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  * Authors:     Ross Biro
   9  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  10  *
  11  * Fixes:
  12  *              Alan Cox        :       verify_area() fixed up
  13  *              Alan Cox        :       ICMP error handling
  14  *              Alan Cox        :       EMSGSIZE if you send too big a packet
  15  *              Alan Cox        :       Now uses generic datagrams and shared
  16  *                                      skbuff library. No more peek crashes,
  17  *                                      no more backlogs
  18  *              Alan Cox        :       Checks sk->broadcast.
  19  *              Alan Cox        :       Uses skb_free_datagram/skb_copy_datagram
  20  *              Alan Cox        :       Raw passes ip options too
  21  *              Alan Cox        :       Setsocketopt added
  22  *              Alan Cox        :       Fixed error return for broadcasts
  23  *              Alan Cox        :       Removed wake_up calls
  24  *              Alan Cox        :       Use ttl/tos
  25  *              Alan Cox        :       Cleaned up old debugging
  26  *              Alan Cox        :       Use new kernel side addresses
  27  *      Arnt Gulbrandsen        :       Fixed MSG_DONTROUTE in raw sockets.
  28  *              Alan Cox        :       BSD style RAW socket demultiplexing.
  29  *              Alan Cox        :       Beginnings of mrouted support.
  30  *              Alan Cox        :       Added IP_HDRINCL option.
  31  *              Alan Cox        :       Skip broadcast check if BSDism set.
  32  *              David S. Miller :       New socket lookup architecture.
  33  *
  34  *              This program is free software; you can redistribute it and/or
  35  *              modify it under the terms of the GNU General Public License
  36  *              as published by the Free Software Foundation; either version
  37  *              2 of the License, or (at your option) any later version.
  38  */
  39
  40 #include <linux/types.h>
  41 #include <asm/atomic.h>
  42 #include <asm/byteorder.h>
  43 #include <asm/current.h>
  44 #include <asm/uaccess.h>
  45 #include <asm/ioctls.h>
  46 #include <linux/stddef.h>
  47 #include <linux/slab.h>
  48 #include <linux/errno.h>
  49 #include <linux/aio.h>
  50 #include <linux/kernel.h>
  51 #include <linux/spinlock.h>
  52 #include <linux/sockios.h>
  53 #include <linux/socket.h>
  54 #include <linux/in.h>
  55 #include <linux/mroute.h>
  56 #include <linux/netdevice.h>
  57 #include <linux/in_route.h>
  58 #include <linux/route.h>
  59 #include <linux/skbuff.h>
  60 #include <net/net_namespace.h>
  61 #include <net/dst.h>
  62 #include <net/sock.h>
  63 #include <linux/ip.h>
  64 #include <linux/net.h>
  65 #include <net/ip.h>
  66 #include <net/icmp.h>
  67 #include <net/udp.h>
  68 #include <net/raw.h>
  69 #include <net/snmp.h>
  70 #include <net/tcp_states.h>
  71 #include <net/inet_common.h>
  72 #include <net/checksum.h>
  73 #include <net/xfrm.h>
  74 #include <linux/rtnetlink.h>
  75 #include <linux/proc_fs.h>
  76 #include <linux/seq_file.h>
  77 #include <linux/netfilter.h>
  78 #include <linux/netfilter_ipv4.h>
  79 #include <linux/compat.h>
  80
  81 static struct raw_hashinfo raw_v4_hashinfo = {
  82         .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
  83 };
  84
  85 void raw_hash_sk(struct sock *sk)
  86 {
  87         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
  88         struct hlist_head *head;
  89
  90         head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
  91
  92         write_lock_bh(&h->lock);
  93         sk_add_node(sk, head);
  94         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
  95         write_unlock_bh(&h->lock);
  96 }
  97 EXPORT_SYMBOL_GPL(raw_hash_sk);
  98
  99 void raw_unhash_sk(struct sock *sk)
 100 {
 101         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
 102
 103         write_lock_bh(&h->lock);
 104         if (sk_del_node_init(sk))
