server/sock.c

   1 /*
   2  * Server-side socket management
   3  *
   4  * Copyright (C) 1999 Marcus Meissner, Ove Kåven
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
  19  *
  20  * FIXME: we use read|write access in all cases. Shouldn't we depend that
  21  * on the access of the current handle?
  22  */
  23
  24 #include "config.h"
  25
  26 #include <assert.h>
  27 #include <fcntl.h>
  28 #include <stdarg.h>
  29 #include <stdio.h>
  30 #include <string.h>
  31 #include <stdlib.h>
  32 #include <errno.h>
  33 #ifdef HAVE_NETINET_IN_H
  34 # include <netinet/in.h>
  35 #endif
  36 #ifdef HAVE_POLL_H
  37 # include <poll.h>
  38 #endif
  39 #include <sys/time.h>
  40 #include <sys/types.h>
  41 #ifdef HAVE_SYS_SOCKET_H
  42 # include <sys/socket.h>
  43 #endif
  44 #ifdef HAVE_SYS_IOCTL_H
  45 #include <sys/ioctl.h>
  46 #endif
  47 #ifdef HAVE_SYS_FILIO_H
  48 # include <sys/filio.h>
  49 #endif
  50 #include <time.h>
  51 #include <unistd.h>
  52 #include <limits.h>
  53 #ifdef HAVE_LINUX_RTNETLINK_H
  54 # include <linux/rtnetlink.h>
  55 #endif
  56
  57 #ifdef HAVE_NETIPX_IPX_H
  58 # include <netipx/ipx.h>
  59 #elif defined(HAVE_LINUX_IPX_H)
  60 # ifdef HAVE_ASM_TYPES_H
  61 #  include <asm/types.h>
  62 # endif
  63 # ifdef HAVE_LINUX_TYPES_H
  64 #  include <linux/types.h>
  65 # endif
  66 # include <linux/ipx.h>
  67 #endif
  68 #if defined(SOL_IPX) || defined(SO_DEFAULT_HEADERS)
  69 # define HAS_IPX
  70 #endif
  71
  72 #ifdef HAVE_LINUX_IRDA_H
  73 # ifdef HAVE_LINUX_TYPES_H
  74 #  include <linux/types.h>
  75 # endif
  76 # include <linux/irda.h>
  77 # define HAS_IRDA
  78 #endif
  79
  80 #include "ntstatus.h"
  81 #define WIN32_NO_STATUS
  82 #include "windef.h"
  83 #include "winternl.h"
  84 #include "winerror.h"
  85 #define USE_WS_PREFIX
  86 #include "winsock2.h"
  87 #include "wsipx.h"
  88 #include "wine/afd.h"
  89
  90 #include "process.h"
  91 #include "file.h"
  92 #include "handle.h"
  93 #include "thread.h"
  94 #include "request.h"
  95 #include "user.h"
  96
  97 /* From winsock.h */
  98 #define FD_MAX_EVENTS              10
  99 #define FD_READ_BIT                0
 100 #define FD_WRITE_BIT               1
 101 #define FD_OOB_BIT                 2
 102 #define FD_ACCEPT_BIT              3
 103 #define FD_CONNECT_BIT             4
 104 #define FD_CLOSE_BIT               5
 105
 106 /*
 107  * Define flags to be used with the WSAAsyncSelect() call.
 108  */
 109 #define FD_READ                    0x00000001
 110 #define FD_WRITE                   0x00000002
 111 #define FD_OOB                     0x00000004
 112 #define FD_ACCEPT                  0x00000008
 113 #define FD_CONNECT                 0x00000010
 114 #define FD_CLOSE                   0x00000020
 115
 116 /* internal per-socket flags */
 117 #define FD_WINE_LISTENING          0x10000000
 118 #define FD_WINE_NONBLOCKING        0x20000000
 119 #define FD_WINE_CONNECTED          0x40000000
 120 #define FD_WINE_RAW                0x80000000
 121 #define FD_WINE_INTERNAL           0xFFFF0000
 122
 123 struct sock
 124 {
 125     struct object       obj;         /* object header */
 126     struct fd          *fd;          /* socket file descriptor */
 127     unsigned int        state;       /* status bits */
 128     unsigned int        mask;        /* event mask */
 129     unsigned int        hmask;       /* held (blocked) events */
 130     unsigned int        pmask;       /* pending events */
 131     unsigned int        flags;       /* socket flags */
 132     int                 polling;     /* is socket being polled? */
 133     unsigned short      proto;       /* socket protocol */
 134     unsigned short      type;        /* socket type */
 135     unsigned short      family;      /* socket family */
 136     struct event       *event;       /* event object */
 137     user_handle_t       window;      /* window to send the message to */
 138     unsigned int        message;     /* message to send */
 139     obj_handle_t        wparam;      /* message wparam (socket handle) */
 140     unsigned int        errors[FD_MAX_EVENTS]; /* event errors */
 141     timeout_t           connect_time;/* time the socket was connected */
 142     struct sock        *deferred;    /* socket that waits for a deferred accept */
 143     struct async_queue  read_q;      /* queue for asynchronous reads */
 144     struct async_queue  write_q;     /* queue for asynchronous writes */
 145     struct async_queue  ifchange_q;  /* queue for interface change notifications */
 146     struct object      *ifchange_obj; /* the interface change notification object */
 147     struct list         ifchange_entry; /* entry in ifchange notification list */
 148 };
 149
 150 static void sock_dump( struct object *obj, int verbose );
 151 static int sock_signaled( struct object *obj, struct wait_queue_entry *entry );
 152 static struct fd *sock_get_fd( struct object *obj );
 153 static void sock_destroy( struct object *obj );
 154 static struct object *sock_get_ifchange( struct sock *sock );
 155 static void sock_release_ifchange( struct sock *sock );
 156
 157 static int sock_get_poll_events( struct fd *fd );
 158 static void sock_poll_event( struct fd *fd, int event );
 159 static enum server_fd_type sock_get_fd_type( struct fd *fd );
 160 static int sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async );
 161 static void sock_queue_async( struct fd *fd, struct async *async, int type, int count );
 162 static void sock_reselect_async( struct fd *fd, struct async_queue *queue );
 163
 164 static int sock_get_ntstatus( int err );
 165 static unsigned int sock_get_error( int err );
 166
 167 static const struct object_ops sock_ops =
 168 {
 169     sizeof(struct sock),          /* size */
 170     sock_dump,                    /* dump */
 171     no_get_type,                  /* get_type */
 172     add_queue,                    /* add_queue */
 173     remove_queue,                 /* remove_queue */
 174     sock_signaled,                /* signaled */
 175     no_satisfied,                 /* satisfied */
 176     no_signal,                    /* signal */
 177     sock_get_fd,                  /* get_fd */
 178     default_fd_map_access,        /* map_access */
 179     default_get_sd,               /* get_sd */
 180     default_set_sd,               /* set_sd */
 181     no_get_full_name,             /* get_full_name */
 182     no_lookup_name,               /* lookup_name */
 183     no_link_name,                 /* link_name */
 184     NULL,                         /* unlink_name */
 185     no_open_file,                 /* open_file */
 186     no_kernel_obj_list,           /* get_kernel_obj_list */
 187     fd_close_handle,              /* close_handle */
 188     sock_destroy                  /* destroy */
 189 };
 190
 191 static const struct fd_ops sock_fd_ops =
 192 {
 193     sock_get_poll_events,         /* get_poll_events */
 194     sock_poll_event,              /* poll_event */
 195     sock_get_fd_type,             /* get_fd_type */
 196     no_fd_read,                   /* read */
 197     no_fd_write,                  /* write */
 198     no_fd_flush,                  /* flush */
 199     default_fd_get_file_info,     /* get_file_info */
 200     no_fd_get_volume_info,        /* get_volume_info */
 201     sock_ioctl,                   /* ioctl */
 202     sock_queue_async,             /* queue_async */
 203     sock_reselect_async           /* reselect_async */
 204 };
 205
 206
 207 /* Permutation of 0..FD_MAX_EVENTS - 1 representing the order in which
 208  * we post messages if there are multiple events.  Used to send
 209  * messages.  The problem is if there is both a FD_CONNECT event and,
 210  * say, an FD_READ event available on the same socket, we want to
 211  * notify the app of the connect event first.  Otherwise it may
 212  * discard the read event because it thinks it hasn't connected yet.
