2 * Copyright (c) 2005 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1989, 1991, 1993
36 * The Regents of the University of California. All rights reserved.
37 * (c) UNIX System Laboratories, Inc.
38 * All or some portions of this file are derived from material licensed
39 * to the University of California by American Telephone and Telegraph
40 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
41 * the permission of UNIX System Laboratories, Inc.
43 * Redistribution and use in source and binary forms, with or without
44 * modification, are permitted provided that the following conditions
46 * 1. Redistributions of source code must retain the above copyright
47 * notice, this list of conditions and the following disclaimer.
48 * 2. Redistributions in binary form must reproduce the above copyright
49 * notice, this list of conditions and the following disclaimer in the
50 * documentation and/or other materials provided with the distribution.
51 * 3. All advertising materials mentioning features or use of this software
52 * must display the following acknowledgement:
53 * This product includes software developed by the University of
54 * California, Berkeley and its contributors.
55 * 4. Neither the name of the University nor the names of its contributors
56 * may be used to endorse or promote products derived from this software
57 * without specific prior written permission.
59 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
71 * @(#)kern_descrip.c 8.6 (Berkeley) 4/19/94
72 * $FreeBSD: src/sys/kern/kern_descrip.c,v 1.81.2.19 2004/02/28 00:43:31 tegge Exp $
73 * $DragonFly: src/sys/kern/kern_descrip.c,v 1.46 2005/07/13 01:38:50 dillon Exp $
76 #include "opt_compat.h"
77 #include <sys/param.h>
78 #include <sys/systm.h>
79 #include <sys/malloc.h>
80 #include <sys/sysproto.h>
82 #include <sys/filedesc.h>
83 #include <sys/kernel.h>
84 #include <sys/sysctl.h>
85 #include <sys/vnode.h>
87 #include <sys/nlookup.h>
90 #include <sys/filio.h>
91 #include <sys/fcntl.h>
92 #include <sys/unistd.h>
93 #include <sys/resourcevar.h>
94 #include <sys/event.h>
95 #include <sys/kern_syscall.h>
96 #include <sys/kcore.h>
97 #include <sys/kinfo.h>
100 #include <vm/vm_extern.h>
102 #include <sys/thread2.h>
103 #include <sys/file2.h>
105 static MALLOC_DEFINE(M_FILEDESC, "file desc", "Open file descriptor table");
106 static MALLOC_DEFINE(M_FILEDESC_TO_LEADER, "file desc to leader",
107 "file desc to leader structures");
108 MALLOC_DEFINE(M_FILE, "file", "Open file structure");
109 static MALLOC_DEFINE(M_SIGIO, "sigio", "sigio structures");
111 static d_open_t fdopen;
114 #define CDEV_MAJOR 22
115 static struct cdevsw fildesc_cdevsw = {
117 /* maj */ CDEV_MAJOR,
129 /* strategy */ nostrategy,
134 static int badfo_readwrite (struct file *fp, struct uio *uio,
135 struct ucred *cred, int flags, struct thread *td);
136 static int badfo_ioctl (struct file *fp, u_long com, caddr_t data,
138 static int badfo_poll (struct file *fp, int events,
139 struct ucred *cred, struct thread *td);
140 static int badfo_kqfilter (struct file *fp, struct knote *kn);
141 static int badfo_stat (struct file *fp, struct stat *sb, struct thread *td);
142 static int badfo_close (struct file *fp, struct thread *td);
143 static int badfo_shutdown (struct file *fp, int how, struct thread *td);
146 * Descriptor management.
148 struct filelist filehead; /* head of list of open files */
149 int nfiles; /* actual number of open files */
153 * System calls on descriptors.
157 getdtablesize(struct getdtablesize_args *uap)
159 struct proc *p = curproc;
162 min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfilesperproc);
167 * Duplicate a file descriptor to a particular value.
169 * note: keep in mind that a potential race condition exists when closing
170 * descriptors from a shared descriptor table (via rfork).
174 dup2(struct dup2_args *uap)
178 error = kern_dup(DUP_FIXED, uap->from, uap->to, uap->sysmsg_fds);
184 * Duplicate a file descriptor.
188 dup(struct dup_args *uap)
192 error = kern_dup(DUP_VARIABLE, uap->fd, 0, uap->sysmsg_fds);
198 kern_fcntl(int fd, int cmd, union fcntl_dat *dat)
200 struct thread *td = curthread;
201 struct proc *p = td->td_proc;
202 struct filedesc *fdp = p->p_fd;
207 int tmp, error, flg = F_POSIX;
211 if ((unsigned)fd >= fdp->fd_nfiles ||
212 (fp = fdp->fd_files[fd].fp) == NULL)
214 pop = &fdp->fd_files[fd].fileflags;
219 if (newmin >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
220 newmin > maxfilesperproc)
222 error = kern_dup(DUP_VARIABLE, fd, newmin, &dat->fc_fd);
226 dat->fc_cloexec = (*pop & UF_EXCLOSE) ? FD_CLOEXEC : 0;
230 *pop = (*pop &~ UF_EXCLOSE) |
231 (dat->fc_cloexec & FD_CLOEXEC ? UF_EXCLOSE : 0);
235 dat->fc_flags = OFLAGS(fp->f_flag);
240 fp->f_flag &= ~FCNTLFLAGS;
241 fp->f_flag |= FFLAGS(dat->fc_flags & ~O_ACCMODE) & FCNTLFLAGS;
242 tmp = fp->f_flag & FNONBLOCK;
243 error = fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, td);
248 tmp = fp->f_flag & FASYNC;
249 error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, td);
254 fp->f_flag &= ~FNONBLOCK;
256 fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, td);
262 error = fo_ioctl(fp, FIOGETOWN, (caddr_t)&dat->fc_owner, td);
268 error = fo_ioctl(fp, FIOSETOWN, (caddr_t)&dat->fc_owner, td);
274 /* Fall into F_SETLK */
277 if (fp->f_type != DTYPE_VNODE)
279 vp = (struct vnode *)fp->f_data;
282 * copyin/lockop may block
285 if (dat->fc_flock.l_whence == SEEK_CUR)
286 dat->fc_flock.l_start += fp->f_offset;
288 switch (dat->fc_flock.l_type) {
290 if ((fp->f_flag & FREAD) == 0) {
294 p->p_leader->p_flag |= P_ADVLOCK;
295 error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_SETLK,
296 &dat->fc_flock, flg);
299 if ((fp->f_flag & FWRITE) == 0) {
303 p->p_leader->p_flag |= P_ADVLOCK;
304 error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_SETLK,
305 &dat->fc_flock, flg);
308 error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_UNLCK,
309 &dat->fc_flock, F_POSIX);
315 /* Check for race with close */
316 if ((unsigned) fd >= fdp->fd_nfiles ||
317 fp != fdp->fd_files[fd].fp) {
318 dat->fc_flock.l_whence = SEEK_SET;
319 dat->fc_flock.l_start = 0;
320 dat->fc_flock.l_len = 0;
321 dat->fc_flock.l_type = F_UNLCK;
322 (void) VOP_ADVLOCK(vp, (caddr_t)p->p_leader,
323 F_UNLCK, &dat->fc_flock, F_POSIX);
329 if (fp->f_type != DTYPE_VNODE)
331 vp = (struct vnode *)fp->f_data;
333 * copyin/lockop may block
336 if (dat->fc_flock.l_type != F_RDLCK &&
337 dat->fc_flock.l_type != F_WRLCK &&
338 dat->fc_flock.l_type != F_UNLCK) {
342 if (dat->fc_flock.l_whence == SEEK_CUR)
343 dat->fc_flock.l_start += fp->f_offset;
344 error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_GETLK,
345 &dat->fc_flock, F_POSIX);
355 * The file control system call.