 105                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 106         write_unlock_bh(&h->lock);
 107 }
 108 EXPORT_SYMBOL_GPL(raw_unhash_sk);
 109
 110 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
 111                 unsigned short num, __be32 raddr, __be32 laddr, int dif)
 112 {
 113         struct hlist_node *node;
 114
 115         sk_for_each_from(sk, node) {
 116                 struct inet_sock *inet = inet_sk(sk);
 117
 118                 if (net_eq(sock_net(sk), net) && inet->inet_num == num  &&
 119                     !(inet->inet_daddr && inet->inet_daddr != raddr)    &&
 120                     !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
 121                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 122                         goto found; /* gotcha */
 123         }
 124         sk = NULL;
 125 found:
 126         return sk;
 127 }
 128
 129 /*
 130  *      0 - deliver
 131  *      1 - block
 132  */
 133 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 134 {
 135         int type;
 136
 137         if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
 138                 return 1;
 139
 140         type = icmp_hdr(skb)->type;
 141         if (type < 32) {
 142                 __u32 data = raw_sk(sk)->filter.data;
 143
 144                 return ((1 << type) & data) != 0;
 145         }
 146
 147         /* Do not block unknown ICMP types */
 148         return 0;
 149 }
 150
 151 /* IP input processing comes here for RAW socket delivery.
 152  * Caller owns SKB, so we must make clones.
 153  *
 154  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 155  * -> It does. And not only TOS, but all IP header.
 156  */
 157 static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 158 {
 159         struct sock *sk;
 160         struct hlist_head *head;
 161         int delivered = 0;
 162         struct net *net;
 163
 164         read_lock(&raw_v4_hashinfo.lock);
 165         head = &raw_v4_hashinfo.ht[hash];
 166         if (hlist_empty(head))
 167                 goto out;
 168
 169         net = dev_net(skb->dev);
 170         sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
 171                              iph->saddr, iph->daddr,
 172                              skb->dev->ifindex);
 173
 174         while (sk) {
 175                 delivered = 1;
 176                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 177                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 178
 179                         /* Not releasing hash table! */
 180                         if (clone)
 181                                 raw_rcv(sk, clone);
 182                 }
 183                 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
 184                                      iph->saddr, iph->daddr,
 185                                      skb->dev->ifindex);
 186         }
 187 out:
 188         read_unlock(&raw_v4_hashinfo.lock);
 189         return delivered;
 190 }
 191
 192 int raw_local_deliver(struct sk_buff *skb, int protocol)
 193 {
 194         int hash;
 195         struct sock *raw_sk;
 196
 197         hash = protocol & (RAW_HTABLE_SIZE - 1);
 198         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 199
 200         /* If there maybe a raw socket we must check - if not we
 201          * don't care less
 202          */
 203         if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
 204                 raw_sk = NULL;
 205
 206         return raw_sk != NULL;
 207
 208 }
 209
 210 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
 211 {
 212         struct inet_sock *inet = inet_sk(sk);
 213         const int type = icmp_hdr(skb)->type;
 214         const int code = icmp_hdr(skb)->code;
 215         int err = 0;
 216         int harderr = 0;
 217
 218         /* Report error on raw socket, if:
 219            1. User requested ip_recverr.
 220            2. Socket is connected (otherwise the error indication
 221               is useless without ip_recverr and error is hard.