 213  */
 214 static const int event_bitorder[FD_MAX_EVENTS] =
 215 {
 216     FD_CONNECT_BIT,
 217     FD_ACCEPT_BIT,
 218     FD_OOB_BIT,
 219     FD_WRITE_BIT,
 220     FD_READ_BIT,
 221     FD_CLOSE_BIT,
 222     6, 7, 8, 9  /* leftovers */
 223 };
 224
 225 /* Flags that make sense only for SOCK_STREAM sockets */
 226 #define STREAM_FLAG_MASK ((unsigned int) (FD_CONNECT | FD_ACCEPT | FD_WINE_LISTENING | FD_WINE_CONNECTED))
 227
 228 typedef enum {
 229     SOCK_SHUTDOWN_ERROR = -1,
 230     SOCK_SHUTDOWN_EOF = 0,
 231     SOCK_SHUTDOWN_POLLHUP = 1
 232 } sock_shutdown_t;
 233
 234 static sock_shutdown_t sock_shutdown_type = SOCK_SHUTDOWN_ERROR;
 235
 236 static sock_shutdown_t sock_check_pollhup(void)
 237 {
 238     sock_shutdown_t ret = SOCK_SHUTDOWN_ERROR;
 239     int fd[2], n;
 240     struct pollfd pfd;
 241     char dummy;
 242
 243     if ( socketpair( AF_UNIX, SOCK_STREAM, 0, fd ) ) return ret;
 244     if ( shutdown( fd[0], 1 ) ) goto out;
 245
 246     pfd.fd = fd[1];
 247     pfd.events = POLLIN;
 248     pfd.revents = 0;
 249
 250     /* Solaris' poll() sometimes returns nothing if given a 0ms timeout here */
 251     n = poll( &pfd, 1, 1 );
 252     if ( n != 1 ) goto out; /* error or timeout */
 253     if ( pfd.revents & POLLHUP )
 254         ret = SOCK_SHUTDOWN_POLLHUP;
 255     else if ( pfd.revents & POLLIN &&
 256               read( fd[1], &dummy, 1 ) == 0 )
 257         ret = SOCK_SHUTDOWN_EOF;
 258
 259 out:
 260     close( fd[0] );
 261     close( fd[1] );
 262     return ret;
 263 }
 264
 265 void sock_init(void)
 266 {
 267     sock_shutdown_type = sock_check_pollhup();
 268
 269     switch ( sock_shutdown_type )
 270     {
 271     case SOCK_SHUTDOWN_EOF:
 272         if (debug_level) fprintf( stderr, "sock_init: shutdown() causes EOF\n" );
 273         break;
 274     case SOCK_SHUTDOWN_POLLHUP:
 275         if (debug_level) fprintf( stderr, "sock_init: shutdown() causes POLLHUP\n" );
 276         break;
 277     default:
 278         fprintf( stderr, "sock_init: ERROR in sock_check_pollhup()\n" );
 279         sock_shutdown_type = SOCK_SHUTDOWN_EOF;
 280     }
 281 }
 282
 283 static int sock_reselect( struct sock *sock )
 284 {
 285     int ev = sock_get_poll_events( sock->fd );
 286
 287     if (debug_level)
 288         fprintf(stderr,"sock_reselect(%p): new mask %x\n", sock, ev);
 289
 290     if (!sock->polling)  /* FIXME: should find a better way to do this */
 291     {
 292         /* previously unconnected socket, is this reselect supposed to connect it? */
 293         if (!(sock->state & ~FD_WINE_NONBLOCKING)) return 0;
 294         /* ok, it is, attach it to the wineserver's main poll loop */
 295         sock->polling = 1;
 296         allow_fd_caching( sock->fd );
 297     }
 298     /* update condition mask */
 299     set_fd_events( sock->fd, ev );
 300     return ev;
 301 }
 302
 303 /* wake anybody waiting on the socket event or send the associated message */
 304 static void sock_wake_up( struct sock *sock )
 305 {
 306     unsigned int events = sock->pmask & sock->mask;
 307     int i;
 308
 309     if ( !events ) return;
 310
 311     if (sock->event)
 312     {
 313         if (debug_level) fprintf(stderr, "signalling events %x ptr %p\n", events, sock->event );
 314         set_event( sock->event );
 315     }
 316     if (sock->window)
 317     {
 318         if (debug_level) fprintf(stderr, "signalling events %x win %08x\n", events, sock->window );
 319         for (i = 0; i < FD_MAX_EVENTS; i++)
 320         {
 321             int event = event_bitorder[i];
 322             if (sock->pmask & (1 << event))
 323             {
 324                 lparam_t lparam = (1 << event) | (sock->errors[event] << 16);
 325                 post_message( sock->window, sock->message, sock->wparam, lparam );
 326             }
 327         }
 328         sock->pmask = 0;
 329         sock_reselect( sock );
 330     }
 331 }
 332
 333 static inline int sock_error( struct fd *fd )
 334 {
 335     unsigned int optval = 0;
 336     socklen_t optlen = sizeof(optval);
 337
 338     getsockopt( get_unix_fd(fd), SOL_SOCKET, SO_ERROR, (void *) &optval, &optlen);
 339     return optval;
 340 }
 341
 342 static int sock_dispatch_asyncs( struct sock *sock, int event, int error )
 343 {
 344     if (is_fd_overlapped( sock->fd ))
 345     {
 346         if (event & (POLLIN|POLLPRI) && async_waiting( &sock->read_q ))
 347         {
 348             if (debug_level) fprintf( stderr, "activating read queue for socket %p\n", sock );
 349             async_wake_up( &sock->read_q, STATUS_ALERTED );
 350             event &= ~(POLLIN|POLLPRI);
 351         }
 352         if (event & POLLOUT && async_waiting( &sock->write_q ))
 353         {
 354             if (debug_level) fprintf( stderr, "activating write queue for socket %p\n", sock );
 355             async_wake_up( &sock->write_q, STATUS_ALERTED );
 356             event &= ~POLLOUT;
 357         }
 358         if ( event & (POLLERR|POLLHUP) )
 359         {
 360             int status = sock_get_ntstatus( error );
 361
 362             if ( !(sock->state & FD_READ) )
 363                 async_wake_up( &sock->read_q, status );
 364             if ( !(sock->state & FD_WRITE) )
 365                 async_wake_up( &sock->write_q, status );
 366         }
 367     }
 368     return event;
 369 }
 370
 371 static void sock_dispatch_events( struct sock *sock, int prevstate, int event, int error )
 372 {
 373     if (prevstate & FD_CONNECT)
 374     {
 375         sock->pmask |= FD_CONNECT;
 376         sock->hmask |= FD_CONNECT;
 377         sock->errors[FD_CONNECT_BIT] = sock_get_error( error );
 378         goto end;
 379     }
 380     if (prevstate & FD_WINE_LISTENING)
 381     {
 382         sock->pmask |= FD_ACCEPT;
 383         sock->hmask |= FD_ACCEPT;
 384         sock->errors[FD_ACCEPT_BIT] = sock_get_error( error );
 385         goto end;
 386     }
 387
 388     if (event & POLLIN)
 389     {
 390         sock->pmask |= FD_READ;
 391         sock->hmask |= FD_READ;
 392         sock->errors[FD_READ_BIT] = 0;
 393     }
 394
 395     if (event & POLLOUT)
 396     {
 397         sock->pmask |= FD_WRITE;
 398         sock->hmask |= FD_WRITE;
 399         sock->errors[FD_WRITE_BIT] = 0;
 400     }
 401
 402     if (event & POLLPRI)
 403     {
 404         sock->pmask |= FD_OOB;
 405         sock->hmask |= FD_OOB;
 406         sock->errors[FD_OOB_BIT] = 0;
 407     }
 408
 409     if (event & (POLLERR|POLLHUP))
 410     {
 411         sock->pmask |= FD_CLOSE;
 412         sock->hmask |= FD_CLOSE;
 413         sock->errors[FD_CLOSE_BIT] = sock_get_error( error );
 414     }
 415 end:
 416     sock_wake_up( sock );
 417 }
 418
 419 static void sock_poll_event( struct fd *fd, int event )
 420 {
 421     struct sock *sock = get_fd_user( fd );
 422     int hangup_seen = 0;
 423     int prevstate = sock->state;
 424     int error = 0;
 425
 426     assert( sock->obj.ops == &sock_ops );
 427     if (debug_level)
