net/ipv4/raw.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              RAW - implementation of IP "raw" sockets.
   7  *
   8  * Version:     $Id: raw.c,v 1.64 2002/02/01 22:01:04 davem Exp $
   9  *
  10  * Authors:     Ross Biro
  11  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  12  *
  13  * Fixes:
  14  *              Alan Cox        :       verify_area() fixed up
  15  *              Alan Cox        :       ICMP error handling
  16  *              Alan Cox        :       EMSGSIZE if you send too big a packet
  17  *              Alan Cox        :       Now uses generic datagrams and shared
  18  *                                      skbuff library. No more peek crashes,
  19  *                                      no more backlogs
  20  *              Alan Cox        :       Checks sk->broadcast.
  21  *              Alan Cox        :       Uses skb_free_datagram/skb_copy_datagram
  22  *              Alan Cox        :       Raw passes ip options too
  23  *              Alan Cox        :       Setsocketopt added
  24  *              Alan Cox        :       Fixed error return for broadcasts
  25  *              Alan Cox        :       Removed wake_up calls
  26  *              Alan Cox        :       Use ttl/tos
  27  *              Alan Cox        :       Cleaned up old debugging
  28  *              Alan Cox        :       Use new kernel side addresses
  29  *      Arnt Gulbrandsen        :       Fixed MSG_DONTROUTE in raw sockets.
  30  *              Alan Cox        :       BSD style RAW socket demultiplexing.
  31  *              Alan Cox        :       Beginnings of mrouted support.
  32  *              Alan Cox        :       Added IP_HDRINCL option.
  33  *              Alan Cox        :       Skip broadcast check if BSDism set.
  34  *              David S. Miller :       New socket lookup architecture.
  35  *
  36  *              This program is free software; you can redistribute it and/or
  37  *              modify it under the terms of the GNU General Public License
  38  *              as published by the Free Software Foundation; either version
  39  *              2 of the License, or (at your option) any later version.
  40  */
  41
  42 #include <linux/types.h>
  43 #include <asm/atomic.h>
  44 #include <asm/byteorder.h>
  45 #include <asm/current.h>
  46 #include <asm/uaccess.h>
  47 #include <asm/ioctls.h>
  48 #include <linux/stddef.h>
  49 #include <linux/slab.h>
  50 #include <linux/errno.h>
  51 #include <linux/aio.h>
  52 #include <linux/kernel.h>
  53 #include <linux/spinlock.h>
  54 #include <linux/sockios.h>
  55 #include <linux/socket.h>
  56 #include <linux/in.h>
  57 #include <linux/mroute.h>
  58 #include <linux/netdevice.h>
  59 #include <linux/in_route.h>
  60 #include <linux/route.h>
  61 #include <linux/skbuff.h>
  62 #include <net/net_namespace.h>
  63 #include <net/dst.h>
  64 #include <net/sock.h>
  65 #include <linux/gfp.h>
  66 #include <linux/ip.h>
  67 #include <linux/net.h>
  68 #include <net/ip.h>
  69 #include <net/icmp.h>
  70 #include <net/udp.h>
  71 #include <net/raw.h>
  72 #include <net/snmp.h>
  73 #include <net/tcp_states.h>
  74 #include <net/inet_common.h>
  75 #include <net/checksum.h>
  76 #include <net/xfrm.h>
  77 #include <linux/rtnetlink.h>
  78 #include <linux/proc_fs.h>
  79 #include <linux/seq_file.h>
  80 #include <linux/netfilter.h>
  81 #include <linux/netfilter_ipv4.h>
  82
  83 static struct raw_hashinfo raw_v4_hashinfo = {
  84         .lock = __RW_LOCK_UNLOCKED(),
  85 };
  86
  87 static void raw_v4_hash(struct sock *sk)
  88 {
  89         struct hlist_head *head = &raw_v4_hashinfo.ht[inet_sk(sk)->num &
  90                                                  (RAW_HTABLE_SIZE - 1)];
  91
  92         write_lock_bh(&raw_v4_hashinfo.lock);
  93         sk_add_node(sk, head);
  94         sock_prot_inc_use(sk->sk_prot);
  95         write_unlock_bh(&raw_v4_hashinfo.lock);
  96 }
  97
  98 static void raw_v4_unhash(struct sock *sk)
  99 {
 100         write_lock_bh(&raw_v4_hashinfo.lock);
