net/socket.c

   1 /*
   2  * NET          An implementation of the SOCKET network access protocol.
   3  *
   4  * Version:     @(#)socket.c    1.1.93  18/02/95
   5  *
   6  * Authors:     Orest Zborowski, <obz@Kodak.COM>
   7  *              Ross Biro, <bir7@leland.Stanford.Edu>
   8  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
   9  *
  10  * Fixes:
  11  *              Anonymous       :       NOTSOCK/BADF cleanup. Error fix in
  12  *                                      shutdown()
  13  *              Alan Cox        :       verify_area() fixes
  14  *              Alan Cox        :       Removed DDI
  15  *              Jonathan Kamens :       SOCK_DGRAM reconnect bug
  16  *              Alan Cox        :       Moved a load of checks to the very
  17  *                                      top level.
  18  *              Alan Cox        :       Move address structures to/from user
  19  *                                      mode above the protocol layers.
  20  *              Rob Janssen     :       Allow 0 length sends.
  21  *              Alan Cox        :       Asynchronous I/O support (cribbed from the
  22  *                                      tty drivers).
  23  *              Niibe Yutaka    :       Asynchronous I/O for writes (4.4BSD style)
  24  *              Jeff Uphoff     :       Made max number of sockets command-line
  25  *                                      configurable.
  26  *              Matti Aarnio    :       Made the number of sockets dynamic,
  27  *                                      to be allocated when needed, and mr.
  28  *                                      Uphoff's max is used as max to be
  29  *                                      allowed to allocate.
  30  *              Linus           :       Argh. removed all the socket allocation
  31  *                                      altogether: it's in the inode now.
  32  *              Alan Cox        :       Made sock_alloc()/sock_release() public
  33  *                                      for NetROM and future kernel nfsd type
  34  *                                      stuff.
  35  *              Alan Cox        :       sendmsg/recvmsg basics.
  36  *              Tom Dyas        :       Export net symbols.
  37  *              Marcin Dalecki  :       Fixed problems with CONFIG_NET="n".
  38  *              Alan Cox        :       Added thread locking to sys_* calls
  39  *                                      for sockets. May have errors at the
  40  *                                      moment.
  41  *              Kevin Buhr      :       Fixed the dumb errors in the above.
  42  *              Andi Kleen      :       Some small cleanups, optimizations,
  43  *                                      and fixed a copy_from_user() bug.
  44  *              Tigran Aivazian :       sys_send(args) calls sys_sendto(args, NULL, 0)
  45  *
  46  *
  47  *              This program is free software; you can redistribute it and/or
  48  *              modify it under the terms of the GNU General Public License
  49  *              as published by the Free Software Foundation; either version
  50  *              2 of the License, or (at your option) any later version.
  51  *
  52  *
  53  *      This module is effectively the top level interface to the BSD socket
  54  *      paradigm.
  55  *
  56  */
  57
  58 #include <linux/config.h>
  59 #include <linux/mm.h>
  60 #include <linux/smp_lock.h>
  61 #include <linux/socket.h>
  62 #include <linux/file.h>
  63 #include <linux/net.h>
  64 #include <linux/interrupt.h>
  65 #include <linux/netdevice.h>
  66 #include <linux/proc_fs.h>
  67 #include <linux/firewall.h>
  68 #include <linux/wanrouter.h>
  69 #include <linux/init.h>
  70 #include <linux/poll.h>
  71
  72 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
  73 #include <linux/kmod.h>
  74 #endif
  75
  76 #include <asm/uaccess.h>
  77
  78 #include <linux/inet.h>
  79 #include <net/ip.h>
  80 #include <net/sock.h>
  81 #include <net/rarp.h>
  82 #include <net/tcp.h>
  83 #include <net/udp.h>
  84 #include <net/scm.h>
  85
  86 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
  87 static long long sock_lseek(struct file *file, long long offset, int whence);
  88 static ssize_t sock_read(struct file *file, char *buf,
  89                          size_t size, loff_t *ppos);
  90 static ssize_t sock_write(struct file *file, const char *buf,
  91                           size_t size, loff_t *ppos);
  92
  93 static int sock_close(struct inode *inode, struct file *file);
  94 static unsigned int sock_poll(struct file *file,
  95                               struct poll_table_struct *wait);
  96 static int sock_ioctl(struct inode *inode, struct file *file,
  97                       unsigned int cmd, unsigned long arg);
  98 static int sock_fasync(int fd, struct file *filp, int on);
  99
 100
 101 /*
 102  *      Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
 103  *      in the operation structures but are done directly via the socketcall() multiplexor.
 104  */
 105
 106 static struct file_operations socket_file_ops = {
 107         sock_lseek,
 108         sock_read,
 109         sock_write,
 110         NULL,                   /* readdir */
 111         sock_poll,
 112         sock_ioctl,
 113         NULL,                   /* mmap */
 114         sock_no_open,           /* special open code to disallow open via /proc */
 115         NULL,                   /* flush */
 116         sock_close,
 117         NULL,                   /* no fsync */
 118         sock_fasync
 119 };
 120
 121 /*
 122  *      The protocol list. Each protocol is registered in here.
 123  */
 124
 125 struct net_proto_family *net_families[NPROTO];
 126
 127 /*
 128  *      Statistics counters of the socket lists
 129  */
 130
 131 static int sockets_in_use  = 0;
 132
 133 /*
 134  *      Support routines. Move socket addresses back and forth across the kernel/user
 135  *      divide and look after the messy bits.
 136  */
 137
 138 #define MAX_SOCK_ADDR   128             /* 108 for Unix domain -
 139                                            16 for IP, 16 for IPX,
 140                                            24 for IPv6,
 141                                            about 80 for AX.25
 142                                            must be at least one bigger than
 143                                            the AF_UNIX size (see net/unix/af_unix.c
 144                                            :unix_mkname()).