358 fcntl(struct fcntl_args *uap)
365 dat.fc_fd = uap->arg;
368 dat.fc_cloexec = uap->arg;
371 dat.fc_flags = uap->arg;
374 dat.fc_owner = uap->arg;
379 error = copyin((caddr_t)uap->arg, &dat.fc_flock,
380 sizeof(struct flock));
386 error = kern_fcntl(uap->fd, uap->cmd, &dat);
391 uap->sysmsg_result = dat.fc_fd;
394 uap->sysmsg_result = dat.fc_cloexec;
397 uap->sysmsg_result = dat.fc_flags;
400 uap->sysmsg_result = dat.fc_owner;
402 error = copyout(&dat.fc_flock, (caddr_t)uap->arg,
403 sizeof(struct flock));
412 * Common code for dup, dup2, and fcntl(F_DUPFD).
414 * The type flag can be either DUP_FIXED or DUP_VARIABLE. DUP_FIXED tells
415 * kern_dup() to destructively dup over an existing file descriptor if new
416 * is already open. DUP_VARIABLE tells kern_dup() to find the lowest
417 * unused file descriptor that is greater than or equal to new.
420 kern_dup(enum dup_type type, int old, int new, int *res)
422 struct thread *td = curthread;
423 struct proc *p = td->td_proc;
424 struct filedesc *fdp = p->p_fd;
428 boolean_t fdalloced = FALSE;
432 * Verify that we have a valid descriptor to dup from and
433 * possibly to dup to.
435 if (old < 0 || new < 0 || new > p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
436 new >= maxfilesperproc)
438 if (old >= fdp->fd_nfiles || fdp->fd_files[old].fp == NULL)
440 if (type == DUP_FIXED && old == new) {
444 fp = fdp->fd_files[old].fp;
448 * Expand the table for the new descriptor if needed. This may
449 * block and drop and reacquire the fidedesc lock.
451 if (type == DUP_VARIABLE || new >= fdp->fd_nfiles) {
452 error = fdalloc(p, new, &newfd);
459 if (type == DUP_VARIABLE)
463 * If the old file changed out from under us then treat it as a
464 * bad file descriptor. Userland should do its own locking to
467 if (fdp->fd_files[old].fp != fp) {
468 if (fdp->fd_files[new].fp == NULL) {
470 fdreserve(fdp, newfd, -1);
471 if (new < fdp->fd_freefile)
472 fdp->fd_freefile = new;
473 while (fdp->fd_lastfile > 0 &&
474 fdp->fd_files[fdp->fd_lastfile].fp == NULL)
480 KASSERT(old != new, ("new fd is same as old"));
483 * Save info on the descriptor being overwritten. We have
484 * to do the unmap now, but we cannot close it without
485 * introducing an ownership race for the slot.
487 delfp = fdp->fd_files[new].fp;
488 if (delfp != NULL && p->p_fdtol != NULL) {
490 * Ask fdfree() to sleep to ensure that all relevant
491 * process leaders can be traversed in closef().
493 fdp->fd_holdleaderscount++;
497 KASSERT(delfp == NULL || type == DUP_FIXED,
498 ("dup() picked an open file"));
500 if (delfp && (fdp->fd_files[new].fileflags & UF_MAPPED))
501 (void) munmapfd(p, new);
505 * Duplicate the source descriptor, update lastfile
507 if (new > fdp->fd_lastfile)
508 fdp->fd_lastfile = new;
509 if (!fdalloced && fdp->fd_files[new].fp == NULL)
510 fdreserve(fdp, new, 1);
511 fdp->fd_files[new].fp = fp;
512 fdp->fd_files[new].fileflags =
513 fdp->fd_files[old].fileflags & ~UF_EXCLOSE;
517 * If we dup'd over a valid file, we now own the reference to it
518 * and must dispose of it using closef() semantics (as if a
519 * close() were performed on it).
