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)->inet_num & (RAW_HTABLE_SIZE - 1)];
  91
  92         write_lock_bh(&h->lock);
  93         sk_add_node(sk, head);
  94         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
  95         write_unlock_bh(&h->lock);
  96 }
  97 EXPORT_SYMBOL_GPL(raw_hash_sk);
  98
  99 void raw_unhash_sk(struct sock *sk)
 100 {
 101         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
 102
 103         write_lock_bh(&h->lock);
 104         if (sk_del_node_init(sk))
 105                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 106         write_unlock_bh(&h->lock);
 107 }
 108 EXPORT_SYMBOL_GPL(raw_unhash_sk);
 109
 110 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
 111                 unsigned short num, __be32 raddr, __be32 laddr, int dif)
 112 {
 113         struct hlist_node *node;
 114
 115         sk_for_each_from(sk, node) {
 116                 struct inet_sock *inet = inet_sk(sk);
 117
 118                 if (net_eq(sock_net(sk), net) && inet->inet_num == num  &&
 119                     !(inet->inet_daddr && inet->inet_daddr != raddr)    &&
 120                     !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
 121                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 122                         goto found; /* gotcha */
 123         }
 124         sk = NULL;
 125 found:
 126         return sk;
 127 }
 128
 129 /*
 130  *      0 - deliver
 131  *      1 - block
 132  */
 133 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 134 {
 135         int type;
 136
 137         if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
 138                 return 1;
 139
 140         type = icmp_hdr(skb)->type;
 141         if (type < 32) {
 142                 __u32 data = raw_sk(sk)->filter.data;
 143
 144                 return ((1 << type) & data) != 0;
 145         }
 146
 147         /* Do not block unknown ICMP types */
 148         return 0;
 149 }
 150
 151 /* IP input processing comes here for RAW socket delivery.
 152  * Caller owns SKB, so we must make clones.
 153  *
 154  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 155  * -> It does. And not only TOS, but all IP header.
 156  */
 157 static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 158 {
 159         struct sock *sk;
 160         struct hlist_head *head;
 161         int delivered = 0;
 162         struct net *net;
 163
 164         read_lock(&raw_v4_hashinfo.lock);
 165         head = &raw_v4_hashinfo.ht[hash];
 166         if (hlist_empty(head))
 167                 goto out;
 168
 169         net = dev_net(skb->dev);
 170         sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
 171                              iph->saddr, iph->daddr,
 172                              skb->dev->ifindex);
 173
 174         while (sk) {
 175                 delivered = 1;
 176                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 177                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 178
 179                         /* Not releasing hash table! */
 180                         if (clone)
 181                                 raw_rcv(sk, clone);
 182                 }
 183                 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
 184                                      iph->saddr, iph->daddr,
 185                                      skb->dev->ifindex);
 186         }
 187 out:
 188         read_unlock(&raw_v4_hashinfo.lock);
 189         return delivered;
 190 }
 191
 192 int raw_local_deliver(struct sk_buff *skb, int protocol)
 193 {
 194         int hash;
 195         struct sock *raw_sk;
 196
 197         hash = protocol & (RAW_HTABLE_SIZE - 1);
 198         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 199
 200         /* If there maybe a raw socket we must check - if not we
 201          * don't care less
 202          */
 203         if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
 204                 raw_sk = NULL;
 205
 206         return raw_sk != NULL;
 207
 208 }
 209
 210 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
 211 {
 212         struct inet_sock *inet = inet_sk(sk);
 213         const int type = icmp_hdr(skb)->type;
 214         const int code = icmp_hdr(skb)->code;
 215         int err = 0;
 216         int harderr = 0;
 217
 218         /* Report error on raw socket, if:
 219            1. User requested ip_recverr.
 220            2. Socket is connected (otherwise the error indication
 221               is useless without ip_recverr and error is hard.