 145                                          */
 146
 147 int move_addr_to_kernel(void *uaddr, int ulen, void *kaddr)
 148 {
 149         if(ulen<0||ulen>MAX_SOCK_ADDR)
 150                 return -EINVAL;
 151         if(ulen==0)
 152                 return 0;
 153         if(copy_from_user(kaddr,uaddr,ulen))
 154                 return -EFAULT;
 155         return 0;
 156 }
 157
 158 int move_addr_to_user(void *kaddr, int klen, void *uaddr, int *ulen)
 159 {
 160         int err;
 161         int len;
 162
 163         if((err=get_user(len, ulen)))
 164                 return err;
 165         if(len>klen)
 166                 len=klen;
 167         if(len<0 || len> MAX_SOCK_ADDR)
 168                 return -EINVAL;
 169         if(len)
 170         {
 171                 if(copy_to_user(uaddr,kaddr,len))
 172                         return -EFAULT;
 173         }
 174         /*
 175          *      "fromlen shall refer to the value before truncation.."
 176          *                      1003.1g
 177          */
 178         return __put_user(klen, ulen);
 179 }
 180
 181 /*
 182  *      Obtains the first available file descriptor and sets it up for use.
 183  */
 184
 185 static int get_fd(struct inode *inode)
 186 {
 187         int fd;
 188
 189         /*
 190          *      Find a file descriptor suitable for return to the user.
 191          */
 192
 193         fd = get_unused_fd();
 194         if (fd >= 0) {
 195                 struct file *file = get_empty_filp();
 196
 197                 if (!file) {
 198                         put_unused_fd(fd);
 199                         return -ENFILE;
 200                 }
 201
 202                 file->f_dentry = d_alloc_root(inode, NULL);
 203                 if (!file->f_dentry) {
 204                         put_filp(file);
 205                         put_unused_fd(fd);
 206                         return -ENOMEM;
 207                 }
 208
 209                 /*
 210                  * The socket maintains a reference to the inode, so we
 211                  * have to increment the count.
 212                  */
 213                 inode->i_count++;
 214
 215                 fd_install(fd, file);
 216                 file->f_op = &socket_file_ops;
 217                 file->f_mode = 3;
 218                 file->f_flags = O_RDWR;
 219                 file->f_pos = 0;
 220         }
 221         return fd;
 222 }
 223
 224 extern __inline__ struct socket *socki_lookup(struct inode *inode)
 225 {
 226         return &inode->u.socket_i;
 227 }
 228
 229 /*
 230  *      Go from a file number to its socket slot.
 231  */
 232
 233 extern struct socket *sockfd_lookup(int fd, int *err)
 234 {
 235         struct file *file;
 236         struct inode *inode;
 237         struct socket *sock;
 238
 239         if (!(file = fget(fd)))
 240         {
 241                 *err = -EBADF;
 242                 return NULL;
 243         }
 244
 245         inode = file->f_dentry->d_inode;
 246         if (!inode || !inode->i_sock || !(sock = socki_lookup(inode)))
 247         {
 248                 *err = -ENOTSOCK;
 249                 fput(file);
 250                 return NULL;
 251         }
 252
 253         if (sock->file != file) {
 254                 printk(KERN_ERR "socki_lookup: socket file changed!\n");
 255                 sock->file = file;
 256         }
 257         return sock;
 258 }
 259
 260 extern __inline__ void sockfd_put(struct socket *sock)
 261 {
 262         fput(sock->file);
 263 }
 264
 265 /*
 266  *      Allocate a socket.
 267  */
 268
 269 struct socket *sock_alloc(void)
 270 {
 271         struct inode * inode;
 272         struct socket * sock;
 273
 274         inode = get_empty_inode();
 275         if (!inode)
 276                 return NULL;
 277
 278         sock = socki_lookup(inode);
 279
 280         inode->i_mode = S_IFSOCK|S_IRWXUGO;
 281         inode->i_sock = 1;
 282         inode->i_uid = current->fsuid;
 283         inode->i_gid = current->fsgid;
 284
 285         sock->inode = inode;
 286         init_waitqueue_head(&sock->wait);
 287         sock->fasync_list = NULL;
 288         sock->state = SS_UNCONNECTED;
 289         sock->flags = 0;
 290         sock->ops = NULL;
 291         sock->sk = NULL;
 292         sock->file = NULL;
 293
 294         sockets_in_use++;
 295         return sock;
 296 }
 297
 298 /*
 299  *      In theory you can't get an open on this inode, but /proc provides
 300  *      a back door. Remember to keep it shut otherwise you'll let the
 301  *      creepy crawlies in.
 302  */
 303
 304 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 305 {
 306         return -ENXIO;
 307 }
 308
 309 void sock_release(struct socket *sock)
 310 {
 311         if (sock->state != SS_UNCONNECTED)
 312                 sock->state = SS_DISCONNECTING;
 313
 314         if (sock->ops)
 315                 sock->ops->release(sock, NULL);
 316
 317         if (sock->fasync_list)
 318                 printk(KERN_ERR "sock_release: fasync list not empty!\n");
 319
 320         --sockets_in_use;       /* Bookkeeping.. */
 321         sock->file=NULL;
 322         iput(sock->inode);
 323 }
 324
 325 int sock_sendmsg(struct socket *sock, struct msghdr *msg, int size)
 326 {
 327         int err;
 328         struct scm_cookie scm;
 329
 330         err = scm_send(sock, msg, &scm);
 331         if (err >= 0) {
 332                 err = sock->ops->sendmsg(sock, msg, size, &scm);
 333                 scm_destroy(&scm);
 334         }
 335         return err;
 336 }
 337
 338 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int size, int flags)
 339 {
 340         struct scm_cookie scm;
 341
 342         memset(&scm, 0, sizeof(scm));
 343
 344         size = sock->ops->recvmsg(sock, msg, size, flags, &scm);
 345         if (size >= 0)
 346                 scm_recv(sock, msg, &scm, flags);
 347
 348         return size;
 349 }
 350
 351
 352 /*
 353  *      Sockets are not seekable.
