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/gfp.h>
  64 #include <linux/ip.h>
  65 #include <linux/net.h>
  66 #include <net/ip.h>
  67 #include <net/icmp.h>
  68 #include <net/udp.h>
  69 #include <net/raw.h>
  70 #include <net/snmp.h>
  71 #include <net/tcp_states.h>
  72 #include <net/inet_common.h>
  73 #include <net/checksum.h>
  74 #include <net/xfrm.h>
  75 #include <linux/rtnetlink.h>
  76 #include <linux/proc_fs.h>
  77 #include <linux/seq_file.h>
  78 #include <linux/netfilter.h>
  79 #include <linux/netfilter_ipv4.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)->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->num == num       &&
 119                     !(inet->daddr && inet->daddr != raddr)              &&
 120                     !(inet->rcv_saddr && 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 (sock_queue_rcv_skb(sk, skb) < 0) {
 295                 atomic_inc(&sk->sk_drops);
 296                 kfree_skb(skb);
 297                 return NET_RX_DROP;
 298         }
 299
 300         return NET_RX_SUCCESS;
 301 }
 302
 303 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 304 {
 305         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 306                 atomic_inc(&sk->sk_drops);
 307                 kfree_skb(skb);
 308                 return NET_RX_DROP;
 309         }
 310         nf_reset(skb);
 311
 312         skb_push(skb, skb->data - skb_network_header(skb));
 313
 314         raw_rcv_skb(sk, skb);
 315         return 0;
 316 }
 317
 318 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 319                         struct rtable *rt,
 320                         unsigned int flags)
 321 {
 322         struct inet_sock *inet = inet_sk(sk);
 323         struct net *net = sock_net(sk);
 324         struct iphdr *iph;
 325         struct sk_buff *skb;
 326         unsigned int iphlen;
 327         int err;
 328
 329         if (length > rt->u.dst.dev->mtu) {
 330                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
 331                                rt->u.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->u.dst.dev) + 15,
 339                                   flags & MSG_DONTWAIT, &err);
 340         if (skb == NULL)
 341                 goto error;
 342         skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
 343
 344         skb->priority = sk->sk_priority;
 345         skb->mark = sk->sk_mark;
 346         skb_dst_set(skb, dst_clone(&rt->u.dst));
 347
 348         skb_reset_network_header(skb);
 349         iph = ip_hdr(skb);
 350         skb_put(skb, length);
 351
 352         skb->ip_summed = CHECKSUM_NONE;
 353
 354         skb->transport_header = skb->network_header;
 355         err = memcpy_fromiovecend((void *)iph, from, 0, length);
 356         if (err)
 357                 goto error_fault;
 358
 359         /* We don't modify invalid header */
 360         iphlen = iph->ihl * 4;
 361         if (iphlen >= sizeof(*iph) && iphlen <= length) {
 362                 if (!iph->saddr)
 363                         iph->saddr = rt->rt_src;
 364                 iph->check   = 0;
 365                 iph->tot_len = htons(length);
 366                 if (!iph->id)
 367                         ip_select_ident(iph, &rt->u.dst, NULL);
 368
 369                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 370         }
 371         if (iph->protocol == IPPROTO_ICMP)
 372                 icmp_out_count(net, ((struct icmphdr *)
 373                         skb_transport_header(skb))->type);
 374
 375         err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 376                       dst_output);
 377         if (err > 0)
 378                 err = inet->recverr ? net_xmit_errno(err) : 0;
 379         if (err)
 380                 goto error;
 381 out:
 382         return 0;
 383
 384 error_fault:
 385         err = -EFAULT;
 386         kfree_skb(skb);
 387 error:
 388         IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
 389         return err;
 390 }
 391
 392 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 393 {
 394         struct iovec *iov;
 395         u8 __user *type = NULL;
 396         u8 __user *code = NULL;
 397         int probed = 0;
 398         unsigned int i;
 399
 400         if (!msg->msg_iov)
 401                 return 0;
 402
 403         for (i = 0; i < msg->msg_iovlen; i++) {
 404                 iov = &msg->msg_iov[i];
 405                 if (!iov)
 406                         continue;
 407
 408                 switch (fl->proto) {
 409                 case IPPROTO_ICMP:
 410                         /* check if one-byte field is readable or not. */
 411                         if (iov->iov_base && iov->iov_len < 1)
 412                                 break;
 413
 414                         if (!type) {
 415                                 type = iov->iov_base;
 416                                 /* check if code field is readable or not. */
 417                                 if (iov->iov_len > 1)
 418                                         code = type + 1;
 419                         } else if (!code)
 420                                 code = iov->iov_base;
 421
 422                         if (type && code) {
 423                                 if (get_user(fl->fl_icmp_type, type) ||
 424                                     get_user(fl->fl_icmp_code, code))
 425                                         return -EFAULT;
 426                                 probed = 1;
 427                         }
 428                         break;
 429                 default:
 430                         probed = 1;
 431                         break;
 432                 }
 433                 if (probed)
 434                         break;
 435         }
 436         return 0;
 437 }
 438
 439 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 440                        size_t len)
 441 {
 442         struct inet_sock *inet = inet_sk(sk);
 443         struct ipcm_cookie ipc;
 444         struct rtable *rt = NULL;
 445         int free = 0;
 446         __be32 daddr;
 447         __be32 saddr;
 448         u8  tos;
 449         int err;
 450
 451         err = -EMSGSIZE;
 452         if (len > 0xFFFF)
 453                 goto out;
 454
 455         /*
 456          *      Check the flags.
