2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
40 #include "opt_compat.h"
41 #include "opt_kdtrace.h"
42 #include "opt_ktrace.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
48 #include <sys/sysent.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/mutex.h>
52 #include <sys/sysproto.h>
53 #include <sys/namei.h>
54 #include <sys/filedesc.h>
55 #include <sys/kernel.h>
56 #include <sys/fcntl.h>
58 #include <sys/filio.h>
59 #include <sys/limits.h>
60 #include <sys/linker.h>
64 #include <sys/unistd.h>
65 #include <sys/vnode.h>
68 #include <sys/dirent.h>
70 #include <sys/syscallsubr.h>
71 #include <sys/sysctl.h>
73 #include <sys/ktrace.h>
76 #include <machine/stdarg.h>
78 #include <security/audit/audit.h>
79 #include <security/mac/mac_framework.h>
82 #include <vm/vm_object.h>
83 #include <vm/vm_page.h>
86 SDT_PROVIDER_DEFINE(vfs);
87 SDT_PROBE_DEFINE(vfs, , stat, mode);
88 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
89 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
90 SDT_PROBE_DEFINE(vfs, , stat, reg);
91 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
92 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
94 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
95 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
96 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
97 static int setfmode(struct thread *td, struct vnode *, int);
98 static int setfflags(struct thread *td, struct vnode *, int);
99 static int setutimes(struct thread *td, struct vnode *,
100 const struct timespec *, int, int);
101 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105 * The module initialization routine for POSIX asynchronous I/O will
106 * set this to the version of AIO that it implements. (Zero means
107 * that it is not implemented.) This value is used here by pathconf()
108 * and in kern_descrip.c by fpathconf().
110 int async_io_version;
113 static int syncprt = 0;
114 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
118 * Sync each mounted filesystem.
120 #ifndef _SYS_SYSPROTO_H_
129 struct sync_args *uap;
131 struct mount *mp, *nmp;
134 mtx_lock(&mountlist_mtx);
135 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
136 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
137 nmp = TAILQ_NEXT(mp, mnt_list);
140 vfslocked = VFS_LOCK_GIANT(mp);
141 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
142 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
145 mp->mnt_kern_flag &= ~MNTK_ASYNC;
147 vfs_msync(mp, MNT_NOWAIT);
148 VFS_SYNC(mp, MNT_NOWAIT);
151 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
152 mp->mnt_noasync == 0)
153 mp->mnt_kern_flag |= MNTK_ASYNC;
155 vn_finished_write(mp);
157 VFS_UNLOCK_GIANT(vfslocked);
158 mtx_lock(&mountlist_mtx);
159 nmp = TAILQ_NEXT(mp, mnt_list);
162 mtx_unlock(&mountlist_mtx);
167 * Change filesystem quotas.
169 #ifndef _SYS_SYSPROTO_H_
170 struct quotactl_args {
180 register struct quotactl_args /* {
192 AUDIT_ARG_CMD(uap->cmd);
193 AUDIT_ARG_UID(uap->uid);
194 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
196 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
197 UIO_USERSPACE, uap->path, td);
198 if ((error = namei(&nd)) != 0)
200 vfslocked = NDHASGIANT(&nd);
201 NDFREE(&nd, NDF_ONLY_PNBUF);
202 mp = nd.ni_vp->v_mount;
205 error = vfs_busy(mp, 0);
208 VFS_UNLOCK_GIANT(vfslocked);
211 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
213 VFS_UNLOCK_GIANT(vfslocked);
218 * Used by statfs conversion routines to scale the block size up if
219 * necessary so that all of the block counts are <= 'max_size'. Note
220 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
224 statfs_scale_blocks(struct statfs *sf, long max_size)
229 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
232 * Attempt to scale the block counts to give a more accurate
233 * overview to userland of the ratio of free space to used
234 * space. To do this, find the largest block count and compute
235 * a divisor that lets it fit into a signed integer <= max_size.
237 if (sf->f_bavail < 0)
238 count = -sf->f_bavail;
240 count = sf->f_bavail;
241 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
242 if (count <= max_size)
245 count >>= flsl(max_size);
252 sf->f_bsize <<= shift;
253 sf->f_blocks >>= shift;
254 sf->f_bfree >>= shift;
255 sf->f_bavail >>= shift;
259 * Get filesystem statistics.
