net/ipv4/af

/* */
This source file includes following definitions.
sk_inuse
get_new_socknum
put_sock
remove_sock
destroy_sock
inet_fcntl
inet_setsockopt
inet_getsockopt
inet_autobind
inet_listen
def_callback1
def_callback2
def_callback3
inet_create
inet_dup
closing
inet_release
inet_bind
inet_error
inet_connect
inet_socketpair
inet_accept
inet_getname
inet_recvfrom
inet_recv
inet_read
inet_send
inet_write
inet_sendto
inet_shutdown
inet_select
inet_ioctl
get_sock
get_sock_raw
get_sock_mcast
inet_proto_init
   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  *              AF_INET protocol family socket handler.
   7  *
   8  * Version:     @(#)af_inet.c   (from sock.c) 1.0.17    06/02/93
   9  *
  10  * Authors:     Ross Biro, <bir7@leland.Stanford.Edu>
  11  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  12  *              Florian La Roche, <flla@stud.uni-sb.de>
  13  *              Alan Cox, <A.Cox@swansea.ac.uk>
  14  *
  15  * Changes (see also sock.c)
  16  *
  17  *              A.N.Kuznetsov   :       Socket death error in accept().
  18  *              John Richardson :       Fix non blocking error in connect()
  19  *                                      so sockets that fail to connect
  20  *                                      don't return -EINPROGRESS.
  21  *              Alan Cox        :       Asynchronous I/O support
  22  *              Alan Cox        :       Keep correct socket pointer on sock structures
  23  *                                      when accept() ed
  24  *              Alan Cox        :       Semantics of SO_LINGER aren't state moved
  25  *                                      to close when you look carefully. With
  26  *                                      this fixed and the accept bug fixed 
  27  *                                      some RPC stuff seems happier.
  28  *              Niibe Yutaka    :       4.4BSD style write async I/O
  29  *              Alan Cox, 
  30  *              Tony Gale       :       Fixed reuse semantics.
  31  *              Alan Cox        :       bind() shouldn't abort existing but dead
  32  *                                      sockets. Stops FTP netin:.. I hope.
  33  *              Alan Cox        :       bind() works correctly for RAW sockets. Note
  34  *                                      that FreeBSD at least is broken in this respect
  35  *                                      so be careful with compatibility tests...
  36  *              Alan Cox        :       routing cache support
  37  *              Alan Cox        :       memzero the socket structure for compactness.
  38  *              Matt Day        :       nonblock connect error handler
  39  *              Alan Cox        :       Allow large numbers of pending sockets
  40  *                                      (eg for big web sites), but only if
  41  *                                      specifically application requested.
  42  *              Alan Cox        :       New buffering throughout IP. Used dumbly.
  43  *              Alan Cox        :       New buffering now used smartly.
  44  *              Alan Cox        :       BSD rather than common sense interpretation of
  45  *                                      listen.
  46  *              Germano Caronni :       Assorted small races.
  47  *
  48  *              This program is free software; you can redistribute it and/or
  49  *              modify it under the terms of the GNU General Public License
  50  *              as published by the Free Software Foundation; either version
  51  *              2 of the License, or (at your option) any later version.
  52  */
  53 
  54 #include <linux/config.h>
  55 #include <linux/errno.h>
  56 #include <linux/types.h>
  57 #include <linux/socket.h>
  58 #include <linux/in.h>
  59 #include <linux/kernel.h>
  60 #include <linux/major.h>
  61 #include <linux/sched.h>
  62 #include <linux/timer.h>
  63 #include <linux/string.h>
  64 #include <linux/sockios.h>
  65 #include <linux/net.h>
  66 #include <linux/fcntl.h>
  67 #include <linux/mm.h>
  68 #include <linux/interrupt.h>
  69 #include <linux/proc_fs.h>
  70 #include <linux/stat.h>
  71 
  72 #include <asm/segment.h>
  73 #include <asm/system.h>
  74 
  75 #include <linux/inet.h>
  76 #include <linux/netdevice.h>
  77 #include <net/ip.h>
  78 #include <net/protocol.h>
  79 #include <net/arp.h>
  80 #include <net/rarp.h>
  81 #include <net/route.h>
  82 #include <net/tcp.h>
  83 #include <net/udp.h>
  84 #include <linux/skbuff.h>
  85 #include <net/sock.h>
  86 #include <net/raw.h>
  87 #include <net/icmp.h>
  88 #include <linux/ip_fw.h>
  89 
  90 #define min(a,b)        ((a)<(b)?(a):(b))
  91 
  92 extern struct proto packet_prot;
  93 extern int raw_get_info(char *, char **, off_t, int, int);
  94 extern int snmp_get_info(char *, char **, off_t, int, int);
  95 extern int afinet_get_info(char *, char **, off_t, int, int);
  96 extern int tcp_get_info(char *, char **, off_t, int, int);
  97 extern int udp_get_info(char *, char **, off_t, int, int);
  98 
  99 int (*rarp_ioctl_hook)(unsigned int,void*) = NULL;
 100 
 101 /*
 102  *      See if a socket number is in use.
 103  */
 104  
 105 static int sk_inuse(struct proto *prot, int num)
     /*  */
 106 {
 107         struct sock *sk;
 108 
 109         for(sk = prot->sock_array[num & (SOCK_ARRAY_SIZE -1 )];
 110                 sk != NULL;  sk=sk->next) 
 111         {
 112                 if (sk->num == num) 
 113                         return(1);
 114         }
 115         return(0);
 116 }
 117 
 118 
 119 /*
 120  *      Pick a new socket number
 121  */
 122 
 123 unsigned short get_new_socknum(struct proto *prot, unsigned short base)
     /*  */
 124 {
 125         static int start=0;
 126 
 127         /*
 128          * Used to cycle through the port numbers so the
 129          * chances of a confused connection drop.
 130          */
 131          
 132         int i, j;
 133         int best = 0;
 134         int size = 32767; /* a big num. */
 135         struct sock *sk;
 136 
 137         if (base == 0) 
 138                 base = PROT_SOCK+1+(start % 1024);
 139         if (base <= PROT_SOCK) 
 140         {
 141                 base += PROT_SOCK+(start % 1024);
 142         }
 143 
 144         /* Now look through the entire array and try to find an empty ptr. */
 145         for(i=0; i < SOCK_ARRAY_SIZE; i++) 
 146         {
 147                 j = 0;
 148                 sk = prot->sock_array[(i+base+1) &(SOCK_ARRAY_SIZE -1)];
 149                 while(sk != NULL) 
 150                 {
 151                         sk = sk->next;
 152                         j++;
 153                 }
 154                 if (j == 0) 
 155                 {
 156                         start =(i+1+start )%1024;
 157                         return(i+base+1);
 158                 }
 159                 if (j < size) 
 160                 {
 161                         best = i;
 162                         size = j;
 163                 }
 164         }
 165 
 166         /* Now make sure the one we want is not in use. */
 167 
 168         while(sk_inuse(prot, base +best+1)) 
 169         {
 170                 best += SOCK_ARRAY_SIZE;
 171         }
 172         return(best+base+1);
 173 }
 174 
 175 /*
 176  *      Add a socket into the socket tables by number.