 457          */
 458
 459         err = -EOPNOTSUPP;
 460         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 461                 goto out;               /* compatibility */
 462
 463         /*
 464          *      Get and verify the address.
 465          */
 466
 467         if (msg->msg_namelen) {
 468                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
 469                 err = -EINVAL;
 470                 if (msg->msg_namelen < sizeof(*usin))
 471                         goto out;
 472                 if (usin->sin_family != AF_INET) {
 473                         static int complained;
 474                         if (!complained++)
 475                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 476                                                  "raw sendmsg. Fix it!\n",
 477                                                  current->comm);
 478                         err = -EAFNOSUPPORT;
 479                         if (usin->sin_family)
 480                                 goto out;
 481                 }
 482                 daddr = usin->sin_addr.s_addr;
 483                 /* ANK: I did not forget to get protocol from port field.
 484                  * I just do not know, who uses this weirdness.
 485                  * IP_HDRINCL is much more convenient.
 486                  */
 487         } else {
 488                 err = -EDESTADDRREQ;
 489                 if (sk->sk_state != TCP_ESTABLISHED)
 490                         goto out;
 491                 daddr = inet->daddr;
 492         }
 493
 494         ipc.addr = inet->saddr;
 495         ipc.opt = NULL;
 496         ipc.shtx.flags = 0;
 497         ipc.oif = sk->sk_bound_dev_if;
 498
 499         if (msg->msg_controllen) {
 500                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
 501                 if (err)
 502                         goto out;
 503                 if (ipc.opt)
 504                         free = 1;
 505         }
 506
 507         saddr = ipc.addr;
 508         ipc.addr = daddr;
 509
 510         if (!ipc.opt)
 511                 ipc.opt = inet->opt;
 512
 513         if (ipc.opt) {
 514                 err = -EINVAL;
 515                 /* Linux does not mangle headers on raw sockets,
 516                  * so that IP options + IP_HDRINCL is non-sense.
 517                  */
 518                 if (inet->hdrincl)
 519                         goto done;
 520                 if (ipc.opt->srr) {
 521                         if (!daddr)
 522                                 goto done;
 523                         daddr = ipc.opt->faddr;
 524                 }
 525         }
 526         tos = RT_CONN_FLAGS(sk);
 527         if (msg->msg_flags & MSG_DONTROUTE)
 528                 tos |= RTO_ONLINK;
 529
 530         if (ipv4_is_multicast(daddr)) {
 531                 if (!ipc.oif)
 532                         ipc.oif = inet->mc_index;
 533                 if (!saddr)
 534                         saddr = inet->mc_addr;
 535         }
 536
 537         {
 538                 struct flowi fl = { .oif = ipc.oif,
 539                                     .mark = sk->sk_mark,
 540                                     .nl_u = { .ip4_u =
 541                                               { .daddr = daddr,
 542                                                 .saddr = saddr,
 543                                                 .tos = tos } },
 544                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 545                                                              sk->sk_protocol,
 546                                   };
 547                 if (!inet->hdrincl) {
 548                         err = raw_probe_proto_opt(&fl, msg);
 549                         if (err)
 550                                 goto done;
 551                 }
 552
 553                 security_sk_classify_flow(sk, &fl);
 554                 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 1);
 555         }
 556         if (err)
 557                 goto done;
 558
 559         err = -EACCES;
 560         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 561                 goto done;
 562
 563         if (msg->msg_flags & MSG_CONFIRM)
 564                 goto do_confirm;
 565 back_from_confirm:
 566
 567         if (inet->hdrincl)
 568                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 569                                         rt, msg->msg_flags);
 570
 571          else {
 572                 if (!ipc.addr)
 573                         ipc.addr = rt->rt_dst;
 574                 lock_sock(sk);
 575                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 576                                         &ipc, &rt, msg->msg_flags);
 577                 if (err)
 578                         ip_flush_pending_frames(sk);
 579                 else if (!(msg->msg_flags & MSG_MORE))
 580                         err = ip_push_pending_frames(sk);
 581                 release_sock(sk);
 582         }
 583 done:
 584         if (free)
 585                 kfree(ipc.opt);
 586         ip_rt_put(rt);
 587
 588 out:
 589         if (err < 0)
 590                 return err;
 591         return len;
 592
 593 do_confirm:
 594         dst_confirm(&rt->u.dst);
 595         if (!(msg->msg_flags & MSG_PROBE) || len)
 596                 goto back_from_confirm;
 597         err = 0;
 598         goto done;
 599 }
 600
 601 static void raw_close(struct sock *sk, long timeout)
 602 {
 603         /*
 604          * Raw sockets may have direct kernel refereneces. Kill them.