261 #ifndef _SYS_SYSPROTO_H_
270 register struct statfs_args /* {
278 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
280 error = copyout(&sf, uap->buf, sizeof(sf));
285 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
289 struct statfs *sp, sb;
294 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
295 AUDITVNODE1, pathseg, path, td);
299 vfslocked = NDHASGIANT(&nd);
300 mp = nd.ni_vp->v_mount;
302 NDFREE(&nd, NDF_ONLY_PNBUF);
304 error = vfs_busy(mp, 0);
307 VFS_UNLOCK_GIANT(vfslocked);
311 error = mac_mount_check_stat(td->td_ucred, mp);
316 * Set these in case the underlying filesystem fails to do so.
319 sp->f_version = STATFS_VERSION;
320 sp->f_namemax = NAME_MAX;
321 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
322 error = VFS_STATFS(mp, sp);
325 if (priv_check(td, PRIV_VFS_GENERATION)) {
326 bcopy(sp, &sb, sizeof(sb));
327 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
328 prison_enforce_statfs(td->td_ucred, mp, &sb);
334 VFS_UNLOCK_GIANT(vfslocked);
339 * Get filesystem statistics.
341 #ifndef _SYS_SYSPROTO_H_
342 struct fstatfs_args {
350 register struct fstatfs_args /* {
358 error = kern_fstatfs(td, uap->fd, &sf);
360 error = copyout(&sf, uap->buf, sizeof(sf));
365 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
369 struct statfs *sp, sb;
375 error = getvnode(td->td_proc->p_fd, fd, &fp);
379 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
380 vn_lock(vp, LK_SHARED | LK_RETRY);
382 AUDIT_ARG_VNODE1(vp);
393 error = vfs_busy(mp, 0);
396 VFS_UNLOCK_GIANT(vfslocked);
400 error = mac_mount_check_stat(td->td_ucred, mp);
405 * Set these in case the underlying filesystem fails to do so.
408 sp->f_version = STATFS_VERSION;
409 sp->f_namemax = NAME_MAX;
410 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
411 error = VFS_STATFS(mp, sp);
414 if (priv_check(td, PRIV_VFS_GENERATION)) {
415 bcopy(sp, &sb, sizeof(sb));
416 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
417 prison_enforce_statfs(td->td_ucred, mp, &sb);
424 VFS_UNLOCK_GIANT(vfslocked);
429 * Get statistics on all filesystems.
431 #ifndef _SYS_SYSPROTO_H_
432 struct getfsstat_args {
441 register struct getfsstat_args /* {
448 return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
453 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
454 * The caller is responsible for freeing memory which will be allocated
458 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
459 enum uio_seg bufseg, int flags)
461 struct mount *mp, *nmp;
462 struct statfs *sfsp, *sp, sb;
463 size_t count, maxcount;
467 maxcount = bufsize / sizeof(struct statfs);
470 else if (bufseg == UIO_USERSPACE)
472 else /* if (bufseg == UIO_SYSSPACE) */ {
474 mtx_lock(&mountlist_mtx);
475 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
478 mtx_unlock(&mountlist_mtx);
479 if (maxcount > count)
481 sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
485 mtx_lock(&mountlist_mtx);
486 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
487 if (prison_canseemount(td->td_ucred, mp) != 0) {
488 nmp = TAILQ_NEXT(mp, mnt_list);
492 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
493 nmp = TAILQ_NEXT(mp, mnt_list);
497 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
498 nmp = TAILQ_NEXT(mp, mnt_list);
501 vfslocked = VFS_LOCK_GIANT(mp);
502 if (sfsp && count < maxcount) {
505 * Set these in case the underlying filesystem
508 sp->f_version = STATFS_VERSION;
509 sp->f_namemax = NAME_MAX;
510 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
512 * If MNT_NOWAIT or MNT_LAZY is specified, do not
513 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
514 * overrides MNT_WAIT.
