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
  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/types.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/stddef.h>
  49 #include <linux/slab.h>
  50 #include <linux/errno.h>
  51 #include <linux/aio.h>
  52 #include <linux/kernel.h>
  53 #include <linux/spinlock.h>
  54 #include <linux/sockios.h>
  55 #include <linux/socket.h>
  56 #include <linux/in.h>
  57 #include <linux/mroute.h>
  58 #include <linux/netdevice.h>
  59 #include <linux/in_route.h>
  60 #include <linux/route.h>
  61 #include <linux/skbuff.h>
  62 #include <net/dst.h>
  63 #include <net/sock.h>
  64 #include <linux/gfp.h>
  65 #include <linux/ip.h>
  66 #include <linux/net.h>
  67 #include <net/ip.h>
  68 #include <net/icmp.h>
  69 #include <net/udp.h>
  70 #include <net/raw.h>
  71 #include <net/snmp.h>
  72 #include <net/tcp_states.h>
  73 #include <net/inet_common.h>
  74 #include <net/checksum.h>
  75 #include <net/xfrm.h>
  76 #include <linux/rtnetlink.h>
  77 #include <linux/proc_fs.h>
  78 #include <linux/seq_file.h>
  79 #include <linux/netfilter.h>
  80 #include <linux/netfilter_ipv4.h>
  81
  82 struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE];
  83 DEFINE_RWLOCK(raw_v4_lock);
  84
  85 static void raw_v4_hash(struct sock *sk)
  86 {
  87         struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num &
  88                                                  (RAWV4_HTABLE_SIZE - 1)];
  89
  90         write_lock_bh(&raw_v4_lock);
  91         sk_add_node(sk, head);
  92         sock_prot_inc_use(sk->sk_prot);
  93         write_unlock_bh(&raw_v4_lock);
  94 }
  95
  96 static void raw_v4_unhash(struct sock *sk)
  97 {
  98         write_lock_bh(&raw_v4_lock);
  99         if (sk_del_node_init(sk))
 100                 sock_prot_dec_use(sk->sk_prot);
 101         write_unlock_bh(&raw_v4_lock);
 102 }
 103
 104 struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
 105                              __be32 raddr, __be32 laddr,
 106                              int dif)
 107 {
 108         struct hlist_node *node;
 109
 110         sk_for_each_from(sk, node) {
 111                 struct inet_sock *inet = inet_sk(sk);
 112
 113                 if (inet->num == num                                    &&
 114                     !(inet->daddr && inet->daddr != raddr)              &&
 115                     !(inet->rcv_saddr && inet->rcv_saddr != laddr)      &&
 116                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 117                         goto found; /* gotcha */
 118         }
 119         sk = NULL;
 120 found:
 121         return sk;
 122 }
 123
 124 /*
 125  *      0 - deliver
 126  *      1 - block
 127  */
 128 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 129 {
 130         int type;
 131
 132         if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
 133                 return 1;
 134
 135         type = icmp_hdr(skb)->type;
 136         if (type < 32) {
 137                 __u32 data = raw_sk(sk)->filter.data;
 138
 139                 return ((1 << type) & data) != 0;
 140         }
 141
 142         /* Do not block unknown ICMP types */
 143         return 0;
 144 }
 145
 146 /* IP input processing comes here for RAW socket delivery.
 147  * Caller owns SKB, so we must make clones.
 148  *
 149  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 150  * -> It does. And not only TOS, but all IP header.
 151  */
 152 int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 153 {
 154         struct sock *sk;
 155         struct hlist_head *head;
 156         int delivered = 0;
 157
 158         read_lock(&raw_v4_lock);
 159         head = &raw_v4_htable[hash];
 160         if (hlist_empty(head))
 161                 goto out;
 162         sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
 163                              iph->saddr, iph->daddr,
 164                              skb->dev->ifindex);
 165
 166         while (sk) {
 167                 delivered = 1;
 168                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 169                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 170
 171                         /* Not releasing hash table! */
 172                         if (clone)
 173                                 raw_rcv(sk, clone);
 174                 }
 175                 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
 176                                      iph->saddr, iph->daddr,
 177                                      skb->dev->ifindex);
 178         }
 179 out:
 180         read_unlock(&raw_v4_lock);
 181         return delivered;
 182 }
 183
 184 void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
 185 {
 186         struct inet_sock *inet = inet_sk(sk);
 187         const int type = icmp_hdr(skb)->type;
 188         const int code = icmp_hdr(skb)->code;
 189         int err = 0;
 190         int harderr = 0;
 191
 192         /* Report error on raw socket, if:
 193            1. User requested ip_recverr.