 101         if (sk_del_node_init(sk))
 102                 sock_prot_dec_use(sk->sk_prot);
 103         write_unlock_bh(&raw_v4_hashinfo.lock);
 104 }
 105
 106 static struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
 107                              __be32 raddr, __be32 laddr,
 108                              int dif)
 109 {
 110         struct hlist_node *node;
 111
 112         sk_for_each_from(sk, node) {
 113                 struct inet_sock *inet = inet_sk(sk);
 114
 115                 if (inet->num == num                                    &&
 116                     !(inet->daddr && inet->daddr != raddr)              &&
 117                     !(inet->rcv_saddr && inet->rcv_saddr != laddr)      &&
 118                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 119                         goto found; /* gotcha */
 120         }
 121         sk = NULL;
 122 found:
 123         return sk;
 124 }
 125
 126 /*
 127  *      0 - deliver
 128  *      1 - block
 129  */
 130 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 131 {
 132         int type;
 133
 134         if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
 135                 return 1;
 136
 137         type = icmp_hdr(skb)->type;
 138         if (type < 32) {
 139                 __u32 data = raw_sk(sk)->filter.data;
 140
 141                 return ((1 << type) & data) != 0;
 142         }
 143
 144         /* Do not block unknown ICMP types */
 145         return 0;
 146 }
 147
 148 /* IP input processing comes here for RAW socket delivery.
 149  * Caller owns SKB, so we must make clones.
 150  *
 151  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 152  * -> It does. And not only TOS, but all IP header.
 153  */
 154 static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 155 {
 156         struct sock *sk;
 157         struct hlist_head *head;
 158         int delivered = 0;
 159
 160         read_lock(&raw_v4_hashinfo.lock);
 161         head = &raw_v4_hashinfo.ht[hash];
 162         if (hlist_empty(head))
 163                 goto out;
 164         sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
 165                              iph->saddr, iph->daddr,
 166                              skb->dev->ifindex);
 167
 168         while (sk) {
 169                 delivered = 1;
 170                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 171                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 172
 173                         /* Not releasing hash table! */
 174                         if (clone)
 175                                 raw_rcv(sk, clone);
 176                 }
 177                 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
 178                                      iph->saddr, iph->daddr,
 179                                      skb->dev->ifindex);
 180         }
 181 out:
 182         read_unlock(&raw_v4_hashinfo.lock);
 183         return delivered;
 184 }
 185
 186 int raw_local_deliver(struct sk_buff *skb, int protocol)
 187 {
 188         int hash;
 189         struct sock *raw_sk;
 190
 191         hash = protocol & (RAW_HTABLE_SIZE - 1);
 192         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 193
 194         /* If there maybe a raw socket we must check - if not we
 195          * don't care less
 196          */
 197         if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
 198                 raw_sk = NULL;
 199
 200         return raw_sk != NULL;
 201
 202 }
 203
 204 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
 205 {
 206         struct inet_sock *inet = inet_sk(sk);
 207         const int type = icmp_hdr(skb)->type;
 208         const int code = icmp_hdr(skb)->code;
 209         int err = 0;
 210         int harderr = 0;
 211
 212         /* Report error on raw socket, if:
 213            1. User requested ip_recverr.
 214            2. Socket is connected (otherwise the error indication
 215               is useless without ip_recverr and error is hard.
