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                 /* 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_network_header(skb));
 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_reset_network_header(skb);
 296         iph = ip_hdr(skb);
 297         skb_put(skb, length);
 298
 299         skb->ip_summed = CHECKSUM_NONE;
 300
 301         skb->transport_header = skb->network_header;
 302         err = memcpy_fromiovecend((void *)iph, from, 0, length);
 303         if (err)
 304                 goto error_fault;
 305
 306         /* We don't modify invalid header */
 307         if (length >= sizeof(*iph) && iph->ihl * 4U <= length) {
 308                 if (!iph->saddr)
 309                         iph->saddr = rt->rt_src;
 310                 iph->check   = 0;
 311                 iph->tot_len = htons(length);
 312                 if (!iph->id)
 313                         ip_select_ident(iph, &rt->u.dst, NULL);
 314
 315                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 316         }
 317         if (iph->protocol == IPPROTO_ICMP)
 318                 icmp_out_count(((struct icmphdr *)
 319                         skb_transport_header(skb))->type);
 320
 321         err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 322                       dst_output);
 323         if (err > 0)
 324                 err = inet->recverr ? net_xmit_errno(err) : 0;
 325         if (err)
 326                 goto error;
 327 out:
 328         return 0;
 329
 330 error_fault:
 331         err = -EFAULT;
 332         kfree_skb(skb);
 333 error:
 334         IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
 335         return err;
 336 }
 337
 338 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 339 {
 340         struct iovec *iov;
 341         u8 __user *type = NULL;
 342         u8 __user *code = NULL;
 343         int probed = 0;
 344         unsigned int i;
 345
 346         if (!msg->msg_iov)
 347                 return 0;
 348
 349         for (i = 0; i < msg->msg_iovlen; i++) {
 350                 iov = &msg->msg_iov[i];
 351                 if (!iov)
 352                         continue;
 353
 354                 switch (fl->proto) {
 355                 case IPPROTO_ICMP:
 356                         /* check if one-byte field is readable or not. */
 357                         if (iov->iov_base && iov->iov_len < 1)
 358                                 break;
 359
 360                         if (!type) {
 361                                 type = iov->iov_base;
 362                                 /* check if code field is readable or not. */
 363                                 if (iov->iov_len > 1)
 364                                         code = type + 1;
 365                         } else if (!code)
 366                                 code = iov->iov_base;
 367
 368                         if (type && code) {
 369                                 if (get_user(fl->fl_icmp_type, type) ||
 370                                     get_user(fl->fl_icmp_code, code))
 371                                         return -EFAULT;
 372                                 probed = 1;
 373                         }
 374                         break;
 375                 default:
 376                         probed = 1;
 377                         break;
 378                 }
 379                 if (probed)
 380                         break;
 381         }
 382         return 0;
 383 }
 384
 385 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 386                        size_t len)
 387 {
 388         struct inet_sock *inet = inet_sk(sk);
 389         struct ipcm_cookie ipc;
 390         struct rtable *rt = NULL;
 391         int free = 0;
 392         __be32 daddr;
 393         __be32 saddr;
 394         u8  tos;
 395         int err;
 396
 397         err = -EMSGSIZE;
 398         if (len > 0xFFFF)
 399                 goto out;
 400
 401         /*
 402          *      Check the flags.
 403          */
 404
 405         err = -EOPNOTSUPP;
 406         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 407                 goto out;               /* compatibility */
 408
 409         /*
 410          *      Get and verify the address.
 411          */
 412
 413         if (msg->msg_namelen) {
 414                 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
 415                 err = -EINVAL;
 416                 if (msg->msg_namelen < sizeof(*usin))
 417                         goto out;
 418                 if (usin->sin_family != AF_INET) {
 419                         static int complained;
 420                         if (!complained++)
 421                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 422                                                  "raw sendmsg. Fix it!\n",
 423                                                  current->comm);
 424                         err = -EAFNOSUPPORT;
 425                         if (usin->sin_family)
 426                                 goto out;
 427                 }
 428                 daddr = usin->sin_addr.s_addr;
 429                 /* ANK: I did not forget to get protocol from port field.