516 if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
517 (flags & MNT_WAIT)) &&
518 (error = VFS_STATFS(mp, sp))) {
519 VFS_UNLOCK_GIANT(vfslocked);
520 mtx_lock(&mountlist_mtx);
521 nmp = TAILQ_NEXT(mp, mnt_list);
525 if (priv_check(td, PRIV_VFS_GENERATION)) {
526 bcopy(sp, &sb, sizeof(sb));
527 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
528 prison_enforce_statfs(td->td_ucred, mp, &sb);
531 if (bufseg == UIO_SYSSPACE)
532 bcopy(sp, sfsp, sizeof(*sp));
533 else /* if (bufseg == UIO_USERSPACE) */ {
534 error = copyout(sp, sfsp, sizeof(*sp));
537 VFS_UNLOCK_GIANT(vfslocked);
543 VFS_UNLOCK_GIANT(vfslocked);
545 mtx_lock(&mountlist_mtx);
546 nmp = TAILQ_NEXT(mp, mnt_list);
549 mtx_unlock(&mountlist_mtx);
550 if (sfsp && count > maxcount)
551 td->td_retval[0] = maxcount;
553 td->td_retval[0] = count;
557 #ifdef COMPAT_FREEBSD4
559 * Get old format filesystem statistics.
561 static void cvtstatfs(struct statfs *, struct ostatfs *);
563 #ifndef _SYS_SYSPROTO_H_
564 struct freebsd4_statfs_args {
570 freebsd4_statfs(td, uap)
572 struct freebsd4_statfs_args /* {
581 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
584 cvtstatfs(&sf, &osb);
585 return (copyout(&osb, uap->buf, sizeof(osb)));
589 * Get filesystem statistics.
591 #ifndef _SYS_SYSPROTO_H_
592 struct freebsd4_fstatfs_args {
598 freebsd4_fstatfs(td, uap)
600 struct freebsd4_fstatfs_args /* {
609 error = kern_fstatfs(td, uap->fd, &sf);
612 cvtstatfs(&sf, &osb);
613 return (copyout(&osb, uap->buf, sizeof(osb)));
617 * Get statistics on all filesystems.
619 #ifndef _SYS_SYSPROTO_H_
620 struct freebsd4_getfsstat_args {
627 freebsd4_getfsstat(td, uap)
629 register struct freebsd4_getfsstat_args /* {
635 struct statfs *buf, *sp;
640 count = uap->bufsize / sizeof(struct ostatfs);
641 size = count * sizeof(struct statfs);
642 error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
644 count = td->td_retval[0];
646 while (count > 0 && error == 0) {
648 error = copyout(&osb, uap->buf, sizeof(osb));
659 * Implement fstatfs() for (NFS) file handles.
661 #ifndef _SYS_SYSPROTO_H_
662 struct freebsd4_fhstatfs_args {
663 struct fhandle *u_fhp;
668 freebsd4_fhstatfs(td, uap)
670 struct freebsd4_fhstatfs_args /* {
671 struct fhandle *u_fhp;
680 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
683 error = kern_fhstatfs(td, fh, &sf);
686 cvtstatfs(&sf, &osb);
687 return (copyout(&osb, uap->buf, sizeof(osb)));
691 * Convert a new format statfs structure to an old format statfs structure.
699 statfs_scale_blocks(nsp, LONG_MAX);
700 bzero(osp, sizeof(*osp));
701 osp->f_bsize = nsp->f_bsize;
702 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
703 osp->f_blocks = nsp->f_blocks;
704 osp->f_bfree = nsp->f_bfree;
705 osp->f_bavail = nsp->f_bavail;
706 osp->f_files = MIN(nsp->f_files, LONG_MAX);
707 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
708 osp->f_owner = nsp->f_owner;
709 osp->f_type = nsp->f_type;
710 osp->f_flags = nsp->f_flags;
711 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
712 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
713 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
714 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
715 strlcpy(osp->f_fstypename, nsp->f_fstypename,
716 MIN(MFSNAMELEN, OMFSNAMELEN));
717 strlcpy(osp->f_mntonname, nsp->f_mntonname,
718 MIN(MNAMELEN, OMNAMELEN));
719 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
720 MIN(MNAMELEN, OMNAMELEN));
721 osp->f_fsid = nsp->f_fsid;
723 #endif /* COMPAT_FREEBSD4 */
726 * Change current working directory to a given file descriptor.