 194            2. Socket is connected (otherwise the error indication
 195               is useless without ip_recverr and error is hard.
 196          */
 197         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 198                 return;
 199
 200         switch (type) {
 201         default:
 202         case ICMP_TIME_EXCEEDED:
 203                 err = EHOSTUNREACH;
 204                 break;
 205         case ICMP_SOURCE_QUENCH:
 206                 return;
 207         case ICMP_PARAMETERPROB:
 208                 err = EPROTO;
 209                 harderr = 1;
 210                 break;
 211         case ICMP_DEST_UNREACH:
 212                 err = EHOSTUNREACH;
 213                 if (code > NR_ICMP_UNREACH)
 214                         break;
 215                 err = icmp_err_convert[code].errno;
 216                 harderr = icmp_err_convert[code].fatal;
 217                 if (code == ICMP_FRAG_NEEDED) {
 218                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 219                         err = EMSGSIZE;
 220                 }
 221         }
 222
 223         if (inet->recverr) {
 224                 struct iphdr *iph = (struct iphdr*)skb->data;
 225                 u8 *payload = skb->data + (iph->ihl << 2);
 226
 227                 if (inet->hdrincl)
 228                         payload = skb->data;
 229                 ip_icmp_error(sk, skb, err, 0, info, payload);
 230         }
 231
 232         if (inet->recverr || harderr) {
 233                 sk->sk_err = err;
 234                 sk->sk_error_report(sk);
 235         }
 236 }
 237
 238 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 239 {
 240         /* Charge it to the socket. */
 241
 242         if (sock_queue_rcv_skb(sk, skb) < 0) {
 243                 /* FIXME: increment a raw drops counter here */
 244                 kfree_skb(skb);
 245                 return NET_RX_DROP;
 246         }
 247
 248         return NET_RX_SUCCESS;
 249 }
 250
 251 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 252 {
 253         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 254                 kfree_skb(skb);
 255                 return NET_RX_DROP;
 256         }
 257         nf_reset(skb);
 258
 259         skb_push(skb, skb->data - skb_network_header(skb));
 260
 261         raw_rcv_skb(sk, skb);
 262         return 0;
 263 }
 264
 265 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 266                         struct rtable *rt,
 267                         unsigned int flags)
 268 {
 269         struct inet_sock *inet = inet_sk(sk);
 270         int hh_len;
 271         struct iphdr *iph;
 272         struct sk_buff *skb;
 273         int err;
 274
 275         if (length > rt->u.dst.dev->mtu) {
 276                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
 277                                rt->u.dst.dev->mtu);
 278                 return -EMSGSIZE;
 279         }
 280         if (flags&MSG_PROBE)
 281                 goto out;
 282
 283         hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
 284
 285         skb = sock_alloc_send_skb(sk, length+hh_len+15,
 286                                   flags&MSG_DONTWAIT, &err);
 287         if (skb == NULL)
 288                 goto error;
 289         skb_reserve(skb, hh_len);
 290
 291         skb->priority = sk->sk_priority;
 292         skb->dst = dst_clone(&rt->u.dst);
 293
 294         skb_reset_network_header(skb);
 295         iph = ip_hdr(skb);
 296         skb_put(skb, length);
 297
 298         skb->ip_summed = CHECKSUM_NONE;
 299
 300         skb->transport_header = skb->network_header;
 301         err = memcpy_fromiovecend((void *)iph, from, 0, length);
 302         if (err)
 303                 goto error_fault;
 304
 305         /* We don't modify invalid header */
 306         if (length >= sizeof(*iph) && iph->ihl * 4U <= length) {
 307                 if (!iph->saddr)
 308                         iph->saddr = rt->rt_src;
 309                 iph->check   = 0;
 310                 iph->tot_len = htons(length);
 311                 if (!iph->id)
 312                         ip_select_ident(iph, &rt->u.dst, NULL);
 313
 314                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 315         }
 316
 317         err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 318                       dst_output);
 319         if (err > 0)
 320                 err = inet->recverr ? net_xmit_errno(err) : 0;
 321         if (err)
 322                 goto error;
 323 out:
 324         return 0;
 325
 326 error_fault:
 327         err = -EFAULT;
 328         kfree_skb(skb);
 329 error:
