fd.c 70 KB
Newer Older
1 2 3
/*
 * Server-side file descriptor management
 *
4
 * Copyright (C) 2000, 2003 Alexandre Julliard
5 6 7 8 9 10 11 12 13 14 15 16 17
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * 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 21 22
 */


#include "config.h"
23
#include "wine/port.h"
24 25

#include <assert.h>
26
#include <errno.h>
27
#include <fcntl.h>
28
#include <limits.h>
29
#include <signal.h>
30
#include <stdarg.h>
31 32 33
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
34 35 36
#ifdef HAVE_POLL_H
#include <poll.h>
#endif
37 38 39
#ifdef HAVE_SYS_POLL_H
#include <sys/poll.h>
#endif
40 41 42 43 44 45 46
#ifdef HAVE_LINUX_MAJOR_H
#include <linux/major.h>
#endif
#ifdef HAVE_SYS_STATVFS_H
#include <sys/statvfs.h>
#endif
#ifdef HAVE_SYS_VFS_H
47 48 49 50 51 52 53 54 55 56 57
/*
 * Solaris defines its system list in sys/list.h.
 * This need to be workaround it here.
 */
#define list SYSLIST
#define list_next SYSLIST_NEXT
#define list_prev SYSLIST_PREV
#define list_head SYSLIST_HEAD
#define list_tail SYSLIST_TAIL
#define list_move_tail SYSLIST_MOVE_TAIL
#define list_remove SYSLIST_REMOVE
58
#include <sys/vfs.h>
59 60 61 62 63 64 65
#undef list
#undef list_next
#undef list_prev
#undef list_head
#undef list_tail
#undef list_move_tail
#undef list_remove
66
#endif
67 68 69
#ifdef HAVE_SYS_PARAM_H
#include <sys/param.h>
#endif
70 71 72 73 74 75
#ifdef HAVE_SYS_MOUNT_H
#include <sys/mount.h>
#endif
#ifdef HAVE_SYS_STATFS_H
#include <sys/statfs.h>
#endif
76 77 78
#ifdef HAVE_SYS_SYSCTL_H
#include <sys/sysctl.h>
#endif
79 80 81 82 83
#ifdef HAVE_SYS_EVENT_H
#include <sys/event.h>
#undef LIST_INIT
#undef LIST_ENTRY
#endif
84 85 86
#ifdef HAVE_STDINT_H
#include <stdint.h>
#endif
87
#include <sys/stat.h>
88 89
#include <sys/time.h>
#include <sys/types.h>
90 91
#include <unistd.h>

92 93
#include "ntstatus.h"
#define WIN32_NO_STATUS
94 95 96 97 98 99
#include "object.h"
#include "file.h"
#include "handle.h"
#include "process.h"
#include "request.h"

100
#include "winternl.h"
101
#include "winioctl.h"
102

103
#if defined(HAVE_SYS_EPOLL_H) && defined(HAVE_EPOLL_CREATE)
104
# include <sys/epoll.h>
105
# define USE_EPOLL
106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131
#elif defined(linux) && defined(__i386__) && defined(HAVE_STDINT_H)
# define USE_EPOLL
# define EPOLLIN POLLIN
# define EPOLLOUT POLLOUT
# define EPOLLERR POLLERR
# define EPOLLHUP POLLHUP
# define EPOLL_CTL_ADD 1
# define EPOLL_CTL_DEL 2
# define EPOLL_CTL_MOD 3

typedef union epoll_data
{
  void *ptr;
  int fd;
  uint32_t u32;
  uint64_t u64;
} epoll_data_t;

struct epoll_event
{
  uint32_t events;
  epoll_data_t data;
};

static inline int epoll_create( int size )
{
132
    return syscall( 254 /*NR_epoll_create*/, size );
133 134 135 136
}

static inline int epoll_ctl( int epfd, int op, int fd, const struct epoll_event *event )
{
137
    return syscall( 255 /*NR_epoll_ctl*/, epfd, op, fd, event );
138 139 140 141
}

static inline int epoll_wait( int epfd, struct epoll_event *events, int maxevents, int timeout )
{
142
    return syscall( 256 /*NR_epoll_wait*/, epfd, events, maxevents, timeout );
143 144 145 146
}

#endif /* linux && __i386__ && HAVE_STDINT_H */

147 148 149 150
#if defined(HAVE_PORT_H) && defined(HAVE_PORT_CREATE)
# include <port.h>
# define USE_EVENT_PORTS
#endif /* HAVE_PORT_H && HAVE_PORT_CREATE */
151

152 153 154 155 156
/* Because of the stupid Posix locking semantics, we need to keep
 * track of all file descriptors referencing a given file, and not
 * close a single one until all the locks are gone (sigh).
 */

157 158 159 160 161
/* file descriptor object */

/* closed_fd is used to keep track of the unix fd belonging to a closed fd object */
struct closed_fd
{
162
    struct list entry;       /* entry in inode closed list */
163
    int         unix_fd;     /* the unix file descriptor */
164
    char        unlink[1];   /* name to unlink on close (if any) */
165 166
};

167 168
struct fd
{
169 170 171 172 173 174
    struct object        obj;         /* object header */
    const struct fd_ops *fd_ops;      /* file descriptor operations */
    struct inode        *inode;       /* inode that this fd belongs to */
    struct list          inode_entry; /* entry in inode fd list */
    struct closed_fd    *closed;      /* structure to store the unix fd at destroy time */
    struct object       *user;        /* object using this file descriptor */
175
    struct list          locks;       /* list of locks on this fd */
176
    unsigned int         access;      /* file access (FILE_READ_DATA etc.) */
177
    unsigned int         options;     /* file options (FILE_DELETE_ON_CLOSE, FILE_SYNCHRONOUS...) */
178
    unsigned int         sharing;     /* file sharing mode */
179
    char                *unix_name;   /* unix file name */
180
    int                  unix_fd;     /* unix file descriptor */
181
    unsigned int         no_fd_status;/* status to return when unix_fd is -1 */
182
    unsigned int         cacheable :1;/* can the fd be cached on the client side? */
183 184
    unsigned int         signaled :1; /* is the fd signaled? */
    unsigned int         fs_locks :1; /* can we use filesystem locks for this fd? */
185
    int                  poll_index;  /* index of fd in poll array */
186 187 188
    struct async_queue  *read_q;      /* async readers of this fd */
    struct async_queue  *write_q;     /* async writers of this fd */
    struct async_queue  *wait_q;      /* other async waiters of this fd */
189
    struct completion   *completion;  /* completion object attached to this fd */
190
    apc_param_t          comp_key;    /* completion key to set in completion events */
191 192 193 194 195 196 197 198 199
};

static void fd_dump( struct object *obj, int verbose );
static void fd_destroy( struct object *obj );

static const struct object_ops fd_ops =
{
    sizeof(struct fd),        /* size */
    fd_dump,                  /* dump */
200
    no_get_type,              /* get_type */
201 202 203 204
    no_add_queue,             /* add_queue */
    NULL,                     /* remove_queue */
    NULL,                     /* signaled */
    NULL,                     /* satisfied */
205
    no_signal,                /* signal */
206
    no_get_fd,                /* get_fd */
207
    no_map_access,            /* map_access */
208 209
    default_get_sd,           /* get_sd */
    default_set_sd,           /* set_sd */
210
    no_lookup_name,           /* lookup_name */
211
    no_open_file,             /* open_file */
212
    no_close_handle,          /* close_handle */
213 214 215
    fd_destroy                /* destroy */
};

216 217 218 219 220 221 222 223 224 225
/* device object */

#define DEVICE_HASH_SIZE 7
#define INODE_HASH_SIZE 17

struct device
{
    struct object       obj;        /* object header */
    struct list         entry;      /* entry in device hash list */
    dev_t               dev;        /* device number */
226
    int                 removable;  /* removable device? (or -1 if unknown) */
227 228 229 230 231 232 233 234 235 236
    struct list         inode_hash[INODE_HASH_SIZE];  /* inodes hash table */
};

static void device_dump( struct object *obj, int verbose );
static void device_destroy( struct object *obj );

static const struct object_ops device_ops =
{
    sizeof(struct device),    /* size */
    device_dump,              /* dump */
237
    no_get_type,              /* get_type */
238 239 240 241 242 243
    no_add_queue,             /* add_queue */
    NULL,                     /* remove_queue */
    NULL,                     /* signaled */
    NULL,                     /* satisfied */
    no_signal,                /* signal */
    no_get_fd,                /* get_fd */
244
    no_map_access,            /* map_access */
245 246
    default_get_sd,           /* get_sd */
    default_set_sd,           /* set_sd */
247
    no_lookup_name,           /* lookup_name */
248
    no_open_file,             /* open_file */
249 250 251 252
    no_close_handle,          /* close_handle */
    device_destroy            /* destroy */
};

253 254 255 256 257 258
/* inode object */

struct inode
{
    struct object       obj;        /* object header */
    struct list         entry;      /* inode hash list entry */
259
    struct device      *device;     /* device containing this inode */
260 261
    ino_t               ino;        /* inode number */
    struct list         open;       /* list of open file descriptors */
262
    struct list         locks;      /* list of file locks */
263
    struct list         closed;     /* list of file descriptors to close at destroy time */
264 265 266 267 268 269 270 271 272
};

static void inode_dump( struct object *obj, int verbose );
static void inode_destroy( struct object *obj );

static const struct object_ops inode_ops =
{
    sizeof(struct inode),     /* size */
    inode_dump,               /* dump */
273
    no_get_type,              /* get_type */
274 275 276 277
    no_add_queue,             /* add_queue */
    NULL,                     /* remove_queue */
    NULL,                     /* signaled */
    NULL,                     /* satisfied */
278
    no_signal,                /* signal */
279
    no_get_fd,                /* get_fd */
280
    no_map_access,            /* map_access */
281 282
    default_get_sd,           /* get_sd */
    default_set_sd,           /* set_sd */
283
    no_lookup_name,           /* lookup_name */
284
    no_open_file,             /* open_file */
285
    no_close_handle,          /* close_handle */
286 287 288
    inode_destroy             /* destroy */
};

