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 struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE];
  84 DEFINE_RWLOCK(raw_v4_lock);
  85
  86 static void raw_v4_hash(struct sock *sk)
  87 {
  88         struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num &
  89                                                  (RAWV4_HTABLE_SIZE - 1)];
  90
  91         write_lock_bh(&raw_v4_lock);
  92         sk_add_node(sk, head);
  93         sock_prot_inc_use(sk->sk_prot);
  94         write_unlock_bh(&raw_v4_lock);
  95 }
  96
  97 static void raw_v4_unhash(struct sock *sk)
  98 {
  99         write_lock_bh(&raw_v4_lock);
 100         if (sk_del_node_init(sk))
 101                 sock_prot_dec_use(sk->sk_prot);
 102         write_unlock_bh(&raw_v4_lock);
 103 }
 104
 105 struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
 106                              __be32 raddr, __be32 laddr,
 107                              int dif)
 108 {
 109         struct hlist_node *node;
 110
 111         sk_for_each_from(sk, node) {
 112                 struct inet_sock *inet = inet_sk(sk);
 113
 114                 if (inet->num == num                                    &&
 115                     !(inet->daddr && inet->daddr != raddr)              &&
 116                     !(inet->rcv_saddr && inet->rcv_saddr != laddr)      &&
 117                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
 118                         goto found; /* gotcha */
 119         }
 120         sk = NULL;
 121 found:
 122         return sk;
 123 }
 124
 125 /*
 126  *      0 - deliver
 127  *      1 - block
 128  */
 129 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
 130 {
 131         int type;
 132
 133         if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
 134                 return 1;
 135
 136         type = icmp_hdr(skb)->type;
 137         if (type < 32) {
 138                 __u32 data = raw_sk(sk)->filter.data;
 139
 140                 return ((1 << type) & data) != 0;
 141         }
 142
 143         /* Do not block unknown ICMP types */
 144         return 0;
 145 }
 146
 147 /* IP input processing comes here for RAW socket delivery.
 148  * Caller owns SKB, so we must make clones.
 149  *
 150  * RFC 1122: SHOULD pass TOS value up to the transport layer.
 151  * -> It does. And not only TOS, but all IP header.
 152  */
 153 int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
 154 {
 155         struct sock *sk;
 156         struct hlist_head *head;
 157         int delivered = 0;
 158
 159         read_lock(&raw_v4_lock);
 160         head = &raw_v4_htable[hash];
 161         if (hlist_empty(head))
 162                 goto out;
 163         sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
 164                              iph->saddr, iph->daddr,
 165                              skb->dev->ifindex);
 166
 167         while (sk) {
 168                 delivered = 1;
 169                 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
 170                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
 171
 172                         /* Not releasing hash table! */
 173                         if (clone)
 174                                 raw_rcv(sk, clone);
 175                 }
 176                 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
 177                                      iph->saddr, iph->daddr,
 178                                      skb->dev->ifindex);
 179         }
 180 out:
 181         read_unlock(&raw_v4_lock);
 182         return delivered;
 183 }
 184
 185 void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
 186 {
 187         struct inet_sock *inet = inet_sk(sk);
 188         const int type = icmp_hdr(skb)->type;
 189         const int code = icmp_hdr(skb)->code;
 190         int err = 0;
 191         int harderr = 0;
 192
 193         /* Report error on raw socket, if:
 194            1. User requested ip_recverr.
 195            2. Socket is connected (otherwise the error indication
 196               is useless without ip_recverr and error is hard.