 216          */
 217         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 218                 return;
 219
 220         switch (type) {
 221         default:
 222         case ICMP_TIME_EXCEEDED:
 223                 err = EHOSTUNREACH;
 224                 break;
 225         case ICMP_SOURCE_QUENCH:
 226                 return;
 227         case ICMP_PARAMETERPROB:
 228                 err = EPROTO;
 229                 harderr = 1;
 230                 break;
 231         case ICMP_DEST_UNREACH:
 232                 err = EHOSTUNREACH;
 233                 if (code > NR_ICMP_UNREACH)
 234                         break;
 235                 err = icmp_err_convert[code].errno;
 236                 harderr = icmp_err_convert[code].fatal;
 237                 if (code == ICMP_FRAG_NEEDED) {
 238                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 239                         err = EMSGSIZE;
 240                 }
 241         }
 242
 243         if (inet->recverr) {
 244                 struct iphdr *iph = (struct iphdr*)skb->data;
 245                 u8 *payload = skb->data + (iph->ihl << 2);
 246
 247                 if (inet->hdrincl)
 248                         payload = skb->data;
 249                 ip_icmp_error(sk, skb, err, 0, info, payload);
 250         }
 251
 252         if (inet->recverr || harderr) {
 253                 sk->sk_err = err;
 254                 sk->sk_error_report(sk);
 255         }
 256 }
 257
 258 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
 259 {
 260         int hash;
 261         struct sock *raw_sk;
 262         struct iphdr *iph;
 263
 264         hash = protocol & (RAW_HTABLE_SIZE - 1);
 265
 266         read_lock(&raw_v4_hashinfo.lock);
 267         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
 268         if (raw_sk != NULL) {
 269                 iph = (struct iphdr *)skb->data;
 270                 while ((raw_sk = __raw_v4_lookup(raw_sk, protocol, iph->daddr,
 271                                                 iph->saddr,
 272                                                 skb->dev->ifindex)) != NULL) {
 273                         raw_err(raw_sk, skb, info);
 274                         raw_sk = sk_next(raw_sk);
 275                         iph = (struct iphdr *)skb->data;
 276                 }
 277         }
 278         read_unlock(&raw_v4_hashinfo.lock);
 279 }
 280
 281 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 282 {
 283         /* Charge it to the socket. */
 284
 285         if (sock_queue_rcv_skb(sk, skb) < 0) {
 286                 atomic_inc(&sk->sk_drops);
 287                 kfree_skb(skb);
 288                 return NET_RX_DROP;
 289         }
 290
 291         return NET_RX_SUCCESS;
 292 }
 293
 294 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 295 {
 296         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 297                 atomic_inc(&sk->sk_drops);
 298                 kfree_skb(skb);
 299                 return NET_RX_DROP;
 300         }
 301         nf_reset(skb);
 302
 303         skb_push(skb, skb->data - skb_network_header(skb));
 304
 305         raw_rcv_skb(sk, skb);
 306         return 0;
 307 }
 308
 309 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 310                         struct rtable *rt,
 311                         unsigned int flags)
 312 {
 313         struct inet_sock *inet = inet_sk(sk);
 314         int hh_len;
 315         struct iphdr *iph;
 316         struct sk_buff *skb;
 317         unsigned int iphlen;
 318         int err;
 319
 320         if (length > rt->u.dst.dev->mtu) {
 321                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
 322                                rt->u.dst.dev->mtu);
 323                 return -EMSGSIZE;
 324         }
 325         if (flags&MSG_PROBE)
 326                 goto out;
 327
 328         hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
 329
 330         skb = sock_alloc_send_skb(sk, length+hh_len+15,
 331                                   flags&MSG_DONTWAIT, &err);
 332         if (skb == NULL)
 333                 goto error;
 334         skb_reserve(skb, hh_len);
 335
 336         skb->priority = sk->sk_priority;
 337         skb->dst = dst_clone(&rt->u.dst);
 338
 339         skb_reset_network_header(skb);
 340         iph = ip_hdr(skb);
 341         skb_put(skb, length);
 342
 343         skb->ip_summed = CHECKSUM_NONE;
 344
 345         skb->transport_header = skb->network_header;
 346         err = memcpy_fromiovecend((void *)iph, from, 0, length);
 347         if (err)
 348                 goto error_fault;
 349
 350         /* We don't modify invalid header */