 222          */
 223         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 224                 return;
 225
 226         switch (type) {
 227         default:
 228         case ICMP_TIME_EXCEEDED:
 229                 err = EHOSTUNREACH;
 230                 break;
 231         case ICMP_SOURCE_QUENCH:
 232                 return;
 233         case ICMP_PARAMETERPROB:
 234                 err = EPROTO;
 235                 harderr = 1;
 236                 break;
 237         case ICMP_DEST_UNREACH:
 238                 err = EHOSTUNREACH;
 239                 if (code > NR_ICMP_UNREACH)
 240                         break;
 241                 err = icmp_err_convert[code].errno;
 242                 harderr = icmp_err_convert[code].fatal;
 243                 if (code == ICMP_FRAG_NEEDED) {
 244                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 245                         err = EMSGSIZE;
 246                 }
 247         }
 248
 249         if (inet->recverr) {
 250                 struct iphdr *iph = (struct iphdr *)skb->data;
 251                 u8 *payload = skb->data + (iph->ihl << 2);
 252
 253                 if (inet->hdrincl)
 254                         payload = skb->data;
 255                 ip_icmp_error(sk, skb, err, 0, info, payload);
 256         }
 257
 258         if (inet->recverr || harderr) {
 259                 sk->sk_err = err;
 260                 sk->sk_error_report(sk);
 261         }
 262 }
 263
 264 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
 265 {
 266         int hash;
 267         struct sock *raw_sk;
 268         struct iphdr *iph;
 269         struct net *net;
 270
 271         hash = protocol & (RAW_HTABLE_SIZE - 1);
 272
 273         read_lock(&raw_v4_hashinfo.lock);
 274         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 275         if (raw_sk != NULL) {
 276                 iph = (struct iphdr *)skb->data;
 277                 net = dev_net(skb->dev);
 278
 279                 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
 280                                                 iph->daddr, iph->saddr,
 281                                                 skb->dev->ifindex)) != NULL) {
 282                         raw_err(raw_sk, skb, info);
 283                         raw_sk = sk_next(raw_sk);
 284                         iph = (struct iphdr *)skb->data;
 285                 }
 286         }
 287         read_unlock(&raw_v4_hashinfo.lock);
 288 }
 289
 290 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 291 {
 292         /* Charge it to the socket. */
 293
 294         if (ip_queue_rcv_skb(sk, skb) < 0) {
 295                 kfree_skb(skb);
 296                 return NET_RX_DROP;
 297         }
 298
 299         return NET_RX_SUCCESS;
 300 }
 301
 302 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 303 {
 304         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 305                 atomic_inc(&sk->sk_drops);
 306                 kfree_skb(skb);
 307                 return NET_RX_DROP;
 308         }
 309         nf_reset(skb);
 310
 311         skb_push(skb, skb->data - skb_network_header(skb));
 312
 313         raw_rcv_skb(sk, skb);
 314         return 0;
 315 }
 316
 317 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 318                         struct rtable **rtp,
 319                         unsigned int flags)
 320 {
 321         struct inet_sock *inet = inet_sk(sk);
 322         struct net *net = sock_net(sk);
 323         struct iphdr *iph;
 324         struct sk_buff *skb;
 325         unsigned int iphlen;
 326         int err;
 327         struct rtable *rt = *rtp;
 328
 329         if (length > rt->dst.dev->mtu) {
 330                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->inet_dport,
 331                                rt->dst.dev->mtu);
 332                 return -EMSGSIZE;
 333         }
 334         if (flags&MSG_PROBE)
 335                 goto out;
 336
 337         skb = sock_alloc_send_skb(sk,
 338                                   length + LL_ALLOCATED_SPACE(rt->dst.dev) + 15,
 339                                   flags & MSG_DONTWAIT, &err);
 340         if (skb == NULL)
 341                 goto error;
 342         skb_reserve(skb, LL_RESERVED_SPACE(rt->dst.dev));
 343
 344         skb->priority = sk->sk_priority;
 345         skb->mark = sk->sk_mark;
 346         skb_dst_set(skb, &rt->dst);
 347         *rtp = NULL;
 348
 349         skb_reset_network_header(skb);
 350         iph = ip_hdr(skb);
 351         skb_put(skb, length);
 352
 353         skb->ip_summed = CHECKSUM_NONE;
 354
 355         skb->transport_header = skb->network_header;
 356         err = -EFAULT;
 357         if (memcpy_fromiovecend((void *)iph, from, 0, length))
 358                 goto error_free;
 359
 360         iphlen = iph->ihl * 4;
 361
 362         /*
 363          * We don't want to modify the ip header, but we do need to
 364          * be sure that it won't cause problems later along the network