289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
/* file lock object */

struct file_lock
{
    struct object       obj;         /* object header */
    struct fd          *fd;          /* fd owning this lock */
    struct list         fd_entry;    /* entry in list of locks on a given fd */
    struct list         inode_entry; /* entry in inode list of locks */
    int                 shared;      /* shared lock? */
    file_pos_t          start;       /* locked region is interval [start;end) */
    file_pos_t          end;
    struct process     *process;     /* process owning this lock */
    struct list         proc_entry;  /* entry in list of locks owned by the process */
};

static void file_lock_dump( struct object *obj, int verbose );
static int file_lock_signaled( struct object *obj, struct thread *thread );

static const struct object_ops file_lock_ops =
{
    sizeof(struct file_lock),   /* size */
    file_lock_dump,             /* dump */
311
    no_get_type,                /* get_type */
312 313 314 315
    add_queue,                  /* add_queue */
    remove_queue,               /* remove_queue */
    file_lock_signaled,         /* signaled */
    no_satisfied,               /* satisfied */
316
    no_signal,                  /* signal */
317
    no_get_fd,                  /* get_fd */
318
    no_map_access,              /* map_access */
319 320
    default_get_sd,             /* get_sd */
    default_set_sd,             /* set_sd */
321
    no_lookup_name,             /* lookup_name */
322
    no_open_file,               /* open_file */
323
    no_close_handle,            /* close_handle */
324 325 326 327 328 329 330 331 332
    no_destroy                  /* destroy */
};


#define OFF_T_MAX       (~((file_pos_t)1 << (8*sizeof(off_t)-1)))
#define FILE_POS_T_MAX  (~(file_pos_t)0)

static file_pos_t max_unix_offset = OFF_T_MAX;

333 334
#define DUMP_LONG_LONG(val) do { \
    if (sizeof(val) > sizeof(unsigned long) && (val) > ~0UL) \
335
        fprintf( stderr, "%lx%08lx", (unsigned long)((unsigned long long)(val) >> 32), (unsigned long)(val) ); \
336 337 338 339 340
    else \
        fprintf( stderr, "%lx", (unsigned long)(val) ); \
  } while (0)


341

342 343 344 345 346
/****************************************************************/
/* timeouts support */

struct timeout_user
{
347
    struct list           entry;      /* entry in sorted timeout list */
348
    timeout_t             when;       /* timeout expiry (absolute time) */
349 350 351 352
    timeout_callback      callback;   /* callback function */
    void                 *private;    /* callback private data */
};

353
static struct list timeout_list = LIST_INIT(timeout_list);   /* sorted timeouts list */
354 355 356 357 358 359 360 361 362
timeout_t current_time;

static inline void set_current_time(void)
{
    static const timeout_t ticks_1601_to_1970 = (timeout_t)86400 * (369 * 365 + 89) * TICKS_PER_SEC;
    struct timeval now;
    gettimeofday( &now, NULL );
    current_time = (timeout_t)now.tv_sec * TICKS_PER_SEC + now.tv_usec * 10 + ticks_1601_to_1970;
}
363 364

/* add a timeout user */
365
struct timeout_user *add_timeout_user( timeout_t when, timeout_callback func, void *private )
366 367
{
    struct timeout_user *user;
368
    struct list *ptr;
369 370

    if (!(user = mem_alloc( sizeof(*user) ))) return NULL;
371
    user->when     = (when > 0) ? when : current_time - when;
372 373 374 375 376
    user->callback = func;
    user->private  = private;

    /* Now insert it in the linked list */

377
    LIST_FOR_EACH( ptr, &timeout_list )
378
    {
379
        struct timeout_user *timeout = LIST_ENTRY( ptr, struct timeout_user, entry );
380
        if (timeout->when >= user->when) break;
381
    }
382
    list_add_before( ptr, &user->entry );
383 384 385 386 387 388
    return user;
}

/* remove a timeout user */
void remove_timeout_user( struct timeout_user *user )
{
389
    list_remove( &user->entry );
390 391 392
    free( user );
}

393 394
/* return a text description of a timeout for debugging purposes */
const char *get_timeout_str( timeout_t timeout )
395
{
396 397 398 399 400 401 402 403 404 405 406 407 408
    static char buffer[64];
    long secs, nsecs;

    if (!timeout) return "0";
    if (timeout == TIMEOUT_INFINITE) return "infinite";

    if (timeout < 0)  /* relative */
    {
        secs = -timeout / TICKS_PER_SEC;
        nsecs = -timeout % TICKS_PER_SEC;
        sprintf( buffer, "+%ld.%07ld", secs, nsecs );
    }
    else  /* absolute */
409
    {
410 411 412
        secs = (timeout - current_time) / TICKS_PER_SEC;
        nsecs = (timeout - current_time) % TICKS_PER_SEC;
        if (nsecs < 0)
413
        {
414 415
            nsecs += TICKS_PER_SEC;
            secs--;
416
        }
417 418 419 420 421 422 423
        if (secs >= 0)
            sprintf( buffer, "%x%08x (+%ld.%07ld)",
                     (unsigned int)(timeout >> 32), (unsigned int)timeout, secs, nsecs );
        else
            sprintf( buffer, "%x%08x (-%ld.%07ld)",
                     (unsigned int)(timeout >> 32), (unsigned int)timeout,
                     -(secs + 1), TICKS_PER_SEC - nsecs );
424
    }
425
    return buffer;
426 427 428 429 430 431 432 433 434 435 436 437 438
}


/****************************************************************/
/* poll support */

static struct fd **poll_users;              /* users array */
static struct pollfd *pollfd;               /* poll fd array */
static int nb_users;                        /* count of array entries actually in use */
static int active_users;                    /* current number of active users */
static int allocated_users;                 /* count of allocated entries in the array */
static struct fd **freelist;                /* list of free entries in the array */

439 440
static int get_next_timeout(void);

441 442 443 444 445
static inline void fd_poll_event( struct fd *fd, int event )
{
    fd->fd_ops->poll_event( fd, event );
}

446 447
#ifdef USE_EPOLL

448
static int epoll_fd = -1;
449 450 451 452 453

static inline void init_epoll(void)
{
    epoll_fd = epoll_create( 128 );
}
454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479

/* set the events that epoll waits for on this fd; helper for set_fd_events */
static inline void set_fd_epoll_events( struct fd *fd, int user, int events )
{
    struct epoll_event ev;
    int ctl;

    if (epoll_fd == -1) return;

    if (events == -1)  /* stop waiting on this fd completely */
    {
        if (pollfd[user].fd == -1) return;  /* already removed */
        ctl = EPOLL_CTL_DEL;
    }
    else if (pollfd[user].fd == -1)
    {
        if (pollfd[user].events) return;  /* stopped waiting on it, don't restart */
        ctl = EPOLL_CTL_ADD;
    }
    else
    {
        if (pollfd[user].events == events) return;  /* nothing to do */
        ctl = EPOLL_CTL_MOD;
    }

    ev.events = events;
480
    memset(&ev.data, 0, sizeof(ev.data));
481 482 483 484 485 486 487 488 489 490 491 492 493
    ev.data.u32 = user;

    if (epoll_ctl( epoll_fd, ctl, fd->unix_fd, &ev ) == -1)
    {
        if (errno == ENOMEM)  /* not enough memory, give up on epoll */
        {
            close( epoll_fd );
            epoll_fd = -1;
        }
        else perror( "epoll_ctl" );  /* should not happen */
    }
}

494 495 496
static inline void remove_epoll_user( struct fd *fd, int user )
{
    if (epoll_fd == -1) return;
497

498 499 500 501 502 503 504 505
    if (pollfd[user].fd != -1)
    {
        struct epoll_event dummy;
        epoll_ctl( epoll_fd, EPOLL_CTL_DEL, fd->unix_fd, &dummy );
    }
}

static inline void main_loop_epoll(void)
506
{
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
    int i, ret, timeout;
    struct epoll_event events[128];

    assert( POLLIN == EPOLLIN );
    assert( POLLOUT == EPOLLOUT );
    assert( POLLERR == EPOLLERR );
    assert( POLLHUP == EPOLLHUP );

    if (epoll_fd == -1) return;

    while (active_users)
    {
        timeout = get_next_timeout();

        if (!active_users) break;  /* last user removed by a timeout */
        if (epoll_fd == -1) break;  /* an error occurred with epoll */

        ret = epoll_wait( epoll_fd, events, sizeof(events)/sizeof(events[0]), timeout );
525
        set_current_time();
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540

        /* put the events into the pollfd array first, like poll does */
        for (i = 0; i < ret; i++)
        {
            int user = events[i].data.u32;
            pollfd[user].revents = events[i].events;
        }

        /* read events from the pollfd array, as set_fd_events may modify them */
        for (i = 0; i < ret; i++)
        {
            int user = events[i].data.u32;
            if (pollfd[user].revents) fd_poll_event( poll_users[user], pollfd[user].revents );
        }
    }
541 542
}

543 544 545 546 547 548
#elif defined(HAVE_KQUEUE)

static int kqueue_fd = -1;

static inline void init_epoll(void)
{
549 550 551 552 553 554 555 556 557
#ifdef __APPLE__ /* kqueue support is broken in Mac OS < 10.5 */
    int mib[2];
    char release[32];
    size_t len = sizeof(release);

    mib[0] = CTL_KERN;
    mib[1] = KERN_OSRELEASE;
    if (sysctl( mib, 2, release, &len, NULL, 0 ) == -1) return;
    if (atoi(release) < 9) return;
558
#endif
559
    kqueue_fd = kqueue();
560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638
}

static inline void set_fd_epoll_events( struct fd *fd, int user, int events )
{
    struct kevent ev[2];

    if (kqueue_fd == -1) return;

    EV_SET( &ev[0], fd->unix_fd, EVFILT_READ, 0, NOTE_LOWAT, 1, (void *)user );
    EV_SET( &ev[1], fd->unix_fd, EVFILT_WRITE, 0, NOTE_LOWAT, 1, (void *)user );

    if (events == -1)  /* stop waiting on this fd completely */
    {
        if (pollfd[user].fd == -1) return;  /* already removed */
        ev[0].flags |= EV_DELETE;
        ev[1].flags |= EV_DELETE;
    }
    else if (pollfd[user].fd == -1)
    {
        if (pollfd[user].events) return;  /* stopped waiting on it, don't restart */
        ev[0].flags |= EV_ADD | ((events & POLLIN) ? EV_ENABLE : EV_DISABLE);
        ev[1].flags |= EV_ADD | ((events & POLLOUT) ? EV_ENABLE : EV_DISABLE);
    }
    else
    {
        if (pollfd[user].events == events) return;  /* nothing to do */
        ev[0].flags |= (events & POLLIN) ? EV_ENABLE : EV_DISABLE;
        ev[1].flags |= (events & POLLOUT) ? EV_ENABLE : EV_DISABLE;
    }

    if (kevent( kqueue_fd, ev, 2, NULL, 0, NULL ) == -1)
    {
        if (errno == ENOMEM)  /* not enough memory, give up on kqueue */
        {
            close( kqueue_fd );
            kqueue_fd = -1;
        }
        else perror( "kevent" );  /* should not happen */
    }
}

static inline void remove_epoll_user( struct fd *fd, int user )
{
    if (kqueue_fd == -1) return;

    if (pollfd[user].fd != -1)
    {
        struct kevent ev[2];

