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
  80 static struct raw_hashinfo raw_v4_hashinfo = {
  81         .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
  82 };
  83
  84 void raw_hash_sk(struct sock *sk)
  85 {
  86         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
  87         struct hlist_head *head;
  88
  89         head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
  90
  91         write_lock_bh(&h->lock);
  92         sk_add_node(sk, head);
  93         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
  94         write_unlock_bh(&h->lock);
  95 }
  96 EXPORT_SYMBOL_GPL(raw_hash_sk);
  97
  98 void raw_unhash_sk(struct sock *sk)
  99 {
 100         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
 101
 102         write_lock_bh(&h->lock);
 103         if (sk_del_node_init(sk))
 104                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 105         write_unlock_bh(&h->lock);
 106 }
 107 EXPORT_SYMBOL_GPL(raw_unhash_sk);
 108
 109 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
 110                 unsigned short num, __be32 raddr, __be32 laddr, int dif)
 111 {
 112         struct hlist_node *node;
 113
 114         sk_for_each_from(sk, node) {
 115                 struct inet_sock *inet = inet_sk(sk);
 116
 117                 if (net_eq(sock_net(sk), net) && inet->inet_num == num  &&
 118                     !(inet->inet_daddr && inet->inet_daddr != raddr)    &&
 119                     !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
 120                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 121                         goto found; /* gotcha */
 122         }
 123         sk = NULL;
 124 found:
 125         return sk;
 126 }
 127
 128 /*
 129  *      0 - deliver
 130  *      1 - block
 131  */
 132 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 133 {
 134         int type;
 135
 136         if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
 137                 return 1;
 138
 139         type = icmp_hdr(skb)->type;
 140         if (type < 32) {
 141                 __u32 data = raw_sk(sk)->filter.data;
 142
 143                 return ((1 << type) & data) != 0;
 144         }
 145
 146         /* Do not block unknown ICMP types */
 147         return 0;
 148 }
 149
 150 /* IP input processing comes here for RAW socket delivery.
 151  * Caller owns SKB, so we must make clones.
 152  *
 153  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 154  * -> It does. And not only TOS, but all IP header.
 155  */
 156 static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 157 {
 158         struct sock *sk;
 159         struct hlist_head *head;
 160         int delivered = 0;
 161         struct net *net;
 162
 163         read_lock(&raw_v4_hashinfo.lock);
 164         head = &raw_v4_hashinfo.ht[hash];
 165         if (hlist_empty(head))
 166                 goto out;
 167
 168         net = dev_net(skb->dev);
 169         sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
 170                              iph->saddr, iph->daddr,
 171                              skb->dev->ifindex);
 172
 173         while (sk) {
 174                 delivered = 1;
 175                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 176                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 177
 178                         /* Not releasing hash table! */
 179                         if (clone)
 180                                 raw_rcv(sk, clone);
 181                 }
 182                 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
 183                                      iph->saddr, iph->daddr,
 184                                      skb->dev->ifindex);
 185         }
 186 out:
 187         read_unlock(&raw_v4_hashinfo.lock);
 188         return delivered;
 189 }
 190
 191 int raw_local_deliver(struct sk_buff *skb, int protocol)
 192 {
 193         int hash;
 194         struct sock *raw_sk;
 195
 196         hash = protocol & (RAW_HTABLE_SIZE - 1);
 197         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 198
 199         /* If there maybe a raw socket we must check - if not we
 200          * don't care less
 201          */
 202         if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
 203                 raw_sk = NULL;
 204
 205         return raw_sk != NULL;
 206
 207 }
 208
 209 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
 210 {
 211         struct inet_sock *inet = inet_sk(sk);
 212         const int type = icmp_hdr(skb)->type;
 213         const int code = icmp_hdr(skb)->code;
 214         int err = 0;
 215         int harderr = 0;
 216
 217         /* Report error on raw socket, if:
 218            1. User requested ip_recverr.
 219            2. Socket is connected (otherwise the error indication
 220               is useless without ip_recverr and error is hard.