522 (void) closef(delfp, td);
524 fdp->fd_holdleaderscount--;
525 if (fdp->fd_holdleaderscount == 0 &&
526 fdp->fd_holdleaderswakeup != 0) {
527 fdp->fd_holdleaderswakeup = 0;
528 wakeup(&fdp->fd_holdleaderscount);
536 * If sigio is on the list associated with a process or process group,
537 * disable signalling from the device, remove sigio from the list and
541 funsetown(struct sigio *sigio)
546 *(sigio->sio_myref) = NULL;
548 if (sigio->sio_pgid < 0) {
549 SLIST_REMOVE(&sigio->sio_pgrp->pg_sigiolst, sigio,
551 } else /* if ((*sigiop)->sio_pgid > 0) */ {
552 SLIST_REMOVE(&sigio->sio_proc->p_sigiolst, sigio,
555 crfree(sigio->sio_ucred);
556 free(sigio, M_SIGIO);
559 /* Free a list of sigio structures. */
561 funsetownlst(struct sigiolst *sigiolst)
565 while ((sigio = SLIST_FIRST(sigiolst)) != NULL)
570 * This is common code for FIOSETOWN ioctl called by fcntl(fd, F_SETOWN, arg).
572 * After permission checking, add a sigio structure to the sigio list for
573 * the process or process group.
576 fsetown(pid_t pgid, struct sigio **sigiop)
592 * Policy - Don't allow a process to FSETOWN a process
593 * in another session.
595 * Remove this test to allow maximum flexibility or
596 * restrict FSETOWN to the current process or process
597 * group for maximum safety.
599 if (proc->p_session != curproc->p_session)
603 } else /* if (pgid < 0) */ {
604 pgrp = pgfind(-pgid);
609 * Policy - Don't allow a process to FSETOWN a process
610 * in another session.
612 * Remove this test to allow maximum flexibility or
613 * restrict FSETOWN to the current process or process
614 * group for maximum safety.
616 if (pgrp->pg_session != curproc->p_session)
622 sigio = malloc(sizeof(struct sigio), M_SIGIO, M_WAITOK);
624 SLIST_INSERT_HEAD(&proc->p_sigiolst, sigio, sio_pgsigio);
625 sigio->sio_proc = proc;
627 SLIST_INSERT_HEAD(&pgrp->pg_sigiolst, sigio, sio_pgsigio);
628 sigio->sio_pgrp = pgrp;
630 sigio->sio_pgid = pgid;
631 sigio->sio_ucred = crhold(curproc->p_ucred);
632 /* It would be convenient if p_ruid was in ucred. */
633 sigio->sio_ruid = curproc->p_ucred->cr_ruid;
634 sigio->sio_myref = sigiop;
642 * This is common code for FIOGETOWN ioctl called by fcntl(fd, F_GETOWN, arg).
645 fgetown(struct sigio *sigio)
647 return (sigio != NULL ? sigio->sio_pgid : 0);
651 * Close many file descriptors.
656 closefrom(struct closefrom_args *uap)
658 return(kern_closefrom(uap->fd));
662 kern_closefrom(int fd)
664 struct thread *td = curthread;
665 struct proc *p = td->td_proc;
666 struct filedesc *fdp;
671 if (fd < 0 || fd > fdp->fd_lastfile)
675 if (kern_close(fdp->fd_lastfile) == EINTR)
677 } while (fdp->fd_lastfile > fd);
683 * Close a file descriptor.
688 close(struct close_args *uap)
690 return(kern_close(uap->fd));
696 struct thread *td = curthread;
697 struct proc *p = td->td_proc;
698 struct filedesc *fdp;
706 if ((unsigned)fd >= fdp->fd_nfiles ||
707 (fp = fdp->fd_files[fd].fp) == NULL)
710 if (fdp->fd_files[fd].fileflags & UF_MAPPED)
711 (void) munmapfd(p, fd);
715 if (p->p_fdtol != NULL) {
717 * Ask fdfree() to sleep to ensure that all relevant
718 * process leaders can be traversed in closef().
720 fdp->fd_holdleaderscount++;
725 * we now hold the fp reference that used to be owned by the descriptor
728 while (fdp->fd_lastfile > 0 && fdp->fd_files[fdp->fd_lastfile].fp == NULL)
730 if (fd < fdp->fd_knlistsize)
731 knote_fdclose(p, fd);
732 error = closef(fp, td);
734 fdp->fd_holdleaderscount--;
735 if (fdp->fd_holdleaderscount == 0 &&
736 fdp->fd_holdleaderswakeup != 0) {
737 fdp->fd_holdleaderswakeup = 0;
738 wakeup(&fdp->fd_holdleaderscount);
745 * shutdown_args(int fd, int how)
748 kern_shutdown(int fd, int how)
750 struct thread *td = curthread;
751 struct proc *p = td->td_proc;
752 struct filedesc *fdp;
759 if ((unsigned)fd >= fdp->fd_nfiles ||
760 (fp = fdp->fd_files[fd].fp) == NULL)
763 error = fo_shutdown(fp, how, td);
770 shutdown(struct shutdown_args *uap)
774 error = kern_shutdown(uap->s, uap->how);
780 kern_fstat(int fd, struct stat *ub)
782 struct thread *td = curthread;
783 struct proc *p = td->td_proc;
784 struct filedesc *fdp;
791 if ((unsigned)fd >= fdp->fd_nfiles ||
792 (fp = fdp->fd_files[fd].fp) == NULL)
795 error = fo_stat(fp, ub, td);
802 * Return status information about a file descriptor.
805 fstat(struct fstat_args *uap)
810 error = kern_fstat(uap->fd, &st);
813 error = copyout(&st, uap->sb, sizeof(st));
818 * XXX: This is for source compatibility with NetBSD. Probably doesn't
822 nfstat(struct nfstat_args *uap)
828 error = kern_fstat(uap->fd, &st);
832 error = copyout(&nst, uap->sb, sizeof (nst));
838 * Return pathconf information about a file descriptor.