 177  */
 178 
 179 void put_sock(unsigned short num, struct sock *sk)
     /*  */
 180 {
 181         struct sock *sk1;
 182         struct sock *sk2;
 183         int mask;
 184         unsigned long flags;
 185         
 186         if(sk->type==SOCK_PACKET)
 187                 return;
 188 
 189         sk->num = num;
 190         sk->next = NULL;
 191         num = num &(SOCK_ARRAY_SIZE -1);
 192 
 193         /* We can't have an interrupt re-enter here. */
 194         save_flags(flags);
 195         cli();
 196 
 197         sk->prot->inuse += 1;
 198         if (sk->prot->highestinuse < sk->prot->inuse)
 199                 sk->prot->highestinuse = sk->prot->inuse;
 200 
 201         if (sk->prot->sock_array[num] == NULL) 
 202         {
 203                 sk->prot->sock_array[num] = sk;
 204                 restore_flags(flags);
 205                 return;
 206         }
 207         restore_flags(flags);
 208         for(mask = 0xff000000; mask != 0xffffffff; mask = (mask >> 8) | mask) 
 209         {
 210                 if ((mask & sk->saddr) &&
 211                     (mask & sk->saddr) != (mask & 0xffffffff)) 
 212                 {
 213                         mask = mask << 8;
 214                         break;
 215                 }
 216         }
 217         cli();
 218         sk1 = sk->prot->sock_array[num];
 219         for(sk2 = sk1; sk2 != NULL; sk2=sk2->next) 
 220         {
 221                 if (!(sk2->saddr & mask)) 
 222                 {
 223                         if (sk2 == sk1) 
 224                         {
 225                                 sk->next = sk->prot->sock_array[num];
 226                                 sk->prot->sock_array[num] = sk;
 227                                 sti();
 228                                 return;
 229                         }
 230                         sk->next = sk2;
 231                         sk1->next= sk;
 232                         sti();
 233                         return;
 234                 }
 235                 sk1 = sk2;
 236         }
 237 
 238         /* Goes at the end. */
 239         sk->next = NULL;
 240         sk1->next = sk;
 241         sti();
 242 }
 243 
 244 /*
 245  *      Remove a socket from the socket tables.
 246  */
 247 
 248 static void remove_sock(struct sock *sk1)
     /*  */
 249 {
 250         struct sock *sk2;
 251         unsigned long flags;
 252 
 253         if (sk1->type==SOCK_PACKET)
 254                 return;
 255                 
 256         if (!sk1->prot) 
 257         {
 258                 NETDEBUG(printk("sock.c: remove_sock: sk1->prot == NULL\n"));
 259                 return;
 260         }
 261 
 262         /* We can't have this changing out from under us. */
 263         save_flags(flags);
 264         cli();
 265         sk2 = sk1->prot->sock_array[sk1->num &(SOCK_ARRAY_SIZE -1)];
 266         if (sk2 == sk1) 
 267         {
 268                 sk1->prot->inuse -= 1;
 269                 sk1->prot->sock_array[sk1->num &(SOCK_ARRAY_SIZE -1)] = sk1->next;
 270                 restore_flags(flags);
 271                 return;
 272         }
 273 
 274         while(sk2 && sk2->next != sk1) 
 275         {
 276                 sk2 = sk2->next;
 277         }
 278 
 279         if (sk2) 
 280         {
 281                 sk1->prot->inuse -= 1;
 282                 sk2->next = sk1->next;
 283                 restore_flags(flags);
 284                 return;
 285         }
 286         restore_flags(flags);
 287 }
 288 
 289 /*
 290  *      Destroy an AF_INET socket
 291  */
 292  
 293 void destroy_sock(struct sock *sk)
     /*  */
 294 {
 295         struct sk_buff *skb;
 296 
 297         sk->inuse = 1;                  /* just to be safe. */
 298 
 299         /* In case it's sleeping somewhere. */
 300         if (!sk->dead) 
 301                 sk->write_space(sk);
 302 
 303         remove_sock(sk);
 304   
 305         /* Now we can no longer get new packets. */
 306         delete_timer(sk);
 307         /* Nor send them */
 308         del_timer(&sk->retransmit_timer);
 309         
 310         while ((skb = tcp_dequeue_partial(sk)) != NULL) {
 311                 IS_SKB(skb);
 312                 kfree_skb(skb, FREE_WRITE);
 313         }
 314 
 315         /* Cleanup up the write buffer. */
 316         while((skb = skb_dequeue(&sk->write_queue)) != NULL) {
 317                 IS_SKB(skb);
 318                 kfree_skb(skb, FREE_WRITE);
 319         }
 320         
 321         /*
 322          *      Don't discard received data until the user side kills its
 323          *      half of the socket.
 324          */
 325 
 326         if (sk->dead) 
 327         {
 328                 while((skb=skb_dequeue(&sk->receive_queue))!=NULL) 
 329                 {
 330                 /*
 331                  * This will take care of closing sockets that were
 332                  * listening and didn't accept everything.
 333                  */
 334                         if (skb->sk != NULL && skb->sk != sk) 
 335                         {
 336                                 IS_SKB(skb);
 337                                 skb->sk->dead = 1;
 338                                 skb->sk->prot->close(skb->sk, 0);
 339                         }
 340                         IS_SKB(skb);
 341                         kfree_skb(skb, FREE_READ);
 342                 }
 343         }       
 344 
 345         /* Now we need to clean up the send head. */
 346         cli();
 347         for(skb = sk->send_head; skb != NULL; )
 348         {
 349                 struct sk_buff *skb2;
 350 
 351                 /*
 352                  * We need to remove skb from the transmit queue,
 353                  * or maybe the arp queue.
 354                  */
 355                 if (skb->next  && skb->prev) {
 356 /*                      printk("destroy_sock: unlinked skb\n");*/
 357                         IS_SKB(skb);
 358                         skb_unlink(skb);
 359                 }
 360                 skb->dev = NULL;
 361                 skb2 = skb->link3;
 362                 kfree_skb(skb, FREE_WRITE);
 363                 skb = skb2;
 364         }
 365         sk->send_head = NULL;
 366         sti();
 367 
 368         /* And now the backlog. */
 369         while((skb=skb_dequeue(&sk->back_log))!=NULL) 
 370         {
 371                 /* this should never happen. */
 372 /*              printk("cleaning back_log\n");*/
 373                 kfree_skb(skb, FREE_READ);
 374         }
 375 
 376         /* Now if it has a half accepted/ closed socket. */
 377         if (sk->pair) 
 378         {
 379                 sk->pair->dead = 1;
 380                 sk->pair->prot->close(sk->pair, 0);
 381                 sk->pair = NULL;
 382         }
 383 
 384         /*
 385          * Now if everything is gone we can free the socket
 386          * structure, otherwise we need to keep it around until
 387          * everything is gone.
 388          */
 389 
 390           if (sk->dead && sk->rmem_alloc == 0 && sk->wmem_alloc == 0) 
 391           {
 392                 kfree_s((void *)sk,sizeof(*sk));
 393           } 
 394           else 
 395           {
 396                 /* this should never happen. */
 397                 /* actually it can if an ack has just been sent. */
 398                 sk->destroy = 1;
 399                 sk->ack_backlog = 0;
 400                 sk->inuse = 0;
 401                 reset_timer(sk, TIME_DESTROY, SOCK_DESTROY_TIME);
 402         }
 403 }
 404 
 405 /*
 406  *      The routines beyond this point handle the behaviour of an AF_INET
 407  *      socket object. Mostly it punts to the subprotocols of IP to do
 408  *      the work.
 409  */
 410  
 411 static int inet_fcntl(struct socket *sock, unsigned int cmd, unsigned long arg)
     /*  */
 412 {
 413         struct sock *sk;
 414 
 415         sk = (struct sock *) sock->data;
 416 
 417         switch(cmd) 
 418         {
 419                 case F_SETOWN:
 420                         /*
 421                          * This is a little restrictive, but it's the only
 422                          * way to make sure that you can't send a sigurg to
 423                          * another process.
