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 = -EFAULT;
 356         if (memcpy_fromiovecend((void *)iph, from, 0, length))
 357                 goto error_free;
 358
 359         iphlen = iph->ihl * 4;
 360
 361         /*
 362          * We don't want to modify the ip header, but we do need to
 363          * be sure that it won't cause problems later along the network
 364          * stack.  Specifically we want to make sure that iph->ihl is a
 365          * sane value.  If ihl points beyond the length of the buffer passed
 366          * in, reject the frame as invalid
 367          */
 368         err = -EINVAL;
 369         if (iphlen > length)
 370                 goto error_free;
 371
 372         if (iphlen >= sizeof(*iph)) {
 373                 if (!iph->saddr)
 374                         iph->saddr = rt->rt_src;
 375                 iph->check   = 0;
 376                 iph->tot_len = htons(length);
 377                 if (!iph->id)
 378                         ip_select_ident(iph, &rt->u.dst, NULL);
 379
 380                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 381         }
 382         if (iph->protocol == IPPROTO_ICMP)
 383                 icmp_out_count(net, ((struct icmphdr *)
 384                         skb_transport_header(skb))->type);
 385
 386         err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 387                       dst_output);
 388         if (err > 0)
 389                 err = net_xmit_errno(err);
 390         if (err)
 391                 goto error;
 392 out:
 393         return 0;
 394
 395 error_free:
 396         kfree_skb(skb);
 397 error:
 398         IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
 399         if (err == -ENOBUFS && !inet->recverr)
 400                 err = 0;
 401         return err;
 402 }
 403
 404 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 405 {
 406         struct iovec *iov;
 407         u8 __user *type = NULL;
 408         u8 __user *code = NULL;
 409         int probed = 0;
 410         unsigned int i;
 411
 412         if (!msg->msg_iov)
 413                 return 0;
 414
 415         for (i = 0; i < msg->msg_iovlen; i++) {
 416                 iov = &msg->msg_iov[i];
 417                 if (!iov)
 418                         continue;
 419
 420                 switch (fl->proto) {
 421                 case IPPROTO_ICMP:
 422                         /* check if one-byte field is readable or not. */
 423                         if (iov->iov_base && iov->iov_len < 1)
 424                                 break;
 425
 426                         if (!type) {
 427                                 type = iov->iov_base;
 428                                 /* check if code field is readable or not. */
 429                                 if (iov->iov_len > 1)
 430                                         code = type + 1;
 431                         } else if (!code)
 432                                 code = iov->iov_base;
 433
 434                         if (type && code) {
 435                                 if (get_user(fl->fl_icmp_type, type) ||
 436                                     get_user(fl->fl_icmp_code, code))
 437                                         return -EFAULT;
 438                                 probed = 1;
 439                         }
 440                         break;
 441                 default:
 442                         probed = 1;
 443                         break;
 444                 }
 445                 if (probed)
 446                         break;
 447         }
 448         return 0;
 449 }
 450
 451 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 452                        size_t len)
 453 {
 454         struct inet_sock *inet = inet_sk(sk);
 455         struct ipcm_cookie ipc;
 456         struct rtable *rt = NULL;
 457         int free = 0;
 458         __be32 daddr;
 459         __be32 saddr;
 460         u8  tos;
 461         int err;
 462
 463         err = -EMSGSIZE;
 464         if (len > 0xFFFF)
 465                 goto out;
 466
 467         /*
 468          *      Check the flags.
 469          */
 470
 471         err = -EOPNOTSUPP;
 472         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 473                 goto out;               /* compatibility */
 474
 475         /*
 476          *      Get and verify the address.
 477          */
 478
 479         if (msg->msg_namelen) {
 480                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
 481                 err = -EINVAL;
 482                 if (msg->msg_namelen < sizeof(*usin))
 483                         goto out;
 484                 if (usin->sin_family != AF_INET) {
 485                         static int complained;
 486                         if (!complained++)
 487                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 488                                                  "raw sendmsg. Fix it!\n",
 489                                                  current->comm);
 490                         err = -EAFNOSUPPORT;
 491                         if (usin->sin_family)
 492                                 goto out;
 493                 }
 494                 daddr = usin->sin_addr.s_addr;
 495                 /* ANK: I did not forget to get protocol from port field.