 222          */
 223         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 224                 return;
 225
 226         switch (type) {
 227         default:
 228         case ICMP_TIME_EXCEEDED:
 229                 err = EHOSTUNREACH;
 230                 break;
 231         case ICMP_SOURCE_QUENCH:
 232                 return;
 233         case ICMP_PARAMETERPROB:
 234                 err = EPROTO;
 235                 harderr = 1;
 236                 break;
 237         case ICMP_DEST_UNREACH:
 238                 err = EHOSTUNREACH;
 239                 if (code > NR_ICMP_UNREACH)
 240                         break;
 241                 err = icmp_err_convert[code].errno;
 242                 harderr = icmp_err_convert[code].fatal;
 243                 if (code == ICMP_FRAG_NEEDED) {
 244                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 245                         err = EMSGSIZE;
 246                 }
 247         }
 248
 249         if (inet->recverr) {
 250                 struct iphdr *iph = (struct iphdr *)skb->data;
 251                 u8 *payload = skb->data + (iph->ihl << 2);
 252
 253                 if (inet->hdrincl)
 254                         payload = skb->data;
 255                 ip_icmp_error(sk, skb, err, 0, info, payload);
 256         }
 257
 258         if (inet->recverr || harderr) {
 259                 sk->sk_err = err;
 260                 sk->sk_error_report(sk);
 261         }
 262 }
 263
 264 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
 265 {
 266         int hash;
 267         struct sock *raw_sk;
 268         struct iphdr *iph;
 269         struct net *net;
 270
 271         hash = protocol & (RAW_HTABLE_SIZE - 1);
 272
 273         read_lock(&raw_v4_hashinfo.lock);
 274         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 275         if (raw_sk != NULL) {
 276                 iph = (struct iphdr *)skb->data;
 277                 net = dev_net(skb->dev);
 278
 279                 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
 280                                                 iph->daddr, iph->saddr,
 281                                                 skb->dev->ifindex)) != NULL) {
 282                         raw_err(raw_sk, skb, info);
 283                         raw_sk = sk_next(raw_sk);
 284                         iph = (struct iphdr *)skb->data;
 285                 }
 286         }
 287         read_unlock(&raw_v4_hashinfo.lock);
 288 }
 289
 290 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 291 {
 292         /* Charge it to the socket. */
 293
 294         if (sock_queue_rcv_skb(sk, skb) < 0) {
 295                 kfree_skb(skb);
 296                 return NET_RX_DROP;
 297         }
 298
 299         return NET_RX_SUCCESS;
 300 }
 301
 302 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 303 {
 304         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 305                 atomic_inc(&sk->sk_drops);
 306                 kfree_skb(skb);
 307                 return NET_RX_DROP;
 308         }
 309         nf_reset(skb);
 310
 311         skb_push(skb, skb->data - skb_network_header(skb));
 312
 313         raw_rcv_skb(sk, skb);
 314         return 0;
 315 }
 316
 317 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 318                         struct rtable *rt,
 319                         unsigned int flags)
 320 {
 321         struct inet_sock *inet = inet_sk(sk);
 322         struct net *net = sock_net(sk);
 323         struct iphdr *iph;
 324         struct sk_buff *skb;
 325         unsigned int iphlen;
 326         int err;
 327
 328         if (length > rt->u.dst.dev->mtu) {
 329                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->inet_dport,
 330                                rt->u.dst.dev->mtu);
 331                 return -EMSGSIZE;
 332         }
 333         if (flags&MSG_PROBE)
 334                 goto out;
 335
 336         skb = sock_alloc_send_skb(sk,
 337                                   length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15,
 338                                   flags & MSG_DONTWAIT, &err);
 339         if (skb == NULL)
 340                 goto error;
 341         skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
 342
 343         skb->priority = sk->sk_priority;
 344         skb->mark = sk->sk_mark;
 345         skb_dst_set(skb, dst_clone(&rt->u.dst));
 346
 347         skb_reset_network_header(skb);
 348         iph = ip_hdr(skb);
 349         skb_put(skb, length);
 350
 351         skb->ip_summed = CHECKSUM_NONE;
 352
 353         skb->transport_header = skb->network_header;
 354         err = -EFAULT;
 355         if (memcpy_fromiovecend((void *)iph, from, 0, length))
 356                 goto error_free;
 357
 358         iphlen = iph->ihl * 4;
 359
 360         /*
 361          * We don't want to modify the ip header, but we do need to
 362          * be sure that it won't cause problems later along the network
