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 <linux/types.h>
  44 #include <asm/atomic.h>
  45 #include <asm/byteorder.h>
  46 #include <asm/current.h>
  47 #include <asm/uaccess.h>
  48 #include <asm/ioctls.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/skbuff.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 = 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 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         int type = skb->h.icmph->type;
 189         int code = skb->h.icmph->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                 /* FIXME: increment a raw drops counter here */
 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                 kfree_skb(skb);
 256                 return NET_RX_DROP;
 257         }
 258         nf_reset(skb);
 259
 260         skb_push(skb, skb->data - skb->nh.raw);
 261
 262         raw_rcv_skb(sk, skb);
 263         return 0;
 264 }
 265
 266 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 267                         struct rtable *rt,
 268                         unsigned int flags)
 269 {
 270         struct inet_sock *inet = inet_sk(sk);
 271         int hh_len;
 272         struct iphdr *iph;
 273         struct sk_buff *skb;
 274         int err;
 275
 276         if (length > rt->u.dst.dev->mtu) {
 277                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
 278                                rt->u.dst.dev->mtu);
 279                 return -EMSGSIZE;
 280         }
 281         if (flags&MSG_PROBE)
 282                 goto out;
 283
 284         hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
 285
 286         skb = sock_alloc_send_skb(sk, length+hh_len+15,
 287                                   flags&MSG_DONTWAIT, &err);
 288         if (skb == NULL)
 289                 goto error;
 290         skb_reserve(skb, hh_len);
 291
 292         skb->priority = sk->sk_priority;
 293         skb->dst = dst_clone(&rt->u.dst);
 294
 295         skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
 296
 297         skb->ip_summed = CHECKSUM_NONE;
 298
 299         skb->h.raw = skb->nh.raw;
 300         err = memcpy_fromiovecend((void *)iph, from, 0, length);
 301         if (err)
 302                 goto error_fault;
 303
 304         /* We don't modify invalid header */
 305         if (length >= sizeof(*iph) && iph->ihl * 4U <= length) {
 306                 if (!iph->saddr)
 307                         iph->saddr = rt->rt_src;
 308                 iph->check   = 0;
 309                 iph->tot_len = htons(length);
 310                 if (!iph->id)
 311                         ip_select_ident(iph, &rt->u.dst, NULL);
 312
 313                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 314         }
 315
 316         err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 317                       dst_output);
 318         if (err > 0)
 319                 err = inet->recverr ? net_xmit_errno(err) : 0;
 320         if (err)
 321                 goto error;
 322 out:
 323         return 0;
 324
 325 error_fault:
 326         err = -EFAULT;
 327         kfree_skb(skb);
 328 error:
 329         IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
 330         return err;
 331 }
 332
 333 static void raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 334 {
 335         struct iovec *iov;
 336         u8 __user *type = NULL;
 337         u8 __user *code = NULL;
 338         int probed = 0;
 339         unsigned int i;
 340
 341         if (!msg->msg_iov)
 342                 return;
 343
 344         for (i = 0; i < msg->msg_iovlen; i++) {
 345                 iov = &msg->msg_iov[i];
 346                 if (!iov)
 347                         continue;
 348
 349                 switch (fl->proto) {
 350                 case IPPROTO_ICMP:
 351                         /* check if one-byte field is readable or not. */
 352                         if (iov->iov_base && iov->iov_len < 1)
 353                                 break;
 354
 355                         if (!type) {
 356                                 type = iov->iov_base;
 357                                 /* check if code field is readable or not. */
 358                                 if (iov->iov_len > 1)
 359                                         code = type + 1;
 360                         } else if (!code)
 361                                 code = iov->iov_base;
 362
 363                         if (type && code) {
 364                                 get_user(fl->fl_icmp_type, type);
 365                                 get_user(fl->fl_icmp_code, code);
 366                                 probed = 1;
 367                         }
 368                         break;
 369                 default:
 370                         probed = 1;
 371                         break;
 372                 }
 373                 if (probed)
 374                         break;
 375         }
 376 }
 377
 378 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 379                        size_t len)
 380 {
 381         struct inet_sock *inet = inet_sk(sk);
 382         struct ipcm_cookie ipc;
 383         struct rtable *rt = NULL;
 384         int free = 0;
 385         u32 daddr;
 386         u32 saddr;
 387         u8  tos;
 388         int err;
 389
 390         err = -EMSGSIZE;
 391         if (len > 0xFFFF)
 392                 goto out;
 393
 394         /*
 395          *      Check the flags.