 197          */
 198         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
 199                 return;
 200
 201         switch (type) {
 202         default:
 203         case ICMP_TIME_EXCEEDED:
 204                 err = EHOSTUNREACH;
 205                 break;
 206         case ICMP_SOURCE_QUENCH:
 207                 return;
 208         case ICMP_PARAMETERPROB:
 209                 err = EPROTO;
 210                 harderr = 1;
 211                 break;
 212         case ICMP_DEST_UNREACH:
 213                 err = EHOSTUNREACH;
 214                 if (code > NR_ICMP_UNREACH)
 215                         break;
 216                 err = icmp_err_convert[code].errno;
 217                 harderr = icmp_err_convert[code].fatal;
 218                 if (code == ICMP_FRAG_NEEDED) {
 219                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
 220                         err = EMSGSIZE;
 221                 }
 222         }
 223
 224         if (inet->recverr) {
 225                 struct iphdr *iph = (struct iphdr*)skb->data;
 226                 u8 *payload = skb->data + (iph->ihl << 2);
 227
 228                 if (inet->hdrincl)
 229                         payload = skb->data;
 230                 ip_icmp_error(sk, skb, err, 0, info, payload);
 231         }
 232
 233         if (inet->recverr || harderr) {
 234                 sk->sk_err = err;
 235                 sk->sk_error_report(sk);
 236         }
 237 }
 238
 239 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
 240 {
 241         /* Charge it to the socket. */
 242
 243         if (sock_queue_rcv_skb(sk, skb) < 0) {
 244                 /* FIXME: increment a raw drops counter here */
 245                 kfree_skb(skb);
 246                 return NET_RX_DROP;
 247         }
 248
 249         return NET_RX_SUCCESS;
 250 }
 251
 252 int raw_rcv(struct sock *sk, struct sk_buff *skb)
 253 {
 254         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
 255                 kfree_skb(skb);
 256                 return NET_RX_DROP;
 257         }
 258         nf_reset(skb);
 259
 260         skb_push(skb, skb->data - skb_network_header(skb));
 261
 262         raw_rcv_skb(sk, skb);
 263         return 0;
 264 }
 265
 266 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
 267                         struct rtable *rt,
 268                         unsigned int flags)
 269 {
 270         struct inet_sock *inet = inet_sk(sk);
 271         int hh_len;
 272         struct iphdr *iph;
 273         struct sk_buff *skb;
 274         unsigned int iphlen;
 275         int err;
 276
 277         if (length > rt->u.dst.dev->mtu) {
 278                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
 279                                rt->u.dst.dev->mtu);
 280                 return -EMSGSIZE;
 281         }
 282         if (flags&MSG_PROBE)
 283                 goto out;
 284
 285         hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
 286
 287         skb = sock_alloc_send_skb(sk, length+hh_len+15,
 288                                   flags&MSG_DONTWAIT, &err);
 289         if (skb == NULL)
 290                 goto error;
 291         skb_reserve(skb, hh_len);
 292
 293         skb->priority = sk->sk_priority;
 294         skb->dst = dst_clone(&rt->u.dst);
 295
 296         skb_reset_network_header(skb);
 297         iph = ip_hdr(skb);
 298         skb_put(skb, length);
 299
 300         skb->ip_summed = CHECKSUM_NONE;
 301
 302         skb->transport_header = skb->network_header;
 303         err = memcpy_fromiovecend((void *)iph, from, 0, length);
 304         if (err)
 305                 goto error_fault;
 306
 307         /* We don't modify invalid header */
 308         iphlen = iph->ihl * 4;
 309         if (iphlen >= sizeof(*iph) && iphlen <= length) {
 310                 if (!iph->saddr)
 311                         iph->saddr = rt->rt_src;
 312                 iph->check   = 0;
 313                 iph->tot_len = htons(length);
 314                 if (!iph->id)
 315                         ip_select_ident(iph, &rt->u.dst, NULL);
 316
 317                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
 318         }
 319         if (iph->protocol == IPPROTO_ICMP)
 320                 icmp_out_count(((struct icmphdr *)
 321                         skb_transport_header(skb))->type);
 322
 323         err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
 324                       dst_output);
 325         if (err > 0)
 326                 err = inet->recverr ? net_xmit_errno(err) : 0;
 327         if (err)
 328                 goto error;
 329 out:
 330         return 0;
 331
 332 error_fault:
 333         err = -EFAULT;
 334         kfree_skb(skb);
 335 error:
 336         IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
 337         return err;
 338 }
 339
 340 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
 341 {
 342         struct iovec *iov;
 343         u8 __user *type = NULL;
 344         u8 __user *code = NULL;
 345         int probed = 0;
 346         unsigned int i;
 347
 348         if (!msg->msg_iov)
 349                 return 0;
 350
 351         for (i = 0; i < msg->msg_iovlen; i++) {
 352                 iov = &msg->msg_iov[i];
 353                 if (!iov)
 354                         continue;
 355
 356                 switch (fl->proto) {
 357                 case IPPROTO_ICMP:
 358                         /* check if one-byte field is readable or not. */
 359                         if (iov->iov_base && iov->iov_len < 1)
 360                                 break;
 361
 362                         if (!type) {
 363                                 type = iov->iov_base;
 364                                 /* check if code field is readable or not. */
 365                                 if (iov->iov_len > 1)
 366                                         code = type + 1;
 367                         } else if (!code)
 368                                 code = iov->iov_base;
 369
 370                         if (type && code) {
 371                                 if (get_user(fl->fl_icmp_type, type) ||
 372                                     get_user(fl->fl_icmp_code, code))
 373                                         return -EFAULT;
 374                                 probed = 1;
 375                         }
 376                         break;
 377                 default:
 378                         probed = 1;
 379                         break;
 380                 }
 381                 if (probed)
 382                         break;
 383         }
 384         return 0;
 385 }
 386
 387 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 388                        size_t len)
 389 {
 390         struct inet_sock *inet = inet_sk(sk);
 391         struct ipcm_cookie ipc;
 392         struct rtable *rt = NULL;
 393         int free = 0;
 394         __be32 daddr;
 395         __be32 saddr;
 396         u8  tos;
 397         int err;
 398
 399         err = -EMSGSIZE;
 400         if (len > 0xFFFF)
 401                 goto out;
 402
 403         /*
 404          *      Check the flags.