 330         IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
 331         return err;
 332 }
 333
 334 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 335 {
 336         struct iovec *iov;
 337         u8 __user *type = NULL;
 338         u8 __user *code = NULL;
 339         int probed = 0;
 340         unsigned int i;
 341
 342         if (!msg->msg_iov)
 343                 return 0;
 344
 345         for (i = 0; i < msg->msg_iovlen; i++) {
 346                 iov = &msg->msg_iov[i];
 347                 if (!iov)
 348                         continue;
 349
 350                 switch (fl->proto) {
 351                 case IPPROTO_ICMP:
 352                         /* check if one-byte field is readable or not. */
 353                         if (iov->iov_base && iov->iov_len < 1)
 354                                 break;
 355
 356                         if (!type) {
 357                                 type = iov->iov_base;
 358                                 /* check if code field is readable or not. */
 359                                 if (iov->iov_len > 1)
 360                                         code = type + 1;
 361                         } else if (!code)
 362                                 code = iov->iov_base;
 363
 364                         if (type && code) {
 365                                 if (get_user(fl->fl_icmp_type, type) ||
 366                                     get_user(fl->fl_icmp_code, code))
 367                                         return -EFAULT;
 368                                 probed = 1;
 369                         }
 370                         break;
 371                 default:
 372                         probed = 1;
 373                         break;
 374                 }
 375                 if (probed)
 376                         break;
 377         }
 378         return 0;
 379 }
 380
 381 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 382                        size_t len)
 383 {
 384         struct inet_sock *inet = inet_sk(sk);
 385         struct ipcm_cookie ipc;
 386         struct rtable *rt = NULL;
 387         int free = 0;
 388         __be32 daddr;
 389         __be32 saddr;
 390         u8  tos;
 391         int err;
 392
 393         err = -EMSGSIZE;
 394         if (len > 0xFFFF)
 395                 goto out;
 396
 397         /*
 398          *      Check the flags.
 399          */
 400
 401         err = -EOPNOTSUPP;
 402         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 403                 goto out;               /* compatibility */
 404
 405         /*
 406          *      Get and verify the address.
 407          */
 408
 409         if (msg->msg_namelen) {
 410                 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
 411                 err = -EINVAL;
 412                 if (msg->msg_namelen < sizeof(*usin))
 413                         goto out;
 414                 if (usin->sin_family != AF_INET) {
 415                         static int complained;
 416                         if (!complained++)
 417                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 418                                                  "raw sendmsg. Fix it!\n",
 419                                                  current->comm);
 420                         err = -EAFNOSUPPORT;
 421                         if (usin->sin_family)
 422                                 goto out;
 423                 }
 424                 daddr = usin->sin_addr.s_addr;
 425                 /* ANK: I did not forget to get protocol from port field.
 426                  * I just do not know, who uses this weirdness.
 427                  * IP_HDRINCL is much more convenient.
 428                  */
 429         } else {
 430                 err = -EDESTADDRREQ;
 431                 if (sk->sk_state != TCP_ESTABLISHED)
 432                         goto out;
 433                 daddr = inet->daddr;
 434         }
 435
 436         ipc.addr = inet->saddr;
 437         ipc.opt = NULL;
 438         ipc.oif = sk->sk_bound_dev_if;
 439
 440         if (msg->msg_controllen) {
 441                 err = ip_cmsg_send(msg, &ipc);
 442                 if (err)
 443                         goto out;
 444                 if (ipc.opt)
 445                         free = 1;
 446         }
 447
 448         saddr = ipc.addr;
 449         ipc.addr = daddr;
 450
 451         if (!ipc.opt)
 452                 ipc.opt = inet->opt;
 453
 454         if (ipc.opt) {
 455                 err = -EINVAL;
 456                 /* Linux does not mangle headers on raw sockets,
 457                  * so that IP options + IP_HDRINCL is non-sense.