842 fpathconf(struct fpathconf_args *uap)
844 struct thread *td = curthread;
845 struct proc *p = td->td_proc;
846 struct filedesc *fdp;
853 if ((unsigned)uap->fd >= fdp->fd_nfiles ||
854 (fp = fdp->fd_files[uap->fd].fp) == NULL)
859 switch (fp->f_type) {
862 if (uap->name != _PC_PIPE_BUF) {
865 uap->sysmsg_result = PIPE_BUF;
871 vp = (struct vnode *)fp->f_data;
872 error = VOP_PATHCONF(vp, uap->name, uap->sysmsg_fds);
883 SYSCTL_INT(_debug, OID_AUTO, fdexpand, CTLFLAG_RD, &fdexpand, 0, "");
886 fdgrow(struct filedesc *fdp, int want)
888 struct fdnode *newfiles;
889 struct fdnode *oldfiles;
894 /* nf has to be of the form 2^n - 1 */
896 } while (nf <= want);
898 newfiles = malloc(nf * sizeof(struct fdnode), M_FILEDESC, M_WAITOK);
901 * deal with file-table extend race that might have occured
902 * when malloc was blocked.
904 if (fdp->fd_nfiles >= nf) {
905 free(newfiles, M_FILEDESC);
909 * Copy the existing ofile and ofileflags arrays
910 * and zero the new portion of each array.
912 extra = nf - fdp->fd_nfiles;
913 bcopy(fdp->fd_files, newfiles, fdp->fd_nfiles * sizeof(struct fdnode));
914 bzero(&newfiles[fdp->fd_nfiles], extra * sizeof(struct fdnode));
916 oldfiles = fdp->fd_files;
917 fdp->fd_files = newfiles;
920 if (oldfiles != fdp->fd_builtin_files)
921 free(oldfiles, M_FILEDESC);
926 * Number of nodes in right subtree, including the root.
929 right_subtree_size(int n)
931 return (n ^ (n | (n + 1)));
938 right_ancestor(int n)
940 return (n | (n + 1));
949 return ((n & (n + 1)) - 1);
953 fdreserve(struct filedesc *fdp, int fd, int incr)
956 fdp->fd_files[fd].allocated += incr;
957 KKASSERT(fdp->fd_files[fd].allocated >= 0);
958 fd = left_ancestor(fd);
963 * Allocate a file descriptor for the process.
966 fdalloc(struct proc *p, int want, int *result)
968 struct filedesc *fdp = p->p_fd;
969 int fd, rsize, rsum, node, lim;
971 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfilesperproc);
974 if (want >= fdp->fd_nfiles)
978 * Search for a free descriptor starting at the higher
979 * of want or fd_freefile. If that fails, consider
980 * expanding the ofile array.
983 /* move up the tree looking for a subtree with a free node */
984 for (fd = max(want, fdp->fd_freefile); fd < min(fdp->fd_nfiles, lim);
985 fd = right_ancestor(fd)) {
986 if (fdp->fd_files[fd].allocated == 0)
989 rsize = right_subtree_size(fd);
990 if (fdp->fd_files[fd].allocated == rsize)
991 continue; /* right subtree full */
994 * Free fd is in the right subtree of the tree rooted at fd.
995 * Call that subtree R. Look for the smallest (leftmost)
996 * subtree of R with an unallocated fd: continue moving
997 * down the left branch until encountering a full left
998 * subtree, then move to the right.
1000 for (rsum = 0, rsize /= 2; rsize > 0; rsize /= 2) {
1002 rsum += fdp->fd_files[node].allocated;
1003 if (fdp->fd_files[fd].allocated == rsum + rsize) {
1004 fd = node; /* move to the right */
1005 if (fdp->fd_files[node].allocated == 0)
1014 * No space in current array. Expand?
1016 if (fdp->fd_nfiles >= lim)
1022 KKASSERT(fd < fdp->fd_nfiles);
1023 fdp->fd_files[fd].fileflags = 0;
1024 if (fd > fdp->fd_lastfile)
1025 fdp->fd_lastfile = fd;
1026 if (want <= fdp->fd_freefile)
1027 fdp->fd_freefile = fd;
1029 KKASSERT(fdp->fd_files[fd].fp == NULL);
1030 fdreserve(fdp, fd, 1);
1035 * Check to see whether n user file descriptors
1036 * are available to the process p.
1039 fdavail(struct proc *p, int n)
1041 struct filedesc *fdp = p->p_fd;
1042 struct fdnode *fdnode;
1045 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfilesperproc);
1046 if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0)
1049 last = min(fdp->fd_nfiles, lim);
1050 fdnode = &fdp->fd_files[fdp->fd_freefile];
1051 for (i = last - fdp->fd_freefile; --i >= 0; ++fdnode) {
1052 if (fdnode->fp == NULL && --n <= 0)
1060 * Create a new open file structure and allocate a file decriptor
1061 * for the process that refers to it. If p is NULL, no descriptor
1062 * is allocated and the file pointer is returned unassociated with
1063 * any process. resultfd is only used if p is not NULL and may
1064 * separately be NULL indicating that you don't need the returned fd.
1066 * A held file pointer is returned. If a descriptor has been allocated
1067 * an additional hold on the fp will be made due to the fd_files[]
1071 falloc(struct proc *p, struct file **resultfp, int *resultfd)
1073 static struct timeval lastfail;
1081 * Handle filetable full issues and root overfill.
1083 if (nfiles >= maxfiles - maxfilesrootres &&
1084 ((p && p->p_ucred->cr_ruid != 0) || nfiles >= maxfiles)) {
1085 if (ppsratecheck(&lastfail, &curfail, 1)) {
1086 printf("kern.maxfiles limit exceeded by uid %d, please see tuning(7).\n",
1087 (p ? p->p_ucred->cr_ruid : -1));
1094 * Allocate a new file descriptor.
1097 fp = malloc(sizeof(struct file), M_FILE, M_WAITOK | M_ZERO);
1099 fp->f_ops = &badfileops;
1102 fp->f_cred = crhold(p->p_ucred);
1104 fp->f_cred = crhold(proc0.p_ucred);
1105 LIST_INSERT_HEAD(&filehead, fp, f_list);
1107 if ((error = fsetfd(p, fp, resultfd)) != 0) {
1108 fdrop(fp, p->p_thread);
1120 * Associate a file pointer with a file descriptor. On success the fp
1121 * will have an additional ref representing the fd_files[] association.