 605          */
 606         ip_ra_control(sk, 0, NULL);
 607
 608         sk_common_release(sk);
 609 }
 610
 611 static void raw_destroy(struct sock *sk)
 612 {
 613         lock_sock(sk);
 614         ip_flush_pending_frames(sk);
 615         release_sock(sk);
 616 }
 617
 618 /* This gets rid of all the nasties in af_inet. -DaveM */
 619 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 620 {
 621         struct inet_sock *inet = inet_sk(sk);
 622         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 623         int ret = -EINVAL;
 624         int chk_addr_ret;
 625
 626         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 627                 goto out;
 628         chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
 629         ret = -EADDRNOTAVAIL;
 630         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 631             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 632                 goto out;
 633         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 634         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 635                 inet->saddr = 0;  /* Use device */
 636         sk_dst_reset(sk);
 637         ret = 0;
 638 out:    return ret;
 639 }
 640
 641 /*
 642  *      This should be easy, if there is something there
 643  *      we return it, otherwise we block.
 644  */
 645
 646 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 647                        size_t len, int noblock, int flags, int *addr_len)
 648 {
 649         struct inet_sock *inet = inet_sk(sk);
 650         size_t copied = 0;
 651         int err = -EOPNOTSUPP;
 652         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 653         struct sk_buff *skb;
 654
 655         if (flags & MSG_OOB)
 656                 goto out;
 657
 658         if (addr_len)
 659                 *addr_len = sizeof(*sin);
 660
 661         if (flags & MSG_ERRQUEUE) {
 662                 err = ip_recv_error(sk, msg, len);
 663                 goto out;
 664         }
 665
 666         skb = skb_recv_datagram(sk, flags, noblock, &err);
 667         if (!skb)
 668                 goto out;
 669
 670         copied = skb->len;
 671         if (len < copied) {
 672                 msg->msg_flags |= MSG_TRUNC;
 673                 copied = len;
 674         }
 675
 676         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 677         if (err)
 678                 goto done;
 679
 680         sock_recv_timestamp(msg, sk, skb);
 681
 682         /* Copy the address. */
 683         if (sin) {
 684                 sin->sin_family = AF_INET;
 685                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 686                 sin->sin_port = 0;
 687                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 688         }
 689         if (inet->cmsg_flags)
 690                 ip_cmsg_recv(msg, skb);
 691         if (flags & MSG_TRUNC)
 692                 copied = skb->len;
 693 done:
 694         skb_free_datagram(sk, skb);
 695 out:
 696         if (err)
 697                 return err;
 698         return copied;
 699 }
 700
 701 static int raw_init(struct sock *sk)
 702 {
 703         struct raw_sock *rp = raw_sk(sk);
 704
 705         if (inet_sk(sk)->num == IPPROTO_ICMP)
 706                 memset(&rp->filter, 0, sizeof(rp->filter));
 707         return 0;
 708 }
 709
 710 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 711 {
 712         if (optlen > sizeof(struct icmp_filter))
 713                 optlen = sizeof(struct icmp_filter);
 714         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 715                 return -EFAULT;
 716         return 0;
 717 }
 718
 719 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 720 {
 721         int len, ret = -EFAULT;
 722
 723         if (get_user(len, optlen))
 724                 goto out;
 725         ret = -EINVAL;
 726         if (len < 0)
 727                 goto out;
 728         if (len > sizeof(struct icmp_filter))
 729                 len = sizeof(struct icmp_filter);
 730         ret = -EFAULT;
 731         if (put_user(len, optlen) ||
 732             copy_to_user(optval, &raw_sk(sk)->filter, len))
 733                 goto out;
 734         ret = 0;
 735 out:    return ret;
 736 }
 737
 738 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 739                           char __user *optval, int optlen)
 740 {
 741         if (optname == ICMP_FILTER) {
 742                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 743                         return -EOPNOTSUPP;
 744                 else
 745                         return raw_seticmpfilter(sk, optval, optlen);
 746         }
 747         return -ENOPROTOOPT;