 424                          */
 425                         if (!suser() && current->pgrp != -arg &&
 426                                 current->pid != arg) return(-EPERM);
 427                         sk->proc = arg;
 428                         return(0);
 429                 case F_GETOWN:
 430                         return(sk->proc);
 431                 default:
 432                         return(-EINVAL);
 433         }
 434 }
 435 
 436 /*
 437  *      Set socket options on an inet socket.
 438  */
 439  
 440 static int inet_setsockopt(struct socket *sock, int level, int optname,
     /*  */
 441                     char *optval, int optlen)
 442 {
 443         struct sock *sk = (struct sock *) sock->data;  
 444         if (level == SOL_SOCKET)
 445                 return sock_setsockopt(sk,level,optname,optval,optlen);
 446         if (sk->prot->setsockopt==NULL)
 447                 return(-EOPNOTSUPP);
 448         else
 449                 return sk->prot->setsockopt(sk,level,optname,optval,optlen);
 450 }
 451 
 452 /*
 453  *      Get a socket option on an AF_INET socket.
 454  */
 455 
 456 static int inet_getsockopt(struct socket *sock, int level, int optname,
     /*  */
 457                     char *optval, int *optlen)
 458 {
 459         struct sock *sk = (struct sock *) sock->data;   
 460         if (level == SOL_SOCKET) 
 461                 return sock_getsockopt(sk,level,optname,optval,optlen);
 462         if(sk->prot->getsockopt==NULL)          
 463                 return(-EOPNOTSUPP);
 464         else
 465                 return sk->prot->getsockopt(sk,level,optname,optval,optlen);
 466 }
 467 
 468 /*
 469  *      Automatically bind an unbound socket.
 470  */
 471 
 472 static int inet_autobind(struct sock *sk)
     /*  */
 473 {
 474         /* We may need to bind the socket. */
 475         if (sk->num == 0) 
 476         {
 477                 sk->num = get_new_socknum(sk->prot, 0);
 478                 if (sk->num == 0) 
 479                         return(-EAGAIN);
 480                 udp_cache_zap();
 481                 tcp_cache_zap();
 482                 put_sock(sk->num, sk);
 483                 sk->dummy_th.source = ntohs(sk->num);
 484         }
 485         return 0;
 486 }
 487 
 488 /*
 489  *      Move a socket into listening state.
 490  */
 491  
 492 static int inet_listen(struct socket *sock, int backlog)
     /*  */
 493 {
 494         struct sock *sk = (struct sock *) sock->data;
 495 
 496         if(inet_autobind(sk)!=0)
 497                 return -EAGAIN;
 498 
 499         /* We might as well re use these. */ 
 500         /*
 501          * note that the backlog is "unsigned char", so truncate it
 502          * somewhere. We might as well truncate it to what everybody
 503          * else does..
 504          * Now truncate to 128 not 5. 
 505          */
 506         if ((unsigned) backlog == 0)    /* BSDism */
 507                 backlog = 1;
 508         if ((unsigned) backlog > SOMAXCONN)
 509                 backlog = SOMAXCONN;
 510         sk->max_ack_backlog = backlog;
 511         if (sk->state != TCP_LISTEN)
 512         {
 513                 sk->ack_backlog = 0;
 514                 sk->state = TCP_LISTEN;
 515         }
 516         return(0);
 517 }
 518 
 519 /*
 520  *      Default callbacks for user INET sockets. These just wake up
 521  *      the user owning the socket.
 522  */
 523 
 524 static void def_callback1(struct sock *sk)
     /*  */
 525 {
 526         if(!sk->dead)
 527                 wake_up_interruptible(sk->sleep);
 528 }
 529 
 530 static void def_callback2(struct sock *sk,int len)
     /*  */
 531 {
 532         if(!sk->dead)
 533         {
 534                 wake_up_interruptible(sk->sleep);
 535                 sock_wake_async(sk->socket, 1);
 536         }
 537 }
 538 
 539 static void def_callback3(struct sock *sk)
     /*  */
 540 {
 541         if(!sk->dead)
 542         {
 543                 wake_up_interruptible(sk->sleep);
 544                 sock_wake_async(sk->socket, 2);
 545         }
 546 }
 547 
 548 /*
 549  *      Create an inet socket.
 550  *
 551  *      FIXME: Gcc would generate much better code if we set the parameters
 552  *      up in in-memory structure order. Gcc68K even more so
 553  */
 554 
 555 static int inet_create(struct socket *sock, int protocol)
     /*  */
 556 {
 557         struct sock *sk;
 558         struct proto *prot;
 559         int err;
 560 
 561         sk = (struct sock *) kmalloc(sizeof(*sk), GFP_KERNEL);
 562         if (sk == NULL) 
 563                 return(-ENOBUFS);
 564         memset(sk,0,sizeof(*sk));       /* Efficient way to set most fields to zero */
 565 /*      sk->num = 0;
 566  *      sk->reuse = 0;*/
 567         switch(sock->type) 
 568         {
 569                 case SOCK_STREAM:
 570                 case SOCK_SEQPACKET:
 571                         if (protocol && protocol != IPPROTO_TCP) 
 572                         {
 573                                 kfree_s((void *)sk, sizeof(*sk));
 574                                 return(-EPROTONOSUPPORT);
 575                         }
 576                         protocol = IPPROTO_TCP;
 577                         sk->no_check = TCP_NO_CHECK;
 578                         prot = &tcp_prot;
 579                         break;
 580 
 581                 case SOCK_DGRAM:
 582                         if (protocol && protocol != IPPROTO_UDP) 
 583                         {
 584                                 kfree_s((void *)sk, sizeof(*sk));
 585                                 return(-EPROTONOSUPPORT);
 586                         }
 587                         protocol = IPPROTO_UDP;
 588                         sk->no_check = UDP_NO_CHECK;
 589                         prot=&udp_prot;
 590                         break;
 591       
 592                 case SOCK_RAW:
 593                         if (!suser()) 
 594                         {
 595                                 kfree_s((void *)sk, sizeof(*sk));
 596                                 return(-EPERM);
 597                         }
 598                         if (!protocol) 
 599                         {
 600                                 kfree_s((void *)sk, sizeof(*sk));
 601                                 return(-EPROTONOSUPPORT);
 602                         }
 603                         prot = &raw_prot;
 604                         sk->reuse = 1;
 605                         sk->num = protocol;
 606                         break;
 607 
 608                 case SOCK_PACKET:
 609                         if (!suser()) 
 610                         {
 611                                 kfree_s((void *)sk, sizeof(*sk));
 612                                 return(-EPERM);
 613                         }
 614                         if (!protocol) 
 615                         {
 616                                 kfree_s((void *)sk, sizeof(*sk));
 617                                 return(-EPROTONOSUPPORT);
 618                         }
 619                         prot = &packet_prot;
 620                         sk->reuse = 1;
 621                         sk->num = protocol;
 622                         break;
 623 
 624                 default:
 625                         kfree_s((void *)sk, sizeof(*sk));
 626                         return(-ESOCKTNOSUPPORT);
 627         }
 628         sk->socket = sock;
 629 #ifdef CONFIG_TCP_NAGLE_OFF
 630         sk->nonagle = 1;
 631 #else    
 632 /*      sk->nonagle = 0;*/
 633 #endif  
 634         sk->type = sock->type;
 635         sk->protocol = protocol;
 636         sk->sndbuf = SK_WMEM_MAX;
 637         sk->rcvbuf = SK_RMEM_MAX;
 638         sk->rto = TCP_TIMEOUT_INIT;             /*TCP_WRITE_TIME*/
 639         sk->cong_window = 1; /* start with only sending one packet at a time. */
 640         sk->priority = 1;
 641         sk->state = TCP_CLOSE;
 642 #ifdef WHAT_WE_DO_THE_MEMZERO_INSTEAD_OF        
 643         sk->stamp.tv_sec=0;
 644         sk->wmem_alloc = 0;
 645         sk->rmem_alloc = 0;
 646         sk->pair = NULL;
 647         sk->opt = NULL;
 648         sk->write_seq = 0;
 649         sk->acked_seq = 0;
 650         sk->copied_seq = 0;
 651         sk->fin_seq = 0;
 652         sk->urg_seq = 0;
 653         sk->urg_data = 0;
 654         sk->proc = 0;
 655         sk->rtt = 0;                            /*TCP_WRITE_TIME << 3;*/
 656         sk->mdev = 0;
 657         sk->backoff = 0;
 658         sk->packets_out = 0;
 659         sk->cong_count = 0;
 660         sk->ssthresh = 0;
 661         sk->max_window = 0;
 662         sk->urginline = 0;
 663         sk->intr = 0;
 664         sk->linger = 0;
 665         sk->destroy = 0;
 666         sk->shutdown = 0;
 667         sk->keepopen = 0;
 668         sk->zapped = 0;
 669         sk->done = 0;
 670         sk->ack_backlog = 0;
 671         sk->window = 0;
 672         sk->bytes_rcv = 0;
 673         sk->dead = 0;
 674         sk->ack_timed = 0;
 675         sk->partial = NULL;
 676         sk->user_mss = 0;
 677         sk->debug = 0;
 678         /* how many packets we should send before forcing an ack. 