 428         fprintf(stderr, "socket %p select event: %x\n", sock, event);
 429
 430     /* we may change event later, remove from loop here */
 431     if (event & (POLLERR|POLLHUP)) set_fd_events( sock->fd, -1 );
 432
 433     if (sock->state & FD_CONNECT)
 434     {
 435         if (event & (POLLERR|POLLHUP))
 436         {
 437             /* we didn't get connected? */
 438             sock->state &= ~FD_CONNECT;
 439             event &= ~POLLOUT;
 440             error = sock_error( fd );
 441         }
 442         else if (event & POLLOUT)
 443         {
 444             /* we got connected */
 445             sock->state |= FD_WINE_CONNECTED|FD_READ|FD_WRITE;
 446             sock->state &= ~FD_CONNECT;
 447             sock->connect_time = current_time;
 448         }
 449     }
 450     else if (sock->state & FD_WINE_LISTENING)
 451     {
 452         /* listening */
 453         if (event & (POLLERR|POLLHUP))
 454             error = sock_error( fd );
 455     }
 456     else
 457     {
 458         /* normal data flow */
 459         if (sock->type == WS_SOCK_STREAM && (event & POLLIN))
 460         {
 461             char dummy;
 462             int nr;
 463
 464             /* Linux 2.4 doesn't report POLLHUP if only one side of the socket
 465              * has been closed, so we need to check for it explicitly here */
 466             nr  = recv( get_unix_fd( fd ), &dummy, 1, MSG_PEEK );
 467             if ( nr == 0 )
 468             {
 469                 hangup_seen = 1;
 470                 event &= ~POLLIN;
 471             }
 472             else if ( nr < 0 )
 473             {
 474                 event &= ~POLLIN;
 475                 /* EAGAIN can happen if an async recv() falls between the server's poll()
 476                    call and the invocation of this routine */
 477                 if ( errno != EAGAIN )
 478                 {
 479                     error = errno;
 480                     event |= POLLERR;
 481                     if ( debug_level )
 482                         fprintf( stderr, "recv error on socket %p: %d\n", sock, errno );
 483                 }
 484             }
 485         }
 486
 487         if ( (hangup_seen || event & (POLLHUP|POLLERR)) && (sock->state & (FD_READ|FD_WRITE)) )
 488         {
 489             error = error ? error : sock_error( fd );
 490             if ( (event & POLLERR) || ( sock_shutdown_type == SOCK_SHUTDOWN_EOF && (event & POLLHUP) ))
 491                 sock->state &= ~FD_WRITE;
 492             sock->state &= ~FD_READ;
 493
 494             if (debug_level)
 495                 fprintf(stderr, "socket %p aborted by error %d, event: %x\n", sock, error, event);
 496         }
 497
 498         if (hangup_seen)
 499             event |= POLLHUP;
 500     }
 501
 502     event = sock_dispatch_asyncs( sock, event, error );
 503     sock_dispatch_events( sock, prevstate, event, error );
 504
 505     /* if anyone is stupid enough to wait on the socket object itself,
 506      * maybe we should wake them up too, just in case? */
 507     wake_up( &sock->obj, 0 );
 508
 509     sock_reselect( sock );
 510 }
 511
 512 static void sock_dump( struct object *obj, int verbose )
 513 {
 514     struct sock *sock = (struct sock *)obj;
 515     assert( obj->ops == &sock_ops );
 516     fprintf( stderr, "Socket fd=%p, state=%x, mask=%x, pending=%x, held=%x\n",
 517             sock->fd, sock->state,
 518             sock->mask, sock->pmask, sock->hmask );
 519 }
 520
 521 static int sock_signaled( struct object *obj, struct wait_queue_entry *entry )
 522 {
 523     struct sock *sock = (struct sock *)obj;
 524     assert( obj->ops == &sock_ops );
 525
 526     return check_fd_events( sock->fd, sock_get_poll_events( sock->fd ) ) != 0;
 527 }
 528
 529 static int sock_get_poll_events( struct fd *fd )
 530 {
 531     struct sock *sock = get_fd_user( fd );
 532     unsigned int mask = sock->mask & ~sock->hmask;
 533     unsigned int smask = sock->state & mask;
 534     int ev = 0;
 535
 536     assert( sock->obj.ops == &sock_ops );
 537
 538     if (sock->state & FD_CONNECT)
 539         /* connecting, wait for writable */
 540         return POLLOUT;
 541
 542     if (async_queued( &sock->read_q ))
 543     {
 544         if (async_waiting( &sock->read_q )) ev |= POLLIN | POLLPRI;
 545     }
 546     else if (smask & FD_READ || (sock->state & FD_WINE_LISTENING && mask & FD_ACCEPT))
 547         ev |= POLLIN | POLLPRI;
 548     /* We use POLLIN with 0 bytes recv() as FD_CLOSE indication for stream sockets. */
 549     else if (sock->type == WS_SOCK_STREAM && (sock->state & FD_READ) && (mask & FD_CLOSE) &&
 550               !(sock->hmask & FD_READ))
 551         ev |= POLLIN;
 552
 553     if (async_queued( &sock->write_q ))
 554     {
 555         if (async_waiting( &sock->write_q )) ev |= POLLOUT;
 556     }
 557     else if (smask & FD_WRITE)
 558         ev |= POLLOUT;
 559
 560     return ev;
 561 }
 562
 563 static enum server_fd_type sock_get_fd_type( struct fd *fd )
 564 {
 565     return FD_TYPE_SOCKET;
 566 }
 567
 568 static void sock_queue_async( struct fd *fd, struct async *async, int type, int count )
 569 {
 570     struct sock *sock = get_fd_user( fd );
 571     struct async_queue *queue;
 572
 573     assert( sock->obj.ops == &sock_ops );
 574
 575     switch (type)
 576     {
 577     case ASYNC_TYPE_READ:
 578         queue = &sock->read_q;
 579         break;
 580     case ASYNC_TYPE_WRITE:
 581         queue = &sock->write_q;
 582         break;
 583     default:
 584         set_error( STATUS_INVALID_PARAMETER );
 585         return;
 586     }
 587
 588     if ( ( !( sock->state & (FD_READ|FD_CONNECT|FD_WINE_LISTENING) ) && type == ASYNC_TYPE_READ  ) ||
 589          ( !( sock->state & (FD_WRITE|FD_CONNECT) ) && type == ASYNC_TYPE_WRITE ) )
 590     {
 591         set_error( STATUS_PIPE_DISCONNECTED );
 592         return;
 593     }
 594
 595     queue_async( queue, async );
 596     sock_reselect( sock );
 597
 598     set_error( STATUS_PENDING );
 599 }
 600
 601 static void sock_reselect_async( struct fd *fd, struct async_queue *queue )
 602 {
 603     struct sock *sock = get_fd_user( fd );
 604     /* ignore reselect on ifchange queue */
 605     if (&sock->ifchange_q != queue)
 606         sock_reselect( sock );
 607 }
 608
 609 static struct fd *sock_get_fd( struct object *obj )
 610 {
 611     struct sock *sock = (struct sock *)obj;
 612     return (struct fd *)grab_object( sock->fd );
 613 }
 614
 615 static void sock_destroy( struct object *obj )
 616 {
 617     struct sock *sock = (struct sock *)obj;
 618     assert( obj->ops == &sock_ops );
 619
 620     /* FIXME: special socket shutdown stuff? */
 621
 622     if ( sock->deferred )
 623         release_object( sock->deferred );
 624
 625     async_wake_up( &sock->ifchange_q, STATUS_CANCELLED );
 626     sock_release_ifchange( sock );
 627     free_async_queue( &sock->read_q );
 628     free_async_queue( &sock->write_q );
 629     free_async_queue( &sock->ifchange_q );