728 #ifndef _SYS_SYSPROTO_H_
736 struct fchdir_args /* {
740 register struct filedesc *fdp = td->td_proc->p_fd;
741 struct vnode *vp, *tdp, *vpold;
747 AUDIT_ARG_FD(uap->fd);
748 if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
753 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
754 vn_lock(vp, LK_SHARED | LK_RETRY);
755 AUDIT_ARG_VNODE1(vp);
756 error = change_dir(vp, td);
757 while (!error && (mp = vp->v_mountedhere) != NULL) {
761 tvfslocked = VFS_LOCK_GIANT(mp);
762 error = VFS_ROOT(mp, LK_SHARED, &tdp);
765 VFS_UNLOCK_GIANT(tvfslocked);
769 VFS_UNLOCK_GIANT(vfslocked);
771 vfslocked = tvfslocked;
775 VFS_UNLOCK_GIANT(vfslocked);
779 VFS_UNLOCK_GIANT(vfslocked);
781 vpold = fdp->fd_cdir;
783 FILEDESC_XUNLOCK(fdp);
784 vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
786 VFS_UNLOCK_GIANT(vfslocked);
791 * Change current working directory (``.'').
793 #ifndef _SYS_SYSPROTO_H_
801 struct chdir_args /* {
806 return (kern_chdir(td, uap->path, UIO_USERSPACE));
810 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
812 register struct filedesc *fdp = td->td_proc->p_fd;
818 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
819 MPSAFE, pathseg, path, td);
820 if ((error = namei(&nd)) != 0)
822 vfslocked = NDHASGIANT(&nd);
823 if ((error = change_dir(nd.ni_vp, td)) != 0) {
825 VFS_UNLOCK_GIANT(vfslocked);
826 NDFREE(&nd, NDF_ONLY_PNBUF);
829 VOP_UNLOCK(nd.ni_vp, 0);
830 VFS_UNLOCK_GIANT(vfslocked);
831 NDFREE(&nd, NDF_ONLY_PNBUF);
834 fdp->fd_cdir = nd.ni_vp;
835 FILEDESC_XUNLOCK(fdp);
836 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
838 VFS_UNLOCK_GIANT(vfslocked);
843 * Helper function for raised chroot(2) security function: Refuse if
844 * any filedescriptors are open directories.
847 chroot_refuse_vdir_fds(fdp)
848 struct filedesc *fdp;
854 FILEDESC_LOCK_ASSERT(fdp);
856 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
857 fp = fget_locked(fdp, fd);
860 if (fp->f_type == DTYPE_VNODE) {
862 if (vp->v_type == VDIR)
870 * This sysctl determines if we will allow a process to chroot(2) if it
871 * has a directory open:
872 * 0: disallowed for all processes.
873 * 1: allowed for processes that were not already chroot(2)'ed.
874 * 2: allowed for all processes.
877 static int chroot_allow_open_directories = 1;
879 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
880 &chroot_allow_open_directories, 0, "");
883 * Change notion of root (``/'') directory.
885 #ifndef _SYS_SYSPROTO_H_
893 struct chroot_args /* {
901 error = priv_check(td, PRIV_VFS_CHROOT);
904 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
905 AUDITVNODE1, UIO_USERSPACE, uap->path, td);
909 vfslocked = NDHASGIANT(&nd);
910 if ((error = change_dir(nd.ni_vp, td)) != 0)
913 if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
916 VOP_UNLOCK(nd.ni_vp, 0);
917 error = change_root(nd.ni_vp, td);
919 VFS_UNLOCK_GIANT(vfslocked);
920 NDFREE(&nd, NDF_ONLY_PNBUF);
924 VFS_UNLOCK_GIANT(vfslocked);
926 NDFREE(&nd, NDF_ONLY_PNBUF);
931 * Common routine for chroot and chdir. Callers must provide a locked vnode
941 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
942 if (vp->v_type != VDIR)
945 error = mac_vnode_check_chdir(td->td_ucred, vp);
949 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
954 * Common routine for kern_chroot() and jail_attach(). The caller is
955 * responsible for invoking priv_check() and mac_vnode_check_chroot() to
956 * authorize this operation.
963 struct filedesc *fdp;
968 VFS_ASSERT_GIANT(vp->v_mount);
969 fdp = td->td_proc->p_fd;
971 if (chroot_allow_open_directories == 0 ||
972 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
973 error = chroot_refuse_vdir_fds(fdp);
975 FILEDESC_XUNLOCK(fdp);
979 oldvp = fdp->fd_rdir;
986 FILEDESC_XUNLOCK(fdp);
987 vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
989 VFS_UNLOCK_GIANT(vfslocked);
994 * Check permissions, allocate an open file structure, and call the device
995 * open routine if any.