 679            if this is set to zero it is the same as sk->delay_acks = 0 */
 680         sk->max_ack_backlog = 0;
 681         sk->inuse = 0;
 682         sk->delay_acks = 0;
 683         sk->daddr = 0;
 684         sk->saddr = 0 /* ip_my_addr() */;
 685         sk->err = 0;
 686         sk->next = NULL;
 687         sk->pair = NULL;
 688         sk->send_tail = NULL;
 689         sk->send_head = NULL;
 690         sk->timeout = 0;
 691         sk->broadcast = 0;
 692         sk->localroute = 0;
 693         sk->blog = 0;
 694         sk->dummy_th.res1=0;
 695         sk->dummy_th.res2=0;
 696         sk->dummy_th.urg_ptr = 0;
 697         sk->dummy_th.fin = 0;
 698         sk->dummy_th.syn = 0;
 699         sk->dummy_th.rst = 0;
 700         sk->dummy_th.psh = 0;
 701         sk->dummy_th.ack = 0;
 702         sk->dummy_th.urg = 0;
 703         sk->dummy_th.dest = 0;
 704         sk->ip_tos=0;
 705         sk->ip_route_cache=NULL;
 706         sk->ip_hcache_ver= 0;
 707         sk->ip_option_len=0;
 708         sk->ip_option_flen=0;
 709         sk->ip_opt_next_hop=0;
 710         sk->ip_opt_ptr[0]=NULL;
 711         sk->ip_opt_ptr[1]=NULL;
 712 #endif  
 713 
 714         /* this is how many unacked bytes we will accept for this socket.  */
 715         sk->max_unacked = 2048; /* needs to be at most 2 full packets. */
 716 
 717         skb_queue_head_init(&sk->write_queue);
 718         skb_queue_head_init(&sk->receive_queue);
 719         sk->mtu = 576;
 720         sk->prot = prot;
 721         sk->sleep = sock->wait;
 722         init_timer(&sk->timer);
 723         init_timer(&sk->retransmit_timer);
 724         sk->timer.data = (unsigned long)sk;
 725         sk->timer.function = &net_timer;
 726         skb_queue_head_init(&sk->back_log);
 727         sock->data =(void *) sk;
 728         sk->dummy_th.doff = sizeof(sk->dummy_th)/4;
 729         sk->ip_ttl=64;
 730         if(sk->type==SOCK_RAW && protocol==IPPROTO_RAW)
 731                 sk->ip_hdrincl=1;
 732         else
 733                 sk->ip_hdrincl=0;
 734 #ifdef CONFIG_IP_MULTICAST
 735         sk->ip_mc_loop=1;
 736         sk->ip_mc_ttl=1;
 737         *sk->ip_mc_name=0;
 738         sk->ip_mc_list=NULL;
 739 #endif
 740         
 741         sk->state_change = def_callback1;
 742         sk->data_ready = def_callback2;
 743         sk->write_space = def_callback3;
 744         sk->error_report = def_callback1;
 745 
 746         if (sk->num) 
 747         {
 748         /*
 749          * It assumes that any protocol which allows
 750          * the user to assign a number at socket
 751          * creation time automatically
 752          * shares.
 753          */
 754                 put_sock(sk->num, sk);
 755                 sk->dummy_th.source = ntohs(sk->num);
 756         }
 757 
 758         if (sk->prot->init) 
 759         {
 760                 err = sk->prot->init(sk);
 761                 if (err != 0) 
 762                 {
 763                         destroy_sock(sk);
 764                         return(err);
 765                 }
 766         }
 767         return(0);
 768 }
 769 
 770 
 771 /*
 772  *      Duplicate a socket.
 773  */
 774  
 775 static int inet_dup(struct socket *newsock, struct socket *oldsock)
     /*  */
 776 {
 777         return(inet_create(newsock,((struct sock *)(oldsock->data))->protocol));
 778 }
 779 
 780 /*
 781  * Return 1 if we still have things to send in our buffers.
 782  */
 783 static inline int closing(struct sock * sk)
     /*  */
 784 {
 785         switch (sk->state) {
 786                 case TCP_FIN_WAIT1:
 787                 case TCP_CLOSING:
 788                 case TCP_LAST_ACK:
 789                         return 1;
 790         }
 791         return 0;
 792 }
 793 
 794 
 795 /*
 796  *      The peer socket should always be NULL (or else). When we call this
 797  *      function we are destroying the object and from then on nobody
 798  *      should refer to it.
 799  */
 800  
 801 static int inet_release(struct socket *sock, struct socket *peer)
     /*  */
 802 {
 803         struct sock *sk = (struct sock *) sock->data;
 804         if (sk == NULL) 
 805                 return(0);
 806 
 807         sk->state_change(sk);
 808 
 809         /* Start closing the connection.  This may take a while. */
 810 
 811 #ifdef CONFIG_IP_MULTICAST
 812         /* Applications forget to leave groups before exiting */
 813         ip_mc_drop_socket(sk);
 814 #endif
 815         /*
 816          * If linger is set, we don't return until the close
 817          * is complete.  Otherwise we return immediately. The
 818          * actually closing is done the same either way.
 819          *
 820          * If the close is due to the process exiting, we never
 821          * linger..
