net/ipv4/raw.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              RAW - implementation of IP "raw" sockets.
   7  *
   8  * Authors:     Ross Biro
   9  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  10  *
  11  * Fixes:
  12  *              Alan Cox        :       verify_area() fixed up
  13  *              Alan Cox        :       ICMP error handling
  14  *              Alan Cox        :       EMSGSIZE if you send too big a packet
  15  *              Alan Cox        :       Now uses generic datagrams and shared
  16  *                                      skbuff library. No more peek crashes,
  17  *                                      no more backlogs
  18  *              Alan Cox        :       Checks sk->broadcast.
  19  *              Alan Cox        :       Uses skb_free_datagram/skb_copy_datagram
  20  *              Alan Cox        :       Raw passes ip options too
  21  *              Alan Cox        :       Setsocketopt added
  22  *              Alan Cox        :       Fixed error return for broadcasts
  23  *              Alan Cox        :       Removed wake_up calls
  24  *              Alan Cox        :       Use ttl/tos
  25  *              Alan Cox        :       Cleaned up old debugging
  26  *              Alan Cox        :       Use new kernel side addresses
  27  *      Arnt Gulbrandsen        :       Fixed MSG_DONTROUTE in raw sockets.
  28  *              Alan Cox        :       BSD style RAW socket demultiplexing.
  29  *              Alan Cox        :       Beginnings of mrouted support.
  30  *              Alan Cox        :       Added IP_HDRINCL option.
  31  *              Alan Cox        :       Skip broadcast check if BSDism set.
  32  *              David S. Miller :       New socket lookup architecture.
  33  *
  34  *              This program is free software; you can redistribute it and/or
  35  *              modify it under the terms of the GNU General Public License
  36  *              as published by the Free Software Foundation; either version
  37  *              2 of the License, or (at your option) any later version.
  38  */
  39
  40 #include <linux/types.h>
  41 #include <asm/atomic.h>
  42 #include <asm/byteorder.h>
  43 #include <asm/current.h>
  44 #include <asm/uaccess.h>
  45 #include <asm/ioctls.h>
  46 #include <linux/stddef.h>
  47 #include <linux/slab.h>
  48 #include <linux/errno.h>
  49 #include <linux/aio.h>
  50 #include <linux/kernel.h>
  51 #include <linux/spinlock.h>
  52 #include <linux/sockios.h>
  53 #include <linux/socket.h>
  54 #include <linux/in.h>
  55 #include <linux/mroute.h>
  56 #include <linux/netdevice.h>
  57 #include <linux/in_route.h>
  58 #include <linux/route.h>
  59 #include <linux/skbuff.h>
  60 #include <net/net_namespace.h>
  61 #include <net/dst.h>
  62 #include <net/sock.h>
  63 #include <linux/gfp.h>
  64 #include <linux/ip.h>
  65 #include <linux/net.h>
  66 #include <net/ip.h>
  67 #include <net/icmp.h>
  68 #include <net/udp.h>
  69 #include <net/raw.h>
  70 #include <net/snmp.h>
  71 #include <net/tcp_states.h>
  72 #include <net/inet_common.h>
  73 #include <net/checksum.h>
  74 #include <net/xfrm.h>
  75 #include <linux/rtnetlink.h>
  76 #include <linux/proc_fs.h>
  77 #include <linux/seq_file.h>
  78 #include <linux/netfilter.h>
  79 #include <linux/netfilter_ipv4.h>
  80
  81 static struct raw_hashinfo raw_v4_hashinfo = {
  82         .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
  83 };
  84
  85 void raw_hash_sk(struct sock *sk)
  86 {
  87         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
  88         struct hlist_head *head;
  89
  90         head = &h->ht[inet_sk(sk)->num & (RAW_HTABLE_SIZE - 1)];
  91
  92         write_lock_bh(&h->lock);
  93         sk_add_node(sk, head);
  94         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
  95         write_unlock_bh(&h->lock);
  96 }
  97 EXPORT_SYMBOL_GPL(raw_hash_sk);
  98
  99 void raw_unhash_sk(struct sock *sk)
 100 {
 101         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
 102
 103         write_lock_bh(&h->lock);