 365          * stack.  Specifically we want to make sure that iph->ihl is a
 366          * sane value.  If ihl points beyond the length of the buffer passed
 367          * in, reject the frame as invalid
 368          */
 369         err = -EINVAL;
 370         if (iphlen > length)
 371                 goto error_free;
 372
 373         if (iphlen >= sizeof(*iph)) {
 374                 if (!iph->saddr)
 375                         iph->saddr = rt->rt_src;
 376                 iph->check   = 0;
 377                 iph->tot_len = htons(length);
 378                 if (!iph->id)
 379                         ip_select_ident(iph, &rt->dst, NULL);
 380
 381                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 382         }
 383         if (iph->protocol == IPPROTO_ICMP)
 384                 icmp_out_count(net, ((struct icmphdr *)
 385                         skb_transport_header(skb))->type);
 386
 387         err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
 388                       rt->dst.dev, dst_output);
 389         if (err > 0)
 390                 err = net_xmit_errno(err);
 391         if (err)
 392                 goto error;
 393 out:
 394         return 0;
 395
 396 error_free:
 397         kfree_skb(skb);
 398 error:
 399         IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
 400         if (err == -ENOBUFS && !inet->recverr)
 401                 err = 0;
 402         return err;
 403 }
 404
 405 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 406 {
 407         struct iovec *iov;
 408         u8 __user *type = NULL;
 409         u8 __user *code = NULL;
 410         int probed = 0;
 411         unsigned int i;
 412
 413         if (!msg->msg_iov)
 414                 return 0;
 415
 416         for (i = 0; i < msg->msg_iovlen; i++) {
 417                 iov = &msg->msg_iov[i];
 418                 if (!iov)
 419                         continue;
 420
 421                 switch (fl->proto) {
 422                 case IPPROTO_ICMP:
 423                         /* check if one-byte field is readable or not. */
 424                         if (iov->iov_base && iov->iov_len < 1)
 425                                 break;
 426
 427                         if (!type) {
 428                                 type = iov->iov_base;
 429                                 /* check if code field is readable or not. */
 430                                 if (iov->iov_len > 1)
 431                                         code = type + 1;
 432                         } else if (!code)
 433                                 code = iov->iov_base;
 434
 435                         if (type && code) {
 436                                 if (get_user(fl->fl_icmp_type, type) ||
 437                                     get_user(fl->fl_icmp_code, code))
 438                                         return -EFAULT;
 439                                 probed = 1;
 440                         }
 441                         break;
 442                 default:
 443                         probed = 1;
 444                         break;
 445                 }
 446                 if (probed)
 447                         break;
 448         }
 449         return 0;
 450 }
 451
 452 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 453                        size_t len)
 454 {
 455         struct inet_sock *inet = inet_sk(sk);
 456         struct ipcm_cookie ipc;
 457         struct rtable *rt = NULL;
 458         int free = 0;
 459         __be32 daddr;
 460         __be32 saddr;
 461         u8  tos;
 462         int err;
 463
 464         err = -EMSGSIZE;
 465         if (len > 0xFFFF)
 466                 goto out;
 467
 468         /*
 469          *      Check the flags.
 470          */
 471
 472         err = -EOPNOTSUPP;
 473         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 474                 goto out;               /* compatibility */
 475
 476         /*
 477          *      Get and verify the address.
 478          */
 479
 480         if (msg->msg_namelen) {
 481                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
 482                 err = -EINVAL;
 483                 if (msg->msg_namelen < sizeof(*usin))
 484                         goto out;
 485                 if (usin->sin_family != AF_INET) {
 486                         static int complained;
 487                         if (!complained++)
 488                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 489                                                  "raw sendmsg. Fix it!\n",
 490                                                  current->comm);
 491                         err = -EAFNOSUPPORT;
 492                         if (usin->sin_family)
 493                                 goto out;
 494                 }
 495                 daddr = usin->sin_addr.s_addr;
 496                 /* ANK: I did not forget to get protocol from port field.
 497                  * I just do not know, who uses this weirdness.
 498                  * IP_HDRINCL is much more convenient.