 221          */
 222         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 223                 return;
 224
 225         switch (type) {
 226         default:
 227         case ICMP_TIME_EXCEEDED:
 228                 err = EHOSTUNREACH;
 229                 break;
 230         case ICMP_SOURCE_QUENCH:
 231                 return;
 232         case ICMP_PARAMETERPROB:
 233                 err = EPROTO;
 234                 harderr = 1;
 235                 break;
 236         case ICMP_DEST_UNREACH:
 237                 err = EHOSTUNREACH;
 238                 if (code > NR_ICMP_UNREACH)
 239                         break;
 240                 err = icmp_err_convert[code].errno;
 241                 harderr = icmp_err_convert[code].fatal;
 242                 if (code == ICMP_FRAG_NEEDED) {
 243                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 244                         err = EMSGSIZE;
 245                 }
 246         }
 247
 248         if (inet->recverr) {
 249                 struct iphdr *iph = (struct iphdr *)skb->data;
 250                 u8 *payload = skb->data + (iph->ihl << 2);
 251
 252                 if (inet->hdrincl)
 253                         payload = skb->data;
 254                 ip_icmp_error(sk, skb, err, 0, info, payload);
 255         }
 256
 257         if (inet->recverr || harderr) {
 258                 sk->sk_err = err;
 259                 sk->sk_error_report(sk);
 260         }
 261 }
 262
 263 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
 264 {
 265         int hash;
 266         struct sock *raw_sk;
 267         struct iphdr *iph;
 268         struct net *net;
 269
 270         hash = protocol & (RAW_HTABLE_SIZE - 1);
 271
 272         read_lock(&raw_v4_hashinfo.lock);
 273         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 274         if (raw_sk != NULL) {
 275                 iph = (struct iphdr *)skb->data;
 276                 net = dev_net(skb->dev);
 277
 278                 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
 279                                                 iph->daddr, iph->saddr,
 280                                                 skb->dev->ifindex)) != NULL) {
 281                         raw_err(raw_sk, skb, info);
 282                         raw_sk = sk_next(raw_sk);
 283                         iph = (struct iphdr *)skb->data;
 284                 }
 285         }
 286         read_unlock(&raw_v4_hashinfo.lock);
 287 }
 288
 289 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 290 {
 291         /* Charge it to the socket. */
 292
 293         if (sock_queue_rcv_skb(sk, skb) < 0) {
 294                 kfree_skb(skb);
 295                 return NET_RX_DROP;
 296         }
 297
 298         return NET_RX_SUCCESS;
 299 }
 300
 301 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 302 {
 303         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 304                 atomic_inc(&sk->sk_drops);
 305                 kfree_skb(skb);
 306                 return NET_RX_DROP;
 307         }
 308         nf_reset(skb);
 309
 310         skb_push(skb, skb->data - skb_network_header(skb));
 311
 312         raw_rcv_skb(sk, skb);
 313         return 0;
 314 }
 315
 316 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 317                         struct rtable *rt,
 318                         unsigned int flags)
 319 {
 320         struct inet_sock *inet = inet_sk(sk);
 321         struct net *net = sock_net(sk);
 322         struct iphdr *iph;
 323         struct sk_buff *skb;
 324         unsigned int iphlen;
 325         int err;
 326
 327         if (length > rt->u.dst.dev->mtu) {
 328                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->inet_dport,
 329                                rt->u.dst.dev->mtu);
 330                 return -EMSGSIZE;
 331         }
 332         if (flags&MSG_PROBE)
 333                 goto out;
 334
 335         skb = sock_alloc_send_skb(sk,
 336                                   length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15,
 337                                   flags & MSG_DONTWAIT, &err);
 338         if (skb == NULL)
 339                 goto error;
 340         skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
 341
 342         skb->priority = sk->sk_priority;
 343         skb->mark = sk->sk_mark;
 344         skb_dst_set(skb, dst_clone(&rt->u.dst));
 345
 346         skb_reset_network_header(skb);
 347         iph = ip_hdr(skb);
 348         skb_put(skb, length);
 349
 350         skb->ip_summed = CHECKSUM_NONE;
 351
 352         skb->transport_header = skb->network_header;
 353         err = -EFAULT;
 354         if (memcpy_fromiovecend((void *)iph, from, 0, length))
 355                 goto error_free;
 356
 357         iphlen = iph->ihl * 4;
 358
 359         /*
 360          * We don't want to modify the ip header, but we do need to
 361          * be sure that it won't cause problems later along the network