 104         if (sk_del_node_init(sk))
 105                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 106         write_unlock_bh(&h->lock);
 107 }
 108 EXPORT_SYMBOL_GPL(raw_unhash_sk);
 109
 110 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
 111                 unsigned short num, __be32 raddr, __be32 laddr, int dif)
 112 {
 113         struct hlist_node *node;
 114
 115         sk_for_each_from(sk, node) {
 116                 struct inet_sock *inet = inet_sk(sk);
 117
 118                 if (net_eq(sock_net(sk), net) && inet->num == num       &&
 119                     !(inet->daddr && inet->daddr != raddr)              &&
 120                     !(inet->rcv_saddr && inet->rcv_saddr != laddr)      &&
 121                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 122                         goto found; /* gotcha */
 123         }
 124         sk = NULL;
 125 found:
 126         return sk;
 127 }
 128
 129 /*
 130  *      0 - deliver
 131  *      1 - block
 132  */
 133 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 134 {
 135         int type;
 136
 137         if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
 138                 return 1;
 139
 140         type = icmp_hdr(skb)->type;
 141         if (type < 32) {
 142                 __u32 data = raw_sk(sk)->filter.data;
 143
 144                 return ((1 << type) & data) != 0;
 145         }
 146
 147         /* Do not block unknown ICMP types */
 148         return 0;
 149 }
 150
 151 /* IP input processing comes here for RAW socket delivery.
 152  * Caller owns SKB, so we must make clones.
 153  *
 154  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 155  * -> It does. And not only TOS, but all IP header.
 156  */
 157 static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 158 {
 159         struct sock *sk;
 160         struct hlist_head *head;
 161         int delivered = 0;
 162         struct net *net;
 163
 164         read_lock(&raw_v4_hashinfo.lock);
 165         head = &raw_v4_hashinfo.ht[hash];
 166         if (hlist_empty(head))
 167                 goto out;
 168
 169         net = dev_net(skb->dev);
 170         sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
 171                              iph->saddr, iph->daddr,
 172                              skb->dev->ifindex);
 173
 174         while (sk) {
 175                 delivered = 1;
 176                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 177                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 178
 179                         /* Not releasing hash table! */
 180                         if (clone)
 181                                 raw_rcv(sk, clone);
 182                 }
 183                 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
 184                                      iph->saddr, iph->daddr,
 185                                      skb->dev->ifindex);
 186         }
 187 out:
 188         read_unlock(&raw_v4_hashinfo.lock);
 189         return delivered;
 190 }
 191
 192 int raw_local_deliver(struct sk_buff *skb, int protocol)
 193 {
 194         int hash;
 195         struct sock *raw_sk;
 196
 197         hash = protocol & (RAW_HTABLE_SIZE - 1);
 198         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 199
 200         /* If there maybe a raw socket we must check - if not we
 201          * don't care less
 202          */
 203         if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
 204                 raw_sk = NULL;
 205
 206         return raw_sk != NULL;
 207
 208 }
 209
 210 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
 211 {
 212         struct inet_sock *inet = inet_sk(sk);
 213         const int type = icmp_hdr(skb)->type;
 214         const int code = icmp_hdr(skb)->code;
 215         int err = 0;
 216         int harderr = 0;
 217
 218         /* Report error on raw socket, if:
 219            1. User requested ip_recverr.
 220            2. Socket is connected (otherwise the error indication
 221               is useless without ip_recverr and error is hard.