        EV_SET( &ev[0], fd->unix_fd, EVFILT_READ, EV_DELETE, 0, 0, 0 );
        EV_SET( &ev[1], fd->unix_fd, EVFILT_WRITE, EV_DELETE, 0, 0, 0 );
        kevent( kqueue_fd, ev, 2, NULL, 0, NULL );
    }
}

static inline void main_loop_epoll(void)
{
    int i, ret, timeout;
    struct kevent events[128];

    if (kqueue_fd == -1) return;

    while (active_users)
    {
        timeout = get_next_timeout();

        if (!active_users) break;  /* last user removed by a timeout */
        if (kqueue_fd == -1) break;  /* an error occurred with kqueue */

        if (timeout != -1)
        {
            struct timespec ts;

            ts.tv_sec = timeout / 1000;
            ts.tv_nsec = (timeout % 1000) * 1000000;
            ret = kevent( kqueue_fd, NULL, 0, events, sizeof(events)/sizeof(events[0]), &ts );
        }
        else ret = kevent( kqueue_fd, NULL, 0, events, sizeof(events)/sizeof(events[0]), NULL );

639
        set_current_time();
640

641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665
        /* put the events into the pollfd array first, like poll does */
        for (i = 0; i < ret; i++)
        {
            long user = (long)events[i].udata;
            pollfd[user].revents = 0;
        }
        for (i = 0; i < ret; i++)
        {
            long user = (long)events[i].udata;
            if (events[i].filter == EVFILT_READ) pollfd[user].revents |= POLLIN;
            else if (events[i].filter == EVFILT_WRITE) pollfd[user].revents |= POLLOUT;
            if (events[i].flags & EV_EOF) pollfd[user].revents |= POLLHUP;
            if (events[i].flags & EV_ERROR) pollfd[user].revents |= POLLERR;
        }

        /* read events from the pollfd array, as set_fd_events may modify them */
        for (i = 0; i < ret; i++)
        {
            long user = (long)events[i].udata;
            if (pollfd[user].revents) fd_poll_event( poll_users[user], pollfd[user].revents );
            pollfd[user].revents = 0;
        }
    }
}

666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766
#elif defined(USE_EVENT_PORTS)

static int port_fd = -1;

static inline void init_epoll(void)
{
    port_fd = port_create();
}

static inline void set_fd_epoll_events( struct fd *fd, int user, int events )
{
    int ret;

    if (port_fd == -1) return;

    if (events == -1)  /* stop waiting on this fd completely */
    {
        if (pollfd[user].fd == -1) return;  /* already removed */
        port_dissociate( port_fd, PORT_SOURCE_FD, fd->unix_fd );
    }
    else if (pollfd[user].fd == -1)
    {
        if (pollfd[user].events) return;  /* stopped waiting on it, don't restart */
        ret = port_associate( port_fd, PORT_SOURCE_FD, fd->unix_fd, events, (void *)user );
    }
    else
    {
        if (pollfd[user].events == events) return;  /* nothing to do */
        ret = port_associate( port_fd, PORT_SOURCE_FD, fd->unix_fd, events, (void *)user );
    }

    if (ret == -1)
    {
        if (errno == ENOMEM)  /* not enough memory, give up on port_associate */
        {
            close( port_fd );
            port_fd = -1;
        }
        else perror( "port_associate" );  /* should not happen */
    }
}

static inline void remove_epoll_user( struct fd *fd, int user )
{
    if (port_fd == -1) return;

    if (pollfd[user].fd != -1)
    {
        port_dissociate( port_fd, PORT_SOURCE_FD, fd->unix_fd );
    }
}

static inline void main_loop_epoll(void)
{
    int i, nget, ret, timeout;
    port_event_t events[128];

    if (port_fd == -1) return;

    while (active_users)
    {
        timeout = get_next_timeout();
        nget = 1;

        if (!active_users) break;  /* last user removed by a timeout */
        if (port_fd == -1) break;  /* an error occurred with event completion */

        if (timeout != -1)
        {
            struct timespec ts;

            ts.tv_sec = timeout / 1000;
            ts.tv_nsec = (timeout % 1000) * 1000000;
            ret = port_getn( port_fd, events, sizeof(events)/sizeof(events[0]), &nget, &ts );
        }
        else ret = port_getn( port_fd, events, sizeof(events)/sizeof(events[0]), &nget, NULL );

	if (ret == -1) break;  /* an error occurred with event completion */

        set_current_time();

        /* put the events into the pollfd array first, like poll does */
        for (i = 0; i < nget; i++)
        {
            long user = (long)events[i].portev_user;
            pollfd[user].revents = events[i].portev_events;
        }

        /* read events from the pollfd array, as set_fd_events may modify them */
        for (i = 0; i < nget; i++)
        {
            long user = (long)events[i].portev_user;
            if (pollfd[user].revents) fd_poll_event( poll_users[user], pollfd[user].revents );
            /* if we are still interested, reassociate the fd */
            if (pollfd[user].fd != -1) {
                port_associate( port_fd, PORT_SOURCE_FD, pollfd[user].fd, pollfd[user].events, (void *)user );
            }
        }
    }
}

767
#else /* HAVE_KQUEUE */
768 769 770 771 772 773

static inline void init_epoll(void) { }
static inline void set_fd_epoll_events( struct fd *fd, int user, int events ) { }
static inline void remove_epoll_user( struct fd *fd, int user ) { }
static inline void main_loop_epoll(void) { }

774 775 776
#endif /* USE_EPOLL */


777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803
/* add a user in the poll array and return its index, or -1 on failure */
static int add_poll_user( struct fd *fd )
{
    int ret;
    if (freelist)
    {
        ret = freelist - poll_users;
        freelist = (struct fd **)poll_users[ret];
    }
    else
    {
        if (nb_users == allocated_users)
        {
            struct fd **newusers;
            struct pollfd *newpoll;
            int new_count = allocated_users ? (allocated_users + allocated_users / 2) : 16;
            if (!(newusers = realloc( poll_users, new_count * sizeof(*poll_users) ))) return -1;
            if (!(newpoll = realloc( pollfd, new_count * sizeof(*pollfd) )))
            {
                if (allocated_users)
                    poll_users = newusers;
                else
                    free( newusers );
                return -1;
            }
            poll_users = newusers;
            pollfd = newpoll;
804
            if (!allocated_users) init_epoll();
805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821
            allocated_users = new_count;
        }
        ret = nb_users++;
    }
    pollfd[ret].fd = -1;
    pollfd[ret].events = 0;
    pollfd[ret].revents = 0;
    poll_users[ret] = fd;
    active_users++;
    return ret;
}

/* remove a user from the poll list */
static void remove_poll_user( struct fd *fd, int user )
{
    assert( user >= 0 );
    assert( poll_users[user] == fd );
822

823
    remove_epoll_user( fd, user );
824 825 826 827 828 829 830 831
    pollfd[user].fd = -1;
    pollfd[user].events = 0;
    pollfd[user].revents = 0;
    poll_users[user] = (struct fd *)freelist;
    freelist = &poll_users[user];
    active_users--;
}

832 833
/* process pending timeouts and return the time until the next timeout, in milliseconds */
static int get_next_timeout(void)
834
{
835
    if (!list_empty( &timeout_list ))
836
    {
837
        struct list expired_list, *ptr;
838

839
        /* first remove all expired timers from the list */
840

841 842 843 844
        list_init( &expired_list );
        while ((ptr = list_head( &timeout_list )) != NULL)
        {
            struct timeout_user *timeout = LIST_ENTRY( ptr, struct timeout_user, entry );
845

846
            if (timeout->when <= current_time)
847
            {
848
                list_remove( &timeout->entry );
849
                list_add_tail( &expired_list, &timeout->entry );
850
            }
851 852
            else break;
        }
853

854
        /* now call the callback for all the removed timers */
855

856 857 858 859 860 861 862 863 864 865 866
        while ((ptr = list_head( &expired_list )) != NULL)
        {
            struct timeout_user *timeout = LIST_ENTRY( ptr, struct timeout_user, entry );
            list_remove( &timeout->entry );
            timeout->callback( timeout->private );
            free( timeout );
        }

        if ((ptr = list_head( &timeout_list )) != NULL)
        {
            struct timeout_user *timeout = LIST_ENTRY( ptr, struct timeout_user, entry );
867
            int diff = (timeout->when - current_time + 9999) / 10000;
868 869
            if (diff < 0) diff = 0;
            return diff;
870
        }
871 872 873 874 875 876 877 878 879
    }
    return -1;  /* no pending timeouts */
}

/* server main poll() loop */
void main_loop(void)
{
    int i, ret, timeout;

880 881
    set_current_time();
    server_start_time = current_time;
882

883
    main_loop_epoll();
884 885
    /* fall through to normal poll loop */

886 887 888 889 890
    while (active_users)
    {
        timeout = get_next_timeout();

        if (!active_users) break;  /* last user removed by a timeout */
891

892
        ret = poll( pollfd, nb_users, timeout );
893
        set_current_time();
894

895 896 897 898 899 900 901 902 903 904 905 906 907 908
        if (ret > 0)
        {
            for (i = 0; i < nb_users; i++)
            {
                if (pollfd[i].revents)
                {
                    fd_poll_event( poll_users[i], pollfd[i].revents );
                    if (!--ret) break;
                }
            }
        }
    }
}

909 910

/****************************************************************/
911 912 913 914
/* device functions */

static struct list device_hash[DEVICE_HASH_SIZE];

915 916 917 918 919 920 921 922 923 924 925 926
static int is_device_removable( dev_t dev, int unix_fd )
{
#if defined(linux) && defined(HAVE_FSTATFS)
    struct statfs stfs;