997 #ifndef _SYS_SYSPROTO_H_
1007 register struct open_args /* {
1014 return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1017 #ifndef _SYS_SYSPROTO_H_
1018 struct openat_args {
1026 openat(struct thread *td, struct openat_args *uap)
1029 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1034 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1038 return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1042 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1043 int flags, int mode)
1045 struct proc *p = td->td_proc;
1046 struct filedesc *fdp = p->p_fd;
1053 int type, indx, error;
1055 struct nameidata nd;
1058 AUDIT_ARG_FFLAGS(flags);
1059 AUDIT_ARG_MODE(mode);
1060 /* XXX: audit dirfd */
1062 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR may
1065 if (flags & O_EXEC) {
1066 if (flags & O_ACCMODE)
1068 } else if ((flags & O_ACCMODE) == O_ACCMODE)
1071 flags = FFLAGS(flags);
1073 error = falloc(td, &nfp, &indx);
1076 /* An extra reference on `nfp' has been held for us by falloc(). */
1078 /* Set the flags early so the finit in devfs can pick them up. */
1079 fp->f_flag = flags & FMASK;
1080 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1081 NDINIT_AT(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, path, fd,
1083 td->td_dupfd = -1; /* XXX check for fdopen */
1084 error = vn_open(&nd, &flags, cmode, fp);
1087 * If the vn_open replaced the method vector, something
1088 * wonderous happened deep below and we just pass it up
1089 * pretending we know what we do.
1091 if (error == ENXIO && fp->f_ops != &badfileops) {
1093 td->td_retval[0] = indx;
1098 * handle special fdopen() case. bleh. dupfdopen() is
1099 * responsible for dropping the old contents of ofiles[indx]
1102 if ((error == ENODEV || error == ENXIO) &&
1103 td->td_dupfd >= 0 && /* XXX from fdopen */
1105 dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1106 td->td_retval[0] = indx;
1111 * Clean up the descriptor, but only if another thread hadn't
1112 * replaced or closed it.
1114 fdclose(fdp, fp, indx, td);
1117 if (error == ERESTART)
1122 vfslocked = NDHASGIANT(&nd);
1123 NDFREE(&nd, NDF_ONLY_PNBUF);
1126 fp->f_vnode = vp; /* XXX Does devfs need this? */
1128 * If the file wasn't claimed by devfs bind it to the normal
1129 * vnode operations here.
1131 if (fp->f_ops == &badfileops) {
1132 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1134 finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1138 if (flags & (O_EXLOCK | O_SHLOCK)) {
1139 lf.l_whence = SEEK_SET;
1142 if (flags & O_EXLOCK)
1143 lf.l_type = F_WRLCK;
1145 lf.l_type = F_RDLCK;
1147 if ((flags & FNONBLOCK) == 0)
1149 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1152 atomic_set_int(&fp->f_flag, FHASLOCK);
1154 if (flags & O_TRUNC) {
1155 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1159 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1161 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
1164 error = VOP_SETATTR(vp, &vat, td->td_ucred);
1166 vn_finished_write(mp);
1170 VFS_UNLOCK_GIANT(vfslocked);
1172 * Release our private reference, leaving the one associated with
1173 * the descriptor table intact.
1176 td->td_retval[0] = indx;
1179 VFS_UNLOCK_GIANT(vfslocked);
1180 fdclose(fdp, fp, indx, td);
1189 #ifndef _SYS_SYSPROTO_H_
1190 struct ocreat_args {
1198 register struct ocreat_args /* {
1204 return (kern_open(td, uap->path, UIO_USERSPACE,
1205 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1207 #endif /* COMPAT_43 */
1210 * Create a special file.