 222          */
 223         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 224                 return;
 225
 226         switch (type) {
 227         default:
 228         case ICMP_TIME_EXCEEDED:
 229                 err = EHOSTUNREACH;
 230                 break;
 231         case ICMP_SOURCE_QUENCH:
 232                 return;
 233         case ICMP_PARAMETERPROB:
 234                 err = EPROTO;
 235                 harderr = 1;
 236                 break;
 237         case ICMP_DEST_UNREACH:
 238                 err = EHOSTUNREACH;
 239                 if (code > NR_ICMP_UNREACH)
 240                         break;
 241                 err = icmp_err_convert[code].errno;
 242                 harderr = icmp_err_convert[code].fatal;
 243                 if (code == ICMP_FRAG_NEEDED) {
 244                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 245                         err = EMSGSIZE;
 246                 }
 247         }
 248
 249         if (inet->recverr) {
 250                 struct iphdr *iph = (struct iphdr *)skb->data;
 251                 u8 *payload = skb->data + (iph->ihl << 2);
 252
 253                 if (inet->hdrincl)
 254                         payload = skb->data;
 255                 ip_icmp_error(sk, skb, err, 0, info, payload);
 256         }
 257
 258         if (inet->recverr || harderr) {
 259                 sk->sk_err = err;
 260                 sk->sk_error_report(sk);
 261         }
 262 }
 263
 264 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
 265 {
 266         int hash;
 267         struct sock *raw_sk;
 268         struct iphdr *iph;
 269         struct net *net;
 270
 271         hash = protocol & (RAW_HTABLE_SIZE - 1);
 272
 273         read_lock(&raw_v4_hashinfo.lock);
 274         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 275         if (raw_sk != NULL) {
 276                 iph = (struct iphdr *)skb->data;
 277                 net = dev_net(skb->dev);
 278
 279                 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
 280                                                 iph->daddr, iph->saddr,
 281                                                 skb->dev->ifindex)) != NULL) {
 282                         raw_err(raw_sk, skb, info);
 283                         raw_sk = sk_next(raw_sk);
 284                         iph = (struct iphdr *)skb->data;
 285                 }
 286         }
 287         read_unlock(&raw_v4_hashinfo.lock);
 288 }
 289
 290 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 291 {
 292         /* Charge it to the socket. */
 293
 294         if (sock_queue_rcv_skb(sk, skb) < 0) {
 295                 atomic_inc(&sk->sk_drops);
 296                 kfree_skb(skb);
 297                 return NET_RX_DROP;
 298         }
 299
 300         return NET_RX_SUCCESS;
 301 }
 302
 303 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 304 {
 305         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 306                 atomic_inc(&sk->sk_drops);
 307                 kfree_skb(skb);
 308                 return NET_RX_DROP;
 309         }
 310         nf_reset(skb);
 311
 312         skb_push(skb, skb->data - skb_network_header(skb));
 313
 314         raw_rcv_skb(sk, skb);
 315         return 0;
 316 }
 317
 318 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 319                         struct rtable *rt,
 320                         unsigned int flags)
 321 {
 322         struct inet_sock *inet = inet_sk(sk);
 323         struct net *net = sock_net(sk);
 324         struct iphdr *iph;
 325         struct sk_buff *skb;
 326         unsigned int iphlen;
 327         int err;
 328
 329         if (length > rt->u.dst.dev->mtu) {
 330                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
 331                                rt->u.dst.dev->mtu);
 332                 return -EMSGSIZE;
 333         }
 334         if (flags&MSG_PROBE)
 335                 goto out;
 336
 337         skb = sock_alloc_send_skb(sk,
 338                                   length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15,
 339                                   flags & MSG_DONTWAIT, &err);
 340         if (skb == NULL)
 341                 goto error;
 342         skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
 343
 344         skb->priority = sk->sk_priority;
 345         skb->mark = sk->sk_mark;
 346         skb_dst_set(skb, dst_clone(&rt->u.dst));
 347
 348         skb_reset_network_header(skb);
 349         iph = ip_hdr(skb);
 350         skb_put(skb, length);
 351
 352         skb->ip_summed = CHECKSUM_NONE;
 353
 354         skb->transport_header = skb->network_header;
 355         err = memcpy_fromiovecend((void *)iph, from, 0, length);
 356         if (err)
 357                 goto error_fault;
 358
 359         /* We don't modify invalid header */