 405          */
 406
 407         err = -EOPNOTSUPP;
 408         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
 409                 goto out;               /* compatibility */
 410
 411         /*
 412          *      Get and verify the address.
 413          */
 414
 415         if (msg->msg_namelen) {
 416                 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
 417                 err = -EINVAL;
 418                 if (msg->msg_namelen < sizeof(*usin))
 419                         goto out;
 420                 if (usin->sin_family != AF_INET) {
 421                         static int complained;
 422                         if (!complained++)
 423                                 printk(KERN_INFO "%s forgot to set AF_INET in "
 424                                                  "raw sendmsg. Fix it!\n",
 425                                                  current->comm);
 426                         err = -EAFNOSUPPORT;
 427                         if (usin->sin_family)
 428                                 goto out;
 429                 }
 430                 daddr = usin->sin_addr.s_addr;
 431                 /* ANK: I did not forget to get protocol from port field.
 432                  * I just do not know, who uses this weirdness.
 433                  * IP_HDRINCL is much more convenient.
 434                  */
 435         } else {
 436                 err = -EDESTADDRREQ;
 437                 if (sk->sk_state != TCP_ESTABLISHED)
 438                         goto out;
 439                 daddr = inet->daddr;
 440         }
 441
 442         ipc.addr = inet->saddr;
 443         ipc.opt = NULL;
 444         ipc.oif = sk->sk_bound_dev_if;
 445
 446         if (msg->msg_controllen) {
 447                 err = ip_cmsg_send(msg, &ipc);
 448                 if (err)
 449                         goto out;
 450                 if (ipc.opt)
 451                         free = 1;
 452         }
 453
 454         saddr = ipc.addr;
 455         ipc.addr = daddr;
 456
 457         if (!ipc.opt)
 458                 ipc.opt = inet->opt;
 459
 460         if (ipc.opt) {
 461                 err = -EINVAL;
 462                 /* Linux does not mangle headers on raw sockets,
 463                  * so that IP options + IP_HDRINCL is non-sense.
 464                  */
 465                 if (inet->hdrincl)
 466                         goto done;
 467                 if (ipc.opt->srr) {
 468                         if (!daddr)
 469                                 goto done;
 470                         daddr = ipc.opt->faddr;
 471                 }
 472         }
 473         tos = RT_CONN_FLAGS(sk);
 474         if (msg->msg_flags & MSG_DONTROUTE)
 475                 tos |= RTO_ONLINK;
 476
 477         if (MULTICAST(daddr)) {
 478                 if (!ipc.oif)
 479                         ipc.oif = inet->mc_index;
 480                 if (!saddr)
 481                         saddr = inet->mc_addr;
 482         }
 483
 484         {
 485                 struct flowi fl = { .oif = ipc.oif,
 486                                     .nl_u = { .ip4_u =
 487                                               { .daddr = daddr,
 488                                                 .saddr = saddr,
 489                                                 .tos = tos } },
 490                                     .proto = inet->hdrincl ? IPPROTO_RAW :
 491                                                              sk->sk_protocol,
 492                                   };
 493                 if (!inet->hdrincl) {
 494                         err = raw_probe_proto_opt(&fl, msg);
 495                         if (err)
 496                                 goto done;
 497                 }
 498
 499                 security_sk_classify_flow(sk, &fl);
 500                 err = ip_route_output_flow(&rt, &fl, sk, 1);
 501         }
 502         if (err)
 503                 goto done;
 504
 505         err = -EACCES;
 506         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
 507                 goto done;
 508
 509         if (msg->msg_flags & MSG_CONFIRM)
 510                 goto do_confirm;
 511 back_from_confirm:
 512
 513         if (inet->hdrincl)
 514                 err = raw_send_hdrinc(sk, msg->msg_iov, len,
 515                                         rt, msg->msg_flags);
 516
 517          else {
 518                 if (!ipc.addr)
 519                         ipc.addr = rt->rt_dst;
 520                 lock_sock(sk);
 521                 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
 522                                         &ipc, rt, msg->msg_flags);