1212 #ifndef _SYS_SYSPROTO_H_
1222 register struct mknod_args /* {
1229 return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1232 #ifndef _SYS_SYSPROTO_H_
1233 struct mknodat_args {
1241 mknodat(struct thread *td, struct mknodat_args *uap)
1244 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1249 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1253 return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1257 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1265 struct nameidata nd;
1268 AUDIT_ARG_MODE(mode);
1270 switch (mode & S_IFMT) {
1273 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1276 error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1279 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1283 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1293 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1294 pathseg, path, fd, td);
1295 if ((error = namei(&nd)) != 0)
1297 vfslocked = NDHASGIANT(&nd);
1300 NDFREE(&nd, NDF_ONLY_PNBUF);
1301 if (vp == nd.ni_dvp)
1306 VFS_UNLOCK_GIANT(vfslocked);
1310 vattr.va_mode = (mode & ALLPERMS) &
1311 ~td->td_proc->p_fd->fd_cmask;
1312 vattr.va_rdev = dev;
1315 switch (mode & S_IFMT) {
1316 case S_IFMT: /* used by badsect to flag bad sectors */
1317 vattr.va_type = VBAD;
1320 vattr.va_type = VCHR;
1323 vattr.va_type = VBLK;
1329 panic("kern_mknod: invalid mode");
1332 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1333 NDFREE(&nd, NDF_ONLY_PNBUF);
1335 VFS_UNLOCK_GIANT(vfslocked);
1336 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1341 if (error == 0 && !whiteout)
1342 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1343 &nd.ni_cnd, &vattr);
1347 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1349 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1350 &nd.ni_cnd, &vattr);
1355 NDFREE(&nd, NDF_ONLY_PNBUF);
1357 vn_finished_write(mp);
1358 VFS_UNLOCK_GIANT(vfslocked);
1363 * Create a named pipe.
1365 #ifndef _SYS_SYSPROTO_H_
1366 struct mkfifo_args {
1374 register struct mkfifo_args /* {
1380 return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1383 #ifndef _SYS_SYSPROTO_H_
1384 struct mkfifoat_args {
1391 mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1394 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1399 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1402 return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1406 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1412 struct nameidata nd;
1415 AUDIT_ARG_MODE(mode);
1418 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1419 pathseg, path, fd, td);
1420 if ((error = namei(&nd)) != 0)
1422 vfslocked = NDHASGIANT(&nd);
1423 if (nd.ni_vp != NULL) {
1424 NDFREE(&nd, NDF_ONLY_PNBUF);
1425 if (nd.ni_vp == nd.ni_dvp)
1430 VFS_UNLOCK_GIANT(vfslocked);
1433 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1434 NDFREE(&nd, NDF_ONLY_PNBUF);
1436 VFS_UNLOCK_GIANT(vfslocked);
1437 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1442 vattr.va_type = VFIFO;
1443 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1445 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1450 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1457 vn_finished_write(mp);
1458 VFS_UNLOCK_GIANT(vfslocked);
1459 NDFREE(&nd, NDF_ONLY_PNBUF);
1464 * Make a hard file link.
1466 #ifndef _SYS_SYSPROTO_H_
1475 register struct link_args /* {
1481 return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1484 #ifndef _SYS_SYSPROTO_H_
1485 struct linkat_args {
1494 linkat(struct thread *td, struct linkat_args *uap)
1499 if (flag & ~AT_SYMLINK_FOLLOW)
1502 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1503 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1506 int hardlink_check_uid = 0;
1507 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1508 &hardlink_check_uid, 0,
1509 "Unprivileged processes cannot create hard links to files owned by other "
1511 static int hardlink_check_gid = 0;
1512 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1513 &hardlink_check_gid, 0,
1514 "Unprivileged processes cannot create hard links to files owned by other "
1518 can_hardlink(struct vnode *vp, struct ucred *cred)
1523 if (!hardlink_check_uid && !hardlink_check_gid)
1526 error = VOP_GETATTR(vp, &va, cred);
1530 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1531 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1536 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1537 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1546 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1549 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1553 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1554 enum uio_seg segflg, int follow)
1558 struct nameidata nd;
1564 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1567 if ((error = namei(&nd)) != 0)
1569 vfslocked = NDHASGIANT(&nd);
1570 NDFREE(&nd, NDF_ONLY_PNBUF);
1572 if (vp->v_type == VDIR) {
1574 VFS_UNLOCK_GIANT(vfslocked);
1575 return (EPERM); /* POSIX */
1577 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1579 VFS_UNLOCK_GIANT(vfslocked);
1582 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1583 segflg, path2, fd2, td);
1584 if ((error = namei(&nd)) == 0) {
1585 lvfslocked = NDHASGIANT(&nd);
1586 if (nd.ni_vp != NULL) {
1587 if (nd.ni_dvp == nd.ni_vp)
1593 } else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1595 error = can_hardlink(vp, td->td_ucred);
1598 error = mac_vnode_check_link(td->td_ucred,
1599 nd.ni_dvp, vp, &nd.ni_cnd);
1602 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1606 NDFREE(&nd, NDF_ONLY_PNBUF);
1607 VFS_UNLOCK_GIANT(lvfslocked);
1610 vn_finished_write(mp);
1611 VFS_UNLOCK_GIANT(vfslocked);
1616 * Make a symbolic link.