1124 fsetfd(struct proc *p, struct file *fp, int *resultfd)
1129 if ((error = fdalloc(p, 0, &fd)) == 0) {
1131 p->p_fd->fd_files[fd].fp = fp;
1138 funsetfd(struct filedesc *fdp, int fd)
1140 fdp->fd_files[fd].fp = NULL;
1141 fdp->fd_files[fd].fileflags = 0;
1142 fdreserve(fdp, fd, -1);
1143 if (fd < fdp->fd_freefile)
1144 fdp->fd_freefile = fd;
1148 fsetcred(struct file *fp, struct ucred *cr)
1156 * Free a file descriptor.
1159 ffree(struct file *fp)
1161 KASSERT((fp->f_count == 0), ("ffree: fp_fcount not 0!"));
1162 LIST_REMOVE(fp, f_list);
1165 cache_drop(fp->f_ncp);
1173 * Build a new filedesc structure.
1176 fdinit(struct proc *p)
1178 struct filedesc *newfdp;
1179 struct filedesc *fdp = p->p_fd;
1181 newfdp = malloc(sizeof(struct filedesc), M_FILEDESC, M_WAITOK|M_ZERO);
1183 newfdp->fd_cdir = fdp->fd_cdir;
1184 vref(newfdp->fd_cdir);
1185 newfdp->fd_ncdir = cache_hold(fdp->fd_ncdir);
1189 * rdir may not be set in e.g. proc0 or anything vm_fork'd off of
1190 * proc0, but should unconditionally exist in other processes.
1193 newfdp->fd_rdir = fdp->fd_rdir;
1194 vref(newfdp->fd_rdir);
1195 newfdp->fd_nrdir = cache_hold(fdp->fd_nrdir);
1198 newfdp->fd_jdir = fdp->fd_jdir;
1199 vref(newfdp->fd_jdir);
1200 newfdp->fd_njdir = cache_hold(fdp->fd_njdir);
1203 /* Create the file descriptor table. */
1204 newfdp->fd_refcnt = 1;
1205 newfdp->fd_cmask = cmask;
1206 newfdp->fd_files = newfdp->fd_builtin_files;
1207 newfdp->fd_nfiles = NDFILE;
1208 newfdp->fd_knlistsize = -1;
1214 * Share a filedesc structure.
1217 fdshare(struct proc *p)
1219 p->p_fd->fd_refcnt++;
1224 * Copy a filedesc structure.
1227 fdcopy(struct proc *p)
1229 struct filedesc *newfdp, *fdp = p->p_fd;
1230 struct fdnode *fdnode;
1233 /* Certain daemons might not have file descriptors. */
1237 newfdp = malloc(sizeof(struct filedesc), M_FILEDESC, M_WAITOK);
1239 if (newfdp->fd_cdir) {
1240 vref(newfdp->fd_cdir);
1241 newfdp->fd_ncdir = cache_hold(newfdp->fd_ncdir);
1244 * We must check for fd_rdir here, at least for now because
1245 * the init process is created before we have access to the
1246 * rootvode to take a reference to it.
1248 if (newfdp->fd_rdir) {
1249 vref(newfdp->fd_rdir);
1250 newfdp->fd_nrdir = cache_hold(newfdp->fd_nrdir);
1252 if (newfdp->fd_jdir) {
1253 vref(newfdp->fd_jdir);
1254 newfdp->fd_njdir = cache_hold(newfdp->fd_njdir);
1256 newfdp->fd_refcnt = 1;
1259 * If the number of open files fits in the internal arrays
1260 * of the open file structure, use them, otherwise allocate
1261 * additional memory for the number of descriptors currently
1264 if (newfdp->fd_lastfile < NDFILE) {
1265 newfdp->fd_files = newfdp->fd_builtin_files;
1269 * Compute the smallest file table size
1270 * for the file descriptors currently in use,
1271 * allowing the table to shrink.
1273 i = newfdp->fd_nfiles;
1274 while ((i-1)/2 > newfdp->fd_lastfile && (i-1)/2 > NDFILE)
1276 newfdp->fd_files = malloc(i * sizeof(struct fdnode),
1277 M_FILEDESC, M_WAITOK);
1279 newfdp->fd_nfiles = i;
1281 if (fdp->fd_files != fdp->fd_builtin_files ||
1282 newfdp->fd_files != newfdp->fd_builtin_files
1284 bcopy(fdp->fd_files, newfdp->fd_files,
1285 i * sizeof(struct fdnode));
1289 * kq descriptors cannot be copied.
1291 if (newfdp->fd_knlistsize != -1) {
1292 fdnode = &newfdp->fd_files[newfdp->fd_lastfile];
1293 for (i = newfdp->fd_lastfile; i >= 0; i--, fdnode--) {
1294 if (fdnode->fp != NULL && fdnode->fp->f_type == DTYPE_KQUEUE)
1295 funsetfd(newfdp, i); /* nulls out *fpp */
1296 if (fdnode->fp == NULL && i == newfdp->fd_lastfile && i > 0)
1297 newfdp->fd_lastfile--;
1299 newfdp->fd_knlist = NULL;
1300 newfdp->fd_knlistsize = -1;
1301 newfdp->fd_knhash = NULL;
1302 newfdp->fd_knhashmask = 0;
1305 fdnode = newfdp->fd_files;
1306 for (i = newfdp->fd_lastfile; i-- >= 0; fdnode++) {
1307 if (fdnode->fp != NULL)
1314 * Release a filedesc structure.