 396          */
 397
 398         err = -EOPNOTSUPP;
 399         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 400                 goto out;               /* compatibility */
 401
 402         /*
 403          *      Get and verify the address.
 404          */
 405
 406         if (msg->msg_namelen) {
 407                 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
 408                 err = -EINVAL;
 409                 if (msg->msg_namelen < sizeof(*usin))
 410                         goto out;
 411                 if (usin->sin_family != AF_INET) {
 412                         static int complained;
 413                         if (!complained++)
 414                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 415                                                  "raw sendmsg. Fix it!\n",
 416                                                  current->comm);
 417                         err = -EAFNOSUPPORT;
 418                         if (usin->sin_family)
 419                                 goto out;
 420                 }
 421                 daddr = usin->sin_addr.s_addr;
 422                 /* ANK: I did not forget to get protocol from port field.
 423                  * I just do not know, who uses this weirdness.
 424                  * IP_HDRINCL is much more convenient.
 425                  */
 426         } else {
 427                 err = -EDESTADDRREQ;
 428                 if (sk->sk_state != TCP_ESTABLISHED)
 429                         goto out;
 430                 daddr = inet->daddr;
 431         }
 432
 433         ipc.addr = inet->saddr;
 434         ipc.opt = NULL;
 435         ipc.oif = sk->sk_bound_dev_if;
 436
 437         if (msg->msg_controllen) {
 438                 err = ip_cmsg_send(msg, &ipc);
 439                 if (err)
 440                         goto out;
 441                 if (ipc.opt)
 442                         free = 1;
 443         }
 444
 445         saddr = ipc.addr;
 446         ipc.addr = daddr;
 447
 448         if (!ipc.opt)
 449                 ipc.opt = inet->opt;
 450
 451         if (ipc.opt) {
 452                 err = -EINVAL;
 453                 /* Linux does not mangle headers on raw sockets,
 454                  * so that IP options + IP_HDRINCL is non-sense.
 455                  */
 456                 if (inet->hdrincl)
 457                         goto done;
 458                 if (ipc.opt->srr) {
 459                         if (!daddr)
 460                                 goto done;
 461                         daddr = ipc.opt->faddr;
 462                 }
 463         }
 464         tos = RT_CONN_FLAGS(sk);
 465         if (msg->msg_flags & MSG_DONTROUTE)
 466                 tos |= RTO_ONLINK;
 467
 468         if (MULTICAST(daddr)) {
 469                 if (!ipc.oif)
 470                         ipc.oif = inet->mc_index;
 471                 if (!saddr)
 472                         saddr = inet->mc_addr;
 473         }
 474
 475         {
 476                 struct flowi fl = { .oif = ipc.oif,
 477                                     .nl_u = { .ip4_u =
 478                                               { .daddr = daddr,
 479                                                 .saddr = saddr,
 480                                                 .tos = tos } },
 481                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 482                                                              sk->sk_protocol,
 483                                   };
 484                 if (!inet->hdrincl)
 485                         raw_probe_proto_opt(&fl, msg);
 486
 487                 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
 488         }
 489         if (err)
 490                 goto done;
 491
 492         err = -EACCES;
 493         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 494                 goto done;
 495
 496         if (msg->msg_flags & MSG_CONFIRM)
 497                 goto do_confirm;
 498 back_from_confirm:
 499
 500         if (inet->hdrincl)
 501                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 502                                         rt, msg->msg_flags);
 503
 504          else {
 505                 if (!ipc.addr)
 506                         ipc.addr = rt->rt_dst;
 507                 lock_sock(sk);
 508                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 509                                         &ipc, rt, msg->msg_flags);
 510                 if (err)
 511                         ip_flush_pending_frames(sk);
 512                 else if (!(msg->msg_flags & MSG_MORE))
 513                         err = ip_push_pending_frames(sk);
 514                 release_sock(sk);
 515         }
 516 done:
 517         if (free)
 518                 kfree(ipc.opt);
 519         ip_rt_put(rt);
 520
 521 out:
 522         if (err < 0)
 523                 return err;
 524         return len;
 525
 526 do_confirm:
 527         dst_confirm(&rt->u.dst);
 528         if (!(msg->msg_flags & MSG_PROBE) || len)
 529                 goto back_from_confirm;
 530         err = 0;
 531         goto done;
 532 }
 533
 534 static void raw_close(struct sock *sk, long timeout)
 535 {
 536         /*
 537          * Raw sockets may have direct kernel refereneces. Kill them.