 360         iphlen = iph->ihl * 4;
 361         if (iphlen >= sizeof(*iph) && iphlen <= length) {
 362                 if (!iph->saddr)
 363                         iph->saddr = rt->rt_src;
 364                 iph->check   = 0;
 365                 iph->tot_len = htons(length);
 366                 if (!iph->id)
 367                         ip_select_ident(iph, &rt->u.dst, NULL);
 368
 369                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 370         }
 371         if (iph->protocol == IPPROTO_ICMP)
 372                 icmp_out_count(net, ((struct icmphdr *)
 373                         skb_transport_header(skb))->type);
 374
 375         err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 376                       dst_output);
 377         if (err > 0)
 378                 err = net_xmit_errno(err);
 379         if (err)
 380                 goto error;
 381 out:
 382         return 0;
 383
 384 error_fault:
 385         err = -EFAULT;
 386         kfree_skb(skb);
 387 error:
 388         IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
 389         if (err == -ENOBUFS && !inet->recverr)
 390                 err = 0;
 391         return err;
 392 }
 393
 394 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 395 {
 396         struct iovec *iov;
 397         u8 __user *type = NULL;
 398         u8 __user *code = NULL;
 399         int probed = 0;
 400         unsigned int i;
 401
 402         if (!msg->msg_iov)
 403                 return 0;
 404
 405         for (i = 0; i < msg->msg_iovlen; i++) {
 406                 iov = &msg->msg_iov[i];
 407                 if (!iov)
 408                         continue;
 409
 410                 switch (fl->proto) {
 411                 case IPPROTO_ICMP:
 412                         /* check if one-byte field is readable or not. */
 413                         if (iov->iov_base && iov->iov_len < 1)
 414                                 break;
 415
 416                         if (!type) {
 417                                 type = iov->iov_base;
 418                                 /* check if code field is readable or not. */
 419                                 if (iov->iov_len > 1)
 420                                         code = type + 1;
 421                         } else if (!code)
 422                                 code = iov->iov_base;
 423
 424                         if (type && code) {
 425                                 if (get_user(fl->fl_icmp_type, type) ||
 426                                     get_user(fl->fl_icmp_code, code))
 427                                         return -EFAULT;
 428                                 probed = 1;
 429                         }
 430                         break;
 431                 default:
 432                         probed = 1;
 433                         break;
 434                 }
 435                 if (probed)
 436                         break;
 437         }
 438         return 0;
 439 }
 440
 441 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 442                        size_t len)
 443 {
 444         struct inet_sock *inet = inet_sk(sk);
 445         struct ipcm_cookie ipc;
 446         struct rtable *rt = NULL;
 447         int free = 0;
 448         __be32 daddr;
 449         __be32 saddr;
 450         u8  tos;
 451         int err;
 452
 453         err = -EMSGSIZE;
 454         if (len > 0xFFFF)
 455                 goto out;
 456
 457         /*
 458          *      Check the flags.
 459          */
 460
 461         err = -EOPNOTSUPP;
 462         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 463                 goto out;               /* compatibility */
 464
 465         /*
 466          *      Get and verify the address.
 467          */
 468
 469         if (msg->msg_namelen) {
 470                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
 471                 err = -EINVAL;
 472                 if (msg->msg_namelen < sizeof(*usin))
 473                         goto out;
 474                 if (usin->sin_family != AF_INET) {
 475                         static int complained;
 476                         if (!complained++)
 477                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 478                                                  "raw sendmsg. Fix it!\n",
 479                                                  current->comm);
 480                         err = -EAFNOSUPPORT;
 481                         if (usin->sin_family)
 482                                 goto out;
 483                 }
 484                 daddr = usin->sin_addr.s_addr;
 485                 /* ANK: I did not forget to get protocol from port field.
 486                  * I just do not know, who uses this weirdness.
 487                  * IP_HDRINCL is much more convenient.