 363          * stack.  Specifically we want to make sure that iph->ihl is a
 364          * sane value.  If ihl points beyond the length of the buffer passed
 365          * in, reject the frame as invalid
 366          */
 367         err = -EINVAL;
 368         if (iphlen > length)
 369                 goto error_free;
 370
 371         if (iphlen >= sizeof(*iph)) {
 372                 if (!iph->saddr)
 373                         iph->saddr = rt->rt_src;
 374                 iph->check   = 0;
 375                 iph->tot_len = htons(length);
 376                 if (!iph->id)
 377                         ip_select_ident(iph, &rt->u.dst, NULL);
 378
 379                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 380         }
 381         if (iph->protocol == IPPROTO_ICMP)
 382                 icmp_out_count(net, ((struct icmphdr *)
 383                         skb_transport_header(skb))->type);
 384
 385         err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 386                       dst_output);
 387         if (err > 0)
 388                 err = net_xmit_errno(err);
 389         if (err)
 390                 goto error;
 391 out:
 392         return 0;
 393
 394 error_free:
 395         kfree_skb(skb);
 396 error:
 397         IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
 398         if (err == -ENOBUFS && !inet->recverr)
 399                 err = 0;
 400         return err;
 401 }
 402
 403 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 404 {
 405         struct iovec *iov;
 406         u8 __user *type = NULL;
 407         u8 __user *code = NULL;
 408         int probed = 0;
 409         unsigned int i;
 410
 411         if (!msg->msg_iov)
 412                 return 0;
 413
 414         for (i = 0; i < msg->msg_iovlen; i++) {
 415                 iov = &msg->msg_iov[i];
 416                 if (!iov)
 417                         continue;
 418
 419                 switch (fl->proto) {
 420                 case IPPROTO_ICMP:
 421                         /* check if one-byte field is readable or not. */
 422                         if (iov->iov_base && iov->iov_len < 1)
 423                                 break;
 424
 425                         if (!type) {
 426                                 type = iov->iov_base;
 427                                 /* check if code field is readable or not. */
 428                                 if (iov->iov_len > 1)
 429                                         code = type + 1;
 430                         } else if (!code)
 431                                 code = iov->iov_base;
 432
 433                         if (type && code) {
 434                                 if (get_user(fl->fl_icmp_type, type) ||
 435                                     get_user(fl->fl_icmp_code, code))
 436                                         return -EFAULT;
 437                                 probed = 1;
 438                         }
 439                         break;
 440                 default:
 441                         probed = 1;
 442                         break;
 443                 }
 444                 if (probed)
 445                         break;
 446         }
 447         return 0;
 448 }
 449
 450 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 451                        size_t len)
 452 {
 453         struct inet_sock *inet = inet_sk(sk);
 454         struct ipcm_cookie ipc;
 455         struct rtable *rt = NULL;
 456         int free = 0;
 457         __be32 daddr;
 458         __be32 saddr;
 459         u8  tos;
 460         int err;
 461
 462         err = -EMSGSIZE;
 463         if (len > 0xFFFF)
 464                 goto out;
 465
 466         /*
 467          *      Check the flags.
 468          */
 469
 470         err = -EOPNOTSUPP;
 471         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 472                 goto out;               /* compatibility */
 473
 474         /*
 475          *      Get and verify the address.
 476          */
 477
 478         if (msg->msg_namelen) {
 479                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
 480                 err = -EINVAL;
 481                 if (msg->msg_namelen < sizeof(*usin))
 482                         goto out;
 483                 if (usin->sin_family != AF_INET) {
 484                         static int complained;
 485                         if (!complained++)
 486                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 487                                                  "raw sendmsg. Fix it!\n",
 488                                                  current->comm);
 489                         err = -EAFNOSUPPORT;
 490                         if (usin->sin_family)
 491                                 goto out;
 492                 }
 493                 daddr = usin->sin_addr.s_addr;
 494                 /* ANK: I did not forget to get protocol from port field.