 499                  */
 500         } else {
 501                 err = -EDESTADDRREQ;
 502                 if (sk->sk_state != TCP_ESTABLISHED)
 503                         goto out;
 504                 daddr = inet->inet_daddr;
 505         }
 506
 507         ipc.addr = inet->inet_saddr;
 508         ipc.opt = NULL;
 509         ipc.tx_flags = 0;
 510         ipc.oif = sk->sk_bound_dev_if;
 511
 512         if (msg->msg_controllen) {
 513                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
 514                 if (err)
 515                         goto out;
 516                 if (ipc.opt)
 517                         free = 1;
 518         }
 519
 520         saddr = ipc.addr;
 521         ipc.addr = daddr;
 522
 523         if (!ipc.opt)
 524                 ipc.opt = inet->opt;
 525
 526         if (ipc.opt) {
 527                 err = -EINVAL;
 528                 /* Linux does not mangle headers on raw sockets,
 529                  * so that IP options + IP_HDRINCL is non-sense.
 530                  */
 531                 if (inet->hdrincl)
 532                         goto done;
 533                 if (ipc.opt->srr) {
 534                         if (!daddr)
 535                                 goto done;
 536                         daddr = ipc.opt->faddr;
 537                 }
 538         }
 539         tos = RT_CONN_FLAGS(sk);
 540         if (msg->msg_flags & MSG_DONTROUTE)
 541                 tos |= RTO_ONLINK;
 542
 543         if (ipv4_is_multicast(daddr)) {
 544                 if (!ipc.oif)
 545                         ipc.oif = inet->mc_index;
 546                 if (!saddr)
 547                         saddr = inet->mc_addr;
 548         }
 549
 550         {
 551                 struct flowi fl = { .oif = ipc.oif,
 552                                     .mark = sk->sk_mark,
 553                                     .fl4_dst = daddr,
 554                                     .fl4_src = saddr,
 555                                     .fl4_tos = tos,
 556                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 557                                                              sk->sk_protocol,
 558                                     .flags = FLOWI_FLAG_CAN_SLEEP,
 559                 };
 560                 if (!inet->hdrincl) {
 561                         err = raw_probe_proto_opt(&fl, msg);
 562                         if (err)
 563                                 goto done;
 564                 }
 565
 566                 security_sk_classify_flow(sk, &fl);
 567                 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, true);
 568         }
 569         if (err)
 570                 goto done;
 571
 572         err = -EACCES;
 573         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 574                 goto done;
 575
 576         if (msg->msg_flags & MSG_CONFIRM)
 577                 goto do_confirm;
 578 back_from_confirm:
 579
 580         if (inet->hdrincl)
 581                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 582                                         &rt, msg->msg_flags);
 583
 584          else {
 585                 if (!ipc.addr)
 586                         ipc.addr = rt->rt_dst;
 587                 lock_sock(sk);
 588                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 589                                         &ipc, &rt, msg->msg_flags);
 590                 if (err)
 591                         ip_flush_pending_frames(sk);
 592                 else if (!(msg->msg_flags & MSG_MORE)) {
 593                         err = ip_push_pending_frames(sk);
 594                         if (err == -ENOBUFS && !inet->recverr)
 595                                 err = 0;
 596                 }
 597                 release_sock(sk);
 598         }
 599 done:
 600         if (free)
 601                 kfree(ipc.opt);
 602         ip_rt_put(rt);
 603
 604 out:
 605         if (err < 0)
 606                 return err;
 607         return len;
 608
 609 do_confirm:
 610         dst_confirm(&rt->dst);
 611         if (!(msg->msg_flags & MSG_PROBE) || len)
 612                 goto back_from_confirm;
 613         err = 0;
 614         goto done;
 615 }
 616
 617 static void raw_close(struct sock *sk, long timeout)
 618 {
 619         /*
 620          * Raw sockets may have direct kernel refereneces. Kill them.
