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 static struct raw_hashinfo raw_v4_hashinfo = {
  84         .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
  85 };
  86
  87 void raw_hash_sk(struct sock *sk)
  88 {
  89         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
  90         struct hlist_head *head;
  91
  92         head = &h->ht[inet_sk(sk)->num & (RAW_HTABLE_SIZE - 1)];
  93
  94         write_lock_bh(&h->lock);
  95         sk_add_node(sk, head);
  96         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
  97         write_unlock_bh(&h->lock);
  98 }
  99 EXPORT_SYMBOL_GPL(raw_hash_sk);
 100
 101 void raw_unhash_sk(struct sock *sk)
 102 {
 103         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
 104
 105         write_lock_bh(&h->lock);
 106         if (sk_del_node_init(sk))
 107                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 108         write_unlock_bh(&h->lock);
 109 }
 110 EXPORT_SYMBOL_GPL(raw_unhash_sk);
 111
 112 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
 113                 unsigned short num, __be32 raddr, __be32 laddr, int dif)
 114 {
 115         struct hlist_node *node;
 116
 117         sk_for_each_from(sk, node) {
 118                 struct inet_sock *inet = inet_sk(sk);
 119
 120                 if (net_eq(sock_net(sk), net) && inet->num == num       &&
 121                     !(inet->daddr && inet->daddr != raddr)              &&
 122                     !(inet->rcv_saddr && inet->rcv_saddr != laddr)      &&
 123                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 124                         goto found; /* gotcha */
 125         }
 126         sk = NULL;
 127 found:
 128         return sk;
 129 }
 130
 131 /*
 132  *      0 - deliver
 133  *      1 - block
 134  */
 135 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 136 {
 137         int type;
 138
 139         if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
 140                 return 1;
 141
 142         type = icmp_hdr(skb)->type;
 143         if (type < 32) {
 144                 __u32 data = raw_sk(sk)->filter.data;
 145
 146                 return ((1 << type) & data) != 0;
 147         }
 148
 149         /* Do not block unknown ICMP types */
 150         return 0;
 151 }
 152
 153 /* IP input processing comes here for RAW socket delivery.
 154  * Caller owns SKB, so we must make clones.
 155  *
 156  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 157  * -> It does. And not only TOS, but all IP header.
 158  */
 159 static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 160 {
 161         struct sock *sk;
 162         struct hlist_head *head;
 163         int delivered = 0;
 164         struct net *net;
 165
 166         read_lock(&raw_v4_hashinfo.lock);
 167         head = &raw_v4_hashinfo.ht[hash];
 168         if (hlist_empty(head))
 169                 goto out;
 170
 171         net = dev_net(skb->dev);
 172         sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
 173                              iph->saddr, iph->daddr,
 174                              skb->dev->ifindex);
 175
 176         while (sk) {
 177                 delivered = 1;
 178                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 179                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 180
 181                         /* Not releasing hash table! */
 182                         if (clone)
 183                                 raw_rcv(sk, clone);
 184                 }
 185                 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
 186                                      iph->saddr, iph->daddr,
 187                                      skb->dev->ifindex);
 188         }
 189 out:
 190         read_unlock(&raw_v4_hashinfo.lock);
 191         return delivered;
 192 }
 193
 194 int raw_local_deliver(struct sk_buff *skb, int protocol)
 195 {
 196         int hash;
 197         struct sock *raw_sk;
 198
 199         hash = protocol & (RAW_HTABLE_SIZE - 1);
 200         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 201
 202         /* If there maybe a raw socket we must check - if not we
 203          * don't care less
 204          */
 205         if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
 206                 raw_sk = NULL;
 207
 208         return raw_sk != NULL;
 209
 210 }
 211
 212 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
 213 {
 214         struct inet_sock *inet = inet_sk(sk);
 215         const int type = icmp_hdr(skb)->type;
 216         const int code = icmp_hdr(skb)->code;
 217         int err = 0;
 218         int harderr = 0;
 219
 220         /* Report error on raw socket, if:
 221            1. User requested ip_recverr.