 430                  * I just do not know, who uses this weirdness.
 431                  * IP_HDRINCL is much more convenient.
 432                  */
 433         } else {
 434                 err = -EDESTADDRREQ;
 435                 if (sk->sk_state != TCP_ESTABLISHED)
 436                         goto out;
 437                 daddr = inet->daddr;
 438         }
 439
 440         ipc.addr = inet->saddr;
 441         ipc.opt = NULL;
 442         ipc.oif = sk->sk_bound_dev_if;
 443
 444         if (msg->msg_controllen) {
 445                 err = ip_cmsg_send(msg, &ipc);
 446                 if (err)
 447                         goto out;
 448                 if (ipc.opt)
 449                         free = 1;
 450         }
 451
 452         saddr = ipc.addr;
 453         ipc.addr = daddr;
 454
 455         if (!ipc.opt)
 456                 ipc.opt = inet->opt;
 457
 458         if (ipc.opt) {
 459                 err = -EINVAL;
 460                 /* Linux does not mangle headers on raw sockets,
 461                  * so that IP options + IP_HDRINCL is non-sense.
 462                  */
 463                 if (inet->hdrincl)
 464                         goto done;
 465                 if (ipc.opt->srr) {
 466                         if (!daddr)
 467                                 goto done;
 468                         daddr = ipc.opt->faddr;
 469                 }
 470         }
 471         tos = RT_CONN_FLAGS(sk);
 472         if (msg->msg_flags & MSG_DONTROUTE)
 473                 tos |= RTO_ONLINK;
 474
 475         if (MULTICAST(daddr)) {
 476                 if (!ipc.oif)
 477                         ipc.oif = inet->mc_index;
 478                 if (!saddr)
 479                         saddr = inet->mc_addr;
 480         }
 481
 482         {
 483                 struct flowi fl = { .oif = ipc.oif,
 484                                     .nl_u = { .ip4_u =
 485                                               { .daddr = daddr,
 486                                                 .saddr = saddr,
 487                                                 .tos = tos } },
 488                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 489                                                              sk->sk_protocol,
 490                                   };
 491                 if (!inet->hdrincl) {
 492                         err = raw_probe_proto_opt(&fl, msg);
 493                         if (err)
 494                                 goto done;
 495                 }
 496
 497                 security_sk_classify_flow(sk, &fl);
 498                 err = ip_route_output_flow(&rt, &fl, sk, 1);
 499         }
 500         if (err)
 501                 goto done;
 502
 503         err = -EACCES;
 504         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 505                 goto done;
 506
 507         if (msg->msg_flags & MSG_CONFIRM)
 508                 goto do_confirm;
 509 back_from_confirm:
 510
 511         if (inet->hdrincl)
 512                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 513                                         rt, msg->msg_flags);
 514
 515          else {
 516                 if (!ipc.addr)
 517                         ipc.addr = rt->rt_dst;
 518                 lock_sock(sk);
 519                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 520                                         &ipc, rt, msg->msg_flags);
 521                 if (err)
 522                         ip_flush_pending_frames(sk);
 523                 else if (!(msg->msg_flags & MSG_MORE))
 524                         err = ip_push_pending_frames(sk);
 525                 release_sock(sk);
 526         }
 527 done:
 528         if (free)
 529                 kfree(ipc.opt);
 530         ip_rt_put(rt);
 531
 532 out:
 533         if (err < 0)
 534                 return err;
 535         return len;
 536
 537 do_confirm:
 538         dst_confirm(&rt->u.dst);
 539         if (!(msg->msg_flags & MSG_PROBE) || len)
 540                 goto back_from_confirm;
 541         err = 0;
 542         goto done;
 543 }
 544
 545 static void raw_close(struct sock *sk, long timeout)
 546 {
 547         /*
 548          * Raw sockets may have direct kernel refereneces. Kill them.