 495                  * I just do not know, who uses this weirdness.
 496                  * IP_HDRINCL is much more convenient.
 497                  */
 498         } else {
 499                 err = -EDESTADDRREQ;
 500                 if (sk->sk_state != TCP_ESTABLISHED)
 501                         goto out;
 502                 daddr = inet->inet_daddr;
 503         }
 504
 505         ipc.addr = inet->inet_saddr;
 506         ipc.opt = NULL;
 507         ipc.shtx.flags = 0;
 508         ipc.oif = sk->sk_bound_dev_if;
 509
 510         if (msg->msg_controllen) {
 511                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
 512                 if (err)
 513                         goto out;
 514                 if (ipc.opt)
 515                         free = 1;
 516         }
 517
 518         saddr = ipc.addr;
 519         ipc.addr = daddr;
 520
 521         if (!ipc.opt)
 522                 ipc.opt = inet->opt;
 523
 524         if (ipc.opt) {
 525                 err = -EINVAL;
 526                 /* Linux does not mangle headers on raw sockets,
 527                  * so that IP options + IP_HDRINCL is non-sense.
 528                  */
 529                 if (inet->hdrincl)
 530                         goto done;
 531                 if (ipc.opt->srr) {
 532                         if (!daddr)
 533                                 goto done;
 534                         daddr = ipc.opt->faddr;
 535                 }
 536         }
 537         tos = RT_CONN_FLAGS(sk);
 538         if (msg->msg_flags & MSG_DONTROUTE)
 539                 tos |= RTO_ONLINK;
 540
 541         if (ipv4_is_multicast(daddr)) {
 542                 if (!ipc.oif)
 543                         ipc.oif = inet->mc_index;
 544                 if (!saddr)
 545                         saddr = inet->mc_addr;
 546         }
 547
 548         {
 549                 struct flowi fl = { .oif = ipc.oif,
 550                                     .mark = sk->sk_mark,
 551                                     .nl_u = { .ip4_u =
 552                                               { .daddr = daddr,
 553                                                 .saddr = saddr,
 554                                                 .tos = tos } },
 555                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 556                                                              sk->sk_protocol,
 557                                   };
 558                 if (!inet->hdrincl) {
 559                         err = raw_probe_proto_opt(&fl, msg);
 560                         if (err)
 561                                 goto done;
 562                 }
 563
 564                 security_sk_classify_flow(sk, &fl);
 565                 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 1);
 566         }
 567         if (err)
 568                 goto done;
 569
 570         err = -EACCES;
 571         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 572                 goto done;
 573
 574         if (msg->msg_flags & MSG_CONFIRM)
 575                 goto do_confirm;
 576 back_from_confirm:
 577
 578         if (inet->hdrincl)
 579                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 580                                         rt, msg->msg_flags);
 581
 582          else {
 583                 if (!ipc.addr)
 584                         ipc.addr = rt->rt_dst;
 585                 lock_sock(sk);
 586                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 587                                         &ipc, &rt, msg->msg_flags);
 588                 if (err)
 589                         ip_flush_pending_frames(sk);
 590                 else if (!(msg->msg_flags & MSG_MORE)) {
 591                         err = ip_push_pending_frames(sk);
 592                         if (err == -ENOBUFS && !inet->recverr)
 593                                 err = 0;
 594                 }
 595                 release_sock(sk);
 596         }
 597 done:
 598         if (free)
 599                 kfree(ipc.opt);
 600         ip_rt_put(rt);
 601
 602 out:
 603         if (err < 0)
 604                 return err;
 605         return len;
 606
 607 do_confirm:
 608         dst_confirm(&rt->u.dst);
 609         if (!(msg->msg_flags & MSG_PROBE) || len)
 610                 goto back_from_confirm;
 611         err = 0;
 612         goto done;
 613 }
 614
 615 static void raw_close(struct sock *sk, long timeout)
 616 {
 617         /*
 618          * Raw sockets may have direct kernel refereneces. Kill them.