 354  */
 355
 356 static long long sock_lseek(struct file *file,long long offset, int whence)
 357 {
 358         return -ESPIPE;
 359 }
 360
 361 /*
 362  *      Read data from a socket. ubuf is a user mode pointer. We make sure the user
 363  *      area ubuf...ubuf+size-1 is writable before asking the protocol.
 364  */
 365
 366 static ssize_t sock_read(struct file *file, char *ubuf,
 367                          size_t size, loff_t *ppos)
 368 {
 369         struct socket *sock;
 370         struct iovec iov;
 371         struct msghdr msg;
 372
 373         if (ppos != &file->f_pos)
 374                 return -ESPIPE;
 375         if (size==0)            /* Match SYS5 behaviour */
 376                 return 0;
 377
 378         sock = socki_lookup(file->f_dentry->d_inode);
 379
 380         msg.msg_name=NULL;
 381         msg.msg_namelen=0;
 382         msg.msg_iov=&iov;
 383         msg.msg_iovlen=1;
 384         msg.msg_control=NULL;
 385         msg.msg_controllen=0;
 386         iov.iov_base=ubuf;
 387         iov.iov_len=size;
 388
 389         return sock_recvmsg(sock, &msg, size,
 390                             !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT);
 391 }
 392
 393
 394 /*
 395  *      Write data to a socket. We verify that the user area ubuf..ubuf+size-1
 396  *      is readable by the user process.
 397  */
 398
 399 static ssize_t sock_write(struct file *file, const char *ubuf,
 400                           size_t size, loff_t *ppos)
 401 {
 402         struct socket *sock;
 403         struct msghdr msg;
 404         struct iovec iov;
 405
 406         if (ppos != &file->f_pos)
 407                 return -ESPIPE;
 408         if(size==0)             /* Match SYS5 behaviour */
 409                 return 0;
 410
 411         sock = socki_lookup(file->f_dentry->d_inode);
 412
 413         msg.msg_name=NULL;
 414         msg.msg_namelen=0;
 415         msg.msg_iov=&iov;
 416         msg.msg_iovlen=1;
 417         msg.msg_control=NULL;
 418         msg.msg_controllen=0;
 419         msg.msg_flags=!(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
 420         iov.iov_base=(void *)ubuf;
 421         iov.iov_len=size;
 422
 423         return sock_sendmsg(sock, &msg, size);
 424 }
 425
 426 int sock_readv_writev(int type, struct inode * inode, struct file * file,
 427                       const struct iovec * iov, long count, long size)
 428 {
 429         struct msghdr msg;
 430         struct socket *sock;
 431
 432         sock = socki_lookup(inode);
 433
 434         msg.msg_name = NULL;
 435         msg.msg_namelen = 0;
 436         msg.msg_control = NULL;
 437         msg.msg_controllen = 0;
 438         msg.msg_iov = (struct iovec *) iov;
 439         msg.msg_iovlen = count;
 440         msg.msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 441
 442         /* read() does a VERIFY_WRITE */
 443         if (type == VERIFY_WRITE)
 444                 return sock_recvmsg(sock, &msg, size, msg.msg_flags);
 445         return sock_sendmsg(sock, &msg, size);
 446 }
 447
 448
 449 /*
 450  *      With an ioctl arg may well be a user mode pointer, but we don't know what to do
 451  *      with it - that's up to the protocol still.
 452  */
 453
 454 int sock_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
 455            unsigned long arg)
 456 {
 457         struct socket *sock = socki_lookup(inode);
 458         return sock->ops->ioctl(sock, cmd, arg);
 459 }
 460
 461
 462 static unsigned int sock_poll(struct file *file, poll_table * wait)
 463 {
 464         struct socket *sock;
 465
 466         sock = socki_lookup(file->f_dentry->d_inode);
 467
 468         /*
 469          *      We can't return errors to poll, so it's either yes or no.
 470          */
 471
 472         return sock->ops->poll(file, sock, wait);
 473 }
 474
 475
 476 int sock_close(struct inode *inode, struct file *filp)
 477 {
 478         /*
 479          *      It was possible the inode is NULL we were
 480          *      closing an unfinished socket.
 481          */
 482
 483         if (!inode)
 484         {
 485                 printk(KERN_DEBUG "sock_close: NULL inode\n");
 486                 return 0;
 487         }
 488         sock_fasync(-1, filp, 0);
 489         sock_release(socki_lookup(inode));
 490         return 0;
 491 }
 492
 493 /*
 494  *      Update the socket async list
 495  */
 496
 497 static int sock_fasync(int fd, struct file *filp, int on)
 498 {
 499         struct fasync_struct *fa, *fna=NULL, **prev;
 500         struct socket *sock;
 501
 502         if (on)
 503         {
 504                 fna=(struct fasync_struct *)kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
 505                 if(fna==NULL)
 506                         return -ENOMEM;
 507         }
 508
 509         sock = socki_lookup(filp->f_dentry->d_inode);
 510
 511         prev=&(sock->fasync_list);
 512
 513         lock_sock(sock->sk);
 514
 515         for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
 516                 if (fa->fa_file==filp)
 517                         break;
 518
 519         if(on)
 520         {
 521                 if(fa!=NULL)
 522                 {
 523                         fa->fa_fd=fd;
 524                         kfree_s(fna,sizeof(struct fasync_struct));
 525                         release_sock(sock->sk);
 526                         return 0;
 527                 }
 528                 fna->fa_file=filp;
 529                 fna->fa_fd=fd;
 530                 fna->magic=FASYNC_MAGIC;
 531                 fna->fa_next=sock->fasync_list;
 532                 sock->fasync_list=fna;
 533         }
 534         else
 535         {
 536                 if (fa!=NULL)
 537                 {
 538                         *prev=fa->fa_next;
 539                         kfree_s(fa,sizeof(struct fasync_struct));
 540                 }
 541         }
 542
 543         release_sock(sock->sk);
 544         return 0;
 545 }
 546
 547 int sock_wake_async(struct socket *sock, int how)
 548 {
 549         if (!sock || !sock->fasync_list)
 550                 return -1;
 551         switch (how)
 552         {
 553         case 1:
 554                 if (sock->flags & SO_WAITDATA)
 555                         break;
 556                 goto call_kill;
 557         case 2:
 558                 if (!(sock->flags & SO_NOSPACE))
 559                         break;
 560                 sock->flags &= ~SO_NOSPACE;
 561                 /* fall through */
 562         case 0:
 563         call_kill:
 564                 kill_fasync(sock->fasync_list, SIGIO);
 565                 break;
 566         }
 567         return 0;
 568 }
 569
 570
 571 int sock_create(int family, int type, int protocol, struct socket **res)
 572 {
 573         int i;
 574         struct socket *sock;
 575
 576         /*
 577          *      Check protocol is in range
 578          */
 579         if(family<0||family>=NPROTO)
 580                 return -EINVAL;
 581
 582 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
 583         /* Attempt to load a protocol module if the find failed.