1317 fdfree(struct proc *p)
1319 struct thread *td = p->p_thread;
1320 struct filedesc *fdp = p->p_fd;
1321 struct fdnode *fdnode;
1323 struct filedesc_to_leader *fdtol;
1328 /* Certain daemons might not have file descriptors. */
1332 /* Check for special need to clear POSIX style locks */
1334 if (fdtol != NULL) {
1335 KASSERT(fdtol->fdl_refcount > 0,
1336 ("filedesc_to_refcount botch: fdl_refcount=%d",
1337 fdtol->fdl_refcount));
1338 if (fdtol->fdl_refcount == 1 &&
1339 (p->p_leader->p_flag & P_ADVLOCK) != 0) {
1341 fdnode = fdp->fd_files;
1342 for (i = 0; i <= fdp->fd_lastfile; i++, fdnode++) {
1343 if (fdnode->fp == NULL ||
1344 fdnode->fp->f_type != DTYPE_VNODE)
1348 lf.l_whence = SEEK_SET;
1351 lf.l_type = F_UNLCK;
1352 vp = (struct vnode *)fp->f_data;
1353 (void) VOP_ADVLOCK(vp,
1354 (caddr_t)p->p_leader,
1358 fdrop(fp, p->p_thread);
1359 /* reload due to possible reallocation */
1360 fdnode = &fdp->fd_files[i];
1364 if (fdtol->fdl_refcount == 1) {
1365 if (fdp->fd_holdleaderscount > 0 &&
1366 (p->p_leader->p_flag & P_ADVLOCK) != 0) {
1368 * close() or do_dup() has cleared a reference
1369 * in a shared file descriptor table.
1371 fdp->fd_holdleaderswakeup = 1;
1372 tsleep(&fdp->fd_holdleaderscount,
1376 if (fdtol->fdl_holdcount > 0) {
1378 * Ensure that fdtol->fdl_leader
1379 * remains valid in closef().
1381 fdtol->fdl_wakeup = 1;
1382 tsleep(fdtol, 0, "fdlhold", 0);
1386 fdtol->fdl_refcount--;
1387 if (fdtol->fdl_refcount == 0 &&
1388 fdtol->fdl_holdcount == 0) {
1389 fdtol->fdl_next->fdl_prev = fdtol->fdl_prev;
1390 fdtol->fdl_prev->fdl_next = fdtol->fdl_next;
1395 free(fdtol, M_FILEDESC_TO_LEADER);
1397 if (--fdp->fd_refcnt > 0)
1400 * we are the last reference to the structure, we can
1401 * safely assume it will not change out from under us.
1403 for (i = 0; i <= fdp->fd_lastfile; ++i) {
1404 if (fdp->fd_files[i].fp)
1405 closef(fdp->fd_files[i].fp, td);
1407 if (fdp->fd_files != fdp->fd_builtin_files)
1408 free(fdp->fd_files, M_FILEDESC);
1410 cache_drop(fdp->fd_ncdir);
1411 vrele(fdp->fd_cdir);
1414 cache_drop(fdp->fd_nrdir);
1415 vrele(fdp->fd_rdir);
1418 cache_drop(fdp->fd_njdir);
1419 vrele(fdp->fd_jdir);
1422 free(fdp->fd_knlist, M_KQUEUE);
1424 free(fdp->fd_knhash, M_KQUEUE);
1425 free(fdp, M_FILEDESC);
1429 * For setugid programs, we don't want to people to use that setugidness
1430 * to generate error messages which write to a file which otherwise would
1431 * otherwise be off-limits to the process.
1433 * This is a gross hack to plug the hole. A better solution would involve
1434 * a special vop or other form of generalized access control mechanism. We
1435 * go ahead and just reject all procfs file systems accesses as dangerous.
1437 * Since setugidsafety calls this only for fd 0, 1 and 2, this check is
1438 * sufficient. We also don't for check setugidness since we know we are.
1441 is_unsafe(struct file *fp)
1443 if (fp->f_type == DTYPE_VNODE &&
1444 ((struct vnode *)(fp->f_data))->v_tag == VT_PROCFS)
1450 * Make this setguid thing safe, if at all possible.
1453 setugidsafety(struct proc *p)
1455 struct thread *td = p->p_thread;
1456 struct filedesc *fdp = p->p_fd;
1459 /* Certain daemons might not have file descriptors. */
1464 * note: fdp->fd_files may be reallocated out from under us while
1465 * we are blocked in a close. Be careful!
1467 for (i = 0; i <= fdp->fd_lastfile; i++) {
1470 if (fdp->fd_files[i].fp && is_unsafe(fdp->fd_files[i].fp)) {
1474 if ((fdp->fd_files[i].fileflags & UF_MAPPED) != 0)
1475 (void) munmapfd(p, i);
1477 if (i < fdp->fd_knlistsize)
1478 knote_fdclose(p, i);
1480 * NULL-out descriptor prior to close to avoid
1481 * a race while close blocks.
1483 fp = fdp->fd_files[i].fp;
1488 while (fdp->fd_lastfile > 0 && fdp->fd_files[fdp->fd_lastfile].fp == NULL)
1493 * Close any files on exec?
1496 fdcloseexec(struct proc *p)
1498 struct thread *td = p->p_thread;
1499 struct filedesc *fdp = p->p_fd;
1502 /* Certain daemons might not have file descriptors. */
1507 * We cannot cache fd_files since operations may block and rip
1508 * them out from under us.
1510 for (i = 0; i <= fdp->fd_lastfile; i++) {
1511 if (fdp->fd_files[i].fp != NULL &&
1512 (fdp->fd_files[i].fileflags & UF_EXCLOSE)) {
1516 if (fdp->fd_files[i].fileflags & UF_MAPPED)
1517 (void) munmapfd(p, i);
1519 if (i < fdp->fd_knlistsize)
1520 knote_fdclose(p, i);
1522 * NULL-out descriptor prior to close to avoid
1523 * a race while close blocks.
1525 fp = fdp->fd_files[i].fp;
1530 while (fdp->fd_lastfile > 0 && fdp->fd_files[fdp->fd_lastfile].fp == NULL)
1535 * It is unsafe for set[ug]id processes to be started with file
1536 * descriptors 0..2 closed, as these descriptors are given implicit
1537 * significance in the Standard C library. fdcheckstd() will create a
1538 * descriptor referencing /dev/null for each of stdin, stdout, and
1539 * stderr that is not already open.
1542 fdcheckstd(struct proc *p)
1544 struct thread *td = p->p_thread;
1545 struct nlookupdata nd;
1546 struct filedesc *fdp;
1549 int fd, i, error, flags, devnull;
1556 for (i = 0; i < 3; i++) {
1557 if (fdp->fd_files[i].fp != NULL)
1560 if ((error = falloc(p, &fp, NULL)) != 0)
1563 error = nlookup_init(&nd, "/dev/null", UIO_SYSSPACE,
1564 NLC_FOLLOW|NLC_LOCKVP);
1565 flags = FREAD | FWRITE;
1567 error = vn_open(&nd, fp, flags, 0);
1569 error = fsetfd(p, fp, &fd);
1577 error = kern_dup(DUP_FIXED, devnull, i, &retval);
1586 * Internal form of close.