 549          */
 550         ip_ra_control(sk, 0, NULL);
 551
 552         sk_common_release(sk);
 553 }
 554
 555 /* This gets rid of all the nasties in af_inet. -DaveM */
 556 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 557 {
 558         struct inet_sock *inet = inet_sk(sk);
 559         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 560         int ret = -EINVAL;
 561         int chk_addr_ret;
 562
 563         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 564                 goto out;
 565         chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
 566         ret = -EADDRNOTAVAIL;
 567         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 568             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 569                 goto out;
 570         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 571         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 572                 inet->saddr = 0;  /* Use device */
 573         sk_dst_reset(sk);
 574         ret = 0;
 575 out:    return ret;
 576 }
 577
 578 /*
 579  *      This should be easy, if there is something there
 580  *      we return it, otherwise we block.
 581  */
 582
 583 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 584                        size_t len, int noblock, int flags, int *addr_len)
 585 {
 586         struct inet_sock *inet = inet_sk(sk);
 587         size_t copied = 0;
 588         int err = -EOPNOTSUPP;
 589         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 590         struct sk_buff *skb;
 591
 592         if (flags & MSG_OOB)
 593                 goto out;
 594
 595         if (addr_len)
 596                 *addr_len = sizeof(*sin);
 597
 598         if (flags & MSG_ERRQUEUE) {
 599                 err = ip_recv_error(sk, msg, len);
 600                 goto out;
 601         }
 602
 603         skb = skb_recv_datagram(sk, flags, noblock, &err);
 604         if (!skb)
 605                 goto out;
 606
 607         copied = skb->len;
 608         if (len < copied) {
 609                 msg->msg_flags |= MSG_TRUNC;
 610                 copied = len;
 611         }
 612
 613         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 614         if (err)
 615                 goto done;
 616
 617         sock_recv_timestamp(msg, sk, skb);
 618
 619         /* Copy the address. */
 620         if (sin) {
 621                 sin->sin_family = AF_INET;
 622                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 623                 sin->sin_port = 0;
 624                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 625         }
 626         if (inet->cmsg_flags)
 627                 ip_cmsg_recv(msg, skb);
 628         if (flags & MSG_TRUNC)
 629                 copied = skb->len;
 630 done:
 631         skb_free_datagram(sk, skb);
 632 out:
 633         if (err)
 634                 return err;
 635         return copied;
 636 }
 637
 638 static int raw_init(struct sock *sk)
 639 {
 640         struct raw_sock *rp = raw_sk(sk);
 641
 642         if (inet_sk(sk)->num == IPPROTO_ICMP)
 643                 memset(&rp->filter, 0, sizeof(rp->filter));
 644         return 0;
 645 }
 646
 647 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 648 {
 649         if (optlen > sizeof(struct icmp_filter))
 650                 optlen = sizeof(struct icmp_filter);
 651         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 652                 return -EFAULT;
 653         return 0;
 654 }
 655
 656 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 657 {
 658         int len, ret = -EFAULT;
 659
 660         if (get_user(len, optlen))
 661                 goto out;
 662         ret = -EINVAL;
 663         if (len < 0)
 664                 goto out;
 665         if (len > sizeof(struct icmp_filter))
 666                 len = sizeof(struct icmp_filter);
 667         ret = -EFAULT;
 668         if (put_user(len, optlen) ||
 669             copy_to_user(optval, &raw_sk(sk)->filter, len))
 670                 goto out;
 671         ret = 0;
 672 out:    return ret;
 673 }
 674
 675 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 676                           char __user *optval, int optlen)
 677 {
 678         if (optname == ICMP_FILTER) {
 679                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 680                         return -EOPNOTSUPP;
 681                 else
 682                         return raw_seticmpfilter(sk, optval, optlen);