 630     if (sock->event) release_object( sock->event );
 631     if (sock->fd)
 632     {
 633         /* shut the socket down to force pending poll() calls in the client to return */
 634         shutdown( get_unix_fd(sock->fd), SHUT_RDWR );
 635         release_object( sock->fd );
 636     }
 637 }
 638
 639 static struct sock *create_socket(void)
 640 {
 641     struct sock *sock;
 642
 643     if (!(sock = alloc_object( &sock_ops ))) return NULL;
 644     sock->fd      = NULL;
 645     sock->state   = 0;
 646     sock->mask    = 0;
 647     sock->hmask   = 0;
 648     sock->pmask   = 0;
 649     sock->polling = 0;
 650     sock->flags   = 0;
 651     sock->proto   = 0;
 652     sock->type    = 0;
 653     sock->family  = 0;
 654     sock->event   = NULL;
 655     sock->window  = 0;
 656     sock->message = 0;
 657     sock->wparam  = 0;
 658     sock->connect_time = 0;
 659     sock->deferred = NULL;
 660     sock->ifchange_obj = NULL;
 661     init_async_queue( &sock->read_q );
 662     init_async_queue( &sock->write_q );
 663     init_async_queue( &sock->ifchange_q );
 664     memset( sock->errors, 0, sizeof(sock->errors) );
 665     return sock;
 666 }
 667
 668 static int get_unix_family( int family )
 669 {
 670     switch (family)
 671     {
 672         case WS_AF_INET: return AF_INET;
 673         case WS_AF_INET6: return AF_INET6;
 674 #ifdef HAS_IPX
 675         case WS_AF_IPX: return AF_IPX;
 676 #endif
 677 #ifdef AF_IRDA
 678         case WS_AF_IRDA: return AF_IRDA;
 679 #endif
 680         case WS_AF_UNSPEC: return AF_UNSPEC;
 681         default: return -1;
 682     }
 683 }
 684
 685 static int get_unix_type( int type )
 686 {
 687     switch (type)
 688     {
 689         case WS_SOCK_DGRAM: return SOCK_DGRAM;
 690         case WS_SOCK_RAW: return SOCK_RAW;
 691         case WS_SOCK_STREAM: return SOCK_STREAM;
 692         default: return -1;
 693     }
 694 }
 695
 696 static int get_unix_protocol( int protocol )
 697 {
 698     if (protocol >= WS_NSPROTO_IPX && protocol <= WS_NSPROTO_IPX + 255)
 699         return protocol;
 700
 701     switch (protocol)
 702     {
 703         case WS_IPPROTO_ICMP: return IPPROTO_ICMP;
 704         case WS_IPPROTO_IGMP: return IPPROTO_IGMP;
 705         case WS_IPPROTO_IP: return IPPROTO_IP;
 706         case WS_IPPROTO_IPIP: return IPPROTO_IPIP;
 707         case WS_IPPROTO_IPV6: return IPPROTO_IPV6;
 708         case WS_IPPROTO_RAW: return IPPROTO_RAW;
 709         case WS_IPPROTO_TCP: return IPPROTO_TCP;
 710         case WS_IPPROTO_UDP: return IPPROTO_UDP;
 711         default: return -1;
 712     }
 713 }
 714
 715 static void set_dont_fragment( int fd, int level, int value )
 716 {
 717     int optname;
 718
 719     if (level == IPPROTO_IP)
 720     {
 721 #ifdef IP_DONTFRAG
 722         optname = IP_DONTFRAG;
 723 #elif defined(IP_MTU_DISCOVER) && defined(IP_PMTUDISC_DO) && defined(IP_PMTUDISC_DONT)
 724         optname = IP_MTU_DISCOVER;
 725         value = value ? IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 726 #else
 727         return;
 728 #endif
 729     }
 730     else
 731     {
 732 #ifdef IPV6_DONTFRAG
 733         optname = IPV6_DONTFRAG;
 734 #elif defined(IPV6_MTU_DISCOVER) && defined(IPV6_PMTUDISC_DO) && defined(IPV6_PMTUDISC_DONT)
 735         optname = IPV6_MTU_DISCOVER;
 736         value = value ? IPV6_PMTUDISC_DO : IPV6_PMTUDISC_DONT;
 737 #else
 738         return;
 739 #endif
 740     }
 741
 742     setsockopt( fd, level, optname, &value, sizeof(value) );
 743 }
 744
 745 static int init_socket( struct sock *sock, int family, int type, int protocol, unsigned int flags )
 746 {
 747     unsigned int options = 0;
 748     int sockfd, unix_type, unix_family, unix_protocol;
 749
 750     unix_family = get_unix_family( family );
 751     unix_type = get_unix_type( type );
 752     unix_protocol = get_unix_protocol( protocol );
 753
 754     if (unix_protocol < 0)
 755     {
 756         if (type && unix_type < 0)
 757             set_win32_error( WSAESOCKTNOSUPPORT );
 758         else
 759             set_win32_error( WSAEPROTONOSUPPORT );
 760         return -1;
 761     }
 762     if (unix_family < 0)
 763     {
 764         if (family >= 0 && unix_type < 0)
 765             set_win32_error( WSAESOCKTNOSUPPORT );
 766         else
 767             set_win32_error( WSAEAFNOSUPPORT );
 768         return -1;
 769     }
 770
 771     sockfd = socket( unix_family, unix_type, unix_protocol );
 772     if (sockfd == -1)
 773     {
 774         if (errno == EINVAL) set_win32_error( WSAESOCKTNOSUPPORT );
 775         else set_win32_error( sock_get_error( errno ));
 776         return -1;
 777     }
 778     fcntl(sockfd, F_SETFL, O_NONBLOCK); /* make socket nonblocking */
 779
 780     if (family == WS_AF_IPX && protocol >= WS_NSPROTO_IPX && protocol <= WS_NSPROTO_IPX + 255)
 781     {
 782 #ifdef HAS_IPX
 783         int ipx_type = protocol - WS_NSPROTO_IPX;
 784
 785 #ifdef SOL_IPX
 786         setsockopt( sockfd, SOL_IPX, IPX_TYPE, &ipx_type, sizeof(ipx_type) );
 787 #else
 788         struct ipx val;
 789         /* Should we retrieve val using a getsockopt call and then
 790          * set the modified one? */
 791         val.ipx_pt = ipx_type;
 792         setsockopt( sockfd, 0, SO_DEFAULT_HEADERS, &val, sizeof(val) );
 793 #endif
 794 #endif
 795     }
 796
 797     if (unix_family == AF_INET || unix_family == AF_INET6)
 798     {
 799         /* ensure IP_DONTFRAGMENT is disabled for SOCK_DGRAM and SOCK_RAW, enabled for SOCK_STREAM */
 800         if (unix_type == SOCK_DGRAM || unix_type == SOCK_RAW) /* in Linux the global default can be enabled */
 801             set_dont_fragment( sockfd, unix_family == AF_INET6 ? IPPROTO_IPV6 : IPPROTO_IP, FALSE );
 802         else if (unix_type == SOCK_STREAM)
 803             set_dont_fragment( sockfd, unix_family == AF_INET6 ? IPPROTO_IPV6 : IPPROTO_IP, TRUE );
 804     }
 805
 806 #ifdef IPV6_V6ONLY
 807     if (unix_family == AF_INET6)
 808     {
 809         static const int enable = 1;
 810         setsockopt( sockfd, IPPROTO_IPV6, IPV6_V6ONLY, &enable, sizeof(enable) );
 811     }
 812 #endif
 813
 814     sock->state  = (type != SOCK_STREAM) ? (FD_READ|FD_WRITE) : 0;
 815     sock->flags  = flags;
 816     sock->proto  = protocol;
 817     sock->type   = type;
 818     sock->family = family;
 819
 820     if (sock->fd)
 821     {
 822         options = get_fd_options( sock->fd );
 823         release_object( sock->fd );
 824     }
 825
 826     if (!(sock->fd = create_anonymous_fd( &sock_fd_ops, sockfd, &sock->obj, options )))
 827     {
 828         return -1;
 829     }
 830     sock_reselect( sock );
 831     clear_error();
 832     return 0;
 833 }
 834
 835 /* accepts a socket and inits it */
 836 static int accept_new_fd( struct sock *sock )
 837 {
 838
 839     /* Try to accept(2). We can't be safe that this an already connected socket
 840      * or that accept() is allowed on it. In those cases we will get -1/errno
 841      * return.