 621          */
 622         ip_ra_control(sk, 0, NULL);
 623
 624         sk_common_release(sk);
 625 }
 626
 627 static void raw_destroy(struct sock *sk)
 628 {
 629         lock_sock(sk);
 630         ip_flush_pending_frames(sk);
 631         release_sock(sk);
 632 }
 633
 634 /* This gets rid of all the nasties in af_inet. -DaveM */
 635 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 636 {
 637         struct inet_sock *inet = inet_sk(sk);
 638         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 639         int ret = -EINVAL;
 640         int chk_addr_ret;
 641
 642         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 643                 goto out;
 644         chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
 645         ret = -EADDRNOTAVAIL;
 646         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 647             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 648                 goto out;
 649         inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
 650         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 651                 inet->inet_saddr = 0;  /* Use device */
 652         sk_dst_reset(sk);
 653         ret = 0;
 654 out:    return ret;
 655 }
 656
 657 /*
 658  *      This should be easy, if there is something there
 659  *      we return it, otherwise we block.
 660  */
 661
 662 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 663                        size_t len, int noblock, int flags, int *addr_len)
 664 {
 665         struct inet_sock *inet = inet_sk(sk);
 666         size_t copied = 0;
 667         int err = -EOPNOTSUPP;
 668         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 669         struct sk_buff *skb;
 670
 671         if (flags & MSG_OOB)
 672                 goto out;
 673
 674         if (addr_len)
 675                 *addr_len = sizeof(*sin);
 676
 677         if (flags & MSG_ERRQUEUE) {
 678                 err = ip_recv_error(sk, msg, len);
 679                 goto out;
 680         }
 681
 682         skb = skb_recv_datagram(sk, flags, noblock, &err);
 683         if (!skb)
 684                 goto out;
 685
 686         copied = skb->len;
 687         if (len < copied) {
 688                 msg->msg_flags |= MSG_TRUNC;
 689                 copied = len;
 690         }
 691
 692         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 693         if (err)
 694                 goto done;
 695
 696         sock_recv_ts_and_drops(msg, sk, skb);
 697
 698         /* Copy the address. */
 699         if (sin) {
 700                 sin->sin_family = AF_INET;
 701                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 702                 sin->sin_port = 0;
 703                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 704         }
 705         if (inet->cmsg_flags)
 706                 ip_cmsg_recv(msg, skb);
 707         if (flags & MSG_TRUNC)
 708                 copied = skb->len;
 709 done:
 710         skb_free_datagram(sk, skb);
 711 out:
 712         if (err)
 713                 return err;
 714         return copied;
 715 }
 716
 717 static int raw_init(struct sock *sk)
 718 {
 719         struct raw_sock *rp = raw_sk(sk);
 720
 721         if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
 722                 memset(&rp->filter, 0, sizeof(rp->filter));
 723         return 0;
 724 }
 725
 726 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 727 {
 728         if (optlen > sizeof(struct icmp_filter))
 729                 optlen = sizeof(struct icmp_filter);
 730         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 731                 return -EFAULT;
 732         return 0;
 733 }
 734
 735 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 736 {
 737         int len, ret = -EFAULT;
 738
 739         if (get_user(len, optlen))
 740                 goto out;
 741         ret = -EINVAL;
 742         if (len < 0)
 743                 goto out;
 744         if (len > sizeof(struct icmp_filter))
 745                 len = sizeof(struct icmp_filter);
 746         ret = -EFAULT;
 747         if (put_user(len, optlen) ||
 748             copy_to_user(optval, &raw_sk(sk)->filter, len))
 749                 goto out;
 750         ret = 0;
 751 out:    return ret;
 752 }
 753
 754 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 755                           char __user *optval, unsigned int optlen)
 756 {
 757         if (optname == ICMP_FILTER) {
 758                 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
 759                         return -EOPNOTSUPP;
 760                 else
 761                         return raw_seticmpfilter(sk, optval, optlen);
 762         }
 763         return -ENOPROTOOPT;
 764 }
 765
 766 static int raw_setsockopt(struct sock *sk, int level, int optname,
 767                           char __user *optval, unsigned int optlen)