 523                 if (err)
 524                         ip_flush_pending_frames(sk);
 525                 else if (!(msg->msg_flags & MSG_MORE))
 526                         err = ip_push_pending_frames(sk);
 527                 release_sock(sk);
 528         }
 529 done:
 530         if (free)
 531                 kfree(ipc.opt);
 532         ip_rt_put(rt);
 533
 534 out:
 535         if (err < 0)
 536                 return err;
 537         return len;
 538
 539 do_confirm:
 540         dst_confirm(&rt->u.dst);
 541         if (!(msg->msg_flags & MSG_PROBE) || len)
 542                 goto back_from_confirm;
 543         err = 0;
 544         goto done;
 545 }
 546
 547 static void raw_close(struct sock *sk, long timeout)
 548 {
 549         /*
 550          * Raw sockets may have direct kernel refereneces. Kill them.
 551          */
 552         ip_ra_control(sk, 0, NULL);
 553
 554         sk_common_release(sk);
 555 }
 556
 557 /* This gets rid of all the nasties in af_inet. -DaveM */
 558 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 559 {
 560         struct inet_sock *inet = inet_sk(sk);
 561         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 562         int ret = -EINVAL;
 563         int chk_addr_ret;
 564
 565         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
 566                 goto out;
 567         chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
 568         ret = -EADDRNOTAVAIL;
 569         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
 570             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
 571                 goto out;
 572         inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
 573         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
 574                 inet->saddr = 0;  /* Use device */
 575         sk_dst_reset(sk);
 576         ret = 0;
 577 out:    return ret;
 578 }
 579
 580 /*
 581  *      This should be easy, if there is something there
 582  *      we return it, otherwise we block.
 583  */
 584
 585 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 586                        size_t len, int noblock, int flags, int *addr_len)
 587 {
 588         struct inet_sock *inet = inet_sk(sk);
 589         size_t copied = 0;
 590         int err = -EOPNOTSUPP;
 591         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
 592         struct sk_buff *skb;
 593
 594         if (flags & MSG_OOB)
 595                 goto out;
 596
 597         if (addr_len)
 598                 *addr_len = sizeof(*sin);
 599
 600         if (flags & MSG_ERRQUEUE) {
 601                 err = ip_recv_error(sk, msg, len);
 602                 goto out;
 603         }
 604
 605         skb = skb_recv_datagram(sk, flags, noblock, &err);
 606         if (!skb)
 607                 goto out;
 608
 609         copied = skb->len;
 610         if (len < copied) {
 611                 msg->msg_flags |= MSG_TRUNC;
 612                 copied = len;
 613         }
 614
 615         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 616         if (err)
 617                 goto done;
 618
 619         sock_recv_timestamp(msg, sk, skb);
 620
 621         /* Copy the address. */
 622         if (sin) {
 623                 sin->sin_family = AF_INET;
 624                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 625                 sin->sin_port = 0;
 626                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 627         }
 628         if (inet->cmsg_flags)
 629                 ip_cmsg_recv(msg, skb);
 630         if (flags & MSG_TRUNC)
 631                 copied = skb->len;
 632 done:
 633         skb_free_datagram(sk, skb);
 634 out:
 635         if (err)
 636                 return err;
 637         return copied;
 638 }
 639
 640 static int raw_init(struct sock *sk)
 641 {
 642         struct raw_sock *rp = raw_sk(sk);
 643
 644         if (inet_sk(sk)->num == IPPROTO_ICMP)
 645                 memset(&rp->filter, 0, sizeof(rp->filter));
 646         return 0;
 647 }
 648
 649 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
 650 {
 651         if (optlen > sizeof(struct icmp_filter))
 652                 optlen = sizeof(struct icmp_filter);
 653         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
 654                 return -EFAULT;
 655         return 0;
 656 }
 657
 658 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
 659 {
 660         int len, ret = -EFAULT;
 661
 662         if (get_user(len, optlen))
 663                 goto out;
 664         ret = -EINVAL;
 665         if (len < 0)
 666                 goto out;
 667         if (len > sizeof(struct icmp_filter))