 584          *
 585          * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
 586          * requested real, full-featured networking support upon configuration.
 587          * Otherwise module support will break!
 588          */
 589         if (net_families[family]==NULL)
 590         {
 591                 char module_name[30];
 592                 sprintf(module_name,"net-pf-%d",family);
 593                 request_module(module_name);
 594         }
 595 #endif
 596
 597         if (net_families[family]==NULL)
 598                 return -EINVAL;
 599
 600 /*
 601  *      Check that this is a type that we know how to manipulate and
 602  *      the protocol makes sense here. The family can still reject the
 603  *      protocol later.
 604  */
 605
 606         if ((type != SOCK_STREAM && type != SOCK_DGRAM &&
 607              type != SOCK_SEQPACKET && type != SOCK_RAW && type != SOCK_RDM &&
 608 #ifdef CONFIG_XTP
 609                 type != SOCK_WEB  &&
 610 #endif
 611              type != SOCK_PACKET) || protocol < 0)
 612                         return -EINVAL;
 613
 614 /*
 615  *      Allocate the socket and allow the family to set things up. if
 616  *      the protocol is 0, the family is instructed to select an appropriate
 617  *      default.
 618  */
 619
 620         if (!(sock = sock_alloc()))
 621         {
 622                 printk(KERN_WARNING "socket: no more sockets\n");
 623                 return -ENFILE;         /* Not exactly a match, but its the
 624                                            closest posix thing */
 625         }
 626
 627         sock->type   = type;
 628
 629         if ((i = net_families[family]->create(sock, protocol)) < 0)
 630         {
 631                 sock_release(sock);
 632                 return i;
 633         }
 634
 635         *res = sock;
 636         return 0;
 637 }
 638
 639 asmlinkage int sys_socket(int family, int type, int protocol)
 640 {
 641         int retval;
 642         struct socket *sock;
 643
 644         lock_kernel();
 645
 646         retval = sock_create(family, type, protocol, &sock);
 647         if (retval < 0)
 648                 goto out;
 649
 650         retval = get_fd(sock->inode);
 651         if (retval < 0)
 652                 goto out_release;
 653         sock->file = fcheck(retval);
 654
 655 out:
 656         unlock_kernel();
 657         return retval;
 658
 659 out_release:
 660         sock_release(sock);
 661         goto out;
 662 }
 663
 664 /*
 665  *      Create a pair of connected sockets.
 666  */
 667
 668 asmlinkage int sys_socketpair(int family, int type, int protocol, int usockvec[2])
 669 {
 670         struct socket *sock1, *sock2;
 671         int fd1, fd2, err;
 672
 673         lock_kernel();
 674
 675         /*
 676          * Obtain the first socket and check if the underlying protocol
 677          * supports the socketpair call.
 678          */
 679
 680         err = sys_socket(family, type, protocol);
 681         if (err < 0)
 682                 goto out;
 683         fd1 = err;
 684
 685         /*
 686          * Now grab another socket
 687          */
 688         err = -EINVAL;
 689         fd2 = sys_socket(family, type, protocol);
 690         if (fd2 < 0)
 691                 goto out_close1;
 692
 693         /*
 694          * Get the sockets for the two fd's
 695          */
 696         sock1 = sockfd_lookup(fd1, &err);
 697         if (!sock1)
 698                 goto out_close2;
 699         sock2 = sockfd_lookup(fd2, &err);
 700         if (!sock2)
 701                 goto out_put1;
 702
 703         /* try to connect the two sockets together */
 704         err = sock1->ops->socketpair(sock1, sock2);
 705         if (err < 0)
 706                 goto out_put2;
 707
 708         err = put_user(fd1, &usockvec[0]);
 709         if (err)
 710                 goto out_put2;
 711         err = put_user(fd2, &usockvec[1]);
 712
 713 out_put2:
 714         sockfd_put(sock2);
 715 out_put1:
 716         sockfd_put(sock1);
 717
 718         if (err) {
 719         out_close2:
 720                 sys_close(fd2);
 721         out_close1:
 722                 sys_close(fd1);
 723         }
 724 out:
 725         unlock_kernel();
 726         return err;
 727 }
 728
 729
 730 /*
 731  *      Bind a name to a socket. Nothing much to do here since it's
 732  *      the protocol's responsibility to handle the local address.
 733  *
 734  *      We move the socket address to kernel space before we call
 735  *      the protocol layer (having also checked the address is ok).
 736  */
 737
 738 asmlinkage int sys_bind(int fd, struct sockaddr *umyaddr, int addrlen)
 739 {
 740         struct socket *sock;
 741         char address[MAX_SOCK_ADDR];
 742         int err;
 743
 744         lock_kernel();
 745         if((sock = sockfd_lookup(fd,&err))!=NULL)
 746         {
 747                 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0)
 748                         err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
 749                 sockfd_put(sock);
 750         }
 751         unlock_kernel();
 752         return err;
 753 }
 754
 755
 756 /*
 757  *      Perform a listen. Basically, we allow the protocol to do anything
 758  *      necessary for a listen, and if that works, we mark the socket as
 759  *      ready for listening.