 842      */
 843     struct sockaddr saddr;
 844     socklen_t slen = sizeof(saddr);
 845     int acceptfd = accept( get_unix_fd(sock->fd), &saddr, &slen );
 846     if (acceptfd != -1)
 847         fcntl( acceptfd, F_SETFL, O_NONBLOCK );
 848     else
 849         set_win32_error( sock_get_error( errno ));
 850     return acceptfd;
 851 }
 852
 853 /* accept a socket (creates a new fd) */
 854 static struct sock *accept_socket( struct sock *sock )
 855 {
 856     struct sock *acceptsock;
 857     int acceptfd;
 858
 859     if (get_unix_fd( sock->fd ) == -1) return NULL;
 860
 861     if ( sock->deferred )
 862     {
 863         acceptsock = sock->deferred;
 864         sock->deferred = NULL;
 865     }
 866     else
 867     {
 868         if ((acceptfd = accept_new_fd( sock )) == -1) return NULL;
 869         if (!(acceptsock = create_socket()))
 870         {
 871             close( acceptfd );
 872             return NULL;
 873         }
 874
 875         /* newly created socket gets the same properties of the listening socket */
 876         acceptsock->state  = FD_WINE_CONNECTED|FD_READ|FD_WRITE;
 877         if (sock->state & FD_WINE_NONBLOCKING)
 878             acceptsock->state |= FD_WINE_NONBLOCKING;
 879         acceptsock->mask    = sock->mask;
 880         acceptsock->proto   = sock->proto;
 881         acceptsock->type    = sock->type;
 882         acceptsock->family  = sock->family;
 883         acceptsock->window  = sock->window;
 884         acceptsock->message = sock->message;
 885         acceptsock->connect_time = current_time;
 886         if (sock->event) acceptsock->event = (struct event *)grab_object( sock->event );
 887         acceptsock->flags = sock->flags;
 888         if (!(acceptsock->fd = create_anonymous_fd( &sock_fd_ops, acceptfd, &acceptsock->obj,
 889                                                     get_fd_options( sock->fd ) )))
 890         {
 891             release_object( acceptsock );
 892             return NULL;
 893         }
 894     }
 895     clear_error();
 896     sock->pmask &= ~FD_ACCEPT;
 897     sock->hmask &= ~FD_ACCEPT;
 898     sock_reselect( sock );
 899     return acceptsock;
 900 }
 901
 902 static int accept_into_socket( struct sock *sock, struct sock *acceptsock )
 903 {
 904     int acceptfd;
 905     struct fd *newfd;
 906
 907     if (get_unix_fd( sock->fd ) == -1) return FALSE;
 908
 909     if ( sock->deferred )
 910     {
 911         newfd = dup_fd_object( sock->deferred->fd, 0, 0,
 912                                get_fd_options( acceptsock->fd ) );
 913         if ( !newfd )
 914             return FALSE;
 915
 916         set_fd_user( newfd, &sock_fd_ops, &acceptsock->obj );
 917
 918         release_object( sock->deferred );
 919         sock->deferred = NULL;
 920     }
 921     else
 922     {
 923         if ((acceptfd = accept_new_fd( sock )) == -1)
 924             return FALSE;
 925
 926         if (!(newfd = create_anonymous_fd( &sock_fd_ops, acceptfd, &acceptsock->obj,
 927                                             get_fd_options( acceptsock->fd ) )))
 928             return FALSE;
 929     }
 930
 931     acceptsock->state  |= FD_WINE_CONNECTED|FD_READ|FD_WRITE;
 932     acceptsock->hmask   = 0;
 933     acceptsock->pmask   = 0;
 934     acceptsock->polling = 0;
 935     acceptsock->proto   = sock->proto;
 936     acceptsock->type    = sock->type;
 937     acceptsock->family  = sock->family;
 938     acceptsock->wparam  = 0;
 939     acceptsock->deferred = NULL;
 940     acceptsock->connect_time = current_time;
 941     fd_copy_completion( acceptsock->fd, newfd );
 942     release_object( acceptsock->fd );
 943     acceptsock->fd = newfd;
 944
 945     clear_error();
 946     sock->pmask &= ~FD_ACCEPT;
 947     sock->hmask &= ~FD_ACCEPT;
 948     sock_reselect( sock );
 949
 950     return TRUE;
 951 }
 952
 953 /* return an errno value mapped to a WSA error */
 954 static unsigned int sock_get_error( int err )
 955 {
 956     switch (err)
 957     {
 958         case EINTR:             return WSAEINTR;
 959         case EBADF:             return WSAEBADF;
 960         case EPERM:
 961         case EACCES:            return WSAEACCES;
 962         case EFAULT:            return WSAEFAULT;
 963         case EINVAL:            return WSAEINVAL;
 964         case EMFILE:            return WSAEMFILE;
 965         case EWOULDBLOCK:       return WSAEWOULDBLOCK;
 966         case EINPROGRESS:       return WSAEINPROGRESS;
 967         case EALREADY:          return WSAEALREADY;
 968         case ENOTSOCK:          return WSAENOTSOCK;
 969         case EDESTADDRREQ:      return WSAEDESTADDRREQ;
 970         case EMSGSIZE:          return WSAEMSGSIZE;
 971         case EPROTOTYPE:        return WSAEPROTOTYPE;
 972         case ENOPROTOOPT:       return WSAENOPROTOOPT;
 973         case EPROTONOSUPPORT:   return WSAEPROTONOSUPPORT;
 974         case ESOCKTNOSUPPORT:   return WSAESOCKTNOSUPPORT;
 975         case EOPNOTSUPP:        return WSAEOPNOTSUPP;
 976         case EPFNOSUPPORT:      return WSAEPFNOSUPPORT;
 977         case EAFNOSUPPORT:      return WSAEAFNOSUPPORT;
 978         case EADDRINUSE:        return WSAEADDRINUSE;
 979         case EADDRNOTAVAIL:     return WSAEADDRNOTAVAIL;
 980         case ENETDOWN:          return WSAENETDOWN;
 981         case ENETUNREACH:       return WSAENETUNREACH;
 982         case ENETRESET:         return WSAENETRESET;
 983         case ECONNABORTED:      return WSAECONNABORTED;
 984         case EPIPE:
 985         case ECONNRESET:        return WSAECONNRESET;
 986         case ENOBUFS:           return WSAENOBUFS;
 987         case EISCONN:           return WSAEISCONN;
 988         case ENOTCONN:          return WSAENOTCONN;
 989         case ESHUTDOWN:         return WSAESHUTDOWN;
 990         case ETOOMANYREFS:      return WSAETOOMANYREFS;
 991         case ETIMEDOUT:         return WSAETIMEDOUT;
 992         case ECONNREFUSED:      return WSAECONNREFUSED;
 993         case ELOOP:             return WSAELOOP;
 994         case ENAMETOOLONG:      return WSAENAMETOOLONG;
 995         case EHOSTDOWN:         return WSAEHOSTDOWN;
 996         case EHOSTUNREACH:      return WSAEHOSTUNREACH;
 997         case ENOTEMPTY:         return WSAENOTEMPTY;
 998 #ifdef EPROCLIM
 999         case EPROCLIM:          return WSAEPROCLIM;
1000 #endif
1001 #ifdef EUSERS
1002         case EUSERS:            return WSAEUSERS;
1003 #endif
1004 #ifdef EDQUOT
1005         case EDQUOT:            return WSAEDQUOT;
1006 #endif
1007 #ifdef ESTALE
1008         case ESTALE:            return WSAESTALE;
1009 #endif
1010 #ifdef EREMOTE
1011         case EREMOTE:           return WSAEREMOTE;
1012 #endif
1013
1014         case 0:                 return 0;
1015         default:
1016             errno = err;
1017             perror("wineserver: sock_get_error() can't map error");
1018             return WSAEFAULT;
1019     }
1020 }
1021
1022 static int sock_get_ntstatus( int err )