 362          * stack.  Specifically we want to make sure that iph->ihl is a
 363          * sane value.  If ihl points beyond the length of the buffer passed
 364          * in, reject the frame as invalid
 365          */
 366         err = -EINVAL;
 367         if (iphlen > length)
 368                 goto error_free;
 369
 370         if (iphlen >= sizeof(*iph)) {
 371                 if (!iph->saddr)
 372                         iph->saddr = rt->rt_src;
 373                 iph->check   = 0;
 374                 iph->tot_len = htons(length);
 375                 if (!iph->id)
 376                         ip_select_ident(iph, &rt->u.dst, NULL);
 377
 378                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 379         }
 380         if (iph->protocol == IPPROTO_ICMP)
 381                 icmp_out_count(net, ((struct icmphdr *)
 382                         skb_transport_header(skb))->type);
 383
 384         err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
 385                       rt->u.dst.dev, dst_output);
 386         if (err > 0)
 387                 err = net_xmit_errno(err);
 388         if (err)
 389                 goto error;
 390 out:
 391         return 0;
 392
 393 error_free:
 394         kfree_skb(skb);
 395 error:
 396         IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
 397         if (err == -ENOBUFS && !inet->recverr)
 398                 err = 0;
 399         return err;
 400 }
 401
 402 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 403 {
 404         struct iovec *iov;
 405         u8 __user *type = NULL;
 406         u8 __user *code = NULL;
 407         int probed = 0;
 408         unsigned int i;
 409
 410         if (!msg->msg_iov)
 411                 return 0;
 412
 413         for (i = 0; i < msg->msg_iovlen; i++) {
 414                 iov = &msg->msg_iov[i];
 415                 if (!iov)
 416                         continue;
 417
 418                 switch (fl->proto) {
 419                 case IPPROTO_ICMP:
 420                         /* check if one-byte field is readable or not. */
 421                         if (iov->iov_base && iov->iov_len < 1)
 422                                 break;
 423
 424                         if (!type) {
 425                                 type = iov->iov_base;
 426                                 /* check if code field is readable or not. */
 427                                 if (iov->iov_len > 1)
 428                                         code = type + 1;
 429                         } else if (!code)
 430                                 code = iov->iov_base;
 431
 432                         if (type && code) {
 433                                 if (get_user(fl->fl_icmp_type, type) ||
 434                                     get_user(fl->fl_icmp_code, code))
 435                                         return -EFAULT;
 436                                 probed = 1;
 437                         }
 438                         break;
 439                 default:
 440                         probed = 1;
 441                         break;
 442                 }
 443                 if (probed)
 444                         break;
 445         }
 446         return 0;
 447 }
 448
 449 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 450                        size_t len)
 451 {
 452         struct inet_sock *inet = inet_sk(sk);
 453         struct ipcm_cookie ipc;
 454         struct rtable *rt = NULL;
 455         int free = 0;
 456         __be32 daddr;
 457         __be32 saddr;
 458         u8  tos;
 459         int err;
 460
 461         err = -EMSGSIZE;
 462         if (len > 0xFFFF)
 463                 goto out;
 464
 465         /*
 466          *      Check the flags.
 467          */
 468
 469         err = -EOPNOTSUPP;
 470         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 471                 goto out;               /* compatibility */
 472
 473         /*
 474          *      Get and verify the address.
 475          */
 476
 477         if (msg->msg_namelen) {
 478                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
 479                 err = -EINVAL;
 480                 if (msg->msg_namelen < sizeof(*usin))
 481                         goto out;
 482                 if (usin->sin_family != AF_INET) {
 483                         static int complained;
 484                         if (!complained++)
 485                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 486                                                  "raw sendmsg. Fix it!\n",
 487                                                  current->comm);
 488                         err = -EAFNOSUPPORT;
 489                         if (usin->sin_family)
 490                                 goto out;
 491                 }
 492                 daddr = usin->sin_addr.s_addr;
 493                 /* ANK: I did not forget to get protocol from port field.