 683         }
 684         return -ENOPROTOOPT;
 685 }
 686
 687 static int raw_setsockopt(struct sock *sk, int level, int optname,
 688                           char __user *optval, int optlen)
 689 {
 690         if (level != SOL_RAW)
 691                 return ip_setsockopt(sk, level, optname, optval, optlen);
 692         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 693 }
 694
 695 #ifdef CONFIG_COMPAT
 696 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 697                                  char __user *optval, int optlen)
 698 {
 699         if (level != SOL_RAW)
 700                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 701         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 702 }
 703 #endif
 704
 705 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 706                           char __user *optval, int __user *optlen)
 707 {
 708         if (optname == ICMP_FILTER) {
 709                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 710                         return -EOPNOTSUPP;
 711                 else
 712                         return raw_geticmpfilter(sk, optval, optlen);
 713         }
 714         return -ENOPROTOOPT;
 715 }
 716
 717 static int raw_getsockopt(struct sock *sk, int level, int optname,
 718                           char __user *optval, int __user *optlen)
 719 {
 720         if (level != SOL_RAW)
 721                 return ip_getsockopt(sk, level, optname, optval, optlen);
 722         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 723 }
 724
 725 #ifdef CONFIG_COMPAT
 726 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 727                                  char __user *optval, int __user *optlen)
 728 {
 729         if (level != SOL_RAW)
 730                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 731         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 732 }
 733 #endif
 734
 735 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 736 {
 737         switch (cmd) {
 738                 case SIOCOUTQ: {
 739                         int amount = atomic_read(&sk->sk_wmem_alloc);
 740                         return put_user(amount, (int __user *)arg);
 741                 }
 742                 case SIOCINQ: {
 743                         struct sk_buff *skb;
 744                         int amount = 0;
 745
 746                         spin_lock_bh(&sk->sk_receive_queue.lock);
 747                         skb = skb_peek(&sk->sk_receive_queue);
 748                         if (skb != NULL)
 749                                 amount = skb->len;
 750                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 751                         return put_user(amount, (int __user *)arg);
 752                 }
 753
 754                 default:
 755 #ifdef CONFIG_IP_MROUTE
 756                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 757 #else
 758                         return -ENOIOCTLCMD;
 759 #endif
 760         }
 761 }
 762
 763 struct proto raw_prot = {
 764         .name              = "RAW",
 765         .owner             = THIS_MODULE,
 766         .close             = raw_close,
 767         .connect           = ip4_datagram_connect,
 768         .disconnect        = udp_disconnect,
 769         .ioctl             = raw_ioctl,
 770         .init              = raw_init,
 771         .setsockopt        = raw_setsockopt,
 772         .getsockopt        = raw_getsockopt,
 773         .sendmsg           = raw_sendmsg,
 774         .recvmsg           = raw_recvmsg,
 775         .bind              = raw_bind,
 776         .backlog_rcv       = raw_rcv_skb,
 777         .hash              = raw_v4_hash,
 778         .unhash            = raw_v4_unhash,
 779         .obj_size          = sizeof(struct raw_sock),
 780 #ifdef CONFIG_COMPAT
 781         .compat_setsockopt = compat_raw_setsockopt,
 782         .compat_getsockopt = compat_raw_getsockopt,
 783 #endif
 784 };
 785
 786 #ifdef CONFIG_PROC_FS
 787 struct raw_iter_state {
 788         int bucket;
 789 };
 790
 791 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
 792
 793 static struct sock *raw_get_first(struct seq_file *seq)
 794 {
 795         struct sock *sk;
 796         struct raw_iter_state* state = raw_seq_private(seq);
 797
 798         for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
 799                 struct hlist_node *node;
 800