 488                  */
 489         } else {
 490                 err = -EDESTADDRREQ;
 491                 if (sk->sk_state != TCP_ESTABLISHED)
 492                         goto out;
 493                 daddr = inet->daddr;
 494         }
 495
 496         ipc.addr = inet->saddr;
 497         ipc.opt = NULL;
 498         ipc.shtx.flags = 0;
 499         ipc.oif = sk->sk_bound_dev_if;
 500
 501         if (msg->msg_controllen) {
 502                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
 503                 if (err)
 504                         goto out;
 505                 if (ipc.opt)
 506                         free = 1;
 507         }
 508
 509         saddr = ipc.addr;
 510         ipc.addr = daddr;
 511
 512         if (!ipc.opt)
 513                 ipc.opt = inet->opt;
 514
 515         if (ipc.opt) {
 516                 err = -EINVAL;
 517                 /* Linux does not mangle headers on raw sockets,
 518                  * so that IP options + IP_HDRINCL is non-sense.
 519                  */
 520                 if (inet->hdrincl)
 521                         goto done;
 522                 if (ipc.opt->srr) {
 523                         if (!daddr)
 524                                 goto done;
 525                         daddr = ipc.opt->faddr;
 526                 }
 527         }
 528         tos = RT_CONN_FLAGS(sk);
 529         if (msg->msg_flags & MSG_DONTROUTE)
 530                 tos |= RTO_ONLINK;
 531
 532         if (ipv4_is_multicast(daddr)) {
 533                 if (!ipc.oif)
 534                         ipc.oif = inet->mc_index;
 535                 if (!saddr)
 536                         saddr = inet->mc_addr;
 537         }
 538
 539         {
 540                 struct flowi fl = { .oif = ipc.oif,
 541                                     .mark = sk->sk_mark,
 542                                     .nl_u = { .ip4_u =
 543                                               { .daddr = daddr,
 544                                                 .saddr = saddr,
 545                                                 .tos = tos } },
 546                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 547                                                              sk->sk_protocol,
 548                                   };
 549                 if (!inet->hdrincl) {
 550                         err = raw_probe_proto_opt(&fl, msg);
 551                         if (err)
 552                                 goto done;
 553                 }
 554
 555                 security_sk_classify_flow(sk, &fl);
 556                 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 1);
 557         }
 558         if (err)
 559                 goto done;
 560
 561         err = -EACCES;
 562         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 563                 goto done;
 564
 565         if (msg->msg_flags & MSG_CONFIRM)
 566                 goto do_confirm;
 567 back_from_confirm:
 568
 569         if (inet->hdrincl)
 570                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 571                                         rt, msg->msg_flags);
 572
 573          else {
 574                 if (!ipc.addr)
 575                         ipc.addr = rt->rt_dst;
 576                 lock_sock(sk);
 577                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 578                                         &ipc, &rt, msg->msg_flags);
 579                 if (err)
 580                         ip_flush_pending_frames(sk);
 581                 else if (!(msg->msg_flags & MSG_MORE)) {
 582                         err = ip_push_pending_frames(sk);
 583                         if (err == -ENOBUFS && !inet->recverr)
 584                                 err = 0;
 585                 }
 586                 release_sock(sk);
 587         }
 588 done:
 589         if (free)
 590                 kfree(ipc.opt);
 591         ip_rt_put(rt);
 592
 593 out:
 594         if (err < 0)
 595                 return err;
 596         return len;
 597
 598 do_confirm:
 599         dst_confirm(&rt->u.dst);
 600         if (!(msg->msg_flags & MSG_PROBE) || len)
 601                 goto back_from_confirm;
 602         err = 0;
 603         goto done;
 604 }
 605
 606 static void raw_close(struct sock *sk, long timeout)
 607 {
 608         /*
 609          * Raw sockets may have direct kernel refereneces. Kill them.