 351         iphlen = iph->ihl * 4;
 352         if (iphlen >= sizeof(*iph) && iphlen <= length) {
 353                 if (!iph->saddr)
 354                         iph->saddr = rt->rt_src;
 355                 iph->check   = 0;
 356                 iph->tot_len = htons(length);
 357                 if (!iph->id)
 358                         ip_select_ident(iph, &rt->u.dst, NULL);
 359
 360                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 361         }
 362         if (iph->protocol == IPPROTO_ICMP)
 363                 icmp_out_count(((struct icmphdr *)
 364                         skb_transport_header(skb))->type);
 365
 366         err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 367                       dst_output);
 368         if (err > 0)
 369                 err = inet->recverr ? net_xmit_errno(err) : 0;
 370         if (err)
 371                 goto error;
 372 out:
 373         return 0;
 374
 375 error_fault:
 376         err = -EFAULT;
 377         kfree_skb(skb);
 378 error:
 379         IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
 380         return err;
 381 }
 382
 383 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 384 {
 385         struct iovec *iov;
 386         u8 __user *type = NULL;
 387         u8 __user *code = NULL;
 388         int probed = 0;
 389         unsigned int i;
 390
 391         if (!msg->msg_iov)
 392                 return 0;
 393
 394         for (i = 0; i < msg->msg_iovlen; i++) {
 395                 iov = &msg->msg_iov[i];
 396                 if (!iov)
 397                         continue;
 398
 399                 switch (fl->proto) {
 400                 case IPPROTO_ICMP:
 401                         /* check if one-byte field is readable or not. */
 402                         if (iov->iov_base && iov->iov_len < 1)
 403                                 break;
 404
 405                         if (!type) {
 406                                 type = iov->iov_base;
 407                                 /* check if code field is readable or not. */
 408                                 if (iov->iov_len > 1)
 409                                         code = type + 1;
 410                         } else if (!code)
 411                                 code = iov->iov_base;
 412
 413                         if (type && code) {
 414                                 if (get_user(fl->fl_icmp_type, type) ||
 415                                     get_user(fl->fl_icmp_code, code))
 416                                         return -EFAULT;
 417                                 probed = 1;
 418                         }
 419                         break;
 420                 default:
 421                         probed = 1;
 422                         break;
 423                 }
 424                 if (probed)
 425                         break;
 426         }
 427         return 0;
 428 }
 429
 430 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 431                        size_t len)
 432 {
 433         struct inet_sock *inet = inet_sk(sk);
 434         struct ipcm_cookie ipc;
 435         struct rtable *rt = NULL;
 436         int free = 0;
 437         __be32 daddr;
 438         __be32 saddr;
 439         u8  tos;
 440         int err;
 441
 442         err = -EMSGSIZE;
 443         if (len > 0xFFFF)
 444                 goto out;
 445
 446         /*
 447          *      Check the flags.
 448          */
 449
 450         err = -EOPNOTSUPP;
 451         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 452                 goto out;               /* compatibility */
 453
 454         /*
 455          *      Get and verify the address.
 456          */
 457
 458         if (msg->msg_namelen) {
 459                 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
 460                 err = -EINVAL;
 461                 if (msg->msg_namelen < sizeof(*usin))
 462                         goto out;
 463                 if (usin->sin_family != AF_INET) {
 464                         static int complained;
 465                         if (!complained++)
 466                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 467                                                  "raw sendmsg. Fix it!\n",
 468                                                  current->comm);
 469                         err = -EAFNOSUPPORT;
 470                         if (usin->sin_family)
 471                                 goto out;
 472                 }
 473                 daddr = usin->sin_addr.s_addr;
 474                 /* ANK: I did not forget to get protocol from port field.