 222            2. Socket is connected (otherwise the error indication
 223               is useless without ip_recverr and error is hard.
 224          */
 225         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 226                 return;
 227
 228         switch (type) {
 229         default:
 230         case ICMP_TIME_EXCEEDED:
 231                 err = EHOSTUNREACH;
 232                 break;
 233         case ICMP_SOURCE_QUENCH:
 234                 return;
 235         case ICMP_PARAMETERPROB:
 236                 err = EPROTO;
 237                 harderr = 1;
 238                 break;
 239         case ICMP_DEST_UNREACH:
 240                 err = EHOSTUNREACH;
 241                 if (code > NR_ICMP_UNREACH)
 242                         break;
 243                 err = icmp_err_convert[code].errno;
 244                 harderr = icmp_err_convert[code].fatal;
 245                 if (code == ICMP_FRAG_NEEDED) {
 246                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 247                         err = EMSGSIZE;
 248                 }
 249         }
 250
 251         if (inet->recverr) {
 252                 struct iphdr *iph = (struct iphdr*)skb->data;
 253                 u8 *payload = skb->data + (iph->ihl << 2);
 254
 255                 if (inet->hdrincl)
 256                         payload = skb->data;
 257                 ip_icmp_error(sk, skb, err, 0, info, payload);
 258         }
 259
 260         if (inet->recverr || harderr) {
 261                 sk->sk_err = err;
 262                 sk->sk_error_report(sk);
 263         }
 264 }
 265
 266 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
 267 {
 268         int hash;
 269         struct sock *raw_sk;
 270         struct iphdr *iph;
 271         struct net *net;
 272
 273         hash = protocol & (RAW_HTABLE_SIZE - 1);
 274
 275         read_lock(&raw_v4_hashinfo.lock);
 276         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 277         if (raw_sk != NULL) {
 278                 iph = (struct iphdr *)skb->data;
 279                 net = dev_net(skb->dev);
 280
 281                 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
 282                                                 iph->daddr, iph->saddr,
 283                                                 skb->dev->ifindex)) != NULL) {
 284                         raw_err(raw_sk, skb, info);
 285                         raw_sk = sk_next(raw_sk);
 286                         iph = (struct iphdr *)skb->data;
 287                 }
 288         }
 289         read_unlock(&raw_v4_hashinfo.lock);
 290 }
 291
 292 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 293 {
 294         /* Charge it to the socket. */
 295
 296         if (sock_queue_rcv_skb(sk, skb) < 0) {
 297                 atomic_inc(&sk->sk_drops);
 298                 kfree_skb(skb);
 299                 return NET_RX_DROP;
 300         }
 301
 302         return NET_RX_SUCCESS;
 303 }
 304
 305 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 306 {
 307         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 308                 atomic_inc(&sk->sk_drops);
 309                 kfree_skb(skb);
 310                 return NET_RX_DROP;
 311         }
 312         nf_reset(skb);
 313
 314         skb_push(skb, skb->data - skb_network_header(skb));
 315
 316         raw_rcv_skb(sk, skb);
 317         return 0;
 318 }
 319
 320 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 321                         struct rtable *rt,
 322                         unsigned int flags)
 323 {
 324         struct inet_sock *inet = inet_sk(sk);
 325         struct iphdr *iph;
 326         struct sk_buff *skb;
 327         unsigned int iphlen;
 328         int err;
 329
 330         if (length > rt->u.dst.dev->mtu) {
 331                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
 332                                rt->u.dst.dev->mtu);
 333                 return -EMSGSIZE;
 334         }
 335         if (flags&MSG_PROBE)
 336                 goto out;
 337
 338         skb = sock_alloc_send_skb(sk,
 339                                   length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15,
 340                                   flags & MSG_DONTWAIT, &err);
 341         if (skb == NULL)
 342                 goto error;
 343         skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
 344
 345         skb->priority = sk->sk_priority;
 346         skb->mark = sk->sk_mark;
 347         skb->dst = dst_clone(&rt->u.dst);
 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 = memcpy_fromiovecend((void *)iph, from, 0, length);
 357         if (err)
 358                 goto error_fault;
 359
 360         /* We don't modify invalid header */