 619          */
 620         ip_ra_control(sk, 0, NULL);
 621
 622         sk_common_release(sk);
 623 }
 624
 625 static void raw_destroy(struct sock *sk)
 626 {
 627         lock_sock(sk);
 628         ip_flush_pending_frames(sk);
 629         release_sock(sk);
 630 }
 631
 632 /* This gets rid of all the nasties in af_inet. -DaveM */
 633 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 634 {
 635         struct inet_sock *inet = inet_sk(sk);
 636         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 637         int ret = -EINVAL;
 638         int chk_addr_ret;
 639
 640         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 641                 goto out;
 642         chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
 643         ret = -EADDRNOTAVAIL;
 644         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 645             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 646                 goto out;
 647         inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
 648         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 649                 inet->inet_saddr = 0;  /* Use device */
 650         sk_dst_reset(sk);
 651         ret = 0;
 652 out:    return ret;
 653 }
 654
 655 /*
 656  *      This should be easy, if there is something there
 657  *      we return it, otherwise we block.
 658  */
 659
 660 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 661                        size_t len, int noblock, int flags, int *addr_len)
 662 {
 663         struct inet_sock *inet = inet_sk(sk);
 664         size_t copied = 0;
 665         int err = -EOPNOTSUPP;
 666         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 667         struct sk_buff *skb;
 668
 669         if (flags & MSG_OOB)
 670                 goto out;
 671
 672         if (addr_len)
 673                 *addr_len = sizeof(*sin);
 674
 675         if (flags & MSG_ERRQUEUE) {
 676                 err = ip_recv_error(sk, msg, len);
 677                 goto out;
 678         }
 679
 680         skb = skb_recv_datagram(sk, flags, noblock, &err);
 681         if (!skb)
 682                 goto out;
 683
 684         copied = skb->len;
 685         if (len < copied) {
 686                 msg->msg_flags |= MSG_TRUNC;
 687                 copied = len;
 688         }
 689
 690         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 691         if (err)
 692                 goto done;
 693
 694         sock_recv_ts_and_drops(msg, sk, skb);
 695
 696         /* Copy the address. */
 697         if (sin) {
 698                 sin->sin_family = AF_INET;
 699                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 700                 sin->sin_port = 0;
 701                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 702         }
 703         if (inet->cmsg_flags)
 704                 ip_cmsg_recv(msg, skb);
 705         if (flags & MSG_TRUNC)
 706                 copied = skb->len;
 707 done:
 708         skb_free_datagram(sk, skb);
 709 out:
 710         if (err)
 711                 return err;
 712         return copied;
 713 }
 714
 715 static int raw_init(struct sock *sk)
 716 {
 717         struct raw_sock *rp = raw_sk(sk);
 718
 719         if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
 720                 memset(&rp->filter, 0, sizeof(rp->filter));
 721         return 0;
 722 }
 723
 724 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 725 {
 726         if (optlen > sizeof(struct icmp_filter))
 727                 optlen = sizeof(struct icmp_filter);
 728         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 729                 return -EFAULT;
 730         return 0;
 731 }
 732
 733 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 734 {
 735         int len, ret = -EFAULT;
 736
 737         if (get_user(len, optlen))
 738                 goto out;
 739         ret = -EINVAL;
 740         if (len < 0)
 741                 goto out;
 742         if (len > sizeof(struct icmp_filter))
 743                 len = sizeof(struct icmp_filter);
 744         ret = -EFAULT;
 745         if (put_user(len, optlen) ||
 746             copy_to_user(optval, &raw_sk(sk)->filter, len))
 747                 goto out;
 748         ret = 0;
 749 out:    return ret;
 750 }
 751
 752 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 753                           char __user *optval, unsigned int optlen)
 754 {
 755         if (optname == ICMP_FILTER) {
 756                 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
 757                         return -EOPNOTSUPP;
 758                 else
 759                         return raw_seticmpfilter(sk, optval, optlen);
 760         }