 475                  * I just do not know, who uses this weirdness.
 476                  * IP_HDRINCL is much more convenient.
 477                  */
 478         } else {
 479                 err = -EDESTADDRREQ;
 480                 if (sk->sk_state != TCP_ESTABLISHED)
 481                         goto out;
 482                 daddr = inet->daddr;
 483         }
 484
 485         ipc.addr = inet->saddr;
 486         ipc.opt = NULL;
 487         ipc.oif = sk->sk_bound_dev_if;
 488
 489         if (msg->msg_controllen) {
 490                 err = ip_cmsg_send(msg, &ipc);
 491                 if (err)
 492                         goto out;
 493                 if (ipc.opt)
 494                         free = 1;
 495         }
 496
 497         saddr = ipc.addr;
 498         ipc.addr = daddr;
 499
 500         if (!ipc.opt)
 501                 ipc.opt = inet->opt;
 502
 503         if (ipc.opt) {
 504                 err = -EINVAL;
 505                 /* Linux does not mangle headers on raw sockets,
 506                  * so that IP options + IP_HDRINCL is non-sense.
 507                  */
 508                 if (inet->hdrincl)
 509                         goto done;
 510                 if (ipc.opt->srr) {
 511                         if (!daddr)
 512                                 goto done;
 513                         daddr = ipc.opt->faddr;
 514                 }
 515         }
 516         tos = RT_CONN_FLAGS(sk);
 517         if (msg->msg_flags & MSG_DONTROUTE)
 518                 tos |= RTO_ONLINK;
 519
 520         if (MULTICAST(daddr)) {
 521                 if (!ipc.oif)
 522                         ipc.oif = inet->mc_index;
 523                 if (!saddr)
 524                         saddr = inet->mc_addr;
 525         }
 526
 527         {
 528                 struct flowi fl = { .oif = ipc.oif,
 529                                     .nl_u = { .ip4_u =
 530                                               { .daddr = daddr,
 531                                                 .saddr = saddr,
 532                                                 .tos = tos } },
 533                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 534                                                              sk->sk_protocol,
 535                                   };
 536                 if (!inet->hdrincl) {
 537                         err = raw_probe_proto_opt(&fl, msg);
 538                         if (err)
 539                                 goto done;
 540                 }
 541
 542                 security_sk_classify_flow(sk, &fl);
 543                 err = ip_route_output_flow(&rt, &fl, sk, 1);
 544         }
 545         if (err)
 546                 goto done;
 547
 548         err = -EACCES;
 549         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 550                 goto done;
 551
 552         if (msg->msg_flags & MSG_CONFIRM)
 553                 goto do_confirm;
 554 back_from_confirm:
 555
 556         if (inet->hdrincl)
 557                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 558                                         rt, msg->msg_flags);
 559
 560          else {
 561                 if (!ipc.addr)
 562                         ipc.addr = rt->rt_dst;
 563                 lock_sock(sk);
 564                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 565                                         &ipc, rt, msg->msg_flags);
 566                 if (err)
 567                         ip_flush_pending_frames(sk);
 568                 else if (!(msg->msg_flags & MSG_MORE))
 569                         err = ip_push_pending_frames(sk);
 570                 release_sock(sk);
 571         }
 572 done:
 573         if (free)
 574                 kfree(ipc.opt);
 575         ip_rt_put(rt);
 576
 577 out:
 578         if (err < 0)
 579                 return err;
 580         return len;
 581
 582 do_confirm:
 583         dst_confirm(&rt->u.dst);
 584         if (!(msg->msg_flags & MSG_PROBE) || len)
 585                 goto back_from_confirm;
 586         err = 0;
 587         goto done;
 588 }
 589
 590 static void raw_close(struct sock *sk, long timeout)
 591 {
 592         /*
 593          * Raw sockets may have direct kernel refereneces. Kill them.