 496                  * I just do not know, who uses this weirdness.
 497                  * IP_HDRINCL is much more convenient.
 498                  */
 499         } else {
 500                 err = -EDESTADDRREQ;
 501                 if (sk->sk_state != TCP_ESTABLISHED)
 502                         goto out;
 503                 daddr = inet->daddr;
 504         }
 505
 506         ipc.addr = inet->saddr;
 507         ipc.opt = NULL;
 508         ipc.shtx.flags = 0;
 509         ipc.oif = sk->sk_bound_dev_if;
 510
 511         if (msg->msg_controllen) {
 512                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
 513                 if (err)
 514                         goto out;
 515                 if (ipc.opt)
 516                         free = 1;
 517         }
 518
 519         saddr = ipc.addr;
 520         ipc.addr = daddr;
 521
 522         if (!ipc.opt)
 523                 ipc.opt = inet->opt;
 524
 525         if (ipc.opt) {
 526                 err = -EINVAL;
 527                 /* Linux does not mangle headers on raw sockets,
 528                  * so that IP options + IP_HDRINCL is non-sense.
 529                  */
 530                 if (inet->hdrincl)
 531                         goto done;
 532                 if (ipc.opt->srr) {
 533                         if (!daddr)
 534                                 goto done;
 535                         daddr = ipc.opt->faddr;
 536                 }
 537         }
 538         tos = RT_CONN_FLAGS(sk);
 539         if (msg->msg_flags & MSG_DONTROUTE)
 540                 tos |= RTO_ONLINK;
 541
 542         if (ipv4_is_multicast(daddr)) {
 543                 if (!ipc.oif)
 544                         ipc.oif = inet->mc_index;
 545                 if (!saddr)
 546                         saddr = inet->mc_addr;
 547         }
 548
 549         {
 550                 struct flowi fl = { .oif = ipc.oif,
 551                                     .mark = sk->sk_mark,
 552                                     .nl_u = { .ip4_u =
 553                                               { .daddr = daddr,
 554                                                 .saddr = saddr,
 555                                                 .tos = tos } },
 556                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 557                                                              sk->sk_protocol,
 558                                   };
 559                 if (!inet->hdrincl) {
 560                         err = raw_probe_proto_opt(&fl, msg);
 561                         if (err)
 562                                 goto done;
 563                 }
 564
 565                 security_sk_classify_flow(sk, &fl);
 566                 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 1);
 567         }
 568         if (err)
 569                 goto done;
 570
 571         err = -EACCES;
 572         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 573                 goto done;
 574
 575         if (msg->msg_flags & MSG_CONFIRM)
 576                 goto do_confirm;
 577 back_from_confirm:
 578
 579         if (inet->hdrincl)
 580                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 581                                         rt, msg->msg_flags);
 582
 583          else {
 584                 if (!ipc.addr)
 585                         ipc.addr = rt->rt_dst;
 586                 lock_sock(sk);
 587                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 588                                         &ipc, &rt, msg->msg_flags);
 589                 if (err)
 590                         ip_flush_pending_frames(sk);
 591                 else if (!(msg->msg_flags & MSG_MORE)) {
 592                         err = ip_push_pending_frames(sk);
 593                         if (err == -ENOBUFS && !inet->recverr)
 594                                 err = 0;
 595                 }
 596                 release_sock(sk);
 597         }
 598 done:
 599         if (free)
 600                 kfree(ipc.opt);
 601         ip_rt_put(rt);
 602
 603 out:
 604         if (err < 0)
 605                 return err;
 606         return len;
 607
 608 do_confirm:
 609         dst_confirm(&rt->u.dst);
 610         if (!(msg->msg_flags & MSG_PROBE) || len)
 611                 goto back_from_confirm;
 612         err = 0;
 613         goto done;
 614 }
 615
 616 static void raw_close(struct sock *sk, long timeout)
 617 {
 618         /*
 619          * Raw sockets may have direct kernel refereneces. Kill them.