    /* check for floppy disk */
    if (major(dev) == FLOPPY_MAJOR) return 1;

    if (fstatfs( unix_fd, &stfs ) == -1) return 0;
    return (stfs.f_type == 0x9660 ||    /* iso9660 */
            stfs.f_type == 0x9fa1 ||    /* supermount */
            stfs.f_type == 0x15013346); /* udf */
927
#elif defined(__FreeBSD__) || defined(__FreeBSD_kernel__) || defined(__DragonFly__) || defined(__APPLE__)
928 929 930
    struct statfs stfs;

    if (fstatfs( unix_fd, &stfs ) == -1) return 0;
931
    return (!strcmp("cd9660", stfs.f_fstypename) || !strcmp("udf", stfs.f_fstypename));
932 933 934 935
#elif defined(__NetBSD__)
    struct statvfs stfs;

    if (fstatvfs( unix_fd, &stfs ) == -1) return 0;
936
    return (!strcmp("cd9660", stfs.f_fstypename) || !strcmp("udf", stfs.f_fstypename));
937 938 939 940 941 942 943 944 945 946 947 948 949 950 951
#elif defined(sun)
# include <sys/dkio.h>
# include <sys/vtoc.h>
    struct dk_cinfo dkinf;
    if (ioctl( unix_fd, DKIOCINFO, &dkinf ) == -1) return 0;
    return (dkinf.dki_ctype == DKC_CDROM ||
            dkinf.dki_ctype == DKC_NCRFLOPPY ||
            dkinf.dki_ctype == DKC_SMSFLOPPY ||
            dkinf.dki_ctype == DKC_INTEL82072 ||
            dkinf.dki_ctype == DKC_INTEL82077);
#else
    return 0;
#endif
}

952
/* retrieve the device object for a given fd, creating it if needed */
953
static struct device *get_device( dev_t dev, int unix_fd )
954 955 956
{
    struct device *device;
    unsigned int i, hash = dev % DEVICE_HASH_SIZE;
957

958 959 960 961 962 963 964 965
    if (device_hash[hash].next)
    {
        LIST_FOR_EACH_ENTRY( device, &device_hash[hash], struct device, entry )
            if (device->dev == dev) return (struct device *)grab_object( device );
    }
    else list_init( &device_hash[hash] );

    /* not found, create it */
966

967
    if (unix_fd == -1) return NULL;
968 969 970
    if ((device = alloc_object( &device_ops )))
    {
        device->dev = dev;
971
        device->removable = is_device_removable( dev, unix_fd );
972 973 974 975 976 977 978 979 980 981 982 983 984
        for (i = 0; i < INODE_HASH_SIZE; i++) list_init( &device->inode_hash[i] );
        list_add_head( &device_hash[hash], &device->entry );
    }
    return device;
}

static void device_dump( struct object *obj, int verbose )
{
    struct device *device = (struct device *)obj;
    fprintf( stderr, "Device dev=" );
    DUMP_LONG_LONG( device->dev );
    fprintf( stderr, "\n" );
}
985

986 987 988 989 990 991 992 993 994 995 996 997 998 999
static void device_destroy( struct object *obj )
{
    struct device *device = (struct device *)obj;
    unsigned int i;

    for (i = 0; i < INODE_HASH_SIZE; i++)
        assert( list_empty(&device->inode_hash[i]) );

    list_remove( &device->entry );  /* remove it from the hash table */
}


/****************************************************************/
/* inode functions */
1000

1001
/* close all pending file descriptors in the closed list */
1002
static void inode_close_pending( struct inode *inode, int keep_unlinks )
1003
{
1004 1005 1006
    struct list *ptr = list_head( &inode->closed );

    while (ptr)
1007
    {
1008 1009 1010
        struct closed_fd *fd = LIST_ENTRY( ptr, struct closed_fd, entry );
        struct list *next = list_next( &inode->closed, ptr );

1011
        if (fd->unix_fd != -1)
1012
        {
1013 1014
            close( fd->unix_fd );
            fd->unix_fd = -1;
1015
        }
1016
        if (!keep_unlinks || !fd->unlink[0])  /* get rid of it unless there's an unlink pending on that file */
1017 1018 1019 1020 1021
        {
            list_remove( ptr );
            free( fd );
        }
        ptr = next;
1022 1023 1024
    }
}

1025 1026 1027
static void inode_dump( struct object *obj, int verbose )
{
    struct inode *inode = (struct inode *)obj;
1028
    fprintf( stderr, "Inode device=%p ino=", inode->device );
1029 1030 1031 1032 1033 1034 1035
    DUMP_LONG_LONG( inode->ino );
    fprintf( stderr, "\n" );
}

static void inode_destroy( struct object *obj )
{
    struct inode *inode = (struct inode *)obj;
1036
    struct list *ptr;
1037

1038 1039
    assert( list_empty(&inode->open) );
    assert( list_empty(&inode->locks) );
1040 1041

    list_remove( &inode->entry );
1042 1043 1044 1045 1046

    while ((ptr = list_head( &inode->closed )))
    {
        struct closed_fd *fd = LIST_ENTRY( ptr, struct closed_fd, entry );
        list_remove( ptr );
1047
        if (fd->unix_fd != -1) close( fd->unix_fd );
1048 1049 1050 1051
        if (fd->unlink[0])
        {
            /* make sure it is still the same file */
            struct stat st;
1052
            if (!stat( fd->unlink, &st ) && st.st_dev == inode->device->dev && st.st_ino == inode->ino)
1053 1054 1055 1056
            {
                if (S_ISDIR(st.st_mode)) rmdir( fd->unlink );
                else unlink( fd->unlink );
            }
1057 1058 1059
        }
        free( fd );
    }
1060
    release_object( inode->device );
1061 1062 1063
}

/* retrieve the inode object for a given fd, creating it if needed */
1064
static struct inode *get_inode( dev_t dev, ino_t ino, int unix_fd )
1065
{
1066
    struct device *device;
1067
    struct inode *inode;
1068
    unsigned int hash = ino % INODE_HASH_SIZE;
1069

1070
    if (!(device = get_device( dev, unix_fd ))) return NULL;
1071 1072

    LIST_FOR_EACH_ENTRY( inode, &device->inode_hash[hash], struct inode, entry )
1073
    {
1074
        if (inode->ino == ino)
1075
        {
1076 1077
            release_object( device );
            return (struct inode *)grab_object( inode );
1078 1079 1080 1081 1082 1083
        }
    }

    /* not found, create it */
    if ((inode = alloc_object( &inode_ops )))
    {
1084
        inode->device = device;
1085 1086
        inode->ino    = ino;
        list_init( &inode->open );
1087
        list_init( &inode->locks );
1088
        list_init( &inode->closed );
1089
        list_add_head( &device->inode_hash[hash], &inode->entry );
1090
    }
1091 1092
    else release_object( device );

1093 1094 1095
    return inode;
}

1096
/* add fd to the inode list of file descriptors to close */
1097 1098
static void inode_add_closed_fd( struct inode *inode, struct closed_fd *fd )
{
1099 1100
    if (!list_empty( &inode->locks ))
    {
1101
        list_add_head( &inode->closed, &fd->entry );
1102
    }
1103 1104
    else if (fd->unlink[0])  /* close the fd but keep the structure around for unlink */
    {
1105 1106
        if (fd->unix_fd != -1) close( fd->unix_fd );
        fd->unix_fd = -1;
1107 1108 1109
        list_add_head( &inode->closed, &fd->entry );
    }
    else  /* no locks on this inode and no unlink, get rid of the fd */
1110
    {
1111
        if (fd->unix_fd != -1) close( fd->unix_fd );
1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
        free( fd );
    }
}


/****************************************************************/
/* file lock functions */

static void file_lock_dump( struct object *obj, int verbose )
{
    struct file_lock *lock = (struct file_lock *)obj;
    fprintf( stderr, "Lock %s fd=%p proc=%p start=",
             lock->shared ? "shared" : "excl", lock->fd, lock->process );
    DUMP_LONG_LONG( lock->start );
    fprintf( stderr, " end=" );
    DUMP_LONG_LONG( lock->end );
    fprintf( stderr, "\n" );
}

static int file_lock_signaled( struct object *obj, struct thread *thread )
{
    struct file_lock *lock = (struct file_lock *)obj;
    /* lock is signaled if it has lost its owner */
    return !lock->process;
}

/* set (or remove) a Unix lock if possible for the given range */
1139
static int set_unix_lock( struct fd *fd, file_pos_t start, file_pos_t end, int type )
1140 1141 1142
{
    struct flock fl;

1143
    if (!fd->fs_locks) return 1;  /* no fs locks possible for this fd */
1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156
    for (;;)
    {
        if (start == end) return 1;  /* can't set zero-byte lock */
        if (start > max_unix_offset) return 1;  /* ignore it */
        fl.l_type   = type;
        fl.l_whence = SEEK_SET;
        fl.l_start  = start;
        if (!end || end > max_unix_offset) fl.l_len = 0;
        else fl.l_len = end - start;
        if (fcntl( fd->unix_fd, F_SETLK, &fl ) != -1) return 1;

        switch(errno)
        {
1157 1158 1159 1160 1161 1162 1163 1164
        case EACCES:
            /* check whether locks work at all on this file system */
            if (fcntl( fd->unix_fd, F_GETLK, &fl ) != -1)
            {
                set_error( STATUS_FILE_LOCK_CONFLICT );
                return 0;
            }
            /* fall through */
1165 1166 1167
        case EIO:
        case ENOLCK:
            /* no locking on this fs, just ignore it */
1168
            fd->fs_locks = 0;
1169
            return 1;
1170 1171 1172
        case EAGAIN:
            set_error( STATUS_FILE_LOCK_CONFLICT );
            return 0;
1173 1174 1175 1176 1177 1178
        case EBADF:
            /* this can happen if we try to set a write lock on a read-only file */
            /* we just ignore that error */
            if (fl.l_type == F_WRLCK) return 1;
            set_error( STATUS_ACCESS_DENIED );
            return 0;
1179
#ifdef EOVERFLOW
1180
        case EOVERFLOW:
1181
#endif
1182
        case EINVAL:
1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
            /* this can happen if off_t is 64-bit but the kernel only supports 32-bit */
            /* in that case we shrink the limit and retry */
            if (max_unix_offset > INT_MAX)
            {
                max_unix_offset = INT_MAX;
                break;  /* retry */
            }
            /* fall through */
        default:
            file_set_error();
            return 0;
        }
    }
}

/* check if interval [start;end) overlaps the lock */
1199
static inline int lock_overlaps( struct file_lock *lock, file_pos_t start, file_pos_t end )
1200 1201 1202 1203 1204 1205 1206
{
    if (lock->end && start >= lock->end) return 0;
    if (end && lock->start >= end) return 0;
    return 1;
}

/* remove Unix locks for all bytes in the specified area that are no longer locked */
1207
static void remove_unix_locks( struct fd *fd, file_pos_t start, file_pos_t end )
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
{
    struct hole
    {
        struct hole *next;
        struct hole *prev;
        file_pos_t   start;
        file_pos_t   end;
    } *first, *cur, *next, *buffer;

    struct list *ptr;
    int count = 0;

    if (!fd->inode) return;
1221
    if (!fd->fs_locks) return;
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323
    if (start == end || start > max_unix_offset) return;
    if (!end || end > max_unix_offset) end = max_unix_offset + 1;