 538          */
 539         ip_ra_control(sk, 0, NULL);
 540
 541         sk_common_release(sk);
 542 }
 543
 544 /* This gets rid of all the nasties in af_inet. -DaveM */
 545 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 546 {
 547         struct inet_sock *inet = inet_sk(sk);
 548         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 549         int ret = -EINVAL;
 550         int chk_addr_ret;
 551
 552         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 553                 goto out;
 554         chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
 555         ret = -EADDRNOTAVAIL;
 556         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 557             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 558                 goto out;
 559         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 560         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 561                 inet->saddr = 0;  /* Use device */
 562         sk_dst_reset(sk);
 563         ret = 0;
 564 out:    return ret;
 565 }
 566
 567 /*
 568  *      This should be easy, if there is something there
 569  *      we return it, otherwise we block.
 570  */
 571
 572 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 573                        size_t len, int noblock, int flags, int *addr_len)
 574 {
 575         struct inet_sock *inet = inet_sk(sk);
 576         size_t copied = 0;
 577         int err = -EOPNOTSUPP;
 578         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 579         struct sk_buff *skb;
 580
 581         if (flags & MSG_OOB)
 582                 goto out;
 583
 584         if (addr_len)
 585                 *addr_len = sizeof(*sin);
 586
 587         if (flags & MSG_ERRQUEUE) {
 588                 err = ip_recv_error(sk, msg, len);
 589                 goto out;
 590         }
 591
 592         skb = skb_recv_datagram(sk, flags, noblock, &err);
 593         if (!skb)
 594                 goto out;
 595
 596         copied = skb->len;
 597         if (len < copied) {
 598                 msg->msg_flags |= MSG_TRUNC;
 599                 copied = len;
 600         }
 601
 602         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 603         if (err)
 604                 goto done;
 605
 606         sock_recv_timestamp(msg, sk, skb);
 607
 608         /* Copy the address. */
 609         if (sin) {
 610                 sin->sin_family = AF_INET;
 611                 sin->sin_addr.s_addr = skb->nh.iph->saddr;
 612                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 613         }
 614         if (inet->cmsg_flags)
 615                 ip_cmsg_recv(msg, skb);
 616         if (flags & MSG_TRUNC)
 617                 copied = skb->len;
 618 done:
 619         skb_free_datagram(sk, skb);
 620 out:
 621         if (err)
 622                 return err;
 623         return copied;
 624 }
 625
 626 static int raw_init(struct sock *sk)
 627 {
 628         struct raw_sock *rp = raw_sk(sk);
 629
 630         if (inet_sk(sk)->num == IPPROTO_ICMP)
 631                 memset(&rp->filter, 0, sizeof(rp->filter));
 632         return 0;
 633 }
 634
 635 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 636 {
 637         if (optlen > sizeof(struct icmp_filter))
 638                 optlen = sizeof(struct icmp_filter);
 639         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 640                 return -EFAULT;
 641         return 0;
 642 }
 643
 644 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 645 {
 646         int len, ret = -EFAULT;
 647
 648         if (get_user(len, optlen))
 649                 goto out;
 650         ret = -EINVAL;
 651         if (len < 0)
 652                 goto out;
 653         if (len > sizeof(struct icmp_filter))
 654                 len = sizeof(struct icmp_filter);
 655         ret = -EFAULT;
 656         if (put_user(len, optlen) ||
 657             copy_to_user(optval, &raw_sk(sk)->filter, len))
 658                 goto out;
 659         ret = 0;
 660 out:    return ret;
 661 }
 662
 663 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 664                           char __user *optval, int optlen)
 665 {
 666         if (optname == ICMP_FILTER) {
 667                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 668                         return -EOPNOTSUPP;
 669                 else
 670                         return raw_seticmpfilter(sk, optval, optlen);