 610          */
 611         ip_ra_control(sk, 0, NULL);
 612
 613         sk_common_release(sk);
 614 }
 615
 616 static void raw_destroy(struct sock *sk)
 617 {
 618         lock_sock(sk);
 619         ip_flush_pending_frames(sk);
 620         release_sock(sk);
 621 }
 622
 623 /* This gets rid of all the nasties in af_inet. -DaveM */
 624 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 625 {
 626         struct inet_sock *inet = inet_sk(sk);
 627         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 628         int ret = -EINVAL;
 629         int chk_addr_ret;
 630
 631         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 632                 goto out;
 633         chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
 634         ret = -EADDRNOTAVAIL;
 635         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 636             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 637                 goto out;
 638         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 639         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 640                 inet->saddr = 0;  /* Use device */
 641         sk_dst_reset(sk);
 642         ret = 0;
 643 out:    return ret;
 644 }
 645
 646 /*
 647  *      This should be easy, if there is something there
 648  *      we return it, otherwise we block.
 649  */
 650
 651 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 652                        size_t len, int noblock, int flags, int *addr_len)
 653 {
 654         struct inet_sock *inet = inet_sk(sk);
 655         size_t copied = 0;
 656         int err = -EOPNOTSUPP;
 657         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 658         struct sk_buff *skb;
 659
 660         if (flags & MSG_OOB)
 661                 goto out;
 662
 663         if (addr_len)
 664                 *addr_len = sizeof(*sin);
 665
 666         if (flags & MSG_ERRQUEUE) {
 667                 err = ip_recv_error(sk, msg, len);
 668                 goto out;
 669         }
 670
 671         skb = skb_recv_datagram(sk, flags, noblock, &err);
 672         if (!skb)
 673                 goto out;
 674
 675         copied = skb->len;
 676         if (len < copied) {
 677                 msg->msg_flags |= MSG_TRUNC;
 678                 copied = len;
 679         }
 680
 681         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 682         if (err)
 683                 goto done;
 684
 685         sock_recv_ts_and_drops(msg, sk, skb);
 686
 687         /* Copy the address. */
 688         if (sin) {
 689                 sin->sin_family = AF_INET;
 690                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 691                 sin->sin_port = 0;
 692                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 693         }
 694         if (inet->cmsg_flags)
 695                 ip_cmsg_recv(msg, skb);
 696         if (flags & MSG_TRUNC)
 697                 copied = skb->len;
 698 done:
 699         skb_free_datagram(sk, skb);
 700 out:
 701         if (err)
 702                 return err;
 703         return copied;
 704 }
 705
 706 static int raw_init(struct sock *sk)
 707 {
 708         struct raw_sock *rp = raw_sk(sk);
 709
 710         if (inet_sk(sk)->num == IPPROTO_ICMP)
 711                 memset(&rp->filter, 0, sizeof(rp->filter));
 712         return 0;
 713 }
 714
 715 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 716 {
 717         if (optlen > sizeof(struct icmp_filter))
 718                 optlen = sizeof(struct icmp_filter);
 719         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 720                 return -EFAULT;
 721         return 0;
 722 }
 723
 724 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 725 {
 726         int len, ret = -EFAULT;
 727
 728         if (get_user(len, optlen))
 729                 goto out;
 730         ret = -EINVAL;
 731         if (len < 0)
 732                 goto out;
 733         if (len > sizeof(struct icmp_filter))
 734                 len = sizeof(struct icmp_filter);
 735         ret = -EFAULT;
 736         if (put_user(len, optlen) ||
 737             copy_to_user(optval, &raw_sk(sk)->filter, len))
 738                 goto out;
 739         ret = 0;
 740 out:    return ret;
 741 }
 742
 743 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 744                           char __user *optval, unsigned int optlen)
 745 {
 746         if (optname == ICMP_FILTER) {
 747                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 748                         return -EOPNOTSUPP;
 749                 else
 750                         return raw_seticmpfilter(sk, optval, optlen);
 751         }
 752         return -ENOPROTOOPT;
