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