 671         }
 672         return -ENOPROTOOPT;
 673 }
 674
 675 static int raw_setsockopt(struct sock *sk, int level, int optname,
 676                           char __user *optval, int optlen)
 677 {
 678         if (level != SOL_RAW)
 679                 return ip_setsockopt(sk, level, optname, optval, optlen);
 680         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 681 }
 682
 683 #ifdef CONFIG_COMPAT
 684 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 685                                  char __user *optval, int optlen)
 686 {
 687         if (level != SOL_RAW)
 688                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 689         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 690 }
 691 #endif
 692
 693 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 694                           char __user *optval, int __user *optlen)
 695 {
 696         if (optname == ICMP_FILTER) {
 697                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 698                         return -EOPNOTSUPP;
 699                 else
 700                         return raw_geticmpfilter(sk, optval, optlen);
 701         }
 702         return -ENOPROTOOPT;
 703 }
 704
 705 static int raw_getsockopt(struct sock *sk, int level, int optname,
 706                           char __user *optval, int __user *optlen)
 707 {
 708         if (level != SOL_RAW)
 709                 return ip_getsockopt(sk, level, optname, optval, optlen);
 710         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 711 }
 712
 713 #ifdef CONFIG_COMPAT
 714 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 715                                  char __user *optval, int __user *optlen)
 716 {
 717         if (level != SOL_RAW)
 718                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 719         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 720 }
 721 #endif
 722
 723 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 724 {
 725         switch (cmd) {
 726                 case SIOCOUTQ: {
 727                         int amount = atomic_read(&sk->sk_wmem_alloc);
 728                         return put_user(amount, (int __user *)arg);
 729                 }
 730                 case SIOCINQ: {
 731                         struct sk_buff *skb;
 732                         int amount = 0;
 733
 734                         spin_lock_bh(&sk->sk_receive_queue.lock);
 735                         skb = skb_peek(&sk->sk_receive_queue);
 736                         if (skb != NULL)
 737                                 amount = skb->len;
 738                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 739                         return put_user(amount, (int __user *)arg);
 740                 }
 741
 742                 default:
 743 #ifdef CONFIG_IP_MROUTE
 744                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 745 #else
 746                         return -ENOIOCTLCMD;
 747 #endif
 748         }
 749 }
 750
 751 struct proto raw_prot = {
 752         .name              = "RAW",
 753         .owner             = THIS_MODULE,
 754         .close             = raw_close,
 755         .connect           = ip4_datagram_connect,
 756         .disconnect        = udp_disconnect,
 757         .ioctl             = raw_ioctl,
 758         .init              = raw_init,
 759         .setsockopt        = raw_setsockopt,
 760         .getsockopt        = raw_getsockopt,
 761         .sendmsg           = raw_sendmsg,
 762         .recvmsg           = raw_recvmsg,
 763         .bind              = raw_bind,
 764         .backlog_rcv       = raw_rcv_skb,
 765         .hash              = raw_v4_hash,
 766         .unhash            = raw_v4_unhash,
 767         .obj_size          = sizeof(struct raw_sock),
 768 #ifdef CONFIG_COMPAT
 769         .compat_setsockopt = compat_raw_setsockopt,
 770         .compat_getsockopt = compat_raw_getsockopt,
 771 #endif
 772 };
 773
 774 #ifdef CONFIG_PROC_FS
 775 struct raw_iter_state {
 776         int bucket;
 777 };
 778
 779 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
 780
 781 static struct sock *raw_get_first(struct seq_file *seq)
 782 {
 783         struct sock *sk;
 784         struct raw_iter_state* state = raw_seq_private(seq);
 785
 786         for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
 787                 struct hlist_node *node;
 788