 668                 len = sizeof(struct icmp_filter);
 669         ret = -EFAULT;
 670         if (put_user(len, optlen) ||
 671             copy_to_user(optval, &raw_sk(sk)->filter, len))
 672                 goto out;
 673         ret = 0;
 674 out:    return ret;
 675 }
 676
 677 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
 678                           char __user *optval, int optlen)
 679 {
 680         if (optname == ICMP_FILTER) {
 681                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 682                         return -EOPNOTSUPP;
 683                 else
 684                         return raw_seticmpfilter(sk, optval, optlen);
 685         }
 686         return -ENOPROTOOPT;
 687 }
 688
 689 static int raw_setsockopt(struct sock *sk, int level, int optname,
 690                           char __user *optval, int optlen)
 691 {
 692         if (level != SOL_RAW)
 693                 return ip_setsockopt(sk, level, optname, optval, optlen);
 694         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 695 }
 696
 697 #ifdef CONFIG_COMPAT
 698 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
 699                                  char __user *optval, int optlen)
 700 {
 701         if (level != SOL_RAW)
 702                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
 703         return do_raw_setsockopt(sk, level, optname, optval, optlen);
 704 }
 705 #endif
 706
 707 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
 708                           char __user *optval, int __user *optlen)
 709 {
 710         if (optname == ICMP_FILTER) {
 711                 if (inet_sk(sk)->num != IPPROTO_ICMP)
 712                         return -EOPNOTSUPP;
 713                 else
 714                         return raw_geticmpfilter(sk, optval, optlen);
 715         }
 716         return -ENOPROTOOPT;
 717 }
 718
 719 static int raw_getsockopt(struct sock *sk, int level, int optname,
 720                           char __user *optval, int __user *optlen)
 721 {
 722         if (level != SOL_RAW)
 723                 return ip_getsockopt(sk, level, optname, optval, optlen);
 724         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 725 }
 726
 727 #ifdef CONFIG_COMPAT
 728 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
 729                                  char __user *optval, int __user *optlen)
 730 {
 731         if (level != SOL_RAW)
 732                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
 733         return do_raw_getsockopt(sk, level, optname, optval, optlen);
 734 }
 735 #endif
 736
 737 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
 738 {
 739         switch (cmd) {
 740                 case SIOCOUTQ: {
 741                         int amount = atomic_read(&sk->sk_wmem_alloc);
 742                         return put_user(amount, (int __user *)arg);
 743                 }
 744                 case SIOCINQ: {
 745                         struct sk_buff *skb;
 746                         int amount = 0;
 747
 748                         spin_lock_bh(&sk->sk_receive_queue.lock);
 749                         skb = skb_peek(&sk->sk_receive_queue);
 750                         if (skb != NULL)
 751                                 amount = skb->len;
 752                         spin_unlock_bh(&sk->sk_receive_queue.lock);
 753                         return put_user(amount, (int __user *)arg);
 754                 }
 755
 756                 default:
 757 #ifdef CONFIG_IP_MROUTE
 758                         return ipmr_ioctl(sk, cmd, (void __user *)arg);
 759 #else
 760                         return -ENOIOCTLCMD;
 761 #endif
 762         }
 763 }
 764
 765 DEFINE_PROTO_INUSE(raw)
 766
 767 struct proto raw_prot = {
 768         .name              = "RAW",
 769         .owner             = THIS_MODULE,
 770         .close             = raw_close,
 771         .connect           = ip4_datagram_connect,
 772         .disconnect        = udp_disconnect,
 773         .ioctl             = raw_ioctl,
 774         .init              = raw_init,
 775         .setsockopt        = raw_setsockopt,
 776         .getsockopt        = raw_getsockopt,
 777         .sendmsg           = raw_sendmsg,
 778         .recvmsg           = raw_recvmsg,
 779         .bind              = raw_bind,
 780         .backlog_rcv       = raw_rcv_skb,
 781         .hash              = raw_v4_hash,
 782         .unhash            = raw_v4_unhash,
 783         .obj_size          = sizeof(struct raw_sock),
 784 #ifdef CONFIG_COMPAT
 785         .compat_setsockopt = compat_raw_setsockopt,
 786         .compat_getsockopt = compat_raw_getsockopt,