1023 {
1024     switch ( err )
1025     {
1026         case EBADF:             return STATUS_INVALID_HANDLE;
1027         case EBUSY:             return STATUS_DEVICE_BUSY;
1028         case EPERM:
1029         case EACCES:            return STATUS_ACCESS_DENIED;
1030         case EFAULT:            return STATUS_NO_MEMORY;
1031         case EINVAL:            return STATUS_INVALID_PARAMETER;
1032         case ENFILE:
1033         case EMFILE:            return STATUS_TOO_MANY_OPENED_FILES;
1034         case EWOULDBLOCK:       return STATUS_CANT_WAIT;
1035         case EINPROGRESS:       return STATUS_PENDING;
1036         case EALREADY:          return STATUS_NETWORK_BUSY;
1037         case ENOTSOCK:          return STATUS_OBJECT_TYPE_MISMATCH;
1038         case EDESTADDRREQ:      return STATUS_INVALID_PARAMETER;
1039         case EMSGSIZE:          return STATUS_BUFFER_OVERFLOW;
1040         case EPROTONOSUPPORT:
1041         case ESOCKTNOSUPPORT:
1042         case EPFNOSUPPORT:
1043         case EAFNOSUPPORT:
1044         case EPROTOTYPE:        return STATUS_NOT_SUPPORTED;
1045         case ENOPROTOOPT:       return STATUS_INVALID_PARAMETER;
1046         case EOPNOTSUPP:        return STATUS_NOT_SUPPORTED;
1047         case EADDRINUSE:        return STATUS_ADDRESS_ALREADY_ASSOCIATED;
1048         case EADDRNOTAVAIL:     return STATUS_INVALID_PARAMETER;
1049         case ECONNREFUSED:      return STATUS_CONNECTION_REFUSED;
1050         case ESHUTDOWN:         return STATUS_PIPE_DISCONNECTED;
1051         case ENOTCONN:          return STATUS_CONNECTION_DISCONNECTED;
1052         case ETIMEDOUT:         return STATUS_IO_TIMEOUT;
1053         case ENETUNREACH:       return STATUS_NETWORK_UNREACHABLE;
1054         case EHOSTUNREACH:      return STATUS_HOST_UNREACHABLE;
1055         case ENETDOWN:          return STATUS_NETWORK_BUSY;
1056         case EPIPE:
1057         case ECONNRESET:        return STATUS_CONNECTION_RESET;
1058         case ECONNABORTED:      return STATUS_CONNECTION_ABORTED;
1059
1060         case 0:                 return STATUS_SUCCESS;
1061         default:
1062             errno = err;
1063             perror("wineserver: sock_get_ntstatus() can't map error");
1064             return STATUS_UNSUCCESSFUL;
1065     }
1066 }
1067
1068 static int sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async )
1069 {
1070     struct sock *sock = get_fd_user( fd );
1071
1072     assert( sock->obj.ops == &sock_ops );
1073
1074     if (get_unix_fd( fd ) == -1 && code != IOCTL_AFD_CREATE) return 0;
1075
1076     switch(code)
1077     {
1078     case IOCTL_AFD_CREATE:
1079     {
1080         const struct afd_create_params *params = get_req_data();
1081
1082         if (get_req_data_size() != sizeof(*params))
1083         {
1084             set_error( STATUS_INVALID_PARAMETER );
1085             return 0;
1086         }
1087         init_socket( sock, params->family, params->type, params->protocol, params->flags );
1088         return 0;
1089     }
1090
1091     case IOCTL_AFD_ACCEPT:
1092     {
1093         struct sock *acceptsock;
1094         obj_handle_t handle;
1095
1096         if (get_reply_max_size() != sizeof(handle))
1097         {
1098             set_error( STATUS_BUFFER_TOO_SMALL );
1099             return 0;
1100         }
1101
1102         if (!(acceptsock = accept_socket( sock ))) return 0;
1103         handle = alloc_handle( current->process, &acceptsock->obj,
1104                                GENERIC_READ | GENERIC_WRITE | SYNCHRONIZE, OBJ_INHERIT );
1105         acceptsock->wparam = handle;
1106         release_object( acceptsock );
1107         set_reply_data( &handle, sizeof(handle) );
1108         return 0;
1109     }
1110
1111     case IOCTL_AFD_ACCEPT_INTO:
1112     {
1113         static const int access = FILE_READ_ATTRIBUTES | FILE_WRITE_ATTRIBUTES | FILE_READ_DATA;
1114         struct sock *acceptsock;
1115         obj_handle_t handle;
1116
1117         if (get_req_data_size() != sizeof(handle))
1118         {
1119             set_error( STATUS_BUFFER_TOO_SMALL );
1120             return 0;
1121         }
1122         handle = *(obj_handle_t *)get_req_data();
1123
1124         if (!(acceptsock = (struct sock *)get_handle_obj( current->process, handle, access, &sock_ops )))
1125             return 0;
1126         if (accept_into_socket( sock, acceptsock ))
1127             acceptsock->wparam = handle;
1128         release_object( acceptsock );
1129         return 0;
1130     }
1131
1132     case IOCTL_AFD_ADDRESS_LIST_CHANGE:
1133         if ((sock->state & FD_WINE_NONBLOCKING) && async_is_blocking( async ))
1134         {
1135             set_win32_error( WSAEWOULDBLOCK );
1136             return 0;
1137         }
1138         if (!sock_get_ifchange( sock )) return 0;
1139         queue_async( &sock->ifchange_q, async );
1140         set_error( STATUS_PENDING );
1141         return 1;
1142
1143     default:
1144         set_error( STATUS_NOT_SUPPORTED );
1145         return 0;
1146     }
1147 }
1148
1149 #ifdef HAVE_LINUX_RTNETLINK_H
1150
1151 /* only keep one ifchange object around, all sockets waiting for wakeups will look to it */
1152 static struct object *ifchange_object;
1153
1154 static void ifchange_dump( struct object *obj, int verbose );
1155 static struct fd *ifchange_get_fd( struct object *obj );
1156 static void ifchange_destroy( struct object *obj );
1157
1158 static int ifchange_get_poll_events( struct fd *fd );
1159 static void ifchange_poll_event( struct fd *fd, int event );
1160
1161 struct ifchange
1162 {
1163     struct object       obj;     /* object header */
1164     struct fd          *fd;      /* interface change file descriptor */
1165     struct list         sockets; /* list of sockets to send interface change notifications */
1166 };
1167
1168 static const struct object_ops ifchange_ops =
1169 {
1170     sizeof(struct ifchange), /* size */
1171     ifchange_dump,           /* dump */
1172     no_get_type,             /* get_type */
1173     add_queue,               /* add_queue */
1174     NULL,                    /* remove_queue */
1175     NULL,                    /* signaled */
1176     no_satisfied,            /* satisfied */
1177     no_signal,               /* signal */
1178     ifchange_get_fd,         /* get_fd */
1179     default_fd_map_access,   /* map_access */
1180     default_get_sd,          /* get_sd */
1181     default_set_sd,          /* set_sd */
1182     no_get_full_name,        /* get_full_name */
1183     no_lookup_name,          /* lookup_name */
1184     no_link_name,            /* link_name */
1185     NULL,                    /* unlink_name */
1186     no_open_file,            /* open_file */
1187     no_kernel_obj_list,      /* get_kernel_obj_list */
1188     no_close_handle,         /* close_handle */
1189     ifchange_destroy         /* destroy */
1190 };
1191
1192 static const struct fd_ops ifchange_fd_ops =
1193 {
1194     ifchange_get_poll_events, /* get_poll_events */