 760  */
 761
 762 asmlinkage int sys_listen(int fd, int backlog)
 763 {
 764         struct socket *sock;
 765         int err;
 766
 767         lock_kernel();
 768         if((sock = sockfd_lookup(fd, &err))!=NULL)
 769         {
 770                 err=sock->ops->listen(sock, backlog);
 771                 sockfd_put(sock);
 772         }
 773         unlock_kernel();
 774         return err;
 775 }
 776
 777
 778 /*
 779  *      For accept, we attempt to create a new socket, set up the link
 780  *      with the client, wake up the client, then return the new
 781  *      connected fd. We collect the address of the connector in kernel
 782  *      space and move it to user at the very end. This is unclean because
 783  *      we open the socket then return an error.
 784  *
 785  *      1003.1g adds the ability to recvmsg() to query connection pending
 786  *      status to recvmsg. We need to add that support in a way thats
 787  *      clean when we restucture accept also.
 788  */
 789
 790 asmlinkage int sys_accept(int fd, struct sockaddr *upeer_sockaddr, int *upeer_addrlen)
 791 {
 792         struct inode *inode;
 793         struct socket *sock, *newsock;
 794         int err, len;
 795         char address[MAX_SOCK_ADDR];
 796
 797         lock_kernel();
 798         sock = sockfd_lookup(fd, &err);
 799         if (!sock)
 800                 goto out;
 801
 802 restart:
 803         err = -EMFILE;
 804         if (!(newsock = sock_alloc()))
 805                 goto out_put;
 806
 807         inode = newsock->inode;
 808         newsock->type = sock->type;
 809
 810         err = sock->ops->dup(newsock, sock);
 811         if (err < 0)
 812                 goto out_release;
 813
 814         err = newsock->ops->accept(sock, newsock, sock->file->f_flags);
 815         if (err < 0)
 816                 goto out_release;
 817         newsock = socki_lookup(inode);
 818
 819         if ((err = get_fd(inode)) < 0)
 820                 goto out_release;
 821         newsock->file = fcheck(err);
 822
 823         if (upeer_sockaddr)
 824         {
 825                 /* Handle the race where the accept works and we
 826                    then getname after it has closed again */
 827                 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 1)<0)
 828                 {
 829                         sys_close(err);
 830                         goto restart;
 831                 }
 832                 /* N.B. Should check for errors here */
 833                 move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
 834         }
 835
 836 out_put:
 837         sockfd_put(sock);
 838 out:
 839         unlock_kernel();
 840         return err;
 841
 842 out_release:
 843         sock_release(newsock);
 844         goto out_put;
 845 }
 846
 847
 848 /*
 849  *      Attempt to connect to a socket with the server address.  The address
 850  *      is in user space so we verify it is OK and move it to kernel space.
 851  *
 852  *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
 853  *      break bindings
 854  *
 855  *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
 856  *      other SEQPACKET protocols that take time to connect() as it doesn't
 857  *      include the -EINPROGRESS status for such sockets.
 858  */
 859
 860 asmlinkage int sys_connect(int fd, struct sockaddr *uservaddr, int addrlen)
 861 {
 862         struct socket *sock;
 863         char address[MAX_SOCK_ADDR];
 864         int err;
 865
 866         lock_kernel();
 867         sock = sockfd_lookup(fd, &err);
 868         if (!sock)
 869                 goto out;
 870         err = move_addr_to_kernel(uservaddr, addrlen, address);
 871         if (err < 0)
 872                 goto out_put;
 873         err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
 874                                  sock->file->f_flags);
 875 out_put:
 876         sockfd_put(sock);
 877 out:
 878         unlock_kernel();
 879         return err;
 880 }
 881
 882 /*
 883  *      Get the local address ('name') of a socket object. Move the obtained
 884  *      name to user space.
 885  */
 886
 887 asmlinkage int sys_getsockname(int fd, struct sockaddr *usockaddr, int *usockaddr_len)
 888 {
 889         struct socket *sock;
 890         char address[MAX_SOCK_ADDR];
 891         int len, err;
 892
 893         lock_kernel();
 894         sock = sockfd_lookup(fd, &err);
 895         if (!sock)
 896                 goto out;
 897         err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
 898         if (err)
 899                 goto out_put;
 900         err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
 901
 902 out_put:
 903         sockfd_put(sock);
 904 out:
 905         unlock_kernel();
 906         return err;
 907 }
 908
 909 /*
 910  *      Get the remote address ('name') of a socket object. Move the obtained
 911  *      name to user space.
 912  */
 913
 914 asmlinkage int sys_getpeername(int fd, struct sockaddr *usockaddr, int *usockaddr_len)
 915 {
 916         struct socket *sock;
 917         char address[MAX_SOCK_ADDR];
 918         int len, err;
 919
 920         lock_kernel();
 921         if ((sock = sockfd_lookup(fd, &err))!=NULL)
 922         {
 923                 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
 924                 if (!err)
 925                         err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
 926                 sockfd_put(sock);
 927         }
 928         unlock_kernel();
 929         return err;
 930 }
 931
 932 /*
 933  *      Send a datagram to a given address. We move the address into kernel
 934  *      space and check the user space data area is readable before invoking
 935  *      the protocol.