 761         return -ENOPROTOOPT;
 762 }
 763
 764 static int raw_setsockopt(struct sock *sk, int level, int optname,
 765                           char __user *optval, unsigned int optlen)
 766 {
 767         if (level != SOL_RAW)
 768                 return ip_setsockopt(sk, level, optname, optval, optlen);
 769         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 770 }
 771
 772 #ifdef CONFIG_COMPAT
 773 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 774                                  char __user *optval, unsigned int optlen)
 775 {
 776         if (level != SOL_RAW)
 777                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 778         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 779 }
 780 #endif
 781
 782 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 783                           char __user *optval, int __user *optlen)
 784 {
 785         if (optname == ICMP_FILTER) {
 786                 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
 787                         return -EOPNOTSUPP;
 788                 else
 789                         return raw_geticmpfilter(sk, optval, optlen);
 790         }
 791         return -ENOPROTOOPT;
 792 }
 793
 794 static int raw_getsockopt(struct sock *sk, int level, int optname,
 795                           char __user *optval, int __user *optlen)
 796 {
 797         if (level != SOL_RAW)
 798                 return ip_getsockopt(sk, level, optname, optval, optlen);
 799         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 800 }
 801
 802 #ifdef CONFIG_COMPAT
 803 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 804                                  char __user *optval, int __user *optlen)
 805 {
 806         if (level != SOL_RAW)
 807                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 808         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 809 }
 810 #endif
 811
 812 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 813 {
 814         switch (cmd) {
 815                 case SIOCOUTQ: {
 816                         int amount = sk_wmem_alloc_get(sk);
 817
 818                         return put_user(amount, (int __user *)arg);
 819                 }
 820                 case SIOCINQ: {
 821                         struct sk_buff *skb;
 822                         int amount = 0;
 823
 824                         spin_lock_bh(&sk->sk_receive_queue.lock);
 825                         skb = skb_peek(&sk->sk_receive_queue);
 826                         if (skb != NULL)
 827                                 amount = skb->len;
 828                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 829                         return put_user(amount, (int __user *)arg);
 830                 }
 831
 832                 default:
 833 #ifdef CONFIG_IP_MROUTE
 834                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 835 #else
 836                         return -ENOIOCTLCMD;
 837 #endif
 838         }
 839 }
 840
 841 struct proto raw_prot = {
 842         .name              = "RAW",
 843         .owner             = THIS_MODULE,
 844         .close             = raw_close,
 845         .destroy           = raw_destroy,
 846         .connect           = ip4_datagram_connect,
 847         .disconnect        = udp_disconnect,
 848         .ioctl             = raw_ioctl,
 849         .init              = raw_init,
 850         .setsockopt        = raw_setsockopt,
 851         .getsockopt        = raw_getsockopt,
 852         .sendmsg           = raw_sendmsg,
 853         .recvmsg           = raw_recvmsg,
 854         .bind              = raw_bind,
 855         .backlog_rcv       = raw_rcv_skb,
 856         .hash              = raw_hash_sk,
 857         .unhash            = raw_unhash_sk,
 858         .obj_size          = sizeof(struct raw_sock),
 859         .h.raw_hash        = &raw_v4_hashinfo,
 860 #ifdef CONFIG_COMPAT
 861         .compat_setsockopt = compat_raw_setsockopt,
 862         .compat_getsockopt = compat_raw_getsockopt,
 863 #endif
 864 };
 865
 866 #ifdef CONFIG_PROC_FS
 867 static struct sock *raw_get_first(struct seq_file *seq)
 868 {
 869         struct sock *sk;
 870         struct raw_iter_state *state = raw_seq_private(seq);
 871
 872         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
 873                         ++state->bucket) {
 874                 struct hlist_node *node;
 875
 876                 sk_for_each(sk, node, &state->h->ht[state->bucket])
 877                         if (sock_net(sk) == seq_file_net(seq))
 878                                 goto found;
 879         }
 880         sk = NULL;
 881 found:
 882         return sk;
 883 }
 884