 494                  * I just do not know, who uses this weirdness.
 495                  * IP_HDRINCL is much more convenient.
 496                  */
 497         } else {
 498                 err = -EDESTADDRREQ;
 499                 if (sk->sk_state != TCP_ESTABLISHED)
 500                         goto out;
 501                 daddr = inet->inet_daddr;
 502         }
 503
 504         ipc.addr = inet->inet_saddr;
 505         ipc.opt = NULL;
 506         ipc.shtx.flags = 0;
 507         ipc.oif = sk->sk_bound_dev_if;
 508
 509         if (msg->msg_controllen) {
 510                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
 511                 if (err)
 512                         goto out;
 513                 if (ipc.opt)
 514                         free = 1;
 515         }
 516
 517         saddr = ipc.addr;
 518         ipc.addr = daddr;
 519
 520         if (!ipc.opt)
 521                 ipc.opt = inet->opt;
 522
 523         if (ipc.opt) {
 524                 err = -EINVAL;
 525                 /* Linux does not mangle headers on raw sockets,
 526                  * so that IP options + IP_HDRINCL is non-sense.
 527                  */
 528                 if (inet->hdrincl)
 529                         goto done;
 530                 if (ipc.opt->srr) {
 531                         if (!daddr)
 532                                 goto done;
 533                         daddr = ipc.opt->faddr;
 534                 }
 535         }
 536         tos = RT_CONN_FLAGS(sk);
 537         if (msg->msg_flags & MSG_DONTROUTE)
 538                 tos |= RTO_ONLINK;
 539
 540         if (ipv4_is_multicast(daddr)) {
 541                 if (!ipc.oif)
 542                         ipc.oif = inet->mc_index;
 543                 if (!saddr)
 544                         saddr = inet->mc_addr;
 545         }
 546
 547         {
 548                 struct flowi fl = { .oif = ipc.oif,
 549                                     .mark = sk->sk_mark,
 550                                     .nl_u = { .ip4_u =
 551                                               { .daddr = daddr,
 552                                                 .saddr = saddr,
 553                                                 .tos = tos } },
 554                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 555                                                              sk->sk_protocol,
 556                                   };
 557                 if (!inet->hdrincl) {
 558                         err = raw_probe_proto_opt(&fl, msg);
 559                         if (err)
 560                                 goto done;
 561                 }
 562
 563                 security_sk_classify_flow(sk, &fl);
 564                 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 1);
 565         }
 566         if (err)
 567                 goto done;
 568
 569         err = -EACCES;
 570         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 571                 goto done;
 572
 573         if (msg->msg_flags & MSG_CONFIRM)
 574                 goto do_confirm;
 575 back_from_confirm:
 576
 577         if (inet->hdrincl)
 578                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 579                                         rt, msg->msg_flags);
 580
 581          else {
 582                 if (!ipc.addr)
 583                         ipc.addr = rt->rt_dst;
 584                 lock_sock(sk);
 585                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 586                                         &ipc, &rt, msg->msg_flags);
 587                 if (err)
 588                         ip_flush_pending_frames(sk);
 589                 else if (!(msg->msg_flags & MSG_MORE)) {
 590                         err = ip_push_pending_frames(sk);
 591                         if (err == -ENOBUFS && !inet->recverr)
 592                                 err = 0;
 593                 }
 594                 release_sock(sk);
 595         }
 596 done:
 597         if (free)
 598                 kfree(ipc.opt);
 599         ip_rt_put(rt);
 600
 601 out:
 602         if (err < 0)
 603                 return err;
 604         return len;
 605
 606 do_confirm:
 607         dst_confirm(&rt->u.dst);
 608         if (!(msg->msg_flags & MSG_PROBE) || len)
 609                 goto back_from_confirm;
 610         err = 0;
 611         goto done;
 612 }
 613
 614 static void raw_close(struct sock *sk, long timeout)
 615 {
 616         /*
 617          * Raw sockets may have direct kernel refereneces. Kill them.