 458                  */
 459                 if (inet->hdrincl)
 460                         goto done;
 461                 if (ipc.opt->srr) {
 462                         if (!daddr)
 463                                 goto done;
 464                         daddr = ipc.opt->faddr;
 465                 }
 466         }
 467         tos = RT_CONN_FLAGS(sk);
 468         if (msg->msg_flags & MSG_DONTROUTE)
 469                 tos |= RTO_ONLINK;
 470
 471         if (MULTICAST(daddr)) {
 472                 if (!ipc.oif)
 473                         ipc.oif = inet->mc_index;
 474                 if (!saddr)
 475                         saddr = inet->mc_addr;
 476         }
 477
 478         {
 479                 struct flowi fl = { .oif = ipc.oif,
 480                                     .nl_u = { .ip4_u =
 481                                               { .daddr = daddr,
 482                                                 .saddr = saddr,
 483                                                 .tos = tos } },
 484                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 485                                                              sk->sk_protocol,
 486                                   };
 487                 if (!inet->hdrincl) {
 488                         err = raw_probe_proto_opt(&fl, msg);
 489                         if (err)
 490                                 goto done;
 491                 }
 492
 493                 security_sk_classify_flow(sk, &fl);
 494                 err = ip_route_output_flow(&rt, &fl, sk, 1);
 495         }
 496         if (err)
 497                 goto done;
 498
 499         err = -EACCES;
 500         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 501                 goto done;
 502
 503         if (msg->msg_flags & MSG_CONFIRM)
 504                 goto do_confirm;
 505 back_from_confirm:
 506
 507         if (inet->hdrincl)
 508                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 509                                         rt, msg->msg_flags);
 510
 511          else {
 512                 if (!ipc.addr)
 513                         ipc.addr = rt->rt_dst;
 514                 lock_sock(sk);
 515                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 516                                         &ipc, rt, msg->msg_flags);
 517                 if (err)
 518                         ip_flush_pending_frames(sk);
 519                 else if (!(msg->msg_flags & MSG_MORE))
 520                         err = ip_push_pending_frames(sk);
 521                 release_sock(sk);
 522         }
 523 done:
 524         if (free)
 525                 kfree(ipc.opt);
 526         ip_rt_put(rt);
 527
 528 out:
 529         if (err < 0)
 530                 return err;
 531         return len;
 532
 533 do_confirm:
 534         dst_confirm(&rt->u.dst);
 535         if (!(msg->msg_flags & MSG_PROBE) || len)
 536                 goto back_from_confirm;
 537         err = 0;
 538         goto done;
 539 }
 540
 541 static void raw_close(struct sock *sk, long timeout)
 542 {
 543         /*
 544          * Raw sockets may have direct kernel refereneces. Kill them.
 545          */
 546         ip_ra_control(sk, 0, NULL);
 547
 548         sk_common_release(sk);
 549 }
 550
 551 /* This gets rid of all the nasties in af_inet. -DaveM */
 552 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 553 {
 554         struct inet_sock *inet = inet_sk(sk);
 555         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 556         int ret = -EINVAL;
 557         int chk_addr_ret;
 558
 559         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 560                 goto out;
 561         chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
 562         ret = -EADDRNOTAVAIL;
 563         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 564             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 565                 goto out;
 566         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 567         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 568                 inet->saddr = 0;  /* Use device */
 569         sk_dst_reset(sk);
 570         ret = 0;
 571 out:    return ret;
 572 }
 573
 574 /*
 575  *      This should be easy, if there is something there
 576  *      we return it, otherwise we block.
 577  */
 578
 579 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 580                        size_t len, int noblock, int flags, int *addr_len)
 581 {
 582         struct inet_sock *inet = inet_sk(sk);
 583         size_t copied = 0;
 584         int err = -EOPNOTSUPP;
 585         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 586         struct sk_buff *skb;
 587
 588         if (flags & MSG_OOB)
 589                 goto out;
 590
 591         if (addr_len)
 592                 *addr_len = sizeof(*sin);
 593
 594         if (flags & MSG_ERRQUEUE) {
 595                 err = ip_recv_error(sk, msg, len);
 596                 goto out;
 597         }
 598
 599         skb = skb_recv_datagram(sk, flags, noblock, &err);
 600         if (!skb)
 601                 goto out;
 602
 603         copied = skb->len;
 604         if (len < copied) {
 605                 msg->msg_flags |= MSG_TRUNC;
 606                 copied = len;
 607         }
 608
 609         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 610         if (err)
 611                 goto done;
 612
 613         sock_recv_timestamp(msg, sk, skb);
 614