 885 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 886 {
 887         struct raw_iter_state *state = raw_seq_private(seq);
 888
 889         do {
 890                 sk = sk_next(sk);
 891 try_again:
 892                 ;
 893         } while (sk && sock_net(sk) != seq_file_net(seq));
 894
 895         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
 896                 sk = sk_head(&state->h->ht[state->bucket]);
 897                 goto try_again;
 898         }
 899         return sk;
 900 }
 901
 902 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 903 {
 904         struct sock *sk = raw_get_first(seq);
 905
 906         if (sk)
 907                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 908                         --pos;
 909         return pos ? NULL : sk;
 910 }
 911
 912 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 913 {
 914         struct raw_iter_state *state = raw_seq_private(seq);
 915
 916         read_lock(&state->h->lock);
 917         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 918 }
 919 EXPORT_SYMBOL_GPL(raw_seq_start);
 920
 921 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 922 {
 923         struct sock *sk;
 924
 925         if (v == SEQ_START_TOKEN)
 926                 sk = raw_get_first(seq);
 927         else
 928                 sk = raw_get_next(seq, v);
 929         ++*pos;
 930         return sk;
 931 }
 932 EXPORT_SYMBOL_GPL(raw_seq_next);
 933
 934 void raw_seq_stop(struct seq_file *seq, void *v)
 935 {
 936         struct raw_iter_state *state = raw_seq_private(seq);
 937
 938         read_unlock(&state->h->lock);
 939 }
 940 EXPORT_SYMBOL_GPL(raw_seq_stop);
 941
 942 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
 943 {
 944         struct inet_sock *inet = inet_sk(sp);
 945         __be32 dest = inet->inet_daddr,
 946                src = inet->inet_rcv_saddr;
 947         __u16 destp = 0,
 948               srcp  = inet->inet_num;
 949
 950         seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
 951                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
 952                 i, src, srcp, dest, destp, sp->sk_state,
 953                 sk_wmem_alloc_get(sp),
 954                 sk_rmem_alloc_get(sp),
 955                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 956                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
 957 }
 958
 959 static int raw_seq_show(struct seq_file *seq, void *v)
 960 {
 961         if (v == SEQ_START_TOKEN)
 962                 seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
 963                                 "rx_queue tr tm->when retrnsmt   uid  timeout "
 964                                 "inode ref pointer drops\n");
 965         else
 966                 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
 967         return 0;
 968 }
 969
 970 static const struct seq_operations raw_seq_ops = {
 971         .start = raw_seq_start,
 972         .next  = raw_seq_next,
 973         .stop  = raw_seq_stop,
 974         .show  = raw_seq_show,
 975 };
 976
 977 int raw_seq_open(struct inode *ino, struct file *file,
 978                  struct raw_hashinfo *h, const struct seq_operations *ops)
 979 {
 980         int err;
 981         struct raw_iter_state *i;
 982
 983         err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
 984         if (err < 0)
 985                 return err;
 986
 987         i = raw_seq_private((struct seq_file *)file->private_data);
 988         i->h = h;
 989         return 0;
 990 }
 991 EXPORT_SYMBOL_GPL(raw_seq_open);
 992
 993 static int raw_v4_seq_open(struct inode *inode, struct file *file)
 994 {
 995         return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
 996 }
 997
 998 static const struct file_operations raw_seq_fops = {
 999         .owner   = THIS_MODULE,
1000         .open    = raw_v4_seq_open,
1001         .read    = seq_read,
1002         .llseek  = seq_lseek,
1003         .release = seq_release_net,
1004 };
1005
1006 static __net_init int raw_init_net(struct net *net)
1007 {
1008         if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
1009                 return -ENOMEM;
1010
1011         return 0;
1012 }
1013
1014 static __net_exit void raw_exit_net(struct net *net)
1015 {
1016         proc_net_remove(net, "raw");
1017 }
1018
1019 static __net_initdata struct pernet_operations raw_net_ops = {
1020         .init = raw_init_net,
1021         .exit = raw_exit_net,
1022 };
1023
1024 int __init raw_proc_init(void)
1025 {
1026         return register_pernet_subsys(&raw_net_ops);
1027 }
1028
1029 void __init raw_proc_exit(void)
1030 {
1031         unregister_pernet_subsys(&raw_net_ops);
1032 }
1033 #endif /* CONFIG_PROC_FS */