 753 }
 754
 755 static int raw_setsockopt(struct sock *sk, int level, int optname,
 756                           char __user *optval, unsigned int optlen)
 757 {
 758         if (level != SOL_RAW)
 759                 return ip_setsockopt(sk, level, optname, optval, optlen);
 760         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 761 }
 762
 763 #ifdef CONFIG_COMPAT
 764 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 765                                  char __user *optval, unsigned int optlen)
 766 {
 767         if (level != SOL_RAW)
 768                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 769         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 770 }
 771 #endif
 772
 773 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 774                           char __user *optval, int __user *optlen)
 775 {
 776         if (optname == ICMP_FILTER) {
 777                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 778                         return -EOPNOTSUPP;
 779                 else
 780                         return raw_geticmpfilter(sk, optval, optlen);
 781         }
 782         return -ENOPROTOOPT;
 783 }
 784
 785 static int raw_getsockopt(struct sock *sk, int level, int optname,
 786                           char __user *optval, int __user *optlen)
 787 {
 788         if (level != SOL_RAW)
 789                 return ip_getsockopt(sk, level, optname, optval, optlen);
 790         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 791 }
 792
 793 #ifdef CONFIG_COMPAT
 794 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 795                                  char __user *optval, int __user *optlen)
 796 {
 797         if (level != SOL_RAW)
 798                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 799         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 800 }
 801 #endif
 802
 803 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 804 {
 805         switch (cmd) {
 806                 case SIOCOUTQ: {
 807                         int amount = sk_wmem_alloc_get(sk);
 808
 809                         return put_user(amount, (int __user *)arg);
 810                 }
 811                 case SIOCINQ: {
 812                         struct sk_buff *skb;
 813                         int amount = 0;
 814
 815                         spin_lock_bh(&sk->sk_receive_queue.lock);
 816                         skb = skb_peek(&sk->sk_receive_queue);
 817                         if (skb != NULL)
 818                                 amount = skb->len;
 819                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 820                         return put_user(amount, (int __user *)arg);
 821                 }
 822
 823                 default:
 824 #ifdef CONFIG_IP_MROUTE
 825                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 826 #else
 827                         return -ENOIOCTLCMD;
 828 #endif
 829         }
 830 }
 831
 832 struct proto raw_prot = {
 833         .name              = "RAW",
 834         .owner             = THIS_MODULE,
 835         .close             = raw_close,
 836         .destroy           = raw_destroy,
 837         .connect           = ip4_datagram_connect,
 838         .disconnect        = udp_disconnect,
 839         .ioctl             = raw_ioctl,
 840         .init              = raw_init,
 841         .setsockopt        = raw_setsockopt,
 842         .getsockopt        = raw_getsockopt,
 843         .sendmsg           = raw_sendmsg,
 844         .recvmsg           = raw_recvmsg,
 845         .bind              = raw_bind,
 846         .backlog_rcv       = raw_rcv_skb,
 847         .hash              = raw_hash_sk,
 848         .unhash            = raw_unhash_sk,
 849         .obj_size          = sizeof(struct raw_sock),
 850         .h.raw_hash        = &raw_v4_hashinfo,
 851 #ifdef CONFIG_COMPAT
 852         .compat_setsockopt = compat_raw_setsockopt,
 853         .compat_getsockopt = compat_raw_getsockopt,
 854 #endif
 855 };
 856
 857 #ifdef CONFIG_PROC_FS
 858 static struct sock *raw_get_first(struct seq_file *seq)
 859 {
 860         struct sock *sk;
 861         struct raw_iter_state *state = raw_seq_private(seq);
 862
 863         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
 864                         ++state->bucket) {
 865                 struct hlist_node *node;
 866
 867                 sk_for_each(sk, node, &state->h->ht[state->bucket])
 868                         if (sock_net(sk) == seq_file_net(seq))
 869                                 goto found;
 870         }
 871         sk = NULL;
 872 found:
 873         return sk;
 874 }
 875