1587 * Decrement reference count on file structure.
1588 * Note: td and/or p may be NULL when closing a file
1589 * that was being passed in a message.
1592 closef(struct file *fp, struct thread *td)
1596 struct filedesc_to_leader *fdtol;
1603 p = NULL; /* allow no proc association */
1605 p = td->td_proc; /* can also be NULL */
1608 * POSIX record locking dictates that any close releases ALL
1609 * locks owned by this process. This is handled by setting
1610 * a flag in the unlock to free ONLY locks obeying POSIX
1611 * semantics, and not to free BSD-style file locks.
1612 * If the descriptor was in a message, POSIX-style locks
1613 * aren't passed with the descriptor.
1616 fp->f_type == DTYPE_VNODE) {
1617 if ((p->p_leader->p_flag & P_ADVLOCK) != 0) {
1618 lf.l_whence = SEEK_SET;
1621 lf.l_type = F_UNLCK;
1622 vp = (struct vnode *)fp->f_data;
1623 (void) VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_UNLCK,
1627 if (fdtol != NULL) {
1629 * Handle special case where file descriptor table
1630 * is shared between multiple process leaders.
1632 for (fdtol = fdtol->fdl_next;
1633 fdtol != p->p_fdtol;
1634 fdtol = fdtol->fdl_next) {
1635 if ((fdtol->fdl_leader->p_flag &
1638 fdtol->fdl_holdcount++;
1639 lf.l_whence = SEEK_SET;
1642 lf.l_type = F_UNLCK;
1643 vp = (struct vnode *)fp->f_data;
1644 (void) VOP_ADVLOCK(vp,
1645 (caddr_t)p->p_leader,
1646 F_UNLCK, &lf, F_POSIX);
1647 fdtol->fdl_holdcount--;
1648 if (fdtol->fdl_holdcount == 0 &&
1649 fdtol->fdl_wakeup != 0) {
1650 fdtol->fdl_wakeup = 0;
1656 return (fdrop(fp, td));
1660 fdrop(struct file *fp, struct thread *td)
1666 if (--fp->f_count > 0)
1668 if (fp->f_count < 0)
1669 panic("fdrop: count < 0");
1670 if ((fp->f_flag & FHASLOCK) && fp->f_type == DTYPE_VNODE) {
1671 lf.l_whence = SEEK_SET;
1674 lf.l_type = F_UNLCK;
1675 vp = (struct vnode *)fp->f_data;
1676 (void) VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK);
1678 if (fp->f_ops != &badfileops)
1679 error = fo_close(fp, td);
1687 * Apply an advisory lock on a file descriptor.
1689 * Just attempt to get a record lock of the requested type on
1690 * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
1694 flock(struct flock_args *uap)
1696 struct proc *p = curproc;
1697 struct filedesc *fdp = p->p_fd;
1702 if ((unsigned)uap->fd >= fdp->fd_nfiles ||
1703 (fp = fdp->fd_files[uap->fd].fp) == NULL)
1705 if (fp->f_type != DTYPE_VNODE)
1706 return (EOPNOTSUPP);
1707 vp = (struct vnode *)fp->f_data;
1708 lf.l_whence = SEEK_SET;
1711 if (uap->how & LOCK_UN) {
1712 lf.l_type = F_UNLCK;
1713 fp->f_flag &= ~FHASLOCK;
1714 return (VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK));
1716 if (uap->how & LOCK_EX)
1717 lf.l_type = F_WRLCK;
1718 else if (uap->how & LOCK_SH)
1719 lf.l_type = F_RDLCK;
1722 fp->f_flag |= FHASLOCK;
1723 if (uap->how & LOCK_NB)
1724 return (VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK));
1725 return (VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK|F_WAIT));
1729 * File Descriptor pseudo-device driver (/dev/fd/).
1731 * Opening minor device N dup()s the file (if any) connected to file
1732 * descriptor N belonging to the calling process. Note that this driver
1733 * consists of only the ``open()'' routine, because all subsequent
1734 * references to this file will be direct to the other driver.
1738 fdopen(dev_t dev, int mode, int type, struct thread *td)
1740 KKASSERT(td->td_proc != NULL);
1743 * XXX Kludge: set curproc->p_dupfd to contain the value of the
1744 * the file descriptor being sought for duplication. The error
1745 * return ensures that the vnode for this device will be released
1746 * by vn_open. Open will detect this special error and take the
1747 * actions in dupfdopen below. Other callers of vn_open or VOP_OPEN
1748 * will simply report the error.
1750 td->td_proc->p_dupfd = minor(dev);
1755 * Duplicate the specified descriptor to a free descriptor.
1758 dupfdopen(struct filedesc *fdp, int indx, int dfd, int mode, int error)
1764 * If the to-be-dup'd fd number is greater than the allowed number
1765 * of file descriptors, or the fd to be dup'd has already been
1766 * closed, then reject.
1768 if ((u_int)dfd >= fdp->fd_nfiles ||
1769 (wfp = fdp->fd_files[dfd].fp) == NULL) {
1774 * There are two cases of interest here.
1776 * For ENODEV simply dup (dfd) to file descriptor
1777 * (indx) and return.
1779 * For ENXIO steal away the file structure from (dfd) and
1780 * store it in (indx). (dfd) is effectively closed by
1783 * Any other error code is just returned.
1788 * Check that the mode the file is being opened for is a
1789 * subset of the mode of the existing descriptor.