1195     ifchange_poll_event,      /* poll_event */
1196     NULL,                     /* get_fd_type */
1197     no_fd_read,               /* read */
1198     no_fd_write,              /* write */
1199     no_fd_flush,              /* flush */
1200     no_fd_get_file_info,      /* get_file_info */
1201     no_fd_get_volume_info,    /* get_volume_info */
1202     no_fd_ioctl,              /* ioctl */
1203     NULL,                     /* queue_async */
1204     NULL                      /* reselect_async */
1205 };
1206
1207 static void ifchange_dump( struct object *obj, int verbose )
1208 {
1209     assert( obj->ops == &ifchange_ops );
1210     fprintf( stderr, "Interface change\n" );
1211 }
1212
1213 static struct fd *ifchange_get_fd( struct object *obj )
1214 {
1215     struct ifchange *ifchange = (struct ifchange *)obj;
1216     return (struct fd *)grab_object( ifchange->fd );
1217 }
1218
1219 static void ifchange_destroy( struct object *obj )
1220 {
1221     struct ifchange *ifchange = (struct ifchange *)obj;
1222     assert( obj->ops == &ifchange_ops );
1223
1224     release_object( ifchange->fd );
1225
1226     /* reset the global ifchange object so that it will be recreated if it is needed again */
1227     assert( obj == ifchange_object );
1228     ifchange_object = NULL;
1229 }
1230
1231 static int ifchange_get_poll_events( struct fd *fd )
1232 {
1233     return POLLIN;
1234 }
1235
1236 /* wake up all the sockets waiting for a change notification event */
1237 static void ifchange_wake_up( struct object *obj, unsigned int status )
1238 {
1239     struct ifchange *ifchange = (struct ifchange *)obj;
1240     struct list *ptr, *next;
1241     assert( obj->ops == &ifchange_ops );
1242     assert( obj == ifchange_object );
1243
1244     LIST_FOR_EACH_SAFE( ptr, next, &ifchange->sockets )
1245     {
1246         struct sock *sock = LIST_ENTRY( ptr, struct sock, ifchange_entry );
1247
1248         assert( sock->ifchange_obj );
1249         async_wake_up( &sock->ifchange_q, status ); /* issue ifchange notification for the socket */
1250         sock_release_ifchange( sock ); /* remove socket from list and decrement ifchange refcount */
1251     }
1252 }
1253
1254 static void ifchange_poll_event( struct fd *fd, int event )
1255 {
1256     struct object *ifchange = get_fd_user( fd );
1257     unsigned int status = STATUS_PENDING;
1258     char buffer[PIPE_BUF];
1259     int r;
1260
1261     r = recv( get_unix_fd(fd), buffer, sizeof(buffer), MSG_DONTWAIT );
1262     if (r < 0)
1263     {
1264         if (errno == EWOULDBLOCK || (EWOULDBLOCK != EAGAIN && errno == EAGAIN))
1265             return;  /* retry when poll() says the socket is ready */
1266         status = sock_get_ntstatus( errno );
1267     }
1268     else if (r > 0)
1269     {
1270         struct nlmsghdr *nlh;
1271
1272         for (nlh = (struct nlmsghdr *)buffer; NLMSG_OK(nlh, r); nlh = NLMSG_NEXT(nlh, r))
1273         {
1274             if (nlh->nlmsg_type == NLMSG_DONE)
1275                 break;
1276             if (nlh->nlmsg_type == RTM_NEWADDR || nlh->nlmsg_type == RTM_DELADDR)
1277                 status = STATUS_SUCCESS;
1278         }
1279     }
1280     else status = STATUS_CANCELLED;
1281
1282     if (status != STATUS_PENDING) ifchange_wake_up( ifchange, status );
1283 }
1284
1285 #endif
1286
1287 /* we only need one of these interface notification objects, all of the sockets dependent upon
1288  * it will wake up when a notification event occurs */
1289  static struct object *get_ifchange( void )
1290  {
1291 #ifdef HAVE_LINUX_RTNETLINK_H
1292     struct ifchange *ifchange;
1293     struct sockaddr_nl addr;
1294     int unix_fd;
1295
1296     if (ifchange_object)
1297     {
1298         /* increment the refcount for each socket that uses the ifchange object */
1299         return grab_object( ifchange_object );
1300     }
1301
1302     /* create the socket we need for processing interface change notifications */
1303     unix_fd = socket( PF_NETLINK, SOCK_RAW, NETLINK_ROUTE );
1304     if (unix_fd == -1)
1305     {
1306         set_win32_error( sock_get_error( errno ));
1307         return NULL;
1308     }
1309     fcntl( unix_fd, F_SETFL, O_NONBLOCK ); /* make socket nonblocking */
1310     memset( &addr, 0, sizeof(addr) );
1311     addr.nl_family = AF_NETLINK;
1312     addr.nl_groups = RTMGRP_IPV4_IFADDR;
1313     /* bind the socket to the special netlink kernel interface */
1314     if (bind( unix_fd, (struct sockaddr *)&addr, sizeof(addr) ) == -1)
1315     {
1316         close( unix_fd );
1317         set_win32_error( sock_get_error( errno ));
1318         return NULL;
1319     }
1320     if (!(ifchange = alloc_object( &ifchange_ops )))
1321     {
1322         close( unix_fd );
1323         set_error( STATUS_NO_MEMORY );
1324         return NULL;
1325     }
1326     list_init( &ifchange->sockets );
1327     if (!(ifchange->fd = create_anonymous_fd( &ifchange_fd_ops, unix_fd, &ifchange->obj, 0 )))
1328     {
1329         release_object( ifchange );
1330         set_error( STATUS_NO_MEMORY );
1331         return NULL;
1332     }
1333     set_fd_events( ifchange->fd, POLLIN ); /* enable read wakeup on the file descriptor */
1334
1335     /* the ifchange object is now successfully configured */
1336     ifchange_object = &ifchange->obj;
1337     return &ifchange->obj;
1338 #else
1339     set_error( STATUS_NOT_SUPPORTED );
1340     return NULL;
1341 #endif
1342 }
1343
1344 /* add the socket to the interface change notification list */
1345 static void ifchange_add_sock( struct object *obj, struct sock *sock )
1346 {
1347 #ifdef HAVE_LINUX_RTNETLINK_H
1348     struct ifchange *ifchange = (struct ifchange *)obj;
1349
1350     list_add_tail( &ifchange->sockets, &sock->ifchange_entry );
1351 #endif
1352 }
1353
1354 /* create a new ifchange queue for a specific socket or, if one already exists, reuse the existing one */
1355 static struct object *sock_get_ifchange( struct sock *sock )
1356 {
1357     struct object *ifchange;
1358
1359     if (sock->ifchange_obj) /* reuse existing ifchange_obj for this socket */
1360         return sock->ifchange_obj;
1361
1362     if (!(ifchange = get_ifchange()))
1363         return NULL;
1364
1365     /* add the socket to the ifchange notification list */
1366     ifchange_add_sock( ifchange, sock );
1367     sock->ifchange_obj = ifchange;
1368     return ifchange;
1369 }
1370
1371 /* destroy an existing ifchange queue for a specific socket */
1372 static void sock_release_ifchange( struct sock *sock )
1373 {
1374     if (sock->ifchange_obj)
1375     {
1376         list_remove( &sock->ifchange_entry );
1377         release_object( sock->ifchange_obj );
1378         sock->ifchange_obj = NULL;
1379     }
1380 }
1381
1382 static struct object_type *socket_device_get_type( struct object *obj );
1383 static void socket_device_dump( struct object *obj, int verbose );
1384 static struct object *socket_device_lookup_name( struct object *obj, struct unicode_str *name, unsigned int attr );