 936  */
 937
 938 asmlinkage int sys_sendto(int fd, void * buff, size_t len, unsigned flags,
 939            struct sockaddr *addr, int addr_len)
 940 {
 941         struct socket *sock;
 942         char address[MAX_SOCK_ADDR];
 943         int err;
 944         struct msghdr msg;
 945         struct iovec iov;
 946
 947         lock_kernel();
 948         sock = sockfd_lookup(fd, &err);
 949         if (!sock)
 950                 goto out;
 951         iov.iov_base=buff;
 952         iov.iov_len=len;
 953         msg.msg_name=NULL;
 954         msg.msg_iov=&iov;
 955         msg.msg_iovlen=1;
 956         msg.msg_control=NULL;
 957         msg.msg_controllen=0;
 958         msg.msg_namelen=addr_len;
 959         if(addr)
 960         {
 961                 err = move_addr_to_kernel(addr, addr_len, address);
 962                 if (err < 0)
 963                         goto out_put;
 964                 msg.msg_name=address;
 965         }
 966         if (sock->file->f_flags & O_NONBLOCK)
 967                 flags |= MSG_DONTWAIT;
 968         msg.msg_flags = flags;
 969         err = sock_sendmsg(sock, &msg, len);
 970
 971 out_put:
 972         sockfd_put(sock);
 973 out:
 974         unlock_kernel();
 975         return err;
 976 }
 977
 978 /*
 979  *      Send a datagram down a socket.
 980  */
 981
 982 asmlinkage int sys_send(int fd, void * buff, size_t len, unsigned flags)
 983 {
 984         return sys_sendto(fd, buff, len, flags, NULL, 0);
 985 }
 986
 987 /*
 988  *      Receive a frame from the socket and optionally record the address of the
 989  *      sender. We verify the buffers are writable and if needed move the
 990  *      sender address from kernel to user space.
 991  */
 992
 993 asmlinkage int sys_recvfrom(int fd, void * ubuf, size_t size, unsigned flags,
 994              struct sockaddr *addr, int *addr_len)
 995 {
 996         struct socket *sock;
 997         struct iovec iov;
 998         struct msghdr msg;
 999         char address[MAX_SOCK_ADDR];
1000         int err,err2;
1001
1002         lock_kernel();
1003         sock = sockfd_lookup(fd, &err);
1004         if (!sock)
1005                 goto out;
1006
1007         msg.msg_control=NULL;
1008         msg.msg_controllen=0;
1009         msg.msg_iovlen=1;
1010         msg.msg_iov=&iov;
1011         iov.iov_len=size;
1012         iov.iov_base=ubuf;
1013         msg.msg_name=address;
1014         msg.msg_namelen=MAX_SOCK_ADDR;
1015         if (sock->file->f_flags & O_NONBLOCK)
1016                 flags |= MSG_DONTWAIT;
1017         err=sock_recvmsg(sock, &msg, size, flags);
1018
1019         if(err >= 0 && addr != NULL)
1020         {
1021                 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1022                 if(err2<0)
1023                         err=err2;
1024         }
1025         sockfd_put(sock);
1026 out:
1027         unlock_kernel();
1028         return err;
1029 }
1030
1031 /*
1032  *      Receive a datagram from a socket.
1033  */
1034
1035 asmlinkage int sys_recv(int fd, void * ubuf, size_t size, unsigned flags)
1036 {
1037         return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1038 }
1039
1040 /*
1041  *      Set a socket option. Because we don't know the option lengths we have
1042  *      to pass the user mode parameter for the protocols to sort out.
1043  */
1044
1045 asmlinkage int sys_setsockopt(int fd, int level, int optname, char *optval, int optlen)
1046 {
1047         int err;
1048         struct socket *sock;
1049
1050         lock_kernel();
1051         if ((sock = sockfd_lookup(fd, &err))!=NULL)
1052         {
1053                 if (level == SOL_SOCKET)
1054                         err=sock_setsockopt(sock,level,optname,optval,optlen);
1055                 else
1056                         err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1057                 sockfd_put(sock);
1058         }
1059         unlock_kernel();
1060         return err;
1061 }
1062
1063 /*
1064  *      Get a socket option. Because we don't know the option lengths we have
1065  *      to pass a user mode parameter for the protocols to sort out.
1066  */
1067
1068 asmlinkage int sys_getsockopt(int fd, int level, int optname, char *optval, int *optlen)
1069 {
1070         int err;
1071         struct socket *sock;
1072
1073         lock_kernel();
1074         if ((sock = sockfd_lookup(fd, &err))!=NULL)
1075         {
1076                 if (level == SOL_SOCKET)
1077                         err=sock_getsockopt(sock,level,optname,optval,optlen);
1078                 else
1079                         err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1080                 sockfd_put(sock);
1081         }
1082         unlock_kernel();
1083         return err;
1084 }
1085
1086
1087 /*
1088  *      Shutdown a socket.
1089  */
1090
1091 asmlinkage int sys_shutdown(int fd, int how)
1092 {
1093         int err;
1094         struct socket *sock;
1095
1096         lock_kernel();
1097         if ((sock = sockfd_lookup(fd, &err))!=NULL)
1098         {
1099                 err=sock->ops->shutdown(sock, how);
1100                 sockfd_put(sock);
1101         }
1102         unlock_kernel();
1103         return err;
1104 }
1105
1106 /*
1107  *      BSD sendmsg interface
1108  */
1109
1110 asmlinkage int sys_sendmsg(int fd, struct msghdr *msg, unsigned flags)
1111 {
1112         struct socket *sock;
1113         char address[MAX_SOCK_ADDR];
1114         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1115         unsigned char ctl[sizeof(struct cmsghdr) + 20]; /* 20 is size of ipv6_pktinfo */
1116         unsigned char *ctl_buf = ctl;
1117         struct msghdr msg_sys;
1118         int err, ctl_len, iov_size, total_len;
1119
1120         lock_kernel();
1121
1122         err = -EFAULT;
1123         if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1124                 goto out;
1125
1126         sock = sockfd_lookup(fd, &err);
1127         if (!sock)
1128                 goto out;
1129
1130         /* do not move before msg_sys is valid */
1131         err = -EINVAL;
1132         if (msg_sys.msg_iovlen > UIO_MAXIOV)
1133                 goto out_put;
1134
1135         /* Check whether to allocate the iovec area*/
1136         err = -ENOMEM;
1137         iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1138         if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1139                 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1140                 if (!iov)
1141                         goto out_put;
1142         }
1143
1144         /* This will also move the address data into kernel space */
1145         err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1146         if (err < 0)
1147                 goto out_freeiov;
1148         total_len = err;
1149
1150         err = -ENOBUFS;
1151
1152         /* msg_controllen must fit to int */
1153         if (msg_sys.msg_controllen > INT_MAX)
1154                 goto out_freeiov;
1155         ctl_len = msg_sys.msg_controllen;
1156         if (ctl_len)
1157         {
1158                 if (ctl_len > sizeof(ctl))
1159                 {
1160                         /* Suggested by the Advanced Sockets API for IPv6 draft:
1161                          * Limit the msg_controllen size by the SO_SNDBUF size.