 615         /* Copy the address. */
 616         if (sin) {
 617                 sin->sin_family = AF_INET;
 618                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 619                 sin->sin_port = 0;
 620                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 621         }
 622         if (inet->cmsg_flags)
 623                 ip_cmsg_recv(msg, skb);
 624         if (flags & MSG_TRUNC)
 625                 copied = skb->len;
 626 done:
 627         skb_free_datagram(sk, skb);
 628 out:
 629         if (err)
 630                 return err;
 631         return copied;
 632 }
 633
 634 static int raw_init(struct sock *sk)
 635 {
 636         struct raw_sock *rp = raw_sk(sk);
 637
 638         if (inet_sk(sk)->num == IPPROTO_ICMP)
 639                 memset(&rp->filter, 0, sizeof(rp->filter));
 640         return 0;
 641 }
 642
 643 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 644 {
 645         if (optlen > sizeof(struct icmp_filter))
 646                 optlen = sizeof(struct icmp_filter);
 647         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 648                 return -EFAULT;
 649         return 0;
 650 }
 651
 652 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 653 {
 654         int len, ret = -EFAULT;
 655
 656         if (get_user(len, optlen))
 657                 goto out;
 658         ret = -EINVAL;
 659         if (len < 0)
 660                 goto out;
 661         if (len > sizeof(struct icmp_filter))
 662                 len = sizeof(struct icmp_filter);
 663         ret = -EFAULT;
 664         if (put_user(len, optlen) ||
 665             copy_to_user(optval, &raw_sk(sk)->filter, len))
 666                 goto out;
 667         ret = 0;
 668 out:    return ret;
 669 }
 670
 671 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 672                           char __user *optval, int optlen)
 673 {
 674         if (optname == ICMP_FILTER) {
 675                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 676                         return -EOPNOTSUPP;
 677                 else
 678                         return raw_seticmpfilter(sk, optval, optlen);
 679         }
 680         return -ENOPROTOOPT;
 681 }
 682
 683 static int raw_setsockopt(struct sock *sk, int level, int optname,
 684                           char __user *optval, int optlen)
 685 {
 686         if (level != SOL_RAW)
 687                 return ip_setsockopt(sk, level, optname, optval, optlen);
 688         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 689 }
 690
 691 #ifdef CONFIG_COMPAT
 692 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 693                                  char __user *optval, int optlen)
 694 {
 695         if (level != SOL_RAW)
 696                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 697         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 698 }
 699 #endif
 700
 701 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 702                           char __user *optval, int __user *optlen)
 703 {
 704         if (optname == ICMP_FILTER) {
 705                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 706                         return -EOPNOTSUPP;
 707                 else
 708                         return raw_geticmpfilter(sk, optval, optlen);
 709         }
 710         return -ENOPROTOOPT;
 711 }
 712
 713 static int raw_getsockopt(struct sock *sk, int level, int optname,
 714                           char __user *optval, int __user *optlen)
 715 {
 716         if (level != SOL_RAW)
 717                 return ip_getsockopt(sk, level, optname, optval, optlen);
 718         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 719 }
 720
 721 #ifdef CONFIG_COMPAT
 722 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 723                                  char __user *optval, int __user *optlen)
 724 {
 725         if (level != SOL_RAW)
 726                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 727         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 728 }
 729 #endif
 730
 731 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 732 {
 733         switch (cmd) {
 734                 case SIOCOUTQ: {
 735                         int amount = atomic_read(&sk->sk_wmem_alloc);
 736                         return put_user(amount, (int __user *)arg);
 737                 }
 738                 case SIOCINQ: {
 739                         struct sk_buff *skb;
 740                         int amount = 0;
 741
 742                         spin_lock_bh(&sk->sk_receive_queue.lock);
 743                         skb = skb_peek(&sk->sk_receive_queue);
 744                         if (skb != NULL)
 745                                 amount = skb->len;
 746                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 747                         return put_user(amount, (int __user *)arg);
 748                 }
 749
 750                 default:
 751 #ifdef CONFIG_IP_MROUTE
 752                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 753 #else
 754                         return -ENOIOCTLCMD;
 755 #endif
 756         }
 757 }
 758
 759 struct proto raw_prot = {
 760         .name              = "RAW",
 761         .owner             = THIS_MODULE,
 762         .close             = raw_close,
 763         .connect           = ip4_datagram_connect,
 764         .disconnect        = udp_disconnect,
 765         .ioctl             = raw_ioctl,
 766         .init              = raw_init,
