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