    /* count the number of locks overlapping the specified area */

    LIST_FOR_EACH( ptr, &fd->inode->locks )
    {
        struct file_lock *lock = LIST_ENTRY( ptr, struct file_lock, inode_entry );
        if (lock->start == lock->end) continue;
        if (lock_overlaps( lock, start, end )) count++;
    }

    if (!count)  /* no locks at all, we can unlock everything */
    {
        set_unix_lock( fd, start, end, F_UNLCK );
        return;
    }

    /* allocate space for the list of holes */
    /* max. number of holes is number of locks + 1 */

    if (!(buffer = malloc( sizeof(*buffer) * (count+1) ))) return;
    first = buffer;
    first->next  = NULL;
    first->prev  = NULL;
    first->start = start;
    first->end   = end;
    next = first + 1;

    /* build a sorted list of unlocked holes in the specified area */

    LIST_FOR_EACH( ptr, &fd->inode->locks )
    {
        struct file_lock *lock = LIST_ENTRY( ptr, struct file_lock, inode_entry );
        if (lock->start == lock->end) continue;
        if (!lock_overlaps( lock, start, end )) continue;

        /* go through all the holes touched by this lock */
        for (cur = first; cur; cur = cur->next)
        {
            if (cur->end <= lock->start) continue; /* hole is before start of lock */
            if (lock->end && cur->start >= lock->end) break;  /* hole is after end of lock */

            /* now we know that lock is overlapping hole */

            if (cur->start >= lock->start)  /* lock starts before hole, shrink from start */
            {
                cur->start = lock->end;
                if (cur->start && cur->start < cur->end) break;  /* done with this lock */
                /* now hole is empty, remove it */
                if (cur->next) cur->next->prev = cur->prev;
                if (cur->prev) cur->prev->next = cur->next;
                else if (!(first = cur->next)) goto done;  /* no more holes at all */
            }
            else if (!lock->end || cur->end <= lock->end)  /* lock larger than hole, shrink from end */
            {
                cur->end = lock->start;
                assert( cur->start < cur->end );
            }
            else  /* lock is in the middle of hole, split hole in two */
            {
                next->prev = cur;
                next->next = cur->next;
                cur->next = next;
                next->start = lock->end;
                next->end = cur->end;
                cur->end = lock->start;
                assert( next->start < next->end );
                assert( cur->end < next->start );
                next++;
                break;  /* done with this lock */
            }
        }
    }

    /* clear Unix locks for all the holes */

    for (cur = first; cur; cur = cur->next)
        set_unix_lock( fd, cur->start, cur->end, F_UNLCK );

 done:
    free( buffer );
}

/* create a new lock on a fd */
static struct file_lock *add_lock( struct fd *fd, int shared, file_pos_t start, file_pos_t end )
{
    struct file_lock *lock;

    if (!(lock = alloc_object( &file_lock_ops ))) return NULL;
    lock->shared  = shared;
    lock->start   = start;
    lock->end     = end;
    lock->fd      = fd;
    lock->process = current->process;

    /* now try to set a Unix lock */
    if (!set_unix_lock( lock->fd, lock->start, lock->end, lock->shared ? F_RDLCK : F_WRLCK ))
    {
        release_object( lock );
        return NULL;
    }
1324 1325 1326
    list_add_tail( &fd->locks, &lock->fd_entry );
    list_add_tail( &fd->inode->locks, &lock->inode_entry );
    list_add_tail( &lock->process->locks, &lock->proc_entry );
1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
    return lock;
}

/* remove an existing lock */
static void remove_lock( struct file_lock *lock, int remove_unix )
{
    struct inode *inode = lock->fd->inode;

    list_remove( &lock->fd_entry );
    list_remove( &lock->inode_entry );
    list_remove( &lock->proc_entry );
    if (remove_unix) remove_unix_locks( lock->fd, lock->start, lock->end );
1339
    if (list_empty( &inode->locks )) inode_close_pending( inode, 1 );
1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
    lock->process = NULL;
    wake_up( &lock->obj, 0 );
    release_object( lock );
}

/* remove all locks owned by a given process */
void remove_process_locks( struct process *process )
{
    struct list *ptr;

    while ((ptr = list_head( &process->locks )))
    {
        struct file_lock *lock = LIST_ENTRY( ptr, struct file_lock, proc_entry );
        remove_lock( lock, 1 );  /* this removes it from the list */
    }
}

/* remove all locks on a given fd */
static void remove_fd_locks( struct fd *fd )
{
    file_pos_t start = FILE_POS_T_MAX, end = 0;
    struct list *ptr;

    while ((ptr = list_head( &fd->locks )))
    {
        struct file_lock *lock = LIST_ENTRY( ptr, struct file_lock, fd_entry );
        if (lock->start < start) start = lock->start;
        if (!lock->end || lock->end > end) end = lock->end - 1;
        remove_lock( lock, 0 );
    }
    if (start < end) remove_unix_locks( fd, start, end + 1 );
}

/* add a lock on an fd */
/* returns handle to wait on */
obj_handle_t lock_fd( struct fd *fd, file_pos_t start, file_pos_t count, int shared, int wait )
{
    struct list *ptr;
    file_pos_t end = start + count;

1380 1381 1382 1383 1384 1385
    if (!fd->inode)  /* not a regular file */
    {
        set_error( STATUS_INVALID_DEVICE_REQUEST );
        return 0;
    }

1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
    /* don't allow wrapping locks */
    if (end && end < start)
    {
        set_error( STATUS_INVALID_PARAMETER );
        return 0;
    }

    /* check if another lock on that file overlaps the area */
    LIST_FOR_EACH( ptr, &fd->inode->locks )
    {
        struct file_lock *lock = LIST_ENTRY( ptr, struct file_lock, inode_entry );
        if (!lock_overlaps( lock, start, end )) continue;
1398
        if (shared && (lock->shared || lock->fd == fd)) continue;
1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
        /* found one */
        if (!wait)
        {
            set_error( STATUS_FILE_LOCK_CONFLICT );
            return 0;
        }
        set_error( STATUS_PENDING );
        return alloc_handle( current->process, lock, SYNCHRONIZE, 0 );
    }

    /* not found, add it */
    if (add_lock( fd, shared, start, end )) return 0;
    if (get_error() == STATUS_FILE_LOCK_CONFLICT)
    {
        /* Unix lock conflict -> tell client to wait and retry */
        if (wait) set_error( STATUS_PENDING );
    }
    return 0;
}

/* remove a lock on an fd */
void unlock_fd( struct fd *fd, file_pos_t start, file_pos_t count )
{
    struct list *ptr;
    file_pos_t end = start + count;

    /* find an existing lock with the exact same parameters */
    LIST_FOR_EACH( ptr, &fd->locks )
    {
        struct file_lock *lock = LIST_ENTRY( ptr, struct file_lock, fd_entry );
        if ((lock->start == start) && (lock->end == end))
        {
            remove_lock( lock, 1 );
            return;
        }
    }
    set_error( STATUS_FILE_LOCK_CONFLICT );
1436 1437 1438
}


1439 1440 1441
/****************************************************************/
/* file descriptor functions */

1442 1443 1444
static void fd_dump( struct object *obj, int verbose )
{
    struct fd *fd = (struct fd *)obj;
1445
    fprintf( stderr, "Fd unix_fd=%d user=%p options=%08x", fd->unix_fd, fd->user, fd->options );
1446 1447
    if (fd->inode) fprintf( stderr, " inode=%p unlink='%s'", fd->inode, fd->closed->unlink );
    fprintf( stderr, "\n" );
1448 1449 1450 1451 1452
}

static void fd_destroy( struct object *obj )
{
    struct fd *fd = (struct fd *)obj;
1453

1454 1455 1456
    free_async_queue( fd->read_q );
    free_async_queue( fd->write_q );
    free_async_queue( fd->wait_q );
1457

1458
    if (fd->completion) release_object( fd->completion );
1459
    remove_fd_locks( fd );
1460
    free( fd->unix_name );
1461
    list_remove( &fd->inode_entry );
1462
    if (fd->poll_index != -1) remove_poll_user( fd, fd->poll_index );
1463 1464 1465 1466 1467 1468 1469 1470 1471
    if (fd->inode)
    {
        inode_add_closed_fd( fd->inode, fd->closed );
        release_object( fd->inode );
    }
    else  /* no inode, close it right away */
    {
        if (fd->unix_fd != -1) close( fd->unix_fd );
    }
1472 1473
}

1474 1475 1476 1477 1478
/* check if the desired access is possible without violating */
/* the sharing mode of other opens of the same file */
static unsigned int check_sharing( struct fd *fd, unsigned int access, unsigned int sharing,
                                   unsigned int open_flags, unsigned int options )
{
1479 1480 1481 1482 1483
    /* only a few access bits are meaningful wrt sharing */
    const unsigned int read_access = FILE_READ_DATA | FILE_EXECUTE;
    const unsigned int write_access = FILE_WRITE_DATA | FILE_APPEND_DATA;
    const unsigned int all_access = read_access | write_access | DELETE;

1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496
    unsigned int existing_sharing = FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE;
    unsigned int existing_access = 0;
    struct list *ptr;

    fd->access = access;
    fd->sharing = sharing;

    LIST_FOR_EACH( ptr, &fd->inode->open )
    {
        struct fd *fd_ptr = LIST_ENTRY( ptr, struct fd, inode_entry );
        if (fd_ptr != fd)
        {
            /* if access mode is 0, sharing mode is ignored */
1497 1498
            if (fd_ptr->access & all_access) existing_sharing &= fd_ptr->sharing;
            existing_access |= fd_ptr->access;
1499 1500 1501
        }
    }

1502 1503
    if (((access & read_access) && !(existing_sharing & FILE_SHARE_READ)) ||
        ((access & write_access) && !(existing_sharing & FILE_SHARE_WRITE)) ||
1504 1505 1506
        ((access & DELETE) && !(existing_sharing & FILE_SHARE_DELETE)))
        return STATUS_SHARING_VIOLATION;
    if (((existing_access & FILE_MAPPING_WRITE) && !(sharing & FILE_SHARE_WRITE)) ||
1507
        ((existing_access & FILE_MAPPING_IMAGE) && (access & FILE_WRITE_DATA)))
1508 1509 1510 1511 1512
        return STATUS_SHARING_VIOLATION;
    if ((existing_access & FILE_MAPPING_IMAGE) && (options & FILE_DELETE_ON_CLOSE))
        return STATUS_CANNOT_DELETE;
    if ((existing_access & FILE_MAPPING_ACCESS) && (open_flags & O_TRUNC))
        return STATUS_USER_MAPPED_FILE;
1513 1514 1515 1516
    if (!(access & all_access))
        return 0;  /* if access mode is 0, sharing mode is ignored (except for mappings) */
    if (((existing_access & read_access) && !(sharing & FILE_SHARE_READ)) ||
        ((existing_access & write_access) && !(sharing & FILE_SHARE_WRITE)) ||
1517 1518 1519 1520 1521
        ((existing_access & DELETE) && !(sharing & FILE_SHARE_DELETE)))
        return STATUS_SHARING_VIOLATION;
    return 0;
}