1385 static struct object *socket_device_open_file( struct object *obj, unsigned int access,
1386                                                unsigned int sharing, unsigned int options );
1387
1388 static const struct object_ops socket_device_ops =
1389 {
1390     sizeof(struct object),      /* size */
1391     socket_device_dump,         /* dump */
1392     socket_device_get_type,     /* get_type */
1393     no_add_queue,               /* add_queue */
1394     NULL,                       /* remove_queue */
1395     NULL,                       /* signaled */
1396     no_satisfied,               /* satisfied */
1397     no_signal,                  /* signal */
1398     no_get_fd,                  /* get_fd */
1399     default_fd_map_access,      /* map_access */
1400     default_get_sd,             /* get_sd */
1401     default_set_sd,             /* set_sd */
1402     default_get_full_name,      /* get_full_name */
1403     socket_device_lookup_name,  /* lookup_name */
1404     directory_link_name,        /* link_name */
1405     default_unlink_name,        /* unlink_name */
1406     socket_device_open_file,    /* open_file */
1407     no_kernel_obj_list,         /* get_kernel_obj_list */
1408     no_close_handle,            /* close_handle */
1409     no_destroy                  /* destroy */
1410 };
1411
1412 static struct object_type *socket_device_get_type( struct object *obj )
1413 {
1414     static const WCHAR name[] = {'D','e','v','i','c','e'};
1415     static const struct unicode_str str = { name, sizeof(name) };
1416     return get_object_type( &str );
1417 }
1418
1419 static void socket_device_dump( struct object *obj, int verbose )
1420 {
1421     fputs( "Socket device\n", stderr );
1422 }
1423
1424 static struct object *socket_device_lookup_name( struct object *obj, struct unicode_str *name, unsigned int attr )
1425 {
1426     return NULL;
1427 }
1428
1429 static struct object *socket_device_open_file( struct object *obj, unsigned int access,
1430                                                unsigned int sharing, unsigned int options )
1431 {
1432     struct sock *sock;
1433
1434     if (!(sock = create_socket())) return NULL;
1435     if (!(sock->fd = alloc_pseudo_fd( &sock_fd_ops, &sock->obj, options )))
1436     {
1437         release_object( sock );
1438         return NULL;
1439     }
1440     return &sock->obj;
1441 }
1442
1443 struct object *create_socket_device( struct object *root, const struct unicode_str *name,
1444                                      unsigned int attr, const struct security_descriptor *sd )
1445 {
1446     return create_named_object( root, &socket_device_ops, name, attr, sd );
1447 }
1448
1449 /* set socket event parameters */
1450 DECL_HANDLER(set_socket_event)
1451 {
1452     struct sock *sock;
1453     struct event *old_event;
1454
1455     if (!(sock = (struct sock *)get_handle_obj( current->process, req->handle,
1456                                                 FILE_WRITE_ATTRIBUTES, &sock_ops))) return;
1457     if (get_unix_fd( sock->fd ) == -1) return;
1458     old_event = sock->event;
1459     sock->mask    = req->mask;
1460     sock->hmask   &= ~req->mask; /* re-enable held events */
1461     sock->event   = NULL;
1462     sock->window  = req->window;
1463     sock->message = req->msg;
1464     sock->wparam  = req->handle;  /* wparam is the socket handle */
1465     if (req->event) sock->event = get_event_obj( current->process, req->event, EVENT_MODIFY_STATE );
1466
1467     if (debug_level && sock->event) fprintf(stderr, "event ptr: %p\n", sock->event);
1468
1469     sock_reselect( sock );
1470
1471     sock->state |= FD_WINE_NONBLOCKING;
1472
1473     /* if a network event is pending, signal the event object
1474        it is possible that FD_CONNECT or FD_ACCEPT network events has happened
1475        before a WSAEventSelect() was done on it.
1476        (when dealing with Asynchronous socket)  */
1477     sock_wake_up( sock );
1478
1479     if (old_event) release_object( old_event ); /* we're through with it */
1480     release_object( &sock->obj );
1481 }
1482
1483 /* get socket event parameters */
1484 DECL_HANDLER(get_socket_event)
1485 {
1486     struct sock *sock;
1487
1488     if (!(sock = (struct sock *)get_handle_obj( current->process, req->handle,
1489                                                 FILE_READ_ATTRIBUTES, &sock_ops ))) return;
1490     if (get_unix_fd( sock->fd ) == -1) return;
1491     reply->mask  = sock->mask;
1492     reply->pmask = sock->pmask;
1493     reply->state = sock->state;
1494     set_reply_data( sock->errors, min( get_reply_max_size(), sizeof(sock->errors) ));
1495
1496     if (req->service)
1497     {
1498         if (req->c_event)
1499         {
1500             struct event *cevent = get_event_obj( current->process, req->c_event,
1501                                                   EVENT_MODIFY_STATE );
1502             if (cevent)
1503             {
1504                 reset_event( cevent );
1505                 release_object( cevent );
1506             }
1507         }
1508         sock->pmask = 0;
1509         sock_reselect( sock );
1510     }
1511     release_object( &sock->obj );
1512 }
1513
1514 /* re-enable pending socket events */
1515 DECL_HANDLER(enable_socket_event)
1516 {
1517     struct sock *sock;
1518
1519     if (!(sock = (struct sock*)get_handle_obj( current->process, req->handle,
1520                                                FILE_WRITE_ATTRIBUTES, &sock_ops)))
1521         return;
1522
1523     if (get_unix_fd( sock->fd ) == -1) return;
1524
1525     /* for event-based notification, windows erases stale events */
1526     sock->pmask &= ~req->mask;
1527
1528     sock->hmask &= ~req->mask;
1529     sock->state |= req->sstate;
1530     sock->state &= ~req->cstate;
1531     if (sock->type != WS_SOCK_STREAM) sock->state &= ~STREAM_FLAG_MASK;
1532
1533     sock_reselect( sock );
1534
1535     release_object( &sock->obj );
1536 }
1537
1538 DECL_HANDLER(set_socket_deferred)
1539 {
1540     struct sock *sock, *acceptsock;
1541
1542     sock=(struct sock *)get_handle_obj( current->process, req->handle, FILE_WRITE_ATTRIBUTES, &sock_ops );
1543     if ( !sock )
1544         return;
1545
1546     acceptsock = (struct sock *)get_handle_obj( current->process, req->deferred, 0, &sock_ops );
1547     if ( !acceptsock )
1548     {
1549         release_object( sock );
1550         return;
1551     }
1552     sock->deferred = acceptsock;
1553     release_object( sock );
1554 }
1555
1556 DECL_HANDLER(get_socket_info)
1557 {
1558     struct sock *sock;
1559
1560     sock = (struct sock *)get_handle_obj( current->process, req->handle, FILE_READ_ATTRIBUTES, &sock_ops );
1561     if (!sock) return;
1562
1563     if (get_unix_fd( sock->fd ) == -1) return;
1564
1565     reply->family   = sock->family;
1566     reply->type     = sock->type;
1567     reply->protocol = sock->proto;
1568
1569     release_object( &sock->obj );
1570 }