 767         .setsockopt        = raw_setsockopt,
 768         .getsockopt        = raw_getsockopt,
 769         .sendmsg           = raw_sendmsg,
 770         .recvmsg           = raw_recvmsg,
 771         .bind              = raw_bind,
 772         .backlog_rcv       = raw_rcv_skb,
 773         .hash              = raw_v4_hash,
 774         .unhash            = raw_v4_unhash,
 775         .obj_size          = sizeof(struct raw_sock),
 776 #ifdef CONFIG_COMPAT
 777         .compat_setsockopt = compat_raw_setsockopt,
 778         .compat_getsockopt = compat_raw_getsockopt,
 779 #endif
 780 };
 781
 782 #ifdef CONFIG_PROC_FS
 783 struct raw_iter_state {
 784         int bucket;
 785 };
 786
 787 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
 788
 789 static struct sock *raw_get_first(struct seq_file *seq)
 790 {
 791         struct sock *sk;
 792         struct raw_iter_state* state = raw_seq_private(seq);
 793
 794         for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
 795                 struct hlist_node *node;
 796
 797                 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
 798                         if (sk->sk_family == PF_INET)
 799                                 goto found;
 800         }
 801         sk = NULL;
 802 found:
 803         return sk;
 804 }
 805
 806 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 807 {
 808         struct raw_iter_state* state = raw_seq_private(seq);
 809
 810         do {
 811                 sk = sk_next(sk);
 812 try_again:
 813                 ;
 814         } while (sk && sk->sk_family != PF_INET);
 815
 816         if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
 817                 sk = sk_head(&raw_v4_htable[state->bucket]);
 818                 goto try_again;
 819         }
 820         return sk;
 821 }
 822
 823 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 824 {
 825         struct sock *sk = raw_get_first(seq);
 826
 827         if (sk)
 828                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 829                         --pos;
 830         return pos ? NULL : sk;
 831 }
 832
 833 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 834 {
 835         read_lock(&raw_v4_lock);
 836         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 837 }
 838
 839 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 840 {
 841         struct sock *sk;
 842
 843         if (v == SEQ_START_TOKEN)
 844                 sk = raw_get_first(seq);
 845         else
 846                 sk = raw_get_next(seq, v);
 847         ++*pos;
 848         return sk;
 849 }
 850
 851 static void raw_seq_stop(struct seq_file *seq, void *v)
 852 {
 853         read_unlock(&raw_v4_lock);
 854 }
 855
 856 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
 857 {
 858         struct inet_sock *inet = inet_sk(sp);
 859         __be32 dest = inet->daddr,
 860                src = inet->rcv_saddr;
 861         __u16 destp = 0,
 862               srcp  = inet->num;
 863
 864         sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
 865                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
 866                 i, src, srcp, dest, destp, sp->sk_state,
 867                 atomic_read(&sp->sk_wmem_alloc),
 868                 atomic_read(&sp->sk_rmem_alloc),
 869                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 870                 atomic_read(&sp->sk_refcnt), sp);
 871         return tmpbuf;
 872 }
 873
 874 static int raw_seq_show(struct seq_file *seq, void *v)
 875 {
 876         char tmpbuf[129];
 877
 878         if (v == SEQ_START_TOKEN)
 879                 seq_printf(seq, "%-127s\n",
 880                                "  sl  local_address rem_address   st tx_queue "
 881                                "rx_queue tr tm->when retrnsmt   uid  timeout "
 882                                "inode");
 883         else {
 884                 struct raw_iter_state *state = raw_seq_private(seq);
 885
 886                 seq_printf(seq, "%-127s\n",
 887                            get_raw_sock(v, tmpbuf, state->bucket));
 888         }
 889         return 0;
 890 }
 891
 892 static const struct seq_operations raw_seq_ops = {
 893         .start = raw_seq_start,
 894         .next  = raw_seq_next,
 895         .stop  = raw_seq_stop,
 896         .show  = raw_seq_show,
 897 };
 898
 899 static int raw_seq_open(struct inode *inode, struct file *file)
 900 {
 901         struct seq_file *seq;
 902         int rc = -ENOMEM;
 903         struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
 904
 905         if (!s)
 906                 goto out;
 907         rc = seq_open(file, &raw_seq_ops);
 908         if (rc)
 909                 goto out_kfree;
 910
 911         seq = file->private_data;
 912         seq->private = s;
 913         memset(s, 0, sizeof(*s));
 914 out:
 915         return rc;
 916 out_kfree:
 917         kfree(s);
 918         goto out;
 919 }
 920
 921 static const struct file_operations raw_seq_fops = {
 922         .owner   = THIS_MODULE,
 923         .open    = raw_seq_open,
 924         .read    = seq_read,
 925         .llseek  = seq_lseek,
 926         .release = seq_release_private,
 927 };
 928
 929 int __init raw_proc_init(void)
 930 {
 931         if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
 932                 return -ENOMEM;
 933         return 0;
 934 }
 935
 936 void __init raw_proc_exit(void)
 937 {
 938         proc_net_remove("raw");
 939 }
 940 #endif /* CONFIG_PROC_FS */