 618          */
 619         ip_ra_control(sk, 0, NULL);
 620
 621         sk_common_release(sk);
 622 }
 623
 624 static void raw_destroy(struct sock *sk)
 625 {
 626         lock_sock(sk);
 627         ip_flush_pending_frames(sk);
 628         release_sock(sk);
 629 }
 630
 631 /* This gets rid of all the nasties in af_inet. -DaveM */
 632 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 633 {
 634         struct inet_sock *inet = inet_sk(sk);
 635         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 636         int ret = -EINVAL;
 637         int chk_addr_ret;
 638
 639         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 640                 goto out;
 641         chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
 642         ret = -EADDRNOTAVAIL;
 643         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 644             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 645                 goto out;
 646         inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
 647         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 648                 inet->inet_saddr = 0;  /* Use device */
 649         sk_dst_reset(sk);
 650         ret = 0;
 651 out:    return ret;
 652 }
 653
 654 /*
 655  *      This should be easy, if there is something there
 656  *      we return it, otherwise we block.
 657  */
 658
 659 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 660                        size_t len, int noblock, int flags, int *addr_len)
 661 {
 662         struct inet_sock *inet = inet_sk(sk);
 663         size_t copied = 0;
 664         int err = -EOPNOTSUPP;
 665         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 666         struct sk_buff *skb;
 667
 668         if (flags & MSG_OOB)
 669                 goto out;
 670
 671         if (addr_len)
 672                 *addr_len = sizeof(*sin);
 673
 674         if (flags & MSG_ERRQUEUE) {
 675                 err = ip_recv_error(sk, msg, len);
 676                 goto out;
 677         }
 678
 679         skb = skb_recv_datagram(sk, flags, noblock, &err);
 680         if (!skb)
 681                 goto out;
 682
 683         copied = skb->len;
 684         if (len < copied) {
 685                 msg->msg_flags |= MSG_TRUNC;
 686                 copied = len;
 687         }
 688
 689         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 690         if (err)
 691                 goto done;
 692
 693         sock_recv_ts_and_drops(msg, sk, skb);
 694
 695         /* Copy the address. */
 696         if (sin) {
 697                 sin->sin_family = AF_INET;
 698                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 699                 sin->sin_port = 0;
 700                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 701         }
 702         if (inet->cmsg_flags)
 703                 ip_cmsg_recv(msg, skb);
 704         if (flags & MSG_TRUNC)
 705                 copied = skb->len;
 706 done:
 707         skb_free_datagram(sk, skb);
 708 out:
 709         if (err)
 710                 return err;
 711         return copied;
 712 }
 713
 714 static int raw_init(struct sock *sk)
 715 {
 716         struct raw_sock *rp = raw_sk(sk);
 717
 718         if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
 719                 memset(&rp->filter, 0, sizeof(rp->filter));
 720         return 0;
 721 }
 722
 723 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 724 {
 725         if (optlen > sizeof(struct icmp_filter))
 726                 optlen = sizeof(struct icmp_filter);
 727         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 728                 return -EFAULT;
 729         return 0;
 730 }
 731
 732 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 733 {
 734         int len, ret = -EFAULT;
 735
 736         if (get_user(len, optlen))
 737                 goto out;
 738         ret = -EINVAL;
 739         if (len < 0)
 740                 goto out;
 741         if (len > sizeof(struct icmp_filter))
 742                 len = sizeof(struct icmp_filter);
 743         ret = -EFAULT;
 744         if (put_user(len, optlen) ||
 745             copy_to_user(optval, &raw_sk(sk)->filter, len))
 746                 goto out;
 747         ret = 0;
 748 out:    return ret;
 749 }
 750
 751 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 752                           char __user *optval, unsigned int optlen)
 753 {
 754         if (optname == ICMP_FILTER) {
 755                 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
 756                         return -EOPNOTSUPP;
 757                 else
 758                         return raw_seticmpfilter(sk, optval, optlen);
 759         }