 768 {
 769         if (level != SOL_RAW)
 770                 return ip_setsockopt(sk, level, optname, optval, optlen);
 771         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 772 }
 773
 774 #ifdef CONFIG_COMPAT
 775 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 776                                  char __user *optval, unsigned int optlen)
 777 {
 778         if (level != SOL_RAW)
 779                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 780         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 781 }
 782 #endif
 783
 784 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 785                           char __user *optval, int __user *optlen)
 786 {
 787         if (optname == ICMP_FILTER) {
 788                 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
 789                         return -EOPNOTSUPP;
 790                 else
 791                         return raw_geticmpfilter(sk, optval, optlen);
 792         }
 793         return -ENOPROTOOPT;
 794 }
 795
 796 static int raw_getsockopt(struct sock *sk, int level, int optname,
 797                           char __user *optval, int __user *optlen)
 798 {
 799         if (level != SOL_RAW)
 800                 return ip_getsockopt(sk, level, optname, optval, optlen);
 801         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 802 }
 803
 804 #ifdef CONFIG_COMPAT
 805 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 806                                  char __user *optval, int __user *optlen)
 807 {
 808         if (level != SOL_RAW)
 809                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 810         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 811 }
 812 #endif
 813
 814 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 815 {
 816         switch (cmd) {
 817                 case SIOCOUTQ: {
 818                         int amount = sk_wmem_alloc_get(sk);
 819
 820                         return put_user(amount, (int __user *)arg);
 821                 }
 822                 case SIOCINQ: {
 823                         struct sk_buff *skb;
 824                         int amount = 0;
 825
 826                         spin_lock_bh(&sk->sk_receive_queue.lock);
 827                         skb = skb_peek(&sk->sk_receive_queue);
 828                         if (skb != NULL)
 829                                 amount = skb->len;
 830                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 831                         return put_user(amount, (int __user *)arg);
 832                 }
 833
 834                 default:
 835 #ifdef CONFIG_IP_MROUTE
 836                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 837 #else
 838                         return -ENOIOCTLCMD;
 839 #endif
 840         }
 841 }
 842
 843 #ifdef CONFIG_COMPAT
 844 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
 845 {
 846         switch (cmd) {
 847         case SIOCOUTQ:
 848         case SIOCINQ:
 849                 return -ENOIOCTLCMD;
 850         default:
 851 #ifdef CONFIG_IP_MROUTE
 852                 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
 853 #else
 854                 return -ENOIOCTLCMD;
 855 #endif
 856         }
 857 }
 858 #endif
 859
 860 struct proto raw_prot = {
 861         .name              = "RAW",
 862         .owner             = THIS_MODULE,
 863         .close             = raw_close,
 864         .destroy           = raw_destroy,
 865         .connect           = ip4_datagram_connect,
 866         .disconnect        = udp_disconnect,
 867         .ioctl             = raw_ioctl,
 868         .init              = raw_init,
 869         .setsockopt        = raw_setsockopt,
 870         .getsockopt        = raw_getsockopt,
 871         .sendmsg           = raw_sendmsg,
 872         .recvmsg           = raw_recvmsg,
 873         .bind              = raw_bind,
 874         .backlog_rcv       = raw_rcv_skb,
 875         .hash              = raw_hash_sk,
 876         .unhash            = raw_unhash_sk,
 877         .obj_size          = sizeof(struct raw_sock),
 878         .h.raw_hash        = &raw_v4_hashinfo,
 879 #ifdef CONFIG_COMPAT
 880         .compat_setsockopt = compat_raw_setsockopt,
 881         .compat_getsockopt = compat_raw_getsockopt,
 882         .compat_ioctl      = compat_raw_ioctl,
 883 #endif
 884 };
 885
 886 #ifdef CONFIG_PROC_FS
 887 static struct sock *raw_get_first(struct seq_file *seq)
 888 {
 889         struct sock *sk;
 890         struct raw_iter_state *state = raw_seq_private(seq);
 891
 892         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
 893                         ++state->bucket) {
 894                 struct hlist_node *node;
 895
 896                 sk_for_each(sk, node, &state->h->ht[state->bucket])
 897                         if (sock_net(sk) == seq_file_net(seq))