 748 }
 749
 750 static int raw_setsockopt(struct sock *sk, int level, int optname,
 751                           char __user *optval, int optlen)
 752 {
 753         if (level != SOL_RAW)
 754                 return ip_setsockopt(sk, level, optname, optval, optlen);
 755         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 756 }
 757
 758 #ifdef CONFIG_COMPAT
 759 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 760                                  char __user *optval, int optlen)
 761 {
 762         if (level != SOL_RAW)
 763                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 764         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 765 }
 766 #endif
 767
 768 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 769                           char __user *optval, int __user *optlen)
 770 {
 771         if (optname == ICMP_FILTER) {
 772                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 773                         return -EOPNOTSUPP;
 774                 else
 775                         return raw_geticmpfilter(sk, optval, optlen);
 776         }
 777         return -ENOPROTOOPT;
 778 }
 779
 780 static int raw_getsockopt(struct sock *sk, int level, int optname,
 781                           char __user *optval, int __user *optlen)
 782 {
 783         if (level != SOL_RAW)
 784                 return ip_getsockopt(sk, level, optname, optval, optlen);
 785         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 786 }
 787
 788 #ifdef CONFIG_COMPAT
 789 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 790                                  char __user *optval, int __user *optlen)
 791 {
 792         if (level != SOL_RAW)
 793                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 794         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 795 }
 796 #endif
 797
 798 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 799 {
 800         switch (cmd) {
 801                 case SIOCOUTQ: {
 802                         int amount = sk_wmem_alloc_get(sk);
 803
 804                         return put_user(amount, (int __user *)arg);
 805                 }
 806                 case SIOCINQ: {
 807                         struct sk_buff *skb;
 808                         int amount = 0;
 809
 810                         spin_lock_bh(&sk->sk_receive_queue.lock);
 811                         skb = skb_peek(&sk->sk_receive_queue);
 812                         if (skb != NULL)
 813                                 amount = skb->len;
 814                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 815                         return put_user(amount, (int __user *)arg);
 816                 }
 817
 818                 default:
 819 #ifdef CONFIG_IP_MROUTE
 820                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 821 #else
 822                         return -ENOIOCTLCMD;
 823 #endif
 824         }
 825 }
 826
 827 struct proto raw_prot = {
 828         .name              = "RAW",
 829         .owner             = THIS_MODULE,
 830         .close             = raw_close,
 831         .destroy           = raw_destroy,
 832         .connect           = ip4_datagram_connect,
 833         .disconnect        = udp_disconnect,
 834         .ioctl             = raw_ioctl,
 835         .init              = raw_init,
 836         .setsockopt        = raw_setsockopt,
 837         .getsockopt        = raw_getsockopt,
 838         .sendmsg           = raw_sendmsg,
 839         .recvmsg           = raw_recvmsg,
 840         .bind              = raw_bind,
 841         .backlog_rcv       = raw_rcv_skb,
 842         .hash              = raw_hash_sk,
 843         .unhash            = raw_unhash_sk,
 844         .obj_size          = sizeof(struct raw_sock),
 845         .h.raw_hash        = &raw_v4_hashinfo,
 846 #ifdef CONFIG_COMPAT
 847         .compat_setsockopt = compat_raw_setsockopt,
 848         .compat_getsockopt = compat_raw_getsockopt,
 849 #endif
 850 };
 851
 852 #ifdef CONFIG_PROC_FS
 853 static struct sock *raw_get_first(struct seq_file *seq)
 854 {
 855         struct sock *sk;
 856         struct raw_iter_state *state = raw_seq_private(seq);
 857
 858         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
 859                         ++state->bucket) {
 860                 struct hlist_node *node;
 861
 862                 sk_for_each(sk, node, &state->h->ht[state->bucket])
 863                         if (sock_net(sk) == seq_file_net(seq))
 864                                 goto found;
 865         }
 866         sk = NULL;
 867 found:
 868         return sk;
 869 }
 870
 871 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 872 {