 789                 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
 790                         if (sk->sk_family == PF_INET)
 791                                 goto found;
 792         }
 793         sk = NULL;
 794 found:
 795         return sk;
 796 }
 797
 798 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 799 {
 800         struct raw_iter_state* state = raw_seq_private(seq);
 801
 802         do {
 803                 sk = sk_next(sk);
 804 try_again:
 805                 ;
 806         } while (sk && sk->sk_family != PF_INET);
 807
 808         if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
 809                 sk = sk_head(&raw_v4_htable[state->bucket]);
 810                 goto try_again;
 811         }
 812         return sk;
 813 }
 814
 815 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 816 {
 817         struct sock *sk = raw_get_first(seq);
 818
 819         if (sk)
 820                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 821                         --pos;
 822         return pos ? NULL : sk;
 823 }
 824
 825 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 826 {
 827         read_lock(&raw_v4_lock);
 828         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 829 }
 830
 831 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 832 {
 833         struct sock *sk;
 834
 835         if (v == SEQ_START_TOKEN)
 836                 sk = raw_get_first(seq);
 837         else
 838                 sk = raw_get_next(seq, v);
 839         ++*pos;
 840         return sk;
 841 }
 842
 843 static void raw_seq_stop(struct seq_file *seq, void *v)
 844 {
 845         read_unlock(&raw_v4_lock);
 846 }
 847
 848 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
 849 {
 850         struct inet_sock *inet = inet_sk(sp);
 851         unsigned int dest = inet->daddr,
 852                      src = inet->rcv_saddr;
 853         __u16 destp = 0,
 854               srcp  = inet->num;
 855
 856         sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
 857                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
 858                 i, src, srcp, dest, destp, sp->sk_state,
 859                 atomic_read(&sp->sk_wmem_alloc),
 860                 atomic_read(&sp->sk_rmem_alloc),
 861                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 862                 atomic_read(&sp->sk_refcnt), sp);
 863         return tmpbuf;
 864 }
 865
 866 static int raw_seq_show(struct seq_file *seq, void *v)
 867 {
 868         char tmpbuf[129];
 869
 870         if (v == SEQ_START_TOKEN)
 871                 seq_printf(seq, "%-127s\n",
 872                                "  sl  local_address rem_address   st tx_queue "
 873                                "rx_queue tr tm->when retrnsmt   uid  timeout "
 874                                "inode");
 875         else {
 876                 struct raw_iter_state *state = raw_seq_private(seq);
 877
 878                 seq_printf(seq, "%-127s\n",
 879                            get_raw_sock(v, tmpbuf, state->bucket));
 880         }
 881         return 0;
 882 }
 883
 884 static struct seq_operations raw_seq_ops = {
 885         .start = raw_seq_start,
 886         .next  = raw_seq_next,
 887         .stop  = raw_seq_stop,
 888         .show  = raw_seq_show,
 889 };
 890
 891 static int raw_seq_open(struct inode *inode, struct file *file)
 892 {
 893         struct seq_file *seq;
 894         int rc = -ENOMEM;
 895         struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
 896
 897         if (!s)
 898                 goto out;
 899         rc = seq_open(file, &raw_seq_ops);
 900         if (rc)
 901                 goto out_kfree;
 902
 903         seq = file->private_data;
 904         seq->private = s;
 905         memset(s, 0, sizeof(*s));
 906 out:
 907         return rc;
 908 out_kfree:
 909         kfree(s);
 910         goto out;
 911 }
 912
 913 static struct file_operations raw_seq_fops = {
 914         .owner   = THIS_MODULE,
 915         .open    = raw_seq_open,
 916         .read    = seq_read,
 917         .llseek  = seq_lseek,
 918         .release = seq_release_private,
 919 };
 920
 921 int __init raw_proc_init(void)
 922 {
 923         if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
 924                 return -ENOMEM;
 925         return 0;
 926 }
 927
 928 void __init raw_proc_exit(void)
 929 {
 930         proc_net_remove("raw");
 931 }
 932 #endif /* CONFIG_PROC_FS */