 822          */
 823 
 824         if (sk->linger == 0 || (current->flags & PF_EXITING))
 825         {
 826                 sk->prot->close(sk,0);
 827                 sk->dead = 1;
 828         } 
 829         else 
 830         {
 831                 sk->prot->close(sk, 0);
 832                 cli();
 833                 if (sk->lingertime)
 834                         current->timeout = jiffies + HZ*sk->lingertime;
 835                 while(closing(sk) && current->timeout>0) 
 836                 {
 837                         interruptible_sleep_on(sk->sleep);
 838                         if (current->signal & ~current->blocked) 
 839                         {
 840                                 break;
 841 #if 0
 842                                 /* not working now - closes can't be restarted */
 843                                 sti();
 844                                 current->timeout=0;
 845                                 return(-ERESTARTSYS);
 846 #endif
 847                         }
 848                 }
 849                 current->timeout=0;
 850                 sti();
 851                 sk->dead = 1;
 852         }
 853         sk->inuse = 1;
 854 
 855         /* This will destroy it. */
 856         sock->data = NULL;
 857         release_sock(sk);
 858         sk->socket = NULL;
 859         return(0);
 860 }
 861 
 862 
 863 static int inet_bind(struct socket *sock, struct sockaddr *uaddr,
     /*  */
 864                int addr_len)
 865 {
 866         struct sockaddr_in *addr=(struct sockaddr_in *)uaddr;
 867         struct sock *sk=(struct sock *)sock->data, *sk2;
 868         unsigned short snum = 0 /* Stoopid compiler.. this IS ok */;
 869         int chk_addr_ret;
 870 
 871         /* check this error. */
 872         if (sk->state != TCP_CLOSE)
 873                 return(-EIO);
 874         if(addr_len<sizeof(struct sockaddr_in))
 875                 return -EINVAL;
 876                 
 877         if(sock->type != SOCK_RAW)
 878         {
 879                 if (sk->num != 0) 
 880                         return(-EINVAL);
 881 
 882                 snum = ntohs(addr->sin_port);
 883                 
 884 #ifdef CONFIG_IP_MASQUERADE
 885                 /*
 886                  *      The kernel masquerader needs some ports
 887                  */             
 888                 if(snum>=PORT_MASQ_BEGIN && snum<=PORT_MASQ_END)
 889                         return -EADDRINUSE;
 890 #endif           
 891 
 892                 if (snum == 0) 
 893                         snum = get_new_socknum(sk->prot, 0);
 894                 if (snum < PROT_SOCK && !suser()) 
 895                         return(-EACCES);
 896         }
 897         
 898         chk_addr_ret = ip_chk_addr(addr->sin_addr.s_addr);
 899         if (addr->sin_addr.s_addr != 0 && chk_addr_ret != IS_MYADDR && chk_addr_ret != IS_MULTICAST)
 900                 return(-EADDRNOTAVAIL); /* Source address MUST be ours! */
 901                 
 902         if (chk_addr_ret || addr->sin_addr.s_addr == 0)
 903                 sk->saddr = addr->sin_addr.s_addr;
 904         
 905         if(sock->type != SOCK_RAW)
 906         {
 907                 /* Make sure we are allowed to bind here. */
 908                 cli();
 909                 for(sk2 = sk->prot->sock_array[snum & (SOCK_ARRAY_SIZE -1)];
 910                                         sk2 != NULL; sk2 = sk2->next) 
 911                 {
 912                 /* should be below! */
 913                         if (sk2->num != snum) 
 914                                 continue;
 915                         if (!sk->reuse)
 916                         {
 917                                 sti();
 918                                 return(-EADDRINUSE);
 919                         }
 920                         
 921                         if (sk2->num != snum) 
 922                                 continue;               /* more than one */
 923                         if (sk2->saddr != sk->saddr) 
 924                                 continue;       /* socket per slot ! -FB */
 925                         if (!sk2->reuse || sk2->state==TCP_LISTEN) 
 926                         {
 927                                 sti();
 928                                 return(-EADDRINUSE);
 929                         }
 930                 }
 931                 sti();
 932 
 933                 remove_sock(sk);
 934                 if(sock->type==SOCK_DGRAM)
 935                         udp_cache_zap();
 936                 if(sock->type==SOCK_STREAM)
 937                         tcp_cache_zap();
 938                 put_sock(snum, sk);
 939                 sk->dummy_th.source = ntohs(sk->num);
 940                 sk->daddr = 0;
 941                 sk->dummy_th.dest = 0;
 942         }
 943         sk->ip_route_cache=NULL;
 944         return(0);
 945 }
 946 
 947 /*
 948  *      Handle sk->err properly. The cli/sti matter.
 949  */
 950  
 951 static int inet_error(struct sock *sk)
     /*  */
 952 {
 953         unsigned long flags;
 954         int err;
 955         save_flags(flags);
 956         cli();  
 957         err=sk->err;
 958         sk->err=0;
 959         restore_flags(flags);
 960         return -err;
 961 }
 962 
 963 /*
 964  *      Connect to a remote host. There is regrettably still a little
 965  *      TCP 'magic' in here.
 966  */
 967  
 968 static int inet_connect(struct socket *sock, struct sockaddr * uaddr,
     /*  */
 969                   int addr_len, int flags)
 970 {
 971         struct sock *sk=(struct sock *)sock->data;
 972         int err;
 973         sock->conn = NULL;
 974 
 975         if (sock->state == SS_CONNECTING && tcp_connected(sk->state))
 976         {
 977                 sock->state = SS_CONNECTED;
 978                 /* Connection completing after a connect/EINPROGRESS/select/connect */
 979                 return 0;       /* Rock and roll */
 980         }
 981 
 982         if (sock->state == SS_CONNECTING && sk->protocol == IPPROTO_TCP && (flags & O_NONBLOCK))
 983         {
 984                 if(sk->err!=0)
 985                 {
 986                         cli();
 987                         err=sk->err;
 988                         sk->err=0;
 989                         sti();
 990                         return -err;
 991                 }
 992                 return -EALREADY;       /* Connecting is currently in progress */
 993         }
 994         if (sock->state != SS_CONNECTING) 
 995         {
 996                 /* We may need to bind the socket. */
 997                 if(inet_autobind(sk)!=0)
 998                         return(-EAGAIN);
 999                 if (sk->prot->connect == NULL) 
1000                         return(-EOPNOTSUPP);
1001                 err = sk->prot->connect(sk, (struct sockaddr_in *)uaddr, addr_len);
1002                 if (err < 0) 
1003                         return(err);
1004                 sock->state = SS_CONNECTING;
1005         }
1006         
1007         if (sk->state > TCP_FIN_WAIT2 && sock->state==SS_CONNECTING)
1008         {
1009                 sock->state=SS_UNCONNECTED;
1010                 cli();
1011                 err=sk->err;
1012                 sk->err=0;
1013                 sti();
1014                 return -err;
1015         }
1016 
1017         if (sk->state != TCP_ESTABLISHED &&(flags & O_NONBLOCK)) 
1018                 return(-EINPROGRESS);
1019 
1020         cli(); /* avoid the race condition */
1021         while(sk->state == TCP_SYN_SENT || sk->state == TCP_SYN_RECV) 
1022         {
1023                 interruptible_sleep_on(sk->sleep);
1024                 if (current->signal & ~current->blocked) 
1025                 {
1026                         sti();
1027                         return(-ERESTARTSYS);
1028                 }
1029                 /* This fixes a nasty in the tcp/ip code. There is a hideous hassle with
1030                    icmp error packets wanting to close a tcp or udp socket. */
1031                 if(sk->err && sk->protocol == IPPROTO_TCP)
1032                 {
1033                         sock->state = SS_UNCONNECTED;
1034                         err = -sk->err;
1035                         sk->err=0;
1036                         sti();
1037                         return err; /* set by tcp_err() */
1038                 }
1039         }
1040         sti();
1041         sock->state = SS_CONNECTED;
1042 
1043         if (sk->state != TCP_ESTABLISHED && sk->err) 
1044         {
1045                 sock->state = SS_UNCONNECTED;
1046                 cli();
1047                 err=sk->err;
1048                 sk->err=0;
1049                 sti();
1050                 return(-err);
1051         }
1052         return(0);
1053 }
1054 
1055 
1056 static int inet_socketpair(struct socket *sock1, struct socket *sock2)
     /*  */
1057 {
1058          return(-EOPNOTSUPP);
1059 }
1060 
1061 
1062 /*
1063  *      Accept a pending connection. The TCP layer now gives BSD semantics.