 876 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 877 {
 878         struct raw_iter_state *state = raw_seq_private(seq);
 879
 880         do {
 881                 sk = sk_next(sk);
 882 try_again:
 883                 ;
 884         } while (sk && sock_net(sk) != seq_file_net(seq));
 885
 886         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
 887                 sk = sk_head(&state->h->ht[state->bucket]);
 888                 goto try_again;
 889         }
 890         return sk;
 891 }
 892
 893 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 894 {
 895         struct sock *sk = raw_get_first(seq);
 896
 897         if (sk)
 898                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 899                         --pos;
 900         return pos ? NULL : sk;
 901 }
 902
 903 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 904 {
 905         struct raw_iter_state *state = raw_seq_private(seq);
 906
 907         read_lock(&state->h->lock);
 908         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 909 }
 910 EXPORT_SYMBOL_GPL(raw_seq_start);
 911
 912 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 913 {
 914         struct sock *sk;
 915
 916         if (v == SEQ_START_TOKEN)
 917                 sk = raw_get_first(seq);
 918         else
 919                 sk = raw_get_next(seq, v);
 920         ++*pos;
 921         return sk;
 922 }
 923 EXPORT_SYMBOL_GPL(raw_seq_next);
 924
 925 void raw_seq_stop(struct seq_file *seq, void *v)
 926 {
 927         struct raw_iter_state *state = raw_seq_private(seq);
 928
 929         read_unlock(&state->h->lock);
 930 }
 931 EXPORT_SYMBOL_GPL(raw_seq_stop);
 932
 933 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
 934 {
 935         struct inet_sock *inet = inet_sk(sp);
 936         __be32 dest = inet->daddr,
 937                src = inet->rcv_saddr;
 938         __u16 destp = 0,
 939               srcp  = inet->num;
 940
 941         seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
 942                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
 943                 i, src, srcp, dest, destp, sp->sk_state,
 944                 sk_wmem_alloc_get(sp),
 945                 sk_rmem_alloc_get(sp),
 946                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 947                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
 948 }
 949
 950 static int raw_seq_show(struct seq_file *seq, void *v)
 951 {
 952         if (v == SEQ_START_TOKEN)
 953                 seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
 954                                 "rx_queue tr tm->when retrnsmt   uid  timeout "
 955                                 "inode ref pointer drops\n");
 956         else
 957                 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
 958         return 0;
 959 }
 960
 961 static const struct seq_operations raw_seq_ops = {
 962         .start = raw_seq_start,
 963         .next  = raw_seq_next,
 964         .stop  = raw_seq_stop,
 965         .show  = raw_seq_show,
 966 };
 967
 968 int raw_seq_open(struct inode *ino, struct file *file,
 969                  struct raw_hashinfo *h, const struct seq_operations *ops)
 970 {
 971         int err;
 972         struct raw_iter_state *i;
 973
 974         err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
 975         if (err < 0)
 976                 return err;
 977
 978         i = raw_seq_private((struct seq_file *)file->private_data);
 979         i->h = h;
 980         return 0;
 981 }
 982 EXPORT_SYMBOL_GPL(raw_seq_open);
 983
 984 static int raw_v4_seq_open(struct inode *inode, struct file *file)
 985 {
 986         return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
 987 }
 988
 989 static const struct file_operations raw_seq_fops = {
 990         .owner   = THIS_MODULE,
 991         .open    = raw_v4_seq_open,
 992         .read    = seq_read,
 993         .llseek  = seq_lseek,
 994         .release = seq_release_net,
 995 };
 996
 997 static __net_init int raw_init_net(struct net *net)
 998 {
 999         if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
1000                 return -ENOMEM;
1001
1002         return 0;
1003 }
1004
1005 static __net_exit void raw_exit_net(struct net *net)
1006 {
1007         proc_net_remove(net, "raw");
1008 }
1009
1010 static __net_initdata struct pernet_operations raw_net_ops = {
1011         .init = raw_init_net,
1012         .exit = raw_exit_net,
1013 };
1014
1015 int __init raw_proc_init(void)
1016 {
1017         return register_pernet_subsys(&raw_net_ops);
1018 }
1019
1020 void __init raw_proc_exit(void)
1021 {
1022         unregister_pernet_subsys(&raw_net_ops);
1023 }
1024 #endif /* CONFIG_PROC_FS */