1522 1523 1524 1525 1526
/* set the events that select waits for on this fd */
void set_fd_events( struct fd *fd, int events )
{
    int user = fd->poll_index;
    assert( poll_users[user] == fd );
1527 1528 1529

    set_fd_epoll_events( fd, user, events );

1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542
    if (events == -1)  /* stop waiting on this fd completely */
    {
        pollfd[user].fd = -1;
        pollfd[user].events = POLLERR;
        pollfd[user].revents = 0;
    }
    else if (pollfd[user].fd != -1 || !pollfd[user].events)
    {
        pollfd[user].fd = fd->unix_fd;
        pollfd[user].events = events;
    }
}

1543 1544 1545 1546 1547
/* prepare an fd for unmounting its corresponding device */
static inline void unmount_fd( struct fd *fd )
{
    assert( fd->inode );

1548 1549
    async_wake_up( fd->read_q, STATUS_VOLUME_DISMOUNTED );
    async_wake_up( fd->write_q, STATUS_VOLUME_DISMOUNTED );
1550 1551 1552 1553 1554 1555

    if (fd->poll_index != -1) set_fd_events( fd, -1 );

    if (fd->unix_fd != -1) close( fd->unix_fd );

    fd->unix_fd = -1;
1556
    fd->no_fd_status = STATUS_VOLUME_DISMOUNTED;
1557 1558 1559 1560 1561 1562 1563
    fd->closed->unix_fd = -1;
    fd->closed->unlink[0] = 0;

    /* stop using Unix locks on this fd (existing locks have been removed by close) */
    fd->fs_locks = 0;
}

1564
/* allocate an fd object, without setting the unix fd yet */
1565
static struct fd *alloc_fd_object(void)
1566
{
1567
    struct fd *fd = alloc_object( &fd_ops );
1568

1569 1570
    if (!fd) return NULL;

1571 1572
    fd->fd_ops     = NULL;
    fd->user       = NULL;
1573 1574
    fd->inode      = NULL;
    fd->closed     = NULL;
1575
    fd->access     = 0;
1576
    fd->options    = 0;
1577
    fd->sharing    = 0;
1578
    fd->unix_fd    = -1;
1579
    fd->unix_name  = NULL;
1580
    fd->cacheable  = 0;
1581
    fd->signaled   = 1;
1582
    fd->fs_locks   = 1;
1583
    fd->poll_index = -1;
1584 1585 1586
    fd->read_q     = NULL;
    fd->write_q    = NULL;
    fd->wait_q     = NULL;
1587
    fd->completion = NULL;
1588
    list_init( &fd->inode_entry );
1589
    list_init( &fd->locks );
1590 1591 1592 1593 1594 1595 1596 1597 1598

    if ((fd->poll_index = add_poll_user( fd )) == -1)
    {
        release_object( fd );
        return NULL;
    }
    return fd;
}

1599
/* allocate a pseudo fd object, for objects that need to behave like files but don't have a unix fd */
1600
struct fd *alloc_pseudo_fd( const struct fd_ops *fd_user_ops, struct object *user, unsigned int options )
1601 1602 1603 1604 1605 1606 1607 1608 1609 1610
{
    struct fd *fd = alloc_object( &fd_ops );

    if (!fd) return NULL;

    fd->fd_ops     = fd_user_ops;
    fd->user       = user;
    fd->inode      = NULL;
    fd->closed     = NULL;
    fd->access     = 0;
1611
    fd->options    = options;
1612
    fd->sharing    = 0;
1613
    fd->unix_name  = NULL;
1614
    fd->unix_fd    = -1;
1615
    fd->cacheable  = 0;
1616
    fd->signaled   = 0;
1617 1618
    fd->fs_locks   = 0;
    fd->poll_index = -1;
1619 1620 1621
    fd->read_q     = NULL;
    fd->write_q    = NULL;
    fd->wait_q     = NULL;
1622
    fd->completion = NULL;
1623
    fd->no_fd_status = STATUS_BAD_DEVICE_TYPE;
1624 1625 1626 1627 1628
    list_init( &fd->inode_entry );
    list_init( &fd->locks );
    return fd;
}

1629
/* duplicate an fd object for a different user */
1630
struct fd *dup_fd_object( struct fd *orig, unsigned int access, unsigned int sharing, unsigned int options )
1631
{
1632
    unsigned int err;
1633
    struct fd *fd = alloc_fd_object();
1634 1635 1636

    if (!fd) return NULL;

1637
    fd->options    = options;
1638
    fd->cacheable  = orig->cacheable;
1639

1640 1641 1642 1643 1644
    if (orig->unix_name)
    {
        if (!(fd->unix_name = mem_alloc( strlen(orig->unix_name) + 1 ))) goto failed;
        strcpy( fd->unix_name, orig->unix_name );
    }
1645 1646 1647 1648 1649 1650 1651

    if (orig->inode)
    {
        struct closed_fd *closed = mem_alloc( sizeof(*closed) );
        if (!closed) goto failed;
        if ((fd->unix_fd = dup( orig->unix_fd )) == -1)
        {
1652
            file_set_error();
1653 1654 1655 1656 1657 1658 1659 1660
            free( closed );
            goto failed;
        }
        closed->unix_fd = fd->unix_fd;
        closed->unlink[0] = 0;
        fd->closed = closed;
        fd->inode = (struct inode *)grab_object( orig->inode );
        list_add_head( &fd->inode->open, &fd->inode_entry );
1661 1662 1663 1664 1665
        if ((err = check_sharing( fd, access, sharing, 0, options )))
        {
            set_error( err );
            goto failed;
        }
1666
    }
1667 1668 1669 1670 1671
    else if ((fd->unix_fd = dup( orig->unix_fd )) == -1)
    {
        file_set_error();
        goto failed;
    }
1672 1673 1674 1675 1676 1677 1678
    return fd;

failed:
    release_object( fd );
    return NULL;
}

1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692
/* find an existing fd object that can be reused for a mapping */
struct fd *get_fd_object_for_mapping( struct fd *fd, unsigned int access, unsigned int sharing )
{
    struct fd *fd_ptr;

    if (!fd->inode) return NULL;

    LIST_FOR_EACH_ENTRY( fd_ptr, &fd->inode->open, struct fd, inode_entry )
        if (fd_ptr->access == access && fd_ptr->sharing == sharing)
            return (struct fd *)grab_object( fd_ptr );

    return NULL;
}

1693 1694 1695 1696 1697 1698
/* set the status to return when the fd has no associated unix fd */
void set_no_fd_status( struct fd *fd, unsigned int status )
{
    fd->no_fd_status = status;
}

1699 1700 1701 1702 1703 1704 1705 1706
/* sets the user of an fd that previously had no user */
void set_fd_user( struct fd *fd, const struct fd_ops *user_ops, struct object *user )
{
    assert( fd->fd_ops == NULL );
    fd->fd_ops = user_ops;
    fd->user   = user;
}

1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
static char *dup_fd_name( struct fd *root, const char *name )
{
    char *ret;

    if (!root) return strdup( name );
    if (!root->unix_name) return NULL;

    /* skip . prefix */
    if (name[0] == '.' && (!name[1] || name[1] == '/')) name++;

    if ((ret = malloc( strlen(root->unix_name) + strlen(name) + 2 )))
    {
        strcpy( ret, root->unix_name );
        if (name[0] && name[0] != '/') strcat( ret, "/" );
        strcat( ret, name );
    }
    return ret;
}

1726
/* open() wrapper that returns a struct fd with no fd user set */
1727
struct fd *open_fd( struct fd *root, const char *name, int flags, mode_t *mode, unsigned int access,
1728
                    unsigned int sharing, unsigned int options )
1729 1730 1731
{
    struct stat st;
    struct closed_fd *closed_fd;
1732
    struct fd *fd;
1733
    const char *unlink_name = "";
1734
    int root_fd = -1;
1735
    int rw_mode;
1736

1737 1738
    if (((options & FILE_DELETE_ON_CLOSE) && !(access & DELETE)) ||
        ((options & FILE_DIRECTORY_FILE) && (flags & O_TRUNC)))
1739 1740 1741 1742 1743
    {
        set_error( STATUS_INVALID_PARAMETER );
        return NULL;
    }

1744
    if (!(fd = alloc_fd_object())) return NULL;
1745

1746
    fd->options = options;
1747
    if (options & FILE_DELETE_ON_CLOSE) unlink_name = name;
1748
    if (!(closed_fd = mem_alloc( sizeof(*closed_fd) + strlen(unlink_name) )))
1749 1750 1751 1752
    {
        release_object( fd );
        return NULL;
    }
1753

1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764
    if (root)
    {
        if ((root_fd = get_unix_fd( root )) == -1) goto error;
        if (fchdir( root_fd ) == -1)
        {
            file_set_error();
            root_fd = -1;
            goto error;
        }
    }

1765 1766 1767 1768 1769 1770 1771 1772
    /* create the directory if needed */
    if ((options & FILE_DIRECTORY_FILE) && (flags & O_CREAT))
    {
        if (mkdir( name, 0777 ) == -1)
        {
            if (errno != EEXIST || (flags & O_EXCL))
            {
                file_set_error();
1773
                goto error;
1774 1775 1776 1777
            }
        }
        flags &= ~(O_CREAT | O_EXCL | O_TRUNC);
    }
1778 1779 1780 1781 1782 1783 1784 1785

    if ((access & FILE_UNIX_WRITE_ACCESS) && !(options & FILE_DIRECTORY_FILE))
    {
        if (access & FILE_UNIX_READ_ACCESS) rw_mode = O_RDWR;
        else rw_mode = O_WRONLY;
    }
    else rw_mode = O_RDONLY;