 361         iphlen = iph->ihl * 4;
 362         if (iphlen >= sizeof(*iph) && iphlen <= length) {
 363                 if (!iph->saddr)
 364                         iph->saddr = rt->rt_src;
 365                 iph->check   = 0;
 366                 iph->tot_len = htons(length);
 367                 if (!iph->id)
 368                         ip_select_ident(iph, &rt->u.dst, NULL);
 369
 370                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 371         }
 372         if (iph->protocol == IPPROTO_ICMP)
 373                 icmp_out_count(((struct icmphdr *)
 374                         skb_transport_header(skb))->type);
 375
 376         err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 377                       dst_output);
 378         if (err > 0)
 379                 err = inet->recverr ? net_xmit_errno(err) : 0;
 380         if (err)
 381                 goto error;
 382 out:
 383         return 0;
 384
 385 error_fault:
 386         err = -EFAULT;
 387         kfree_skb(skb);
 388 error:
 389         IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
 390         return err;
 391 }
 392
 393 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 394 {
 395         struct iovec *iov;
 396         u8 __user *type = NULL;
 397         u8 __user *code = NULL;
 398         int probed = 0;
 399         unsigned int i;
 400
 401         if (!msg->msg_iov)
 402                 return 0;
 403
 404         for (i = 0; i < msg->msg_iovlen; i++) {
 405                 iov = &msg->msg_iov[i];
 406                 if (!iov)
 407                         continue;
 408
 409                 switch (fl->proto) {
 410                 case IPPROTO_ICMP:
 411                         /* check if one-byte field is readable or not. */
 412                         if (iov->iov_base && iov->iov_len < 1)
 413                                 break;
 414
 415                         if (!type) {
 416                                 type = iov->iov_base;
 417                                 /* check if code field is readable or not. */
 418                                 if (iov->iov_len > 1)
 419                                         code = type + 1;
 420                         } else if (!code)
 421                                 code = iov->iov_base;
 422
 423                         if (type && code) {
 424                                 if (get_user(fl->fl_icmp_type, type) ||
 425                                     get_user(fl->fl_icmp_code, code))
 426                                         return -EFAULT;
 427                                 probed = 1;
 428                         }
 429                         break;
 430                 default:
 431                         probed = 1;
 432                         break;
 433                 }
 434                 if (probed)
 435                         break;
 436         }
 437         return 0;
 438 }
 439
 440 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 441                        size_t len)
 442 {
 443         struct inet_sock *inet = inet_sk(sk);
 444         struct ipcm_cookie ipc;
 445         struct rtable *rt = NULL;
 446         int free = 0;
 447         __be32 daddr;
 448         __be32 saddr;
 449         u8  tos;
 450         int err;
 451
 452         err = -EMSGSIZE;
 453         if (len > 0xFFFF)
 454                 goto out;
 455
 456         /*
 457          *      Check the flags.
 458          */
 459
 460         err = -EOPNOTSUPP;
 461         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 462                 goto out;               /* compatibility */
 463
 464         /*
 465          *      Get and verify the address.
 466          */
 467
 468         if (msg->msg_namelen) {
 469                 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
 470                 err = -EINVAL;
 471                 if (msg->msg_namelen < sizeof(*usin))
 472                         goto out;
 473                 if (usin->sin_family != AF_INET) {
 474                         static int complained;
 475                         if (!complained++)
 476                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 477                                                  "raw sendmsg. Fix it!\n",
 478                                                  current->comm);
 479                         err = -EAFNOSUPPORT;
 480                         if (usin->sin_family)
 481                                 goto out;
 482                 }
 483                 daddr = usin->sin_addr.s_addr;
 484                 /* ANK: I did not forget to get protocol from port field.
 485                  * I just do not know, who uses this weirdness.
 486                  * IP_HDRINCL is much more convenient.