1162                          */
1163                         /* Note - when this code becomes multithreaded on
1164                          * SMP machines you have a race to fix here.
1165                          */
1166                         err = -ENOBUFS;
1167                         ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1168                         if (ctl_buf == NULL)
1169                                 goto out_freeiov;
1170                 }
1171                 err = -EFAULT;
1172                 if (copy_from_user(ctl_buf, msg_sys.msg_control, ctl_len))
1173                         goto out_freectl;
1174                 msg_sys.msg_control = ctl_buf;
1175         }
1176         msg_sys.msg_flags = flags;
1177
1178         if (sock->file->f_flags & O_NONBLOCK)
1179                 msg_sys.msg_flags |= MSG_DONTWAIT;
1180         err = sock_sendmsg(sock, &msg_sys, total_len);
1181
1182 out_freectl:
1183         if (ctl_buf != ctl)
1184                 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1185 out_freeiov:
1186         if (iov != iovstack)
1187                 sock_kfree_s(sock->sk, iov, iov_size);
1188 out_put:
1189         sockfd_put(sock);
1190 out:
1191         unlock_kernel();
1192         return err;
1193 }
1194
1195 /*
1196  *      BSD recvmsg interface
1197  */
1198
1199 asmlinkage int sys_recvmsg(int fd, struct msghdr *msg, unsigned int flags)
1200 {
1201         struct socket *sock;
1202         struct iovec iovstack[UIO_FASTIOV];
1203         struct iovec *iov=iovstack;
1204         struct msghdr msg_sys;
1205         unsigned long cmsg_ptr;
1206         int err, iov_size, total_len, len;
1207
1208         /* kernel mode address */
1209         char addr[MAX_SOCK_ADDR];
1210
1211         /* user mode address pointers */
1212         struct sockaddr *uaddr;
1213         int *uaddr_len;
1214
1215         lock_kernel();
1216         err=-EFAULT;
1217         if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1218                 goto out;
1219
1220         sock = sockfd_lookup(fd, &err);
1221         if (!sock)
1222                 goto out;
1223
1224         err = -EINVAL;
1225         if (msg_sys.msg_iovlen > UIO_MAXIOV)
1226                 goto out_put;
1227
1228         /* Check whether to allocate the iovec area*/
1229         err = -ENOMEM;
1230         iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1231         if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1232                 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1233                 if (!iov)
1234                         goto out_put;
1235         }
1236
1237         /*
1238          *      Save the user-mode address (verify_iovec will change the
1239          *      kernel msghdr to use the kernel address space)
1240          */
1241
1242         uaddr = msg_sys.msg_name;
1243         uaddr_len = &msg->msg_namelen;
1244         err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1245         if (err < 0)
1246                 goto out_freeiov;
1247         total_len=err;
1248
1249         cmsg_ptr = (unsigned long)msg_sys.msg_control;
1250         msg_sys.msg_flags = 0;
1251
1252         if (sock->file->f_flags & O_NONBLOCK)
1253                 flags |= MSG_DONTWAIT;
1254         err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1255         if (err < 0)
1256                 goto out_freeiov;
1257         len = err;
1258
1259         if (uaddr != NULL) {
1260                 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1261                 if (err < 0)
1262                         goto out_freeiov;
1263         }
1264         err = __put_user(msg_sys.msg_flags, &msg->msg_flags);
1265         if (err)
1266                 goto out_freeiov;
1267         err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1268                                                          &msg->msg_controllen);
1269         if (err)
1270                 goto out_freeiov;
1271         err = len;
1272
1273 out_freeiov:
1274         if (iov != iovstack)
1275                 sock_kfree_s(sock->sk, iov, iov_size);
1276 out_put:
1277         sockfd_put(sock);
1278 out:
1279         unlock_kernel();
1280         return err;
1281 }
1282
1283
1284 /*
1285  *      Perform a file control on a socket file descriptor.
1286  *
1287  *      Doesn't aquire a fd lock, because no network fcntl
1288  *      function sleeps currently.
1289  */
1290
1291 int sock_fcntl(struct file *filp, unsigned int cmd, unsigned long arg)
1292 {
1293         struct socket *sock;
1294
1295         sock = socki_lookup (filp->f_dentry->d_inode);
1296         if (sock && sock->ops)
1297                 return sock->ops->fcntl(sock, cmd, arg);
1298         return(-EINVAL);
1299 }
1300
1301 /* Argument list sizes for sys_socketcall */
1302 #define AL(x) ((x) * sizeof(unsigned long))
1303 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1304                                 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1305                                 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1306 #undef AL
1307
1308 /*
1309  *      System call vectors.
1310  *
1311  *      Argument checking cleaned up. Saved 20% in size.
1312  *  This function doesn't need to set the kernel lock because
1313  *  it is set by the callees.