 620          */
 621         ip_ra_control(sk, 0, NULL);
 622
 623         sk_common_release(sk);
 624 }
 625
 626 static void raw_destroy(struct sock *sk)
 627 {
 628         lock_sock(sk);
 629         ip_flush_pending_frames(sk);
 630         release_sock(sk);
 631 }
 632
 633 /* This gets rid of all the nasties in af_inet. -DaveM */
 634 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 635 {
 636         struct inet_sock *inet = inet_sk(sk);
 637         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 638         int ret = -EINVAL;
 639         int chk_addr_ret;
 640
 641         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 642                 goto out;
 643         chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
 644         ret = -EADDRNOTAVAIL;
 645         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 646             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 647                 goto out;
 648         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 649         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 650                 inet->saddr = 0;  /* Use device */
 651         sk_dst_reset(sk);
 652         ret = 0;
 653 out:    return ret;
 654 }
 655
 656 /*
 657  *      This should be easy, if there is something there
 658  *      we return it, otherwise we block.
 659  */
 660
 661 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 662                        size_t len, int noblock, int flags, int *addr_len)
 663 {
 664         struct inet_sock *inet = inet_sk(sk);
 665         size_t copied = 0;
 666         int err = -EOPNOTSUPP;
 667         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 668         struct sk_buff *skb;
 669
 670         if (flags & MSG_OOB)
 671                 goto out;
 672
 673         if (addr_len)
 674                 *addr_len = sizeof(*sin);
 675
 676         if (flags & MSG_ERRQUEUE) {
 677                 err = ip_recv_error(sk, msg, len);
 678                 goto out;
 679         }
 680
 681         skb = skb_recv_datagram(sk, flags, noblock, &err);
 682         if (!skb)
 683                 goto out;
 684
 685         copied = skb->len;
 686         if (len < copied) {
 687                 msg->msg_flags |= MSG_TRUNC;
 688                 copied = len;
 689         }
 690
 691         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 692         if (err)
 693                 goto done;
 694
 695         sock_recv_timestamp(msg, sk, skb);
 696
 697         /* Copy the address. */
 698         if (sin) {
 699                 sin->sin_family = AF_INET;
 700                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 701                 sin->sin_port = 0;
 702                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 703         }
 704         if (inet->cmsg_flags)
 705                 ip_cmsg_recv(msg, skb);
 706         if (flags & MSG_TRUNC)
 707                 copied = skb->len;
 708 done:
 709         skb_free_datagram(sk, skb);
 710 out:
 711         if (err)
 712                 return err;
 713         return copied;
 714 }
 715
 716 static int raw_init(struct sock *sk)
 717 {
 718         struct raw_sock *rp = raw_sk(sk);
 719
 720         if (inet_sk(sk)->num == IPPROTO_ICMP)
 721                 memset(&rp->filter, 0, sizeof(rp->filter));
 722         return 0;
 723 }
 724
 725 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 726 {
 727         if (optlen > sizeof(struct icmp_filter))
 728                 optlen = sizeof(struct icmp_filter);
 729         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 730                 return -EFAULT;
 731         return 0;
 732 }
 733
 734 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 735 {
 736         int len, ret = -EFAULT;
 737
 738         if (get_user(len, optlen))
 739                 goto out;
 740         ret = -EINVAL;
 741         if (len < 0)
 742                 goto out;
 743         if (len > sizeof(struct icmp_filter))
 744                 len = sizeof(struct icmp_filter);
 745         ret = -EFAULT;
 746         if (put_user(len, optlen) ||
 747             copy_to_user(optval, &raw_sk(sk)->filter, len))
 748                 goto out;
 749         ret = 0;
 750 out:    return ret;
 751 }
 752
 753 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 754                           char __user *optval, unsigned int optlen)
 755 {
 756         if (optname == ICMP_FILTER) {
 757                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 758                         return -EOPNOTSUPP;
 759                 else
 760                         return raw_seticmpfilter(sk, optval, optlen);
 761         }
 762         return -ENOPROTOOPT;