 487                  */
 488         } else {
 489                 err = -EDESTADDRREQ;
 490                 if (sk->sk_state != TCP_ESTABLISHED)
 491                         goto out;
 492                 daddr = inet->daddr;
 493         }
 494
 495         ipc.addr = inet->saddr;
 496         ipc.opt = NULL;
 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 /* This gets rid of all the nasties in af_inet. -DaveM */
 612 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 613 {
 614         struct inet_sock *inet = inet_sk(sk);
 615         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 616         int ret = -EINVAL;
 617         int chk_addr_ret;
 618
 619         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 620                 goto out;
 621         chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
 622         ret = -EADDRNOTAVAIL;
 623         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 624             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 625                 goto out;
 626         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 627         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 628                 inet->saddr = 0;  /* Use device */
 629         sk_dst_reset(sk);
 630         ret = 0;
 631 out:    return ret;
 632 }
 633
 634 /*
 635  *      This should be easy, if there is something there
 636  *      we return it, otherwise we block.
 637  */
 638
 639 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 640                        size_t len, int noblock, int flags, int *addr_len)
 641 {
 642         struct inet_sock *inet = inet_sk(sk);
 643         size_t copied = 0;
 644         int err = -EOPNOTSUPP;
 645         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 646         struct sk_buff *skb;
 647
 648         if (flags & MSG_OOB)
 649                 goto out;
 650
 651         if (addr_len)
 652                 *addr_len = sizeof(*sin);
 653
 654         if (flags & MSG_ERRQUEUE) {
 655                 err = ip_recv_error(sk, msg, len);
 656                 goto out;
 657         }
 658
 659         skb = skb_recv_datagram(sk, flags, noblock, &err);
 660         if (!skb)
 661                 goto out;
 662
 663         copied = skb->len;
 664         if (len < copied) {
 665                 msg->msg_flags |= MSG_TRUNC;
 666                 copied = len;
 667         }
 668
 669         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 670         if (err)
 671                 goto done;
 672
 673         sock_recv_timestamp(msg, sk, skb);
 674
 675         /* Copy the address. */
 676         if (sin) {
 677                 sin->sin_family = AF_INET;
 678                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 679                 sin->sin_port = 0;
 680                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 681         }
 682         if (inet->cmsg_flags)
 683                 ip_cmsg_recv(msg, skb);
 684         if (flags & MSG_TRUNC)
 685                 copied = skb->len;
 686 done:
 687         skb_free_datagram(sk, skb);
 688 out:
 689         if (err)
 690                 return err;
 691         return copied;
 692 }
 693
 694 static int raw_init(struct sock *sk)
 695 {
 696         struct raw_sock *rp = raw_sk(sk);
 697
 698         if (inet_sk(sk)->num == IPPROTO_ICMP)
 699                 memset(&rp->filter, 0, sizeof(rp->filter));
 700         return 0;
 701 }
 702
 703 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 704 {
 705         if (optlen > sizeof(struct icmp_filter))
 706                 optlen = sizeof(struct icmp_filter);
 707         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 708                 return -EFAULT;
 709         return 0;
 710 }
 711
 712 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 713 {
 714         int len, ret = -EFAULT;
 715
 716         if (get_user(len, optlen))
 717                 goto out;
 718         ret = -EINVAL;
 719         if (len < 0)
 720                 goto out;
 721         if (len > sizeof(struct icmp_filter))
 722                 len = sizeof(struct icmp_filter);
 723         ret = -EFAULT;
 724         if (put_user(len, optlen) ||
 725             copy_to_user(optval, &raw_sk(sk)->filter, len))
 726                 goto out;
 727         ret = 0;
 728 out:    return ret;
 729 }
 730
 731 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 732                           char __user *optval, int optlen)
 733 {
 734         if (optname == ICMP_FILTER) {
 735                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 736                         return -EOPNOTSUPP;
 737                 else
 738                         return raw_seticmpfilter(sk, optval, optlen);