 787 #endif
 788         REF_PROTO_INUSE(raw)
 789 };
 790
 791 #ifdef CONFIG_PROC_FS
 792 struct raw_iter_state {
 793         int bucket;
 794 };
 795
 796 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
 797
 798 static struct sock *raw_get_first(struct seq_file *seq)
 799 {
 800         struct sock *sk;
 801         struct raw_iter_state* state = raw_seq_private(seq);
 802
 803         for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
 804                 struct hlist_node *node;
 805
 806                 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
 807                         if (sk->sk_family == PF_INET)
 808                                 goto found;
 809         }
 810         sk = NULL;
 811 found:
 812         return sk;
 813 }
 814
 815 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
 816 {
 817         struct raw_iter_state* state = raw_seq_private(seq);
 818
 819         do {
 820                 sk = sk_next(sk);
 821 try_again:
 822                 ;
 823         } while (sk && sk->sk_family != PF_INET);
 824
 825         if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
 826                 sk = sk_head(&raw_v4_htable[state->bucket]);
 827                 goto try_again;
 828         }
 829         return sk;
 830 }
 831
 832 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
 833 {
 834         struct sock *sk = raw_get_first(seq);
 835
 836         if (sk)
 837                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
 838                         --pos;
 839         return pos ? NULL : sk;
 840 }
 841
 842 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
 843 {
 844         read_lock(&raw_v4_lock);
 845         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 846 }
 847
 848 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 849 {
 850         struct sock *sk;
 851
 852         if (v == SEQ_START_TOKEN)
 853                 sk = raw_get_first(seq);
 854         else
 855                 sk = raw_get_next(seq, v);
 856         ++*pos;
 857         return sk;
 858 }
 859
 860 static void raw_seq_stop(struct seq_file *seq, void *v)
 861 {
 862         read_unlock(&raw_v4_lock);
 863 }
 864
 865 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
 866 {
 867         struct inet_sock *inet = inet_sk(sp);
 868         __be32 dest = inet->daddr,
 869                src = inet->rcv_saddr;
 870         __u16 destp = 0,
 871               srcp  = inet->num;
 872
 873         sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
 874                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
 875                 i, src, srcp, dest, destp, sp->sk_state,
 876                 atomic_read(&sp->sk_wmem_alloc),
 877                 atomic_read(&sp->sk_rmem_alloc),
 878                 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
 879                 atomic_read(&sp->sk_refcnt), sp);
 880         return tmpbuf;
 881 }
 882
 883 static int raw_seq_show(struct seq_file *seq, void *v)
 884 {
 885         char tmpbuf[129];
 886
 887         if (v == SEQ_START_TOKEN)
 888                 seq_printf(seq, "%-127s\n",
 889                                "  sl  local_address rem_address   st tx_queue "
 890                                "rx_queue tr tm->when retrnsmt   uid  timeout "
 891                                "inode");
 892         else {
 893                 struct raw_iter_state *state = raw_seq_private(seq);
 894
 895                 seq_printf(seq, "%-127s\n",
 896                            get_raw_sock(v, tmpbuf, state->bucket));
 897         }
 898         return 0;
 899 }
 900
 901 static const struct seq_operations raw_seq_ops = {
 902         .start = raw_seq_start,
 903         .next  = raw_seq_next,
 904         .stop  = raw_seq_stop,
 905         .show  = raw_seq_show,
 906 };
 907
 908 static int raw_seq_open(struct inode *inode, struct file *file)
 909 {
 910         return seq_open_private(file, &raw_seq_ops,
 911                         sizeof(struct raw_iter_state));
 912 }
 913
 914 static const struct file_operations raw_seq_fops = {
 915         .owner   = THIS_MODULE,
 916         .open    = raw_seq_open,
 917         .read    = seq_read,
 918         .llseek  = seq_lseek,
 919         .release = seq_release_private,
 920 };
 921
 922 int __init raw_proc_init(void)
 923 {
 924         if (!proc_net_fops_create(&init_net, "raw", S_IRUGO, &raw_seq_fops))
 925                 return -ENOMEM;
 926         return 0;
 927 }
 928
 929 void __init raw_proc_exit(void)
 930 {
 931         proc_net_remove(&init_net, "raw");
 932 }
 933 #endif /* CONFIG_PROC_FS */