 594          */
 595         ip_ra_control(sk, 0, NULL);
 596
 597         sk_common_release(sk);
 598 }
 599
 600 /* This gets rid of all the nasties in af_inet. -DaveM */
 601 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 602 {
 603         struct inet_sock *inet = inet_sk(sk);
 604         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 605         int ret = -EINVAL;
 606         int chk_addr_ret;
 607
 608         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 609                 goto out;
 610         chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
 611         ret = -EADDRNOTAVAIL;
 612         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 613             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 614                 goto out;
 615         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 616         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 617                 inet->saddr = 0;  /* Use device */
 618         sk_dst_reset(sk);
 619         ret = 0;
 620 out:    return ret;
 621 }
 622
 623 /*
 624  *      This should be easy, if there is something there
 625  *      we return it, otherwise we block.
 626  */
 627
 628 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 629                        size_t len, int noblock, int flags, int *addr_len)
 630 {
 631         struct inet_sock *inet = inet_sk(sk);
 632         size_t copied = 0;
 633         int err = -EOPNOTSUPP;
 634         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 635         struct sk_buff *skb;
 636
 637         if (flags & MSG_OOB)
 638                 goto out;
 639
 640         if (addr_len)
 641                 *addr_len = sizeof(*sin);
 642
 643         if (flags & MSG_ERRQUEUE) {
 644                 err = ip_recv_error(sk, msg, len);
 645                 goto out;
 646         }
 647
 648         skb = skb_recv_datagram(sk, flags, noblock, &err);
 649         if (!skb)
 650                 goto out;
 651
 652         copied = skb->len;
 653         if (len < copied) {
 654                 msg->msg_flags |= MSG_TRUNC;
 655                 copied = len;
 656         }
 657
 658         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 659         if (err)
 660                 goto done;
 661
 662         sock_recv_timestamp(msg, sk, skb);
 663
 664         /* Copy the address. */
 665         if (sin) {
 666                 sin->sin_family = AF_INET;
 667                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 668                 sin->sin_port = 0;
 669                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 670         }
 671         if (inet->cmsg_flags)
 672                 ip_cmsg_recv(msg, skb);
 673         if (flags & MSG_TRUNC)
 674                 copied = skb->len;
 675 done:
 676         skb_free_datagram(sk, skb);
 677 out:
 678         if (err)
 679                 return err;
 680         return copied;
 681 }
 682
 683 static int raw_init(struct sock *sk)
 684 {
 685         struct raw_sock *rp = raw_sk(sk);
 686
 687         if (inet_sk(sk)->num == IPPROTO_ICMP)
 688                 memset(&rp->filter, 0, sizeof(rp->filter));
 689         return 0;
 690 }
 691
 692 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 693 {
 694         if (optlen > sizeof(struct icmp_filter))
 695                 optlen = sizeof(struct icmp_filter);
 696         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 697                 return -EFAULT;
 698         return 0;
 699 }
 700
 701 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 702 {
 703         int len, ret = -EFAULT;
 704
 705         if (get_user(len, optlen))
 706                 goto out;
 707         ret = -EINVAL;
 708         if (len < 0)
 709                 goto out;
 710         if (len > sizeof(struct icmp_filter))
 711                 len = sizeof(struct icmp_filter);
 712         ret = -EFAULT;
 713         if (put_user(len, optlen) ||
 714             copy_to_user(optval, &raw_sk(sk)->filter, len))
 715                 goto out;
 716         ret = 0;
 717 out:    return ret;
 718 }
 719
 720 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 721                           char __user *optval, int optlen)
 722 {
 723         if (optname == ICMP_FILTER) {
 724                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 725                         return -EOPNOTSUPP;
 726                 else