1786
    fd->unix_name = dup_fd_name( root, name );
1787

1788
    if ((fd->unix_fd = open( name, rw_mode | (flags & ~O_TRUNC), *mode )) == -1)
1789
    {
1790
        /* if we tried to open a directory for write access, retry read-only */
1791 1792 1793 1794 1795 1796 1797
        if (errno == EISDIR)
        {
            if ((access & FILE_UNIX_WRITE_ACCESS) || (flags & O_CREAT))
                fd->unix_fd = open( name, O_RDONLY | (flags & ~(O_TRUNC | O_CREAT | O_EXCL)), *mode );
        }

        if (fd->unix_fd == -1)
1798 1799 1800 1801
        {
            file_set_error();
            goto error;
        }
1802
    }
1803

1804
    closed_fd->unix_fd = fd->unix_fd;
1805
    closed_fd->unlink[0] = 0;
1806 1807 1808
    fstat( fd->unix_fd, &st );
    *mode = st.st_mode;

1809 1810
    /* only bother with an inode for normal files and directories */
    if (S_ISREG(st.st_mode) || S_ISDIR(st.st_mode))
1811
    {
1812
        unsigned int err;
1813
        struct inode *inode = get_inode( st.st_dev, st.st_ino, fd->unix_fd );
1814 1815 1816 1817 1818 1819

        if (!inode)
        {
            /* we can close the fd because there are no others open on the same file,
             * otherwise we wouldn't have failed to allocate a new inode
             */
1820
            goto error;
1821 1822 1823
        }
        fd->inode = inode;
        fd->closed = closed_fd;
1824
        fd->cacheable = !inode->device->removable;
1825
        list_add_head( &inode->open, &fd->inode_entry );
1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839

        /* check directory options */
        if ((options & FILE_DIRECTORY_FILE) && !S_ISDIR(st.st_mode))
        {
            release_object( fd );
            set_error( STATUS_NOT_A_DIRECTORY );
            return NULL;
        }
        if ((options & FILE_NON_DIRECTORY_FILE) && S_ISDIR(st.st_mode))
        {
            release_object( fd );
            set_error( STATUS_FILE_IS_A_DIRECTORY );
            return NULL;
        }
1840
        if ((err = check_sharing( fd, access, sharing, flags, options )))
1841 1842
        {
            release_object( fd );
1843
            set_error( err );
1844 1845
            return NULL;
        }
1846
        strcpy( closed_fd->unlink, unlink_name );
1847 1848 1849 1850 1851 1852 1853 1854 1855 1856
        if (flags & O_TRUNC)
        {
            if (S_ISDIR(st.st_mode))
            {
                release_object( fd );
                set_error( STATUS_OBJECT_NAME_COLLISION );
                return NULL;
            }
            ftruncate( fd->unix_fd, 0 );
        }
1857
    }
1858
    else  /* special file */
1859
    {
1860 1861 1862 1863 1864
        if (options & FILE_DIRECTORY_FILE)
        {
            set_error( STATUS_NOT_A_DIRECTORY );
            goto error;
        }
1865 1866 1867
        if (unlink_name[0])  /* we can't unlink special files */
        {
            set_error( STATUS_INVALID_PARAMETER );
1868
            goto error;
1869
        }
1870
        free( closed_fd );
1871
        fd->cacheable = 1;
1872
    }
1873
    return fd;
1874 1875 1876 1877

error:
    release_object( fd );
    free( closed_fd );
1878
    if (root_fd != -1) fchdir( server_dir_fd ); /* go back to the server dir */
1879
    return NULL;
1880 1881
}

1882 1883
/* create an fd for an anonymous file */
/* if the function fails the unix fd is closed */
1884 1885
struct fd *create_anonymous_fd( const struct fd_ops *fd_user_ops, int unix_fd, struct object *user,
                                unsigned int options )
1886
{
1887
    struct fd *fd = alloc_fd_object();
1888 1889 1890

    if (fd)
    {
1891
        set_fd_user( fd, fd_user_ops, user );
1892
        fd->unix_fd = unix_fd;
1893
        fd->options = options;
1894 1895 1896 1897 1898 1899
        return fd;
    }
    close( unix_fd );
    return NULL;
}

1900 1901
/* retrieve the object that is using an fd */
void *get_fd_user( struct fd *fd )
1902
{
1903 1904
    return fd->user;
}
1905

1906 1907 1908 1909 1910 1911
/* retrieve the opening options for the fd */
unsigned int get_fd_options( struct fd *fd )
{
    return fd->options;
}

1912 1913 1914
/* retrieve the unix fd for an object */
int get_unix_fd( struct fd *fd )
{
1915
    if (fd->unix_fd == -1) set_error( fd->no_fd_status );
1916
    return fd->unix_fd;
1917 1918
}

1919 1920 1921 1922 1923 1924
/* check if two file descriptors point to the same file */
int is_same_file_fd( struct fd *fd1, struct fd *fd2 )
{
    return fd1->inode == fd2->inode;
}

1925 1926 1927 1928 1929 1930
/* allow the fd to be cached (can't be reset once set) */
void allow_fd_caching( struct fd *fd )
{
    fd->cacheable = 1;
}

1931 1932 1933 1934 1935 1936
/* check if fd is on a removable device */
int is_fd_removable( struct fd *fd )
{
    return (fd->inode && fd->inode->device->removable);
}

1937 1938 1939 1940 1941 1942 1943
/* set or clear the fd signaled state */
void set_fd_signaled( struct fd *fd, int signaled )
{
    fd->signaled = signaled;
    if (signaled) wake_up( fd->user, 0 );
}

1944 1945 1946 1947 1948 1949
/* set or clear the fd signaled state */
int is_fd_signaled( struct fd *fd )
{
    return fd->signaled;
}

1950 1951 1952 1953 1954 1955
/* handler for close_handle that refuses to close fd-associated handles in other processes */
int fd_close_handle( struct object *obj, struct process *process, obj_handle_t handle )
{
    return (!current || current->process == process);
}

1956 1957 1958 1959 1960
/* check if events are pending and if yes return which one(s) */
int check_fd_events( struct fd *fd, int events )
{
    struct pollfd pfd;

1961
    if (fd->unix_fd == -1) return POLLERR;
1962
    if (fd->inode) return events;  /* regular files are always signaled */
1963

1964 1965 1966 1967
    pfd.fd     = fd->unix_fd;
    pfd.events = events;
    if (poll( &pfd, 1, 0 ) <= 0) return 0;
    return pfd.revents;
1968 1969 1970 1971 1972
}

/* default signaled() routine for objects that poll() on an fd */
int default_fd_signaled( struct object *obj, struct thread *thread )
{
1973
    struct fd *fd = get_obj_fd( obj );
1974
    int ret = fd->signaled;
1975 1976
    release_object( fd );
    return ret;
1977 1978
}

1979 1980 1981 1982 1983 1984 1985 1986 1987 1988
/* default map_access() routine for objects that behave like an fd */
unsigned int default_fd_map_access( struct object *obj, unsigned int access )
{
    if (access & GENERIC_READ)    access |= FILE_GENERIC_READ;
    if (access & GENERIC_WRITE)   access |= FILE_GENERIC_WRITE;
    if (access & GENERIC_EXECUTE) access |= FILE_GENERIC_EXECUTE;
    if (access & GENERIC_ALL)     access |= FILE_ALL_ACCESS;
    return access & ~(GENERIC_READ | GENERIC_WRITE | GENERIC_EXECUTE | GENERIC_ALL);
}

1989 1990 1991 1992
int default_fd_get_poll_events( struct fd *fd )
{
    int events = 0;

1993 1994
    if (async_waiting( fd->read_q )) events |= POLLIN;
    if (async_waiting( fd->write_q )) events |= POLLOUT;
1995 1996 1997
    return events;
}

1998 1999 2000
/* default handler for poll() events */
void default_poll_event( struct fd *fd, int event )
{
2001 2002
    if (event & (POLLIN | POLLERR | POLLHUP)) async_wake_up( fd->read_q, STATUS_ALERTED );
    if (event & (POLLOUT | POLLERR | POLLHUP)) async_wake_up( fd->write_q, STATUS_ALERTED );
2003 2004

    /* if an error occurred, stop polling this fd to avoid busy-looping */
2005
    if (event & (POLLERR | POLLHUP)) set_fd_events( fd, -1 );
2006
    else if (!fd->inode) set_fd_events( fd, fd->fd_ops->get_poll_events( fd ) );
2007 2008
}

2009
struct async *fd_queue_async( struct fd *fd, const async_data_t *data, int type )
2010
{
2011
    struct async_queue *queue;
2012
    struct async *async;
2013 2014 2015 2016

    switch (type)
    {
    case ASYNC_TYPE_READ:
2017
        if (!fd->read_q && !(fd->read_q = create_async_queue( fd ))) return NULL;
2018
        queue = fd->read_q;
2019 2020
        break;
    case ASYNC_TYPE_WRITE:
2021
        if (!fd->write_q && !(fd->write_q = create_async_queue( fd ))) return NULL;
2022
        queue = fd->write_q;
2023
        break;
2024
    case ASYNC_TYPE_WAIT:
2025
        if (!fd->wait_q && !(fd->wait_q = create_async_queue( fd ))) return NULL;
2026
        queue = fd->wait_q;
2027
        break;
2028
    default:
2029
        queue = NULL;
2030
        assert(0);
2031 2032
    }

2033
    if ((async = create_async( current, queue, data )) && type != ASYNC_TYPE_WAIT)
2034 2035 2036 2037
    {
        if (!fd->inode)
            set_fd_events( fd, fd->fd_ops->get_poll_events( fd ) );
        else  /* regular files are always ready for read and write */
2038
            async_wake_up( queue, STATUS_ALERTED );
2039 2040
    }
    return async;
2041 2042
}

2043
void fd_async_wake_up( struct fd *fd, int type, unsigned int status )
2044 2045 2046 2047
{
    switch (type)
    {
    case ASYNC_TYPE_READ:
2048
        async_wake_up( fd->read_q, status );
2049 2050
        break;
    case ASYNC_TYPE_WRITE:
2051
        async_wake_up( fd->write_q, status );
2052 2053
        break;
    case ASYNC_TYPE_WAIT:
2054
        async_wake_up( fd->wait_q, status );
2055 2056 2057 2058 2059 2060
        break;
    default:
        assert(0);
    }
}

2061 2062 2063 2064 2065
void fd_reselect_async( struct fd *fd, struct async_queue *queue )
{
    fd->fd_ops->reselect_async( fd, queue );
}

2066 2067 2068 2069 2070
void no_fd_queue_async( struct fd *fd, const async_data_t *data, int type, int count )
{
    set_error( STATUS_OBJECT_TYPE_MISMATCH );
}

2071
void default_fd_queue_async( struct fd *fd, const async_data_t *data, int type, int count )
2072
{
2073
    struct async *async;
2074

