -EOPNOTSUPP;
 619                 else
 620                         return raw_seticmpfilter(sk, optval, optlen);
 621         }
 622         return -ENOPROTOOPT;
 623 }
 624
 625 static int raw_getsockopt(struct sock *sk, int level, int optname,
 626                           char *optval, int *optlen)
 627 {
 628         if (level != SOL_RAW)
 629                 return ip_getsockopt(sk, level, optname, optval, optlen);
 630
 631         if (optname == ICMP_FILTER) {
 632                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 633                         return -EOPNOTSUPP;
 634                 else
 635                         return raw_geticmpfilter(sk, optval, optlen);
 636         }
 637         return -ENOPROTOOPT;
 638 }
 639
 640 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 641 {
 642         switch (cmd) {
 643                 case SIOCOUTQ: {
 644                         int amount = atomic_read(&sk->sk_wmem_alloc);
 645                         return put_user(amount, (int *)arg);
 646                 }
 647                 case SIOCINQ: {
 648                         struct sk_buff *skb;
 649                         int amount = 0;
 650
 651                         spin_lock_irq(&sk->sk_receive_queue.lock);
 652                         skb = skb_peek(&sk->sk_receive_queue);
 653                         if (skb != NULL)
 654                                 amount = skb->len;
 655                         spin_unlock_irq(&sk->sk_receive_queue.lock);
 656                         return put_user(amount, (int *)arg);
 657                 }
 658
 659                 default:
 660 #ifdef CONFIG_IP_MROUTE
 661                         return ipmr_ioctl(sk, cmd, arg);
 662 #else
 663                         return -ENOIOCTLCMD;
 664 #endif
 665         }
 666 }
 667
 668 struct proto raw_prot = {
 669         .name =         "RAW",
 670         .close =        raw_close,
 671         .connect =      udp_connect,
 672         .disconnect =   udp_disconnect,
 673         .ioctl =        raw_ioctl,
 674         .init =         raw_init,
 675         .setsockopt =   raw_setsockopt,
 676         .getsockopt =   raw_getsockopt,
 677         .sendmsg =      raw_sendmsg,
 678         .recvmsg =      raw_recvmsg,
 679         .bind =         raw_bind,
 680         .backlog_rcv =  raw_rcv_skb,
 681         .hash =         raw_v4_hash,
 682         .unhash =       raw_v4_unhash,
 683 };
 684
 685 #ifdef CONFIG_PROC_FS
 686 struct raw_iter_state {
 687         int bucket;
 688 };
 689
 690 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
 691
 692 static struct sock *raw_get_first(struct seq_file *seq)
 693 {
 694         struct sock *sk;
 695         struct raw_iter_state* state = raw_seq_private(seq);
 696
 697         for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
 698                 struct hlist_node *node;
 699
 700                 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
 701                         if (sk->sk_family == PF_INET)
 702                                 goto found;
 703         }
 704         sk = NULL;
 705 found:
 706         return sk;
 707 }
 708
 709 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 710 {
 711         struct raw_iter_state* state = raw_seq_private(seq);
 712
 713         do {
 714                 sk = sk_next(sk);
 715 try_again:
 716                 ;
 717         } while (sk && sk->sk_family != PF_INET);
 718
 719         if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
 720                 sk = sk_head(&raw_v4_htable[state->bucket]);
 721                 goto try_again;
 722         }
 723         return sk;
 724 }
 725
 726 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 727 {
 728         struct sock *sk = raw_get_first(seq);
 729
 730         if (sk)
 731                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 732                         --pos;
 733         return pos ? NULL : sk;
 734 }
 735
 736 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 737 {
 738         read_lock(&raw_v4_lock);
 739         return *pos ? raw_get_idx(seq, *pos) : (void *)1;
 740 }
 741
 742 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 743 {
 744         struct sock *sk;
 745
 746         if (v == (void *)1)
 747                 sk = raw_get_first(seq);
 748         else
 749                 sk = raw_get_next(seq, v);
 750         ++*pos;
 751         return sk;
 752 }
 753
 754 static void raw_seq_stop(struct seq_file *seq, void *v)
 755 {
 756         read_unlock(&raw_v4_lock);
 757 }
 758
 759 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
 760 {
 761         struct inet_opt *inet = inet_sk(sp);
 762         unsigned int dest = inet->daddr,
 763                      src = inet->rcv_saddr;
 764         __u16 destp = 0,
 765               srcp  = inet->num;
 766
 767         sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
 768                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
 769                 i, src, srcp, dest, destp, sp->sk_state,
 770                 atomic_read(&sp->sk_wmem_alloc),
 771                 atomic_read(&sp->sk_rmem_alloc),
 772                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 773                 atomic_read(&sp->sk_refcnt), sp);
 774         return tmpbuf;
 775 }
 776
 777 static int raw_seq_show(struct seq_file *seq, void *v)
 778 {
 779         char tmpbuf[129];
 780
 781         if (v == (void *)1)
 782                 seq_printf(seq, "%-127s\n",
 783                                "  sl  local_address rem_address   st tx_queue "
 784                                "rx_queue tr tm->when retrnsmt   uid  timeout "
 785                                "inode");
 786         else {
 787                 struct raw_iter_state *state = raw_seq_private(seq);
 788
 789                 seq_printf(seq, "%-127s\n",
 790                            get_raw_sock(v, tmpbuf, state->bucket));
 791         }
 792         return 0;
 793 }
 794
 795 static struct seq_operations raw_seq_ops = {
 796         .start = raw_seq_start,
 797         .next  = raw_seq_next,
 798         .stop  = raw_seq_stop,
 799         .show  = raw_seq_show,
 800 };
 801
 802 static int raw_seq_open(struct inode *inode, struct file *file)
 803 {
 804         struct seq_file *seq;
 805         int rc = -ENOMEM;
 806         struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
 807
 808         if (!s)
 809                 goto out;
 810         rc = seq_open(file, &raw_seq_ops);
 811         if (rc)
 812                 goto out_kfree;
 813
 814         seq = file->private_data;
 815         seq->private = s;
 816         memset(s, 0, sizeof(*s));
 817 out:
 818         return rc;
 819 out_kfree:
 820         kfree(s);
 821         goto out;
 822 }
 823
 824 static struct file_operations raw_seq_fops = {
 825         .owner   = THIS_MODULE,
 826         .open    = raw_seq_open,
 827         .read    = seq_read,
 828         .llseek  = seq_lseek,
 829         .release = seq_release_private,
 830 };
 831
 832 int __init raw_proc_init(void)
 833 {
 834         struct proc_dir_entry *p;
 835         int rc = 0;
 836
 837         p = create_proc_entry("raw", S_IRUGO, proc_net);
 838         if (p)
 839                 p->proc_fops = &raw_seq_fops;
 840         else
 841                 rc = -ENOMEM;
 842         return rc;
 843 }
 844
 845 void __init raw_proc_exit(void)
 846 {
 847         remove_proc_entry("raw", proc_net);
 848 }
 849 #endif /* CONFIG_PROC_FS */