 763 }
 764
 765 static int raw_setsockopt(struct sock *sk, int level, int optname,
 766                           char __user *optval, unsigned int optlen)
 767 {
 768         if (level != SOL_RAW)
 769                 return ip_setsockopt(sk, level, optname, optval, optlen);
 770         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 771 }
 772
 773 #ifdef CONFIG_COMPAT
 774 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 775                                  char __user *optval, unsigned int optlen)
 776 {
 777         if (level != SOL_RAW)
 778                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 779         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 780 }
 781 #endif
 782
 783 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 784                           char __user *optval, int __user *optlen)
 785 {
 786         if (optname == ICMP_FILTER) {
 787                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 788                         return -EOPNOTSUPP;
 789                 else
 790                         return raw_geticmpfilter(sk, optval, optlen);
 791         }
 792         return -ENOPROTOOPT;
 793 }
 794
 795 static int raw_getsockopt(struct sock *sk, int level, int optname,
 796                           char __user *optval, int __user *optlen)
 797 {
 798         if (level != SOL_RAW)
 799                 return ip_getsockopt(sk, level, optname, optval, optlen);
 800         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 801 }
 802
 803 #ifdef CONFIG_COMPAT
 804 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 805                                  char __user *optval, int __user *optlen)
 806 {
 807         if (level != SOL_RAW)
 808                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 809         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 810 }
 811 #endif
 812
 813 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 814 {
 815         switch (cmd) {
 816                 case SIOCOUTQ: {
 817                         int amount = sk_wmem_alloc_get(sk);
 818
 819                         return put_user(amount, (int __user *)arg);
 820                 }
 821                 case SIOCINQ: {
 822                         struct sk_buff *skb;
 823                         int amount = 0;
 824
 825                         spin_lock_bh(&sk->sk_receive_queue.lock);
 826                         skb = skb_peek(&sk->sk_receive_queue);
 827                         if (skb != NULL)
 828                                 amount = skb->len;
 829                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 830                         return put_user(amount, (int __user *)arg);
 831                 }
 832
 833                 default:
 834 #ifdef CONFIG_IP_MROUTE
 835                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 836 #else
 837                         return -ENOIOCTLCMD;
 838 #endif
 839         }
 840 }
 841
 842 struct proto raw_prot = {
 843         .name              = "RAW",
 844         .owner             = THIS_MODULE,
 845         .close             = raw_close,
 846         .destroy           = raw_destroy,
 847         .connect           = ip4_datagram_connect,
 848         .disconnect        = udp_disconnect,
 849         .ioctl             = raw_ioctl,
 850         .init              = raw_init,
 851         .setsockopt        = raw_setsockopt,
 852         .getsockopt        = raw_getsockopt,
 853         .sendmsg           = raw_sendmsg,
 854         .recvmsg           = raw_recvmsg,
 855         .bind              = raw_bind,
 856         .backlog_rcv       = raw_rcv_skb,
 857         .hash              = raw_hash_sk,
 858         .unhash            = raw_unhash_sk,
 859         .obj_size          = sizeof(struct raw_sock),
 860         .h.raw_hash        = &raw_v4_hashinfo,
 861 #ifdef CONFIG_COMPAT
 862         .compat_setsockopt = compat_raw_setsockopt,
 863         .compat_getsockopt = compat_raw_getsockopt,
 864 #endif
 865 };
 866
 867 #ifdef CONFIG_PROC_FS
 868 static struct sock *raw_get_first(struct seq_file *seq)
 869 {
 870         struct sock *sk;
 871         struct raw_iter_state *state = raw_seq_private(seq);
 872
 873         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
 874                         ++state->bucket) {
 875                 struct hlist_node *node;
 876
 877                 sk_for_each(sk, node, &state->h->ht[state->bucket])
 878                         if (sock_net(sk) == seq_file_net(seq))
 879                                 goto found;
 880         }
 881         sk = NULL;
 882 found:
 883         return sk;
 884 }
 885