1064  */
1065 
1066 static int inet_accept(struct socket *sock, struct socket *newsock, int flags)
     /*  */
1067 {
1068         struct sock *sk1, *sk2;
1069         int err;
1070 
1071         sk1 = (struct sock *) sock->data;
1072 
1073         /*
1074          * We've been passed an extra socket.
1075          * We need to free it up because the tcp module creates
1076          * its own when it accepts one.
1077          */
1078         if (newsock->data)
1079         {
1080                 struct sock *sk=(struct sock *)newsock->data;
1081                 newsock->data=NULL;
1082                 sk->dead = 1;
1083                 destroy_sock(sk);
1084         }
1085   
1086         if (sk1->prot->accept == NULL) 
1087                 return(-EOPNOTSUPP);
1088 
1089         /* Restore the state if we have been interrupted, and then returned. */
1090         if (sk1->pair != NULL ) 
1091         {
1092                 sk2 = sk1->pair;
1093                 sk1->pair = NULL;
1094         } 
1095         else
1096         {
1097                 sk2 = sk1->prot->accept(sk1,flags);
1098                 if (sk2 == NULL) 
1099                 {
1100                         err=sk1->err;
1101                         sk1->err=0;
1102                         return(-err);
1103                 }
1104         }
1105         newsock->data = (void *)sk2;
1106         sk2->sleep = newsock->wait;
1107         sk2->socket = newsock;
1108         newsock->conn = NULL;
1109         if (flags & O_NONBLOCK) 
1110                 return(0);
1111 
1112         cli(); /* avoid the race. */
1113         while(sk2->state == TCP_SYN_RECV) 
1114         {
1115                 interruptible_sleep_on(sk2->sleep);
1116                 if (current->signal & ~current->blocked) 
1117                 {
1118                         sti();
1119                         sk1->pair = sk2;
1120                         sk2->sleep = NULL;
1121                         sk2->socket=NULL;
1122                         newsock->data = NULL;
1123                         return(-ERESTARTSYS);
1124                 }
1125         }
1126         sti();
1127 
1128         if (sk2->state != TCP_ESTABLISHED && sk2->err > 0) 
1129         {
1130                 err = -sk2->err;
1131                 sk2->err=0;
1132                 sk2->dead=1;    /* ANK */
1133                 destroy_sock(sk2);
1134                 newsock->data = NULL;
1135                 return(err);
1136         }
1137         newsock->state = SS_CONNECTED;
1138         return(0);
1139 }
1140 
1141 
1142 /*
1143  *      This does both peername and sockname.
1144  */
1145  
1146 static int inet_getname(struct socket *sock, struct sockaddr *uaddr,
     /*  */
1147                  int *uaddr_len, int peer)
1148 {
1149         struct sockaddr_in *sin=(struct sockaddr_in *)uaddr;
1150         struct sock *sk;
1151   
1152         sin->sin_family = AF_INET;
1153         sk = (struct sock *) sock->data;
1154         if (peer) 
1155         {
1156                 if (!tcp_connected(sk->state)) 
1157                         return(-ENOTCONN);
1158                 sin->sin_port = sk->dummy_th.dest;
1159                 sin->sin_addr.s_addr = sk->daddr;
1160         } 
1161         else 
1162         {
1163                 sin->sin_port = sk->dummy_th.source;
1164                 if (sk->saddr == 0) 
1165                         sin->sin_addr.s_addr = ip_my_addr();
1166                 else 
1167                         sin->sin_addr.s_addr = sk->saddr;
1168         }
1169         *uaddr_len = sizeof(*sin);
1170         return(0);
1171 }
1172 
1173 
1174 /*
1175  *      The assorted BSD I/O operations
1176  */
1177 
1178 static int inet_recvfrom(struct socket *sock, void *ubuf, int size, int noblock, 
     /*  */
1179                    unsigned flags, struct sockaddr *sin, int *addr_len )
1180 {
1181         struct sock *sk = (struct sock *) sock->data;
1182         
1183         if (sk->prot->recvfrom == NULL) 
1184                 return(-EOPNOTSUPP);
1185         if(sk->err)
1186                 return inet_error(sk);
1187         /* We may need to bind the socket. */
1188         if(inet_autobind(sk)!=0)
1189                 return(-EAGAIN);
1190         return(sk->prot->recvfrom(sk, (unsigned char *) ubuf, size, noblock, flags,
1191                              (struct sockaddr_in*)sin, addr_len));
1192 }
1193 
1194 
1195 static int inet_recv(struct socket *sock, void *ubuf, int size, int noblock,
     /*  */
1196           unsigned flags)
1197 {
1198         /* BSD explicitly states these are the same - so we do it this way to be sure */
1199         return inet_recvfrom(sock,ubuf,size,noblock,flags,NULL,NULL);
1200 }
1201 
1202 static int inet_read(struct socket *sock, char *ubuf, int size, int noblock)
     /*  */
1203 {
1204         struct sock *sk = (struct sock *) sock->data;
1205         
1206         if(sk->err)
1207                 return inet_error(sk);
1208         /* We may need to bind the socket. */
1209         if(inet_autobind(sk))
1210                 return(-EAGAIN);        
1211         return(sk->prot->read(sk, (unsigned char *) ubuf, size, noblock, 0));
1212 }
1213 
1214 static int inet_send(struct socket *sock, const void *ubuf, int size, int noblock, 
     /*  */
1215                unsigned flags)
1216 {
1217         struct sock *sk = (struct sock *) sock->data;
1218         if (sk->shutdown & SEND_SHUTDOWN) 
1219         {
1220                 send_sig(SIGPIPE, current, 1);
1221                 return(-EPIPE);
1222         }
1223         if(sk->err)
1224                 return inet_error(sk);
1225         /* We may need to bind the socket. */
1226         if(inet_autobind(sk)!=0)
1227                 return(-EAGAIN);
1228         return(sk->prot->write(sk, (const unsigned char *) ubuf, size, noblock, flags));
1229 }
1230 
1231 static int inet_write(struct socket *sock, const char *ubuf, int size, int noblock)
     /*  */
1232 {
1233         return inet_send(sock,ubuf,size,noblock,0);
1234 }
1235 
1236 static int inet_sendto(struct socket *sock, const void *ubuf, int size, int noblock, 
     /*  */
1237             unsigned flags, struct sockaddr *sin, int addr_len)
1238 {
1239         struct sock *sk = (struct sock *) sock->data;
1240         if (sk->shutdown & SEND_SHUTDOWN) 
1241         {
1242                 send_sig(SIGPIPE, current, 1);
1243                 return(-EPIPE);
1244         }
1245         if (sk->prot->sendto == NULL) 
1246                 return(-EOPNOTSUPP);
1247         if(sk->err)
1248                 return inet_error(sk);
1249         /* We may need to bind the socket. */
1250         if(inet_autobind(sk)!=0)
1251                 return -EAGAIN;
1252         return(sk->prot->sendto(sk, (const unsigned char *) ubuf, size, noblock, flags, 
1253                            (struct sockaddr_in *)sin, addr_len));
1254 }
1255 
1256 
1257 static int inet_shutdown(struct socket *sock, int how)
     /*  */
1258 {
1259         struct sock *sk=(struct sock*)sock->data;
1260 
1261         /*
1262          * This should really check to make sure
1263          * the socket is a TCP socket. (WHY AC...)