 801                 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
 802                         if (sk->sk_family == PF_INET)
 803                                 goto found;
 804         }
 805         sk = NULL;
 806 found:
 807         return sk;
 808 }
 809
 810 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 811 {
 812         struct raw_iter_state* state = raw_seq_private(seq);
 813
 814         do {
 815                 sk = sk_next(sk);
 816 try_again:
 817                 ;
 818         } while (sk && sk->sk_family != PF_INET);
 819
 820         if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
 821                 sk = sk_head(&raw_v4_htable[state->bucket]);
 822                 goto try_again;
 823         }
 824         return sk;
 825 }
 826
 827 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 828 {
 829         struct sock *sk = raw_get_first(seq);
 830
 831         if (sk)
 832                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 833                         --pos;
 834         return pos ? NULL : sk;
 835 }
 836
 837 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 838 {
 839         read_lock(&raw_v4_lock);
 840         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 841 }
 842
 843 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 844 {
 845         struct sock *sk;
 846
 847         if (v == SEQ_START_TOKEN)
 848                 sk = raw_get_first(seq);
 849         else
 850                 sk = raw_get_next(seq, v);
 851         ++*pos;
 852         return sk;
 853 }
 854
 855 static void raw_seq_stop(struct seq_file *seq, void *v)
 856 {
 857         read_unlock(&raw_v4_lock);
 858 }
 859
 860 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
 861 {
 862         struct inet_sock *inet = inet_sk(sp);
 863         __be32 dest = inet->daddr,
 864                src = inet->rcv_saddr;
 865         __u16 destp = 0,
 866               srcp  = inet->num;
 867
 868         sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
 869                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
 870                 i, src, srcp, dest, destp, sp->sk_state,
 871                 atomic_read(&sp->sk_wmem_alloc),
 872                 atomic_read(&sp->sk_rmem_alloc),
 873                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 874                 atomic_read(&sp->sk_refcnt), sp);
 875         return tmpbuf;
 876 }
 877
 878 static int raw_seq_show(struct seq_file *seq, void *v)
 879 {
 880         char tmpbuf[129];
 881
 882         if (v == SEQ_START_TOKEN)
 883                 seq_printf(seq, "%-127s\n",
 884                                "  sl  local_address rem_address   st tx_queue "
 885                                "rx_queue tr tm->when retrnsmt   uid  timeout "
 886                                "inode");
 887         else {
 888                 struct raw_iter_state *state = raw_seq_private(seq);
 889
 890                 seq_printf(seq, "%-127s\n",
 891                            get_raw_sock(v, tmpbuf, state->bucket));
 892         }
 893         return 0;
 894 }
 895
 896 static const struct seq_operations raw_seq_ops = {
 897         .start = raw_seq_start,
 898         .next  = raw_seq_next,
 899         .stop  = raw_seq_stop,
 900         .show  = raw_seq_show,
 901 };
 902
 903 static int raw_seq_open(struct inode *inode, struct file *file)
 904 {
 905         struct seq_file *seq;
 906         int rc = -ENOMEM;
 907         struct raw_iter_state *s;
 908
 909         s = kzalloc(sizeof(*s), GFP_KERNEL);
 910         if (!s)
 911                 goto out;
 912         rc = seq_open(file, &raw_seq_ops);
 913         if (rc)
 914                 goto out_kfree;
 915
 916         seq = file->private_data;
 917         seq->private = s;
 918 out:
 919         return rc;
 920 out_kfree:
 921         kfree(s);
 922         goto out;
 923 }
 924
 925 static const struct file_operations raw_seq_fops = {
 926         .owner   = THIS_MODULE,
 927         .open    = raw_seq_open,
 928         .read    = seq_read,
 929         .llseek  = seq_lseek,
 930         .release = seq_release_private,
 931 };
 932
 933 int __init raw_proc_init(void)
 934 {
 935         if (!proc_net_fops_create(&init_net, "raw", S_IRUGO, &raw_seq_fops))
 936                 return -ENOMEM;
 937         return 0;
 938 }
 939
 940 void __init raw_proc_exit(void)
 941 {
 942         proc_net_remove(&init_net, "raw");
 943 }
 944 #endif /* CONFIG_PROC_FS */