 886 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 887 {
 888         struct raw_iter_state *state = raw_seq_private(seq);
 889
 890         do {
 891                 sk = sk_next(sk);
 892 try_again:
 893                 ;
 894         } while (sk && sock_net(sk) != seq_file_net(seq));
 895
 896         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
 897                 sk = sk_head(&state->h->ht[state->bucket]);
 898                 goto try_again;
 899         }
 900         return sk;
 901 }
 902
 903 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 904 {
 905         struct sock *sk = raw_get_first(seq);
 906
 907         if (sk)
 908                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 909                         --pos;
 910         return pos ? NULL : sk;
 911 }
 912
 913 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 914 {
 915         struct raw_iter_state *state = raw_seq_private(seq);
 916
 917         read_lock(&state->h->lock);
 918         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 919 }
 920 EXPORT_SYMBOL_GPL(raw_seq_start);
 921
 922 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 923 {
 924         struct sock *sk;
 925
 926         if (v == SEQ_START_TOKEN)
 927                 sk = raw_get_first(seq);
 928         else
 929                 sk = raw_get_next(seq, v);
 930         ++*pos;
 931         return sk;
 932 }
 933 EXPORT_SYMBOL_GPL(raw_seq_next);
 934
 935 void raw_seq_stop(struct seq_file *seq, void *v)
 936 {
 937         struct raw_iter_state *state = raw_seq_private(seq);
 938
 939         read_unlock(&state->h->lock);
 940 }
 941 EXPORT_SYMBOL_GPL(raw_seq_stop);
 942
 943 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
 944 {
 945         struct inet_sock *inet = inet_sk(sp);
 946         __be32 dest = inet->daddr,
 947                src = inet->rcv_saddr;
 948         __u16 destp = 0,
 949               srcp  = inet->num;
 950
 951         seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
 952                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
 953                 i, src, srcp, dest, destp, sp->sk_state,
 954                 sk_wmem_alloc_get(sp),
 955                 sk_rmem_alloc_get(sp),
 956                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 957                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
 958 }
 959
 960 static int raw_seq_show(struct seq_file *seq, void *v)
 961 {
 962         if (v == SEQ_START_TOKEN)
 963                 seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
 964                                 "rx_queue tr tm->when retrnsmt   uid  timeout "
 965                                 "inode ref pointer drops\n");
 966         else
 967                 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
 968         return 0;
 969 }
 970
 971 static const struct seq_operations raw_seq_ops = {
 972         .start = raw_seq_start,
 973         .next  = raw_seq_next,
 974         .stop  = raw_seq_stop,
 975         .show  = raw_seq_show,
 976 };
 977
 978 int raw_seq_open(struct inode *ino, struct file *file,
 979                  struct raw_hashinfo *h, const struct seq_operations *ops)
 980 {
 981         int err;
 982         struct raw_iter_state *i;
 983
 984         err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
 985         if (err < 0)
 986                 return err;
 987
 988         i = raw_seq_private((struct seq_file *)file->private_data);
 989         i->h = h;
 990         return 0;
 991 }
 992 EXPORT_SYMBOL_GPL(raw_seq_open);
 993
 994 static int raw_v4_seq_open(struct inode *inode, struct file *file)
 995 {
 996         return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
 997 }
 998
 999 static const struct file_operations raw_seq_fops = {
1000         .owner   = THIS_MODULE,
1001         .open    = raw_v4_seq_open,
1002         .read    = seq_read,
1003         .llseek  = seq_lseek,
1004         .release = seq_release_net,
1005 };
1006
1007 static __net_init int raw_init_net(struct net *net)
1008 {
1009         if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
1010                 return -ENOMEM;
1011
1012         return 0;
1013 }
1014
1015 static __net_exit void raw_exit_net(struct net *net)
1016 {
1017         proc_net_remove(net, "raw");
1018 }
1019
1020 static __net_initdata struct pernet_operations raw_net_ops = {
1021         .init = raw_init_net,
1022         .exit = raw_exit_net,
1023 };
1024
1025 int __init raw_proc_init(void)
1026 {
1027         return register_pernet_subsys(&raw_net_ops);
1028 }
1029
1030 void __init raw_proc_exit(void)
1031 {
1032         unregister_pernet_subsys(&raw_net_ops);
1033 }
1034 #endif /* CONFIG_PROC_FS */