1314  */
1315
1316 asmlinkage int sys_socketcall(int call, unsigned long *args)
1317 {
1318         unsigned long a[6];
1319         unsigned long a0,a1;
1320         int err;
1321
1322         if(call<1||call>SYS_RECVMSG)
1323                 return -EINVAL;
1324
1325         /* copy_from_user should be SMP safe. */
1326         if (copy_from_user(a, args, nargs[call]))
1327                 return -EFAULT;
1328
1329         a0=a[0];
1330         a1=a[1];
1331
1332         switch(call)
1333         {
1334                 case SYS_SOCKET:
1335                         err = sys_socket(a0,a1,a[2]);
1336                         break;
1337                 case SYS_BIND:
1338                         err = sys_bind(a0,(struct sockaddr *)a1, a[2]);
1339                         break;
1340                 case SYS_CONNECT:
1341                         err = sys_connect(a0, (struct sockaddr *)a1, a[2]);
1342                         break;
1343                 case SYS_LISTEN:
1344                         err = sys_listen(a0,a1);
1345                         break;
1346                 case SYS_ACCEPT:
1347                         err = sys_accept(a0,(struct sockaddr *)a1, (int *)a[2]);
1348                         break;
1349                 case SYS_GETSOCKNAME:
1350                         err = sys_getsockname(a0,(struct sockaddr *)a1, (int *)a[2]);
1351                         break;
1352                 case SYS_GETPEERNAME:
1353                         err = sys_getpeername(a0, (struct sockaddr *)a1, (int *)a[2]);
1354                         break;
1355                 case SYS_SOCKETPAIR:
1356                         err = sys_socketpair(a0,a1, a[2], (int *)a[3]);
1357                         break;
1358                 case SYS_SEND:
1359                         err = sys_send(a0, (void *)a1, a[2], a[3]);
1360                         break;
1361                 case SYS_SENDTO:
1362                         err = sys_sendto(a0,(void *)a1, a[2], a[3],
1363                                          (struct sockaddr *)a[4], a[5]);
1364                         break;
1365                 case SYS_RECV:
1366                         err = sys_recv(a0, (void *)a1, a[2], a[3]);
1367                         break;
1368                 case SYS_RECVFROM:
1369                         err = sys_recvfrom(a0, (void *)a1, a[2], a[3],
1370                                            (struct sockaddr *)a[4], (int *)a[5]);
1371                         break;
1372                 case SYS_SHUTDOWN:
1373                         err = sys_shutdown(a0,a1);
1374                         break;
1375                 case SYS_SETSOCKOPT:
1376                         err = sys_setsockopt(a0, a1, a[2], (char *)a[3], a[4]);
1377                         break;
1378                 case SYS_GETSOCKOPT:
1379                         err = sys_getsockopt(a0, a1, a[2], (char *)a[3], (int *)a[4]);
1380                         break;
1381                 case SYS_SENDMSG:
1382                         err = sys_sendmsg(a0, (struct msghdr *) a1, a[2]);
1383                         break;
1384                 case SYS_RECVMSG:
1385                         err = sys_recvmsg(a0, (struct msghdr *) a1, a[2]);
1386                         break;
1387                 default:
1388                         err = -EINVAL;
1389                         break;
1390         }
1391         return err;
1392 }
1393
1394 /*
1395  *      This function is called by a protocol handler that wants to
1396  *      advertise its address family, and have it linked into the
1397  *      SOCKET module.
1398  */
1399
1400 int sock_register(struct net_proto_family *ops)
1401 {
1402         if (ops->family >= NPROTO) {
1403                 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
1404                 return -ENOBUFS;
1405         }
1406         net_families[ops->family]=ops;
1407         return 0;
1408 }
1409
1410 /*
1411  *      This function is called by a protocol handler that wants to
1412  *      remove its address family, and have it unlinked from the
1413  *      SOCKET module.
1414  */
1415
1416 int sock_unregister(int family)
1417 {
1418         if (family < 0 || family >= NPROTO)
1419                 return -1;
1420
1421         net_families[family]=NULL;
1422         return 0;
1423 }
1424
1425 void __init proto_init(void)
1426 {
1427         extern struct net_proto protocols[];    /* Network protocols */
1428         struct net_proto *pro;
1429
1430         /* Kick all configured protocols. */
1431         pro = protocols;
1432         while (pro->name != NULL)
1433         {
1434                 (*pro->init_func)(pro);
1435                 pro++;
1436         }
1437         /* We're all done... */
1438 }
1439
1440 extern void sk_init(void);
1441 #ifdef CONFIG_WAN_ROUTER
1442 extern void wanrouter_init(void);
1443 #endif
1444
1445 void __init sock_init(void)
1446 {
1447         int i;
1448
1449         printk(KERN_INFO "Linux NET4.0 for Linux 2.2\n");
1450         printk(KERN_INFO "Based upon Swansea University Computer Society NET3.039\n");
1451
1452         /*
1453          *      Initialize all address (protocol) families.
1454          */
1455
1456         for (i = 0; i < NPROTO; i++)
1457                 net_families[i] = NULL;
1458
1459         /*
1460          *      Initialize sock SLAB cache.
1461          */
1462
1463         sk_init();
1464
1465 #ifdef SLAB_SKB
1466         /*
1467          *      Initialize skbuff SLAB cache
1468          */
1469         skb_init();
1470 #endif
1471
1472
1473         /*
1474          *      Wan router layer.
1475          */
1476
1477 #ifdef CONFIG_WAN_ROUTER
1478         wanrouter_init();
1479 #endif
1480
1481         /*
1482          *      Attach the firewall module if configured
1483          */
1484
1485 #ifdef CONFIG_FIREWALL
1486         fwchain_init();
1487 #endif
1488
1489         /*
1490          *      Initialize the protocols module.
1491          */
1492
1493         proto_init();
1494
1495         /*
1496          *      The netlink device handler may be needed early.
1497          */
1498
1499 #ifdef  CONFIG_RTNETLINK
1500         rtnetlink_init();
1501 #endif
1502 #ifdef CONFIG_NETLINK_DEV
1503         init_netlink();
1504 #endif
1505 }
1506
1507 int socket_get_info(char *buffer, char **start, off_t offset, int length)
1508 {
1509         int len = sprintf(buffer, "sockets: used %d\n", sockets_in_use);
1510         if (offset >= len)
1511         {
1512                 *start = buffer;
1513                 return 0;
1514         }
1515         *start = buffer + offset;
1516         len -= offset;
1517         if (len > length)
1518                 len = length;
1519         return len;
1520 }