1264          */
1265         how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
1266                        1->2 bit 2 snds.
1267                        2->3 */
1268         if ((how & ~SHUTDOWN_MASK) || how==0)   /* MAXINT->0 */
1269                 return(-EINVAL);
1270         if (sock->state == SS_CONNECTING && sk->state == TCP_ESTABLISHED)
1271                 sock->state = SS_CONNECTED;
1272         if (!tcp_connected(sk->state)) 
1273                 return(-ENOTCONN);
1274         sk->shutdown |= how;
1275         if (sk->prot->shutdown)
1276                 sk->prot->shutdown(sk, how);
1277         return(0);
1278 }
1279 
1280 
1281 static int inet_select(struct socket *sock, int sel_type, select_table *wait )
     /*  */
1282 {
1283         struct sock *sk=(struct sock *) sock->data;
1284         if (sk->prot->select == NULL) 
1285         {
1286                 return(0);
1287         }
1288         return(sk->prot->select(sk, sel_type, wait));
1289 }
1290 
1291 /*
1292  *      ioctl() calls you can issue on an INET socket. Most of these are
1293  *      device configuration and stuff and very rarely used. Some ioctls
1294  *      pass on to the socket itself.
1295  *
1296  *      NOTE: I like the idea of a module for the config stuff. ie ifconfig
1297  *      loads the devconfigure module does its configuring and unloads it.
1298  *      There's a good 20K of config code hanging around the kernel.
1299  */
1300 
1301 static int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
     /*  */
1302 {
1303         struct sock *sk=(struct sock *)sock->data;
1304         int err;
1305         int pid;
1306 
1307         switch(cmd) 
1308         {
1309                 case FIOSETOWN:
1310                 case SIOCSPGRP:
1311                         err=verify_area(VERIFY_READ,(int *)arg,sizeof(long));
1312                         if(err)
1313                                 return err;
1314                         pid = get_user((int *) arg);
1315                         /* see inet_fcntl */
1316                         if (current->pid != pid && current->pgrp != -pid && !suser())
1317                                 return -EPERM;
1318                         sk->proc = pid;
1319                         return(0);
1320                 case FIOGETOWN:
1321                 case SIOCGPGRP:
1322                         err=verify_area(VERIFY_WRITE,(void *) arg, sizeof(long));
1323                         if(err)
1324                                 return err;
1325                         put_fs_long(sk->proc,(int *)arg);
1326                         return(0);                      
1327                 case SIOCGSTAMP:
1328                         if(sk->stamp.tv_sec==0)
1329                                 return -ENOENT;
1330                         err=verify_area(VERIFY_WRITE,(void *)arg,sizeof(struct timeval));
1331                         if(err)
1332                                 return err;
1333                         memcpy_tofs((void *)arg,&sk->stamp,sizeof(struct timeval));
1334                         return 0;
1335                 case SIOCADDRT:
1336                 case SIOCDELRT:
1337                         return(ip_rt_ioctl(cmd,(void *) arg));
1338                 case SIOCDARP:
1339                 case SIOCGARP:
1340                 case SIOCSARP:
1341                         return(arp_ioctl(cmd,(void *) arg));
1342                 case SIOCDRARP:
1343                 case SIOCGRARP:
1344                 case SIOCSRARP:
1345                         if (rarp_ioctl_hook != NULL)
1346                                 return(rarp_ioctl_hook(cmd,(void *) arg));
1347                 case SIOCGIFCONF:
1348                 case SIOCGIFFLAGS:
1349                 case SIOCSIFFLAGS:
1350                 case SIOCGIFADDR:
1351                 case SIOCSIFADDR:
1352                 case SIOCADDMULTI:
1353                 case SIOCDELMULTI:
1354                 case SIOCGIFDSTADDR:
1355                 case SIOCSIFDSTADDR:
1356                 case SIOCGIFBRDADDR:
1357                 case SIOCSIFBRDADDR:
1358                 case SIOCGIFNETMASK:
1359                 case SIOCSIFNETMASK:
1360                 case SIOCGIFMETRIC:
1361                 case SIOCSIFMETRIC:
1362                 case SIOCGIFMEM:
1363                 case SIOCSIFMEM:
1364                 case SIOCGIFMTU:
1365                 case SIOCSIFMTU:
1366                 case SIOCSIFLINK:
1367                 case SIOCGIFHWADDR:
1368                 case SIOCSIFHWADDR:
1369                 case OLD_SIOCGIFHWADDR:
1370                 case SIOCSIFMAP:
1371                 case SIOCGIFMAP:
1372                 case SIOCSIFSLAVE:
1373                 case SIOCGIFSLAVE:
1374                         return(dev_ioctl(cmd,(void *) arg));
1375 
1376                 default:
1377                         if ((cmd >= SIOCDEVPRIVATE) &&
1378                            (cmd <= (SIOCDEVPRIVATE + 15)))
1379                                 return(dev_ioctl(cmd,(void *) arg));
1380 
1381                         if (sk->prot->ioctl==NULL) 
1382                                 return(-EINVAL);
1383                         return(sk->prot->ioctl(sk, cmd, arg));
1384         }
1385         /*NOTREACHED*/
1386         return(0);
1387 }
1388 
1389 /*
1390  * This routine must find a socket given a TCP or UDP header.
1391  * Everything is assumed to be in net order.
1392  *
1393  * We give priority to more closely bound ports: if some socket
1394  * is bound to a particular foreign address, it will get the packet
1395  * rather than somebody listening to any address..
1396  */
1397 
1398 struct sock *get_sock(struct proto *prot, unsigned short num,
     /*  */
1399                                 unsigned long raddr,
1400                                 unsigned short rnum, unsigned long laddr)
1401 {
1402         struct sock *s;
1403         struct sock *result = NULL;
1404         int badness = -1;
1405         unsigned short hnum;
1406 
1407         hnum = ntohs(num);
1408 
1409         /*
1410          * SOCK_ARRAY_SIZE must be a power of two.  This will work better
1411          * than a prime unless 3 or more sockets end up using the same
1412          * array entry.  This should not be a problem because most
1413          * well known sockets don't overlap that much, and for
1414          * the other ones, we can just be careful about picking our
1415          * socket number when we choose an arbitrary one.
1416          */
1417 
1418         for(s = prot->sock_array[hnum & (SOCK_ARRAY_SIZE - 1)];
1419                         s != NULL; s = s->next) 
1420         {
1421                 int score = 0;
1422 
1423                 if (s->num != hnum) 
1424                         continue;
1425 
1426                 if(s->dead && (s->state == TCP_CLOSE))
1427                         continue;
1428                 /* local address matches? */
1429                 if (s->saddr) {
1430                         if (s->saddr != laddr)
1431                                 continue;
1432                         score++;
1433                 }
1434                 /* remote address matches? */
1435                 if (s->daddr) {
1436                         if (s->daddr != raddr)
1437                                 continue;
1438                         score++;
1439                 }
1440                 /* remote port matches? */
1441                 if (s->dummy_th.dest) {
1442                         if (s->dummy_th.dest != rnum)
1443                                 continue;
1444                         score++;
1445                 }
1446                 /* perfect match? */
1447                 if (score == 3)
1448                         return s;
1449                 /* no, check if this is the best so far.. */
1450                 if (score <= badness)
1451                         continue;
1452                 result = s;
1453                 badness = score;
1454         }
1455         return result;
1456 }
1457 
1458 /*
1459  *      Deliver a datagram to raw sockets.