 727                         return raw_seticmpfilter(sk, optval, optlen);
 728         }
 729         return -ENOPROTOOPT;
 730 }
 731
 732 static int raw_setsockopt(struct sock *sk, int level, int optname,
 733                           char __user *optval, int optlen)
 734 {
 735         if (level != SOL_RAW)
 736                 return ip_setsockopt(sk, level, optname, optval, optlen);
 737         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 738 }
 739
 740 #ifdef CONFIG_COMPAT
 741 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 742                                  char __user *optval, int optlen)
 743 {
 744         if (level != SOL_RAW)
 745                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 746         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 747 }
 748 #endif
 749
 750 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 751                           char __user *optval, int __user *optlen)
 752 {
 753         if (optname == ICMP_FILTER) {
 754                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 755                         return -EOPNOTSUPP;
 756                 else
 757                         return raw_geticmpfilter(sk, optval, optlen);
 758         }
 759         return -ENOPROTOOPT;
 760 }
 761
 762 static int raw_getsockopt(struct sock *sk, int level, int optname,
 763                           char __user *optval, int __user *optlen)
 764 {
 765         if (level != SOL_RAW)
 766                 return ip_getsockopt(sk, level, optname, optval, optlen);
 767         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 768 }
 769
 770 #ifdef CONFIG_COMPAT
 771 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 772                                  char __user *optval, int __user *optlen)
 773 {
 774         if (level != SOL_RAW)
 775                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 776         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 777 }
 778 #endif
 779
 780 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 781 {
 782         switch (cmd) {
 783                 case SIOCOUTQ: {
 784                         int amount = atomic_read(&sk->sk_wmem_alloc);
 785                         return put_user(amount, (int __user *)arg);
 786                 }
 787                 case SIOCINQ: {
 788                         struct sk_buff *skb;
 789                         int amount = 0;
 790
 791                         spin_lock_bh(&sk->sk_receive_queue.lock);
 792                         skb = skb_peek(&sk->sk_receive_queue);
 793                         if (skb != NULL)
 794                                 amount = skb->len;
 795                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 796                         return put_user(amount, (int __user *)arg);
 797                 }
 798
 799                 default:
 800 #ifdef CONFIG_IP_MROUTE
 801                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 802 #else
 803                         return -ENOIOCTLCMD;
 804 #endif
 805         }
 806 }
 807
 808 DEFINE_PROTO_INUSE(raw)
 809
 810 struct proto raw_prot = {
 811         .name              = "RAW",
 812         .owner             = THIS_MODULE,
 813         .close             = raw_close,
 814         .connect           = ip4_datagram_connect,
 815         .disconnect        = udp_disconnect,
 816         .ioctl             = raw_ioctl,
 817         .init              = raw_init,
 818         .setsockopt        = raw_setsockopt,
 819         .getsockopt        = raw_getsockopt,
 820         .sendmsg           = raw_sendmsg,
 821         .recvmsg           = raw_recvmsg,
 822         .bind              = raw_bind,
 823         .backlog_rcv       = raw_rcv_skb,
 824         .hash              = raw_v4_hash,
 825         .unhash            = raw_v4_unhash,
 826         .obj_size          = sizeof(struct raw_sock),
 827 #ifdef CONFIG_COMPAT
 828         .compat_setsockopt = compat_raw_setsockopt,
 829         .compat_getsockopt = compat_raw_getsockopt,
 830 #endif
 831         REF_PROTO_INUSE(raw)
 832 };
 833
 834 #ifdef CONFIG_PROC_FS
 835 struct raw_iter_state {
 836         int bucket;
 837 };
 838
 839 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
 840
 841 static struct sock *raw_get_first(struct seq_file *seq)
 842 {
 843         struct sock *sk;
 844         struct raw_iter_state* state = raw_seq_private(seq);