 739         }
 740         return -ENOPROTOOPT;
 741 }
 742
 743 static int raw_setsockopt(struct sock *sk, int level, int optname,
 744                           char __user *optval, int optlen)
 745 {
 746         if (level != SOL_RAW)
 747                 return ip_setsockopt(sk, level, optname, optval, optlen);
 748         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 749 }
 750
 751 #ifdef CONFIG_COMPAT
 752 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 753                                  char __user *optval, int optlen)
 754 {
 755         if (level != SOL_RAW)
 756                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 757         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 758 }
 759 #endif
 760
 761 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 762                           char __user *optval, int __user *optlen)
 763 {
 764         if (optname == ICMP_FILTER) {
 765                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 766                         return -EOPNOTSUPP;
 767                 else
 768                         return raw_geticmpfilter(sk, optval, optlen);
 769         }
 770         return -ENOPROTOOPT;
 771 }
 772
 773 static int raw_getsockopt(struct sock *sk, int level, int optname,
 774                           char __user *optval, int __user *optlen)
 775 {
 776         if (level != SOL_RAW)
 777                 return ip_getsockopt(sk, level, optname, optval, optlen);
 778         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 779 }
 780
 781 #ifdef CONFIG_COMPAT
 782 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 783                                  char __user *optval, int __user *optlen)
 784 {
 785         if (level != SOL_RAW)
 786                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 787         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 788 }
 789 #endif
 790
 791 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 792 {
 793         switch (cmd) {
 794                 case SIOCOUTQ: {
 795                         int amount = atomic_read(&sk->sk_wmem_alloc);
 796                         return put_user(amount, (int __user *)arg);
 797                 }
 798                 case SIOCINQ: {
 799                         struct sk_buff *skb;
 800                         int amount = 0;
 801
 802                         spin_lock_bh(&sk->sk_receive_queue.lock);
 803                         skb = skb_peek(&sk->sk_receive_queue);
 804                         if (skb != NULL)
 805                                 amount = skb->len;
 806                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 807                         return put_user(amount, (int __user *)arg);
 808                 }
 809
 810                 default:
 811 #ifdef CONFIG_IP_MROUTE
 812                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 813 #else
 814                         return -ENOIOCTLCMD;
 815 #endif
 816         }
 817 }
 818
 819 struct proto raw_prot = {
 820         .name              = "RAW",
 821         .owner             = THIS_MODULE,
 822         .close             = raw_close,
 823         .connect           = ip4_datagram_connect,
 824         .disconnect        = udp_disconnect,
 825         .ioctl             = raw_ioctl,
 826         .init              = raw_init,
 827         .setsockopt        = raw_setsockopt,
 828         .getsockopt        = raw_getsockopt,
 829         .sendmsg           = raw_sendmsg,
 830         .recvmsg           = raw_recvmsg,
 831         .bind              = raw_bind,
 832         .backlog_rcv       = raw_rcv_skb,
 833         .hash              = raw_hash_sk,
 834         .unhash            = raw_unhash_sk,
 835         .obj_size          = sizeof(struct raw_sock),
 836         .h.raw_hash        = &raw_v4_hashinfo,
 837 #ifdef CONFIG_COMPAT
 838         .compat_setsockopt = compat_raw_setsockopt,
 839         .compat_getsockopt = compat_raw_getsockopt,
 840 #endif
 841 };
 842
 843 #ifdef CONFIG_PROC_FS
 844 static struct sock *raw_get_first(struct seq_file *seq)
 845 {
 846         struct sock *sk;
 847         struct raw_iter_state* state = raw_seq_private(seq);
 848
 849         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
 850                         ++state->bucket) {
 851                 struct hlist_node *node;
 852
 853                 sk_for_each(sk, node, &state->h->ht[state->bucket])
 854                         if (sock_net(sk) == seq_file_net(seq))
 855                                 goto found;
 856         }
 857         sk = NULL;
 858 found:
 859         return sk;
 860 }
 861