 760         return -ENOPROTOOPT;
 761 }
 762
 763 static int raw_setsockopt(struct sock *sk, int level, int optname,
 764                           char __user *optval, unsigned int optlen)
 765 {
 766         if (level != SOL_RAW)
 767                 return ip_setsockopt(sk, level, optname, optval, optlen);
 768         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 769 }
 770
 771 #ifdef CONFIG_COMPAT
 772 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 773                                  char __user *optval, unsigned int optlen)
 774 {
 775         if (level != SOL_RAW)
 776                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 777         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 778 }
 779 #endif
 780
 781 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 782                           char __user *optval, int __user *optlen)
 783 {
 784         if (optname == ICMP_FILTER) {
 785                 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
 786                         return -EOPNOTSUPP;
 787                 else
 788                         return raw_geticmpfilter(sk, optval, optlen);
 789         }
 790         return -ENOPROTOOPT;
 791 }
 792
 793 static int raw_getsockopt(struct sock *sk, int level, int optname,
 794                           char __user *optval, int __user *optlen)
 795 {
 796         if (level != SOL_RAW)
 797                 return ip_getsockopt(sk, level, optname, optval, optlen);
 798         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 799 }
 800
 801 #ifdef CONFIG_COMPAT
 802 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 803                                  char __user *optval, int __user *optlen)
 804 {
 805         if (level != SOL_RAW)
 806                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 807         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 808 }
 809 #endif
 810
 811 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 812 {
 813         switch (cmd) {
 814                 case SIOCOUTQ: {
 815                         int amount = sk_wmem_alloc_get(sk);
 816
 817                         return put_user(amount, (int __user *)arg);
 818                 }
 819                 case SIOCINQ: {
 820                         struct sk_buff *skb;
 821                         int amount = 0;
 822
 823                         spin_lock_bh(&sk->sk_receive_queue.lock);
 824                         skb = skb_peek(&sk->sk_receive_queue);
 825                         if (skb != NULL)
 826                                 amount = skb->len;
 827                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 828                         return put_user(amount, (int __user *)arg);
 829                 }
 830
 831                 default:
 832 #ifdef CONFIG_IP_MROUTE
 833                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 834 #else
 835                         return -ENOIOCTLCMD;
 836 #endif
 837         }
 838 }
 839
 840 struct proto raw_prot = {
 841         .name              = "RAW",
 842         .owner             = THIS_MODULE,
 843         .close             = raw_close,
 844         .destroy           = raw_destroy,
 845         .connect           = ip4_datagram_connect,
 846         .disconnect        = udp_disconnect,
 847         .ioctl             = raw_ioctl,
 848         .init              = raw_init,
 849         .setsockopt        = raw_setsockopt,
 850         .getsockopt        = raw_getsockopt,
 851         .sendmsg           = raw_sendmsg,
 852         .recvmsg           = raw_recvmsg,
 853         .bind              = raw_bind,
 854         .backlog_rcv       = raw_rcv_skb,
 855         .hash              = raw_hash_sk,
 856         .unhash            = raw_unhash_sk,
 857         .obj_size          = sizeof(struct raw_sock),
 858         .h.raw_hash        = &raw_v4_hashinfo,
 859 #ifdef CONFIG_COMPAT
 860         .compat_setsockopt = compat_raw_setsockopt,
 861         .compat_getsockopt = compat_raw_getsockopt,
 862 #endif
 863 };
 864
 865 #ifdef CONFIG_PROC_FS
 866 static struct sock *raw_get_first(struct seq_file *seq)
 867 {
 868         struct sock *sk;
 869         struct raw_iter_state *state = raw_seq_private(seq);
 870
 871         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
 872                         ++state->bucket) {
 873                 struct hlist_node *node;
 874
 875                 sk_for_each(sk, node, &state->h->ht[state->bucket])
 876                         if (sock_net(sk) == seq_file_net(seq))
 877                                 goto found;
 878         }
 879         sk = NULL;
 880 found:
 881         return sk;
 882 }
 883