 898                                 goto found;
 899         }
 900         sk = NULL;
 901 found:
 902         return sk;
 903 }
 904
 905 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 906 {
 907         struct raw_iter_state *state = raw_seq_private(seq);
 908
 909         do {
 910                 sk = sk_next(sk);
 911 try_again:
 912                 ;
 913         } while (sk && sock_net(sk) != seq_file_net(seq));
 914
 915         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
 916                 sk = sk_head(&state->h->ht[state->bucket]);
 917                 goto try_again;
 918         }
 919         return sk;
 920 }
 921
 922 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 923 {
 924         struct sock *sk = raw_get_first(seq);
 925
 926         if (sk)
 927                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 928                         --pos;
 929         return pos ? NULL : sk;
 930 }
 931
 932 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 933 {
 934         struct raw_iter_state *state = raw_seq_private(seq);
 935
 936         read_lock(&state->h->lock);
 937         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 938 }
 939 EXPORT_SYMBOL_GPL(raw_seq_start);
 940
 941 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 942 {
 943         struct sock *sk;
 944
 945         if (v == SEQ_START_TOKEN)
 946                 sk = raw_get_first(seq);
 947         else
 948                 sk = raw_get_next(seq, v);
 949         ++*pos;
 950         return sk;
 951 }
 952 EXPORT_SYMBOL_GPL(raw_seq_next);
 953
 954 void raw_seq_stop(struct seq_file *seq, void *v)
 955 {
 956         struct raw_iter_state *state = raw_seq_private(seq);
 957
 958         read_unlock(&state->h->lock);
 959 }
 960 EXPORT_SYMBOL_GPL(raw_seq_stop);
 961
 962 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
 963 {
 964         struct inet_sock *inet = inet_sk(sp);
 965         __be32 dest = inet->inet_daddr,
 966                src = inet->inet_rcv_saddr;
 967         __u16 destp = 0,
 968               srcp  = inet->inet_num;
 969
 970         seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
 971                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
 972                 i, src, srcp, dest, destp, sp->sk_state,
 973                 sk_wmem_alloc_get(sp),
 974                 sk_rmem_alloc_get(sp),
 975                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 976                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
 977 }
 978
 979 static int raw_seq_show(struct seq_file *seq, void *v)
 980 {
 981         if (v == SEQ_START_TOKEN)
 982                 seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
 983                                 "rx_queue tr tm->when retrnsmt   uid  timeout "
 984                                 "inode ref pointer drops\n");
 985         else
 986                 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
 987         return 0;
 988 }
 989
 990 static const struct seq_operations raw_seq_ops = {
 991         .start = raw_seq_start,
 992         .next  = raw_seq_next,
 993         .stop  = raw_seq_stop,
 994         .show  = raw_seq_show,
 995 };
 996
 997 int raw_seq_open(struct inode *ino, struct file *file,
 998                  struct raw_hashinfo *h, const struct seq_operations *ops)
 999 {
1000         int err;
1001         struct raw_iter_state *i;
1002
1003         err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1004         if (err < 0)
1005                 return err;
1006
1007         i = raw_seq_private((struct seq_file *)file->private_data);
1008         i->h = h;
1009         return 0;
1010 }
1011 EXPORT_SYMBOL_GPL(raw_seq_open);
1012
1013 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1014 {
1015         return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1016 }
1017
1018 static const struct file_operations raw_seq_fops = {
1019         .owner   = THIS_MODULE,
1020         .open    = raw_v4_seq_open,
1021         .read    = seq_read,
1022         .llseek  = seq_lseek,
1023         .release = seq_release_net,
1024 };
1025
1026 static __net_init int raw_init_net(struct net *net)
1027 {
1028         if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
1029                 return -ENOMEM;
1030
1031         return 0;
1032 }
1033
1034 static __net_exit void raw_exit_net(struct net *net)
1035 {
1036         proc_net_remove(net, "raw");
1037 }
1038
1039 static __net_initdata struct pernet_operations raw_net_ops = {
1040         .init = raw_init_net,
1041         .exit = raw_exit_net,
1042 };
1043
1044 int __init raw_proc_init(void)
1045 {
1046         return register_pernet_subsys(&raw_net_ops);
1047 }
1048
1049 void __init raw_proc_exit(void)
1050 {
1051         unregister_pernet_subsys(&raw_net_ops);
1052 }
1053 #endif /* CONFIG_PROC_FS */