 845
 846         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
 847                         ++state->bucket) {
 848                 struct hlist_node *node;
 849
 850                 sk_for_each(sk, node, &raw_v4_hashinfo.ht[state->bucket])
 851                         if (sk->sk_family == PF_INET)
 852                                 goto found;
 853         }
 854         sk = NULL;
 855 found:
 856         return sk;
 857 }
 858
 859 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 860 {
 861         struct raw_iter_state* state = raw_seq_private(seq);
 862
 863         do {
 864                 sk = sk_next(sk);
 865 try_again:
 866                 ;
 867         } while (sk && sk->sk_family != PF_INET);
 868
 869         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
 870                 sk = sk_head(&raw_v4_hashinfo.ht[state->bucket]);
 871                 goto try_again;
 872         }
 873         return sk;
 874 }
 875
 876 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 877 {
 878         struct sock *sk = raw_get_first(seq);
 879
 880         if (sk)
 881                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 882                         --pos;
 883         return pos ? NULL : sk;
 884 }
 885
 886 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 887 {
 888         read_lock(&raw_v4_hashinfo.lock);
 889         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 890 }
 891
 892 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 893 {
 894         struct sock *sk;
 895
 896         if (v == SEQ_START_TOKEN)
 897                 sk = raw_get_first(seq);
 898         else
 899                 sk = raw_get_next(seq, v);
 900         ++*pos;
 901         return sk;
 902 }
 903
 904 static void raw_seq_stop(struct seq_file *seq, void *v)
 905 {
 906         read_unlock(&raw_v4_hashinfo.lock);
 907 }
 908
 909 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
 910 {
 911         struct inet_sock *inet = inet_sk(sp);
 912         __be32 dest = inet->daddr,
 913                src = inet->rcv_saddr;
 914         __u16 destp = 0,
 915               srcp  = inet->num;
 916
 917         sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
 918                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d",
 919                 i, src, srcp, dest, destp, sp->sk_state,
 920                 atomic_read(&sp->sk_wmem_alloc),
 921                 atomic_read(&sp->sk_rmem_alloc),
 922                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 923                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
 924         return tmpbuf;
 925 }
 926
 927 #define TMPSZ 128
 928
 929 static int raw_seq_show(struct seq_file *seq, void *v)
 930 {
 931         char tmpbuf[TMPSZ+1];
 932
 933         if (v == SEQ_START_TOKEN)
 934                 seq_printf(seq, "%-*s\n", TMPSZ-1,
 935                                "  sl  local_address rem_address   st tx_queue "
 936                                "rx_queue tr tm->when retrnsmt   uid  timeout "
 937                                "inode  drops");
 938         else {
 939                 struct raw_iter_state *state = raw_seq_private(seq);
 940
 941                 seq_printf(seq, "%-*s\n", TMPSZ-1,
 942                            get_raw_sock(v, tmpbuf, state->bucket));
 943         }
 944         return 0;
 945 }
 946
 947 static const struct seq_operations raw_seq_ops = {
 948         .start = raw_seq_start,
 949         .next  = raw_seq_next,
 950         .stop  = raw_seq_stop,
 951         .show  = raw_seq_show,
 952 };
 953
 954 static int raw_seq_open(struct inode *inode, struct file *file)
 955 {
 956         return seq_open_private(file, &raw_seq_ops,
 957                         sizeof(struct raw_iter_state));
 958 }
 959
 960 static const struct file_operations raw_seq_fops = {
 961         .owner   = THIS_MODULE,
 962         .open    = raw_seq_open,
 963         .read    = seq_read,
 964         .llseek  = seq_lseek,
 965         .release = seq_release_private,
 966 };
 967
 968 int __init raw_proc_init(void)
 969 {
 970         if (!proc_net_fops_create(&init_net, "raw", S_IRUGO, &raw_seq_fops))
 971                 return -ENOMEM;
 972         return 0;
 973 }
 974
 975 void __init raw_proc_exit(void)
 976 {
 977         proc_net_remove(&init_net, "raw");
 978 }
 979 #endif /* CONFIG_PROC_FS */