 862 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 863 {
 864         struct raw_iter_state* state = raw_seq_private(seq);
 865
 866         do {
 867                 sk = sk_next(sk);
 868 try_again:
 869                 ;
 870         } while (sk && sock_net(sk) != seq_file_net(seq));
 871
 872         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
 873                 sk = sk_head(&state->h->ht[state->bucket]);
 874                 goto try_again;
 875         }
 876         return sk;
 877 }
 878
 879 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 880 {
 881         struct sock *sk = raw_get_first(seq);
 882
 883         if (sk)
 884                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 885                         --pos;
 886         return pos ? NULL : sk;
 887 }
 888
 889 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 890 {
 891         struct raw_iter_state *state = raw_seq_private(seq);
 892
 893         read_lock(&state->h->lock);
 894         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 895 }
 896 EXPORT_SYMBOL_GPL(raw_seq_start);
 897
 898 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 899 {
 900         struct sock *sk;
 901
 902         if (v == SEQ_START_TOKEN)
 903                 sk = raw_get_first(seq);
 904         else
 905                 sk = raw_get_next(seq, v);
 906         ++*pos;
 907         return sk;
 908 }
 909 EXPORT_SYMBOL_GPL(raw_seq_next);
 910
 911 void raw_seq_stop(struct seq_file *seq, void *v)
 912 {
 913         struct raw_iter_state *state = raw_seq_private(seq);
 914
 915         read_unlock(&state->h->lock);
 916 }
 917 EXPORT_SYMBOL_GPL(raw_seq_stop);
 918
 919 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
 920 {
 921         struct inet_sock *inet = inet_sk(sp);
 922         __be32 dest = inet->daddr,
 923                src = inet->rcv_saddr;
 924         __u16 destp = 0,
 925               srcp  = inet->num;
 926
 927         seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
 928                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d",
 929                 i, src, srcp, dest, destp, sp->sk_state,
 930                 atomic_read(&sp->sk_wmem_alloc),
 931                 atomic_read(&sp->sk_rmem_alloc),
 932                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 933                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
 934 }
 935
 936 static int raw_seq_show(struct seq_file *seq, void *v)
 937 {
 938         if (v == SEQ_START_TOKEN)
 939                 seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
 940                                 "rx_queue tr tm->when retrnsmt   uid  timeout "
 941                                 "inode  drops\n");
 942         else
 943                 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
 944         return 0;
 945 }
 946
 947 static const struct seq_operations raw_seq_ops = {
 948         .start = raw_seq_start,
 949         .next  = raw_seq_next,
 950         .stop  = raw_seq_stop,
 951         .show  = raw_seq_show,
 952 };
 953
 954 int raw_seq_open(struct inode *ino, struct file *file,
 955                  struct raw_hashinfo *h, const struct seq_operations *ops)
 956 {
 957         int err;
 958         struct raw_iter_state *i;
 959
 960         err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
 961         if (err < 0)
 962                 return err;
 963
 964         i = raw_seq_private((struct seq_file *)file->private_data);
 965         i->h = h;
 966         return 0;
 967 }
 968 EXPORT_SYMBOL_GPL(raw_seq_open);
 969
 970 static int raw_v4_seq_open(struct inode *inode, struct file *file)
 971 {
 972         return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
 973 }
 974
 975 static const struct file_operations raw_seq_fops = {
 976         .owner   = THIS_MODULE,
 977         .open    = raw_v4_seq_open,
 978         .read    = seq_read,
 979         .llseek  = seq_lseek,
 980         .release = seq_release_net,
 981 };
 982
 983 static __net_init int raw_init_net(struct net *net)
 984 {
 985         if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
 986                 return -ENOMEM;
 987
 988         return 0;
 989 }
 990
 991 static __net_exit void raw_exit_net(struct net *net)
 992 {
 993         proc_net_remove(net, "raw");
 994 }
 995
 996 static __net_initdata struct pernet_operations raw_net_ops = {
 997         .init = raw_init_net,
 998         .exit = raw_exit_net,
 999 };
1000
1001 int __init raw_proc_init(void)
1002 {
1003         return register_pernet_subsys(&raw_net_ops);
1004 }
1005
1006 void __init raw_proc_exit(void)
1007 {
1008         unregister_pernet_subsys(&raw_net_ops);
1009 }
1010 #endif /* CONFIG_PROC_FS */