1460  */
1461  
1462 struct sock *get_sock_raw(struct sock *sk, 
     /*  */
1463                                 unsigned short num,
1464                                 unsigned long raddr,
1465                                 unsigned long laddr)
1466 {
1467         struct sock *s;
1468 
1469         s=sk;
1470 
1471         for(; s != NULL; s = s->next) 
1472         {
1473                 if (s->num != num) 
1474                         continue;
1475                 if(s->dead && (s->state == TCP_CLOSE))
1476                         continue;
1477                 if(s->daddr && s->daddr!=raddr)
1478                         continue;
1479                 if(s->saddr  && s->saddr!=laddr)
1480                         continue;
1481                 return(s);
1482         }
1483         return(NULL);
1484 }
1485 
1486 #ifdef CONFIG_IP_MULTICAST
1487 /*
1488  *      Deliver a datagram to broadcast/multicast sockets.
1489  */
1490  
1491 struct sock *get_sock_mcast(struct sock *sk, 
     /*  */
1492                                 unsigned short num,
1493                                 unsigned long raddr,
1494                                 unsigned short rnum, unsigned long laddr)
1495 {
1496         struct sock *s;
1497         unsigned short hnum;
1498 
1499         hnum = ntohs(num);
1500 
1501         /*
1502          * SOCK_ARRAY_SIZE must be a power of two.  This will work better
1503          * than a prime unless 3 or more sockets end up using the same
1504          * array entry.  This should not be a problem because most
1505          * well known sockets don't overlap that much, and for
1506          * the other ones, we can just be careful about picking our
1507          * socket number when we choose an arbitrary one.
1508          */
1509         
1510         s=sk;
1511 
1512         for(; s != NULL; s = s->next) 
1513         {
1514                 if (s->num != hnum) 
1515                         continue;
1516                 if(s->dead && (s->state == TCP_CLOSE))
1517                         continue;
1518                 if(s->daddr && s->daddr!=raddr)
1519                         continue;
1520                 if (s->dummy_th.dest != rnum && s->dummy_th.dest != 0) 
1521                         continue;
1522                 if(s->saddr  && s->saddr!=laddr)
1523                         continue;
1524                 return(s);
1525         }
1526         return(NULL);
1527 }
1528 
1529 #endif
1530 
1531 static struct proto_ops inet_proto_ops = {
1532         AF_INET,
1533 
1534         inet_create,
1535         inet_dup,
1536         inet_release,
1537         inet_bind,
1538         inet_connect,
1539         inet_socketpair,
1540         inet_accept,
1541         inet_getname, 
1542         inet_read,
1543         inet_write,
1544         inet_select,
1545         inet_ioctl,
1546         inet_listen,
1547         inet_send,
1548         inet_recv,
1549         inet_sendto,
1550         inet_recvfrom,
1551         inet_shutdown,
1552         inet_setsockopt,
1553         inet_getsockopt,
1554         inet_fcntl,
1555 };
1556 
1557 extern unsigned long seq_offset;
1558 
1559 /*
1560  *      Called by socket.c on kernel startup.  
1561  */
1562  
1563 void inet_proto_init(struct net_proto *pro)
     /*  */
1564 {
1565         struct inet_protocol *p;
1566         int i;
1567 
1568 
1569         printk("Swansea University Computer Society TCP/IP for NET3.030 (Snapshot #1)\n");
1570 
1571         /*
1572          *      Tell SOCKET that we are alive... 
1573          */
1574    
1575         (void) sock_register(inet_proto_ops.family, &inet_proto_ops);
1576 
1577         seq_offset = CURRENT_TIME*250;
1578 
1579         /*
1580          *      Add all the protocols. 
1581          */
1582          
1583         for(i = 0; i < SOCK_ARRAY_SIZE; i++) 
1584         {
1585                 tcp_prot.sock_array[i] = NULL;
1586                 udp_prot.sock_array[i] = NULL;
1587                 raw_prot.sock_array[i] = NULL;
1588         }
1589         tcp_prot.inuse = 0;
1590         tcp_prot.highestinuse = 0;
1591         udp_prot.inuse = 0;
1592         udp_prot.highestinuse = 0;
1593         raw_prot.inuse = 0;
1594         raw_prot.highestinuse = 0;
1595 
1596         printk("IP Protocols: ");
1597         for(p = inet_protocol_base; p != NULL;) 
1598         {
1599                 struct inet_protocol *tmp = (struct inet_protocol *) p->next;
1600                 inet_add_protocol(p);
1601                 printk("%s%s",p->name,tmp?", ":"\n");
1602                 p = tmp;
1603         }
1604 
1605         /*
1606          *      Set the ARP module up
1607          */
1608         arp_init();
1609         /*
1610          *      Set the IP module up
1611          */
1612         ip_init();
1613 
1614 #if defined(CONFIG_IP_ACCT)||defined(CONFIG_IP_FIREWALL)|| \
1615     defined(CONFIG_IP_MASQUERADE)
1616         ip_fw_init();
1617 #endif
1618 
1619 #ifdef CONFIG_INET_RARP
1620         proc_net_register(&(struct proc_dir_entry) {
1621                 PROC_NET_RARP, 4, "rarp",
1622                 S_IFREG | S_IRUGO, 1, 0, 0,
1623                 0, &proc_net_inode_operations,
1624                 rarp_get_info
1625         });
1626         rarp_ioctl_hook = rarp_ioctl;
1627 #endif
1628         proc_net_register(&(struct proc_dir_entry) {
1629                 PROC_NET_RAW, 3, "raw",
1630                 S_IFREG | S_IRUGO, 1, 0, 0,
1631                 0, &proc_net_inode_operations,
1632                 raw_get_info
1633         });
1634         proc_net_register(&(struct proc_dir_entry) {
1635                 PROC_NET_SNMP, 4, "snmp",
1636                 S_IFREG | S_IRUGO, 1, 0, 0,
1637                 0, &proc_net_inode_operations,
1638                 snmp_get_info
1639         });
1640         proc_net_register(&(struct proc_dir_entry) {
1641                 PROC_NET_SOCKSTAT, 8, "sockstat",
1642                 S_IFREG | S_IRUGO, 1, 0, 0,
1643                 0, &proc_net_inode_operations,
1644                 afinet_get_info
1645         });
1646         proc_net_register(&(struct proc_dir_entry) {
1647                 PROC_NET_TCP, 3, "tcp",
1648                 S_IFREG | S_IRUGO, 1, 0, 0,
1649                 0, &proc_net_inode_operations,
1650                 tcp_get_info
1651         });
1652         proc_net_register(&(struct proc_dir_entry) {
1653                 PROC_NET_UDP, 3, "udp",
1654                 S_IFREG | S_IRUGO, 1, 0, 0,
1655                 0, &proc_net_inode_operations,
1656                 udp_get_info
1657         });
1658         proc_net_register(&(struct proc_dir_entry) {
1659                 PROC_NET_ROUTE, 5, "route",
1660                 S_IFREG | S_IRUGO, 1, 0, 0,
1661                 0, &proc_net_inode_operations,
1662                 rt_get_info
1663         });
1664 }
/* */
root/net/ipv4/af_inet.c

DEFINITIONS