 884 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 885 {
 886         struct raw_iter_state *state = raw_seq_private(seq);
 887
 888         do {
 889                 sk = sk_next(sk);
 890 try_again:
 891                 ;
 892         } while (sk && sock_net(sk) != seq_file_net(seq));
 893
 894         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
 895                 sk = sk_head(&state->h->ht[state->bucket]);
 896                 goto try_again;
 897         }
 898         return sk;
 899 }
 900
 901 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 902 {
 903         struct sock *sk = raw_get_first(seq);
 904
 905         if (sk)
 906                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 907                         --pos;
 908         return pos ? NULL : sk;
 909 }
 910
 911 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 912 {
 913         struct raw_iter_state *state = raw_seq_private(seq);
 914
 915         read_lock(&state->h->lock);
 916         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 917 }
 918 EXPORT_SYMBOL_GPL(raw_seq_start);
 919
 920 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 921 {
 922         struct sock *sk;
 923
 924         if (v == SEQ_START_TOKEN)
 925                 sk = raw_get_first(seq);
 926         else
 927                 sk = raw_get_next(seq, v);
 928         ++*pos;
 929         return sk;
 930 }
 931 EXPORT_SYMBOL_GPL(raw_seq_next);
 932
 933 void raw_seq_stop(struct seq_file *seq, void *v)
 934 {
 935         struct raw_iter_state *state = raw_seq_private(seq);
 936
 937         read_unlock(&state->h->lock);
 938 }
 939 EXPORT_SYMBOL_GPL(raw_seq_stop);
 940
 941 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
 942 {
 943         struct inet_sock *inet = inet_sk(sp);
 944         __be32 dest = inet->inet_daddr,
 945                src = inet->inet_rcv_saddr;
 946         __u16 destp = 0,
 947               srcp  = inet->inet_num;
 948
 949         seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
 950                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
 951                 i, src, srcp, dest, destp, sp->sk_state,
 952                 sk_wmem_alloc_get(sp),
 953                 sk_rmem_alloc_get(sp),
 954                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 955                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
 956 }
 957
 958 static int raw_seq_show(struct seq_file *seq, void *v)
 959 {
 960         if (v == SEQ_START_TOKEN)
 961                 seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
 962                                 "rx_queue tr tm->when retrnsmt   uid  timeout "
 963                                 "inode ref pointer drops\n");
 964         else
 965                 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
 966         return 0;
 967 }
 968
 969 static const struct seq_operations raw_seq_ops = {
 970         .start = raw_seq_start,
 971         .next  = raw_seq_next,
 972         .stop  = raw_seq_stop,
 973         .show  = raw_seq_show,
 974 };
 975
 976 int raw_seq_open(struct inode *ino, struct file *file,
 977                  struct raw_hashinfo *h, const struct seq_operations *ops)
 978 {
 979         int err;
 980         struct raw_iter_state *i;
 981
 982         err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
 983         if (err < 0)
 984                 return err;
 985
 986         i = raw_seq_private((struct seq_file *)file->private_data);
 987         i->h = h;
 988         return 0;
 989 }
 990 EXPORT_SYMBOL_GPL(raw_seq_open);
 991
 992 static int raw_v4_seq_open(struct inode *inode, struct file *file)
 993 {
 994         return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
 995 }
 996
 997 static const struct file_operations raw_seq_fops = {
 998         .owner   = THIS_MODULE,
 999         .open    = raw_v4_seq_open,
1000         .read    = seq_read,
1001         .llseek  = seq_lseek,
1002         .release = seq_release_net,
1003 };
1004
1005 static __net_init int raw_init_net(struct net *net)
1006 {
1007         if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
1008                 return -ENOMEM;
1009
1010         return 0;
1011 }
1012
1013 static __net_exit void raw_exit_net(struct net *net)
1014 {
1015         proc_net_remove(net, "raw");
1016 }
1017
1018 static __net_initdata struct pernet_operations raw_net_ops = {
1019         .init = raw_init_net,
1020         .exit = raw_exit_net,
1021 };
1022
1023 int __init raw_proc_init(void)
1024 {
1025         return register_pernet_subsys(&raw_net_ops);
1026 }
1027
1028 void __init raw_proc_exit(void)
1029 {
1030         unregister_pernet_subsys(&raw_net_ops);
1031 }
1032 #endif /* CONFIG_PROC_FS */