1791 if (((mode & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag)
1793 fp = fdp->fd_files[indx].fp;
1795 if (fp && fdp->fd_files[indx].fileflags & UF_MAPPED)
1796 (void) munmapfd(p, indx);
1798 fdp->fd_files[indx].fp = wfp;
1799 fdp->fd_files[indx].fileflags = fdp->fd_files[dfd].fileflags;
1801 if (indx > fdp->fd_lastfile)
1802 fdp->fd_lastfile = indx;
1804 * we now own the reference to fp that the ofiles[] array
1805 * used to own. Release it.
1808 fdrop(fp, curthread);
1813 * Steal away the file pointer from dfd, and stuff it into indx.
1815 fp = fdp->fd_files[indx].fp;
1817 if (fp && fdp->fd_files[indx].fileflags & UF_MAPPED)
1818 (void) munmapfd(p, indx);
1820 fdp->fd_files[indx].fp = fdp->fd_files[dfd].fp;
1821 fdp->fd_files[indx].fileflags = fdp->fd_files[dfd].fileflags;
1825 * we now own the reference to fp that the files[] array
1826 * used to own. Release it.
1829 fdrop(fp, curthread);
1831 * Complete the clean up of the filedesc structure by
1832 * recomputing the various hints.
1834 if (indx > fdp->fd_lastfile) {
1835 fdp->fd_lastfile = indx;
1837 while (fdp->fd_lastfile > 0 &&
1838 fdp->fd_files[fdp->fd_lastfile].fp == NULL) {
1851 struct filedesc_to_leader *
1852 filedesc_to_leader_alloc(struct filedesc_to_leader *old,
1853 struct proc *leader)
1855 struct filedesc_to_leader *fdtol;
1857 fdtol = malloc(sizeof(struct filedesc_to_leader),
1858 M_FILEDESC_TO_LEADER, M_WAITOK);
1859 fdtol->fdl_refcount = 1;
1860 fdtol->fdl_holdcount = 0;
1861 fdtol->fdl_wakeup = 0;
1862 fdtol->fdl_leader = leader;
1864 fdtol->fdl_next = old->fdl_next;
1865 fdtol->fdl_prev = old;
1866 old->fdl_next = fdtol;
1867 fdtol->fdl_next->fdl_prev = fdtol;
1869 fdtol->fdl_next = fdtol;
1870 fdtol->fdl_prev = fdtol;
1876 * Get file structures.
1879 sysctl_kern_file(SYSCTL_HANDLER_ARGS)
1881 struct kinfo_file kf;
1882 struct filedesc *fdp;
1890 * Note: because the number of file descriptors is calculated
1891 * in different ways for sizing vs returning the data,
1892 * there is information leakage from the first loop. However,
1893 * it is of a similar order of magnitude to the leakage from
1894 * global system statistics such as kern.openfiles.
1896 * When just doing a count, note that we cannot just count
1897 * the elements and add f_count via the filehead list because
1898 * threaded processes share their descriptor table and f_count might
1899 * still be '1' in that case.
1903 LIST_FOREACH(p, &allproc, p_list) {
1904 if (p->p_stat == SIDL)
1906 if (!PRISON_CHECK(req->td->td_proc->p_ucred, p->p_ucred) != 0)
1908 if ((fdp = p->p_fd) == NULL)
1910 for (n = 0; n < fdp->fd_nfiles; ++n) {
1911 if ((fp = fdp->fd_files[n].fp) == NULL)
1913 if (req->oldptr == NULL) {
1916 kcore_make_file(&kf, fp, p->p_pid,
1917 p->p_ucred->cr_uid, n);
1918 error = SYSCTL_OUT(req, &kf, sizeof(kf));
1928 * When just calculating the size, overestimate a bit to try to
1929 * prevent system activity from causing the buffer-fill call
1932 if (req->oldptr == NULL) {
1933 count = (count + 16) + (count / 10);
1934 error = SYSCTL_OUT(req, NULL, count * sizeof(kf));
1939 SYSCTL_PROC(_kern, KERN_FILE, file, CTLTYPE_OPAQUE|CTLFLAG_RD,
1940 0, 0, sysctl_kern_file, "S,file", "Entire file table");
1942 SYSCTL_INT(_kern, KERN_MAXFILESPERPROC, maxfilesperproc, CTLFLAG_RW,
1943 &maxfilesperproc, 0, "Maximum files allowed open per process");
1945 SYSCTL_INT(_kern, KERN_MAXFILES, maxfiles, CTLFLAG_RW,
1946 &maxfiles, 0, "Maximum number of files");
1948 SYSCTL_INT(_kern, OID_AUTO, maxfilesrootres, CTLFLAG_RW,
1949 &maxfilesrootres, 0, "Descriptors reserved for root use");
1951 SYSCTL_INT(_kern, OID_AUTO, openfiles, CTLFLAG_RD,
1952 &nfiles, 0, "System-wide number of open files");
1955 fildesc_drvinit(void *unused)
1959 cdevsw_add(&fildesc_cdevsw, 0, 0);
1960 for (fd = 0; fd < NUMFDESC; fd++) {
1961 make_dev(&fildesc_cdevsw, fd,
1962 UID_BIN, GID_BIN, 0666, "fd/%d", fd);
1964 make_dev(&fildesc_cdevsw, 0, UID_ROOT, GID_WHEEL, 0666, "stdin");
1965 make_dev(&fildesc_cdevsw, 1, UID_ROOT, GID_WHEEL, 0666, "stdout");
1966 make_dev(&fildesc_cdevsw, 2, UID_ROOT, GID_WHEEL, 0666, "stderr");
1969 struct fileops badfileops = {
1994 badfo_ioctl(struct file *fp, u_long com, caddr_t data, struct thread *td)
2000 badfo_poll(struct file *fp, int events, struct ucred *cred, struct thread *td)
2006 badfo_kqfilter(struct file *fp, struct knote *kn)
2012 badfo_stat(struct file *fp, struct stat *sb, struct thread *td)
2018 badfo_close(struct file *fp, struct thread *td)
2024 badfo_shutdown(struct file *fp, int how, struct thread *td)
2030 nofo_shutdown(struct file *fp, int how, struct thread *td)
2032 return (EOPNOTSUPP);
2035 SYSINIT(fildescdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,
2036 fildesc_drvinit,NULL)