 873         struct raw_iter_state *state = raw_seq_private(seq);
 874
 875         do {
 876                 sk = sk_next(sk);
 877 try_again:
 878                 ;
 879         } while (sk && sock_net(sk) != seq_file_net(seq));
 880
 881         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
 882                 sk = sk_head(&state->h->ht[state->bucket]);
 883                 goto try_again;
 884         }
 885         return sk;
 886 }
 887
 888 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 889 {
 890         struct sock *sk = raw_get_first(seq);
 891
 892         if (sk)
 893                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 894                         --pos;
 895         return pos ? NULL : sk;
 896 }
 897
 898 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 899 {
 900         struct raw_iter_state *state = raw_seq_private(seq);
 901
 902         read_lock(&state->h->lock);
 903         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 904 }
 905 EXPORT_SYMBOL_GPL(raw_seq_start);
 906
 907 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 908 {
 909         struct sock *sk;
 910
 911         if (v == SEQ_START_TOKEN)
 912                 sk = raw_get_first(seq);
 913         else
 914                 sk = raw_get_next(seq, v);
 915         ++*pos;
 916         return sk;
 917 }
 918 EXPORT_SYMBOL_GPL(raw_seq_next);
 919
 920 void raw_seq_stop(struct seq_file *seq, void *v)
 921 {
 922         struct raw_iter_state *state = raw_seq_private(seq);
 923
 924         read_unlock(&state->h->lock);
 925 }
 926 EXPORT_SYMBOL_GPL(raw_seq_stop);
 927
 928 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
 929 {
 930         struct inet_sock *inet = inet_sk(sp);
 931         __be32 dest = inet->daddr,
 932                src = inet->rcv_saddr;
 933         __u16 destp = 0,
 934               srcp  = inet->num;
 935
 936         seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
 937                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
 938                 i, src, srcp, dest, destp, sp->sk_state,
 939                 sk_wmem_alloc_get(sp),
 940                 sk_rmem_alloc_get(sp),
 941                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 942                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
 943 }
 944
 945 static int raw_seq_show(struct seq_file *seq, void *v)
 946 {
 947         if (v == SEQ_START_TOKEN)
 948                 seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
 949                                 "rx_queue tr tm->when retrnsmt   uid  timeout "
 950                                 "inode ref pointer drops\n");
 951         else
 952                 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
 953         return 0;
 954 }
 955
 956 static const struct seq_operations raw_seq_ops = {
 957         .start = raw_seq_start,
 958         .next  = raw_seq_next,
 959         .stop  = raw_seq_stop,
 960         .show  = raw_seq_show,
 961 };
 962
 963 int raw_seq_open(struct inode *ino, struct file *file,
 964                  struct raw_hashinfo *h, const struct seq_operations *ops)
 965 {
 966         int err;
 967         struct raw_iter_state *i;
 968
 969         err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
 970         if (err < 0)
 971                 return err;
 972
 973         i = raw_seq_private((struct seq_file *)file->private_data);
 974         i->h = h;
 975         return 0;
 976 }
 977 EXPORT_SYMBOL_GPL(raw_seq_open);
 978
 979 static int raw_v4_seq_open(struct inode *inode, struct file *file)
 980 {
 981         return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
 982 }
 983
 984 static const struct file_operations raw_seq_fops = {
 985         .owner   = THIS_MODULE,
 986         .open    = raw_v4_seq_open,
 987         .read    = seq_read,
 988         .llseek  = seq_lseek,
 989         .release = seq_release_net,
 990 };
 991
 992 static __net_init int raw_init_net(struct net *net)
 993 {
 994         if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
 995                 return -ENOMEM;
 996
 997         return 0;
 998 }
 999
1000 static __net_exit void raw_exit_net(struct net *net)
1001 {
1002         proc_net_remove(net, "raw");
1003 }
1004
1005 static __net_initdata struct pernet_operations raw_net_ops = {
1006         .init = raw_init_net,
1007         .exit = raw_exit_net,
1008 };
1009
1010 int __init raw_proc_init(void)
1011 {
1012         return register_pernet_subsys(&raw_net_ops);
1013 }
1014
1015 void __init raw_proc_exit(void)
1016 {
1017         unregister_pernet_subsys(&raw_net_ops);
1018 }
1019 #endif /* CONFIG_PROC_FS */