2075
    if ((async = fd_queue_async( fd, data, type )))
2076 2077 2078 2079
    {
        release_object( async );
        set_error( STATUS_PENDING );
    }
2080 2081
}

2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094
/* default reselect_async() fd routine */
void default_fd_reselect_async( struct fd *fd, struct async_queue *queue )
{
    if (queue != fd->wait_q)
    {
        int poll_events = fd->fd_ops->get_poll_events( fd );
        int events = check_fd_events( fd, poll_events );
        if (events) fd->fd_ops->poll_event( fd, events );
        else set_fd_events( fd, poll_events );
    }
}

/* default cancel_async() fd routine */
2095
void default_fd_cancel_async( struct fd *fd, struct process *process, struct thread *thread, client_ptr_t iosb )
2096
{
2097 2098 2099 2100 2101 2102 2103
    int n = 0;

    n += async_wake_up_by( fd->read_q, process, thread, iosb, STATUS_CANCELLED );
    n += async_wake_up_by( fd->write_q, process, thread, iosb, STATUS_CANCELLED );
    n += async_wake_up_by( fd->wait_q, process, thread, iosb, STATUS_CANCELLED );
    if (!n && iosb)
        set_error( STATUS_NOT_FOUND );
2104 2105
}

2106
/* default flush() routine */
2107
void no_flush( struct fd *fd, struct event **event )
2108 2109 2110 2111
{
    set_error( STATUS_OBJECT_TYPE_MISMATCH );
}

2112 2113 2114 2115 2116 2117 2118 2119 2120 2121
static inline int is_valid_mounted_device( struct stat *st )
{
#if defined(linux) || defined(__sun__)
    return S_ISBLK( st->st_mode );
#else
    /* disks are char devices on *BSD */
    return S_ISCHR( st->st_mode );
#endif
}

2122
/* close all Unix file descriptors on a device to allow unmounting it */
2123
static void unmount_device( struct fd *device_fd )
2124 2125
{
    unsigned int i;
2126 2127
    struct stat st;
    struct device *device;
2128 2129
    struct inode *inode;
    struct fd *fd;
2130
    int unix_fd = get_unix_fd( device_fd );
2131

2132 2133
    if (unix_fd == -1) return;

2134
    if (fstat( unix_fd, &st ) == -1 || !is_valid_mounted_device( &st ))
2135 2136 2137 2138 2139
    {
        set_error( STATUS_INVALID_PARAMETER );
        return;
    }

2140
    if (!(device = get_device( st.st_rdev, -1 ))) return;
2141

2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155
    for (i = 0; i < INODE_HASH_SIZE; i++)
    {
        LIST_FOR_EACH_ENTRY( inode, &device->inode_hash[i], struct inode, entry )
        {
            LIST_FOR_EACH_ENTRY( fd, &inode->open, struct fd, inode_entry )
            {
                unmount_fd( fd );
            }
            inode_close_pending( inode, 0 );
        }
    }
    /* remove it from the hash table */
    list_remove( &device->entry );
    list_init( &device->entry );
2156
    release_object( device );
2157 2158
}

2159 2160 2161 2162 2163 2164 2165
obj_handle_t no_fd_ioctl( struct fd *fd, ioctl_code_t code, const async_data_t *async,
                          int blocking, const void *data, data_size_t size )
{
    set_error( STATUS_OBJECT_TYPE_MISMATCH );
    return 0;
}

2166
/* default ioctl() routine */
2167
obj_handle_t default_fd_ioctl( struct fd *fd, ioctl_code_t code, const async_data_t *async,
2168
                               int blocking, const void *data, data_size_t size )
2169 2170 2171 2172 2173
{
    switch(code)
    {
    case FSCTL_DISMOUNT_VOLUME:
        unmount_device( fd );
2174
        return 0;
2175 2176
    default:
        set_error( STATUS_NOT_SUPPORTED );
2177
        return 0;
2178 2179 2180
    }
}

2181 2182 2183 2184 2185 2186 2187
/* same as get_handle_obj but retrieve the struct fd associated to the object */
static struct fd *get_handle_fd_obj( struct process *process, obj_handle_t handle,
                                     unsigned int access )
{
    struct fd *fd = NULL;
    struct object *obj;

2188
    if ((obj = get_handle_obj( process, handle, access, NULL )))
2189
    {
2190
        fd = get_obj_fd( obj );
2191 2192 2193 2194 2195
        release_object( obj );
    }
    return fd;
}

2196
struct completion *fd_get_completion( struct fd *fd, apc_param_t *p_key )
2197
{
2198
    *p_key = fd->comp_key;
2199 2200 2201 2202 2203 2204 2205
    return fd->completion ? (struct completion *)grab_object( fd->completion ) : NULL;
}

void fd_copy_completion( struct fd *src, struct fd *dst )
{
    assert( !dst->completion );
    dst->completion = fd_get_completion( src, &dst->comp_key );
2206 2207
}

2208 2209 2210 2211
/* flush a file buffers */
DECL_HANDLER(flush_file)
{
    struct fd *fd = get_handle_fd_obj( current->process, req->handle, 0 );
2212
    struct event * event = NULL;
2213 2214 2215

    if (fd)
    {
2216
        fd->fd_ops->flush( fd, &event );
2217
        if ( event )
2218 2219 2220
        {
            reply->event = alloc_handle( current->process, event, SYNCHRONIZE, 0 );
        }
2221 2222 2223 2224
        release_object( fd );
    }
}

2225 2226 2227 2228 2229
/* open a file object */
DECL_HANDLER(open_file_object)
{
    struct unicode_str name;
    struct directory *root = NULL;
2230
    struct object *obj, *result;
2231 2232 2233 2234 2235 2236 2237

    get_req_unicode_str( &name );
    if (req->rootdir && !(root = get_directory_obj( current->process, req->rootdir, 0 )))
        return;

    if ((obj = open_object_dir( root, &name, req->attributes, NULL )))
    {
2238
        if ((result = obj->ops->open_file( obj, req->access, req->sharing, req->options )))
2239
        {
2240 2241
            reply->handle = alloc_handle( current->process, result, req->access, req->attributes );
            release_object( result );
2242 2243 2244 2245 2246 2247 2248
        }
        release_object( obj );
    }

    if (root) release_object( root );
}

2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
/* get the Unix name from a file handle */
DECL_HANDLER(get_handle_unix_name)
{
    struct fd *fd;

    if ((fd = get_handle_fd_obj( current->process, req->handle, 0 )))
    {
        if (fd->unix_name)
        {
            data_size_t name_len = strlen( fd->unix_name );
            reply->name_len = name_len;
            if (name_len <= get_reply_max_size()) set_reply_data( fd->unix_name, name_len );
            else set_error( STATUS_BUFFER_OVERFLOW );
        }
2263
        else set_error( STATUS_OBJECT_TYPE_MISMATCH );
2264 2265 2266 2267
        release_object( fd );
    }
}

2268 2269 2270 2271 2272
/* get a Unix fd to access a file */
DECL_HANDLER(get_handle_fd)
{
    struct fd *fd;

2273
    if ((fd = get_handle_fd_obj( current->process, req->handle, 0 )))
2274
    {
2275 2276
        int unix_fd = get_unix_fd( fd );
        if (unix_fd != -1)
2277
        {
2278
            reply->type = fd->fd_ops->get_fd_type( fd );
2279
            reply->cacheable = fd->cacheable;
2280 2281
            reply->options = fd->options;
            reply->access = get_handle_access( current->process, req->handle );
2282
            send_client_fd( current->process, unix_fd, req->handle );
2283
        }
2284 2285 2286 2287
        release_object( fd );
    }
}

2288 2289 2290 2291
/* perform an ioctl on a file */
DECL_HANDLER(ioctl)
{
    unsigned int access = (req->code >> 14) & (FILE_READ_DATA|FILE_WRITE_DATA);
2292
    struct fd *fd = get_handle_fd_obj( current->process, req->async.handle, access );
2293 2294 2295

    if (fd)
    {
2296
        reply->wait = fd->fd_ops->ioctl( fd, req->code, &req->async, req->blocking,
2297 2298
                                         get_req_data(), get_req_data_size() );
        reply->options = fd->options;
2299 2300 2301 2302
        release_object( fd );
    }
}

2303 2304 2305
/* create / reschedule an async I/O */
DECL_HANDLER(register_async)
{
2306 2307
    unsigned int access;
    struct fd *fd;
2308

2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320
    switch(req->type)
    {
    case ASYNC_TYPE_READ:
        access = FILE_READ_DATA;
        break;
    case ASYNC_TYPE_WRITE:
        access = FILE_WRITE_DATA;
        break;
    default:
        set_error( STATUS_INVALID_PARAMETER );
        return;
    }
2321

2322
    if ((fd = get_handle_fd_obj( current->process, req->async.handle, access )))
2323
    {
2324
        if (get_unix_fd( fd ) != -1) fd->fd_ops->queue_async( fd, &req->async, req->type, req->count );
2325 2326 2327
        release_object( fd );
    }
}
2328 2329 2330 2331 2332

/* cancels all async I/O */
DECL_HANDLER(cancel_async)
{
    struct fd *fd = get_handle_fd_obj( current->process, req->handle, 0 );
2333
    struct thread *thread = req->only_thread ? current : NULL;
2334

2335 2336
    if (fd)
    {
2337
        if (get_unix_fd( fd ) != -1) fd->fd_ops->cancel_async( fd, current->process, thread, req->iosb );
2338
        release_object( fd );
2339
    }
2340
}
2341 2342 2343 2344 2345 2346 2347 2348

/* attach completion object to a fd */
DECL_HANDLER(set_completion_info)
{
    struct fd *fd = get_handle_fd_obj( current->process, req->handle, 0 );

    if (fd)
    {
2349
        if (!(fd->options & (FILE_SYNCHRONOUS_IO_ALERT | FILE_SYNCHRONOUS_IO_NONALERT)) && !fd->completion)
2350 2351 2352 2353 2354 2355 2356 2357
        {
            fd->completion = get_completion_obj( current->process, req->chandle, IO_COMPLETION_MODIFY_STATE );
            fd->comp_key = req->ckey;
        }
        else set_error( STATUS_INVALID_PARAMETER );
        release_object( fd );
    }
}
2358 2359 2360 2361 2362 2363 2364

/* push new completion msg into a completion queue attached to the fd */
DECL_HANDLER(add_fd_completion)
{
    struct fd *fd = get_handle_fd_obj( current->process, req->handle, 0 );
    if (fd)
    {
2365 2366
        if (fd->completion)
            add_completion( fd->completion, fd->comp_key, req->cvalue, req->status, req->information );
2367 2368 2369
        release_object( fd );
    }
}