1618 #ifndef _SYS_SYSPROTO_H_
1619 struct symlink_args {
1627 register struct symlink_args /* {
1633 return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1636 #ifndef _SYS_SYSPROTO_H_
1637 struct symlinkat_args {
1644 symlinkat(struct thread *td, struct symlinkat_args *uap)
1647 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1652 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1655 return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1659 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1660 enum uio_seg segflg)
1666 struct nameidata nd;
1669 if (segflg == UIO_SYSSPACE) {
1672 syspath = uma_zalloc(namei_zone, M_WAITOK);
1673 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1676 AUDIT_ARG_TEXT(syspath);
1679 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1680 segflg, path2, fd, td);
1681 if ((error = namei(&nd)) != 0)
1683 vfslocked = NDHASGIANT(&nd);
1685 NDFREE(&nd, NDF_ONLY_PNBUF);
1686 if (nd.ni_vp == nd.ni_dvp)
1691 VFS_UNLOCK_GIANT(vfslocked);
1695 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1696 NDFREE(&nd, NDF_ONLY_PNBUF);
1698 VFS_UNLOCK_GIANT(vfslocked);
1699 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1704 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1706 vattr.va_type = VLNK;
1707 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1712 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1718 NDFREE(&nd, NDF_ONLY_PNBUF);
1720 vn_finished_write(mp);
1721 VFS_UNLOCK_GIANT(vfslocked);
1723 if (segflg != UIO_SYSSPACE)
1724 uma_zfree(namei_zone, syspath);
1729 * Delete a whiteout from the filesystem.
1734 register struct undelete_args /* {
1740 struct nameidata nd;
1745 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1746 UIO_USERSPACE, uap->path, td);
1750 vfslocked = NDHASGIANT(&nd);
1752 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1753 NDFREE(&nd, NDF_ONLY_PNBUF);
1754 if (nd.ni_vp == nd.ni_dvp)
1760 VFS_UNLOCK_GIANT(vfslocked);
1763 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1764 NDFREE(&nd, NDF_ONLY_PNBUF);
1766 VFS_UNLOCK_GIANT(vfslocked);
1767 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1771 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1772 NDFREE(&nd, NDF_ONLY_PNBUF);
1774 vn_finished_write(mp);
1775 VFS_UNLOCK_GIANT(vfslocked);
1780 * Delete a name from the filesystem.
1782 #ifndef _SYS_SYSPROTO_H_
1783 struct unlink_args {
1790 struct unlink_args /* {
1795 return (kern_unlink(td, uap->path, UIO_USERSPACE));
1798 #ifndef _SYS_SYSPROTO_H_
1799 struct unlinkat_args {
1806 unlinkat(struct thread *td, struct unlinkat_args *uap)
1808 int flag = uap->flag;
1810 char *path = uap->path;
1812 if (flag & ~AT_REMOVEDIR)
1815 if (flag & AT_REMOVEDIR)
1816 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1818 return (kern_unlinkat(td, fd, path, UIO_USERSPACE));
1822 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1825 return (kern_unlinkat(td, AT_FDCWD, path, pathseg));
1829 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
1834 struct nameidata nd;
1839 NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1840 pathseg, path, fd, td);
1841 if ((error = namei(&nd)) != 0)
1842 return (error == EINVAL ? EPERM : error);
1843 vfslocked = NDHASGIANT(&nd);
1845 if (vp->v_type == VDIR)
1846 error = EPERM; /* POSIX */
1849 * The root of a mounted filesystem cannot be deleted.
1851 * XXX: can this only be a VDIR case?
1853 if (vp->v_vflag & VV_ROOT)
1857 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1858 NDFREE(&nd, NDF_ONLY_PNBUF);
1860 if (vp == nd.ni_dvp)
1864 VFS_UNLOCK_GIANT(vfslocked);
1865 if ((error = vn_start_write(NULL, &mp,
1866 V_XSLEEP | PCATCH)) != 0)
1871 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1876 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1880 vn_finished_write(mp);
1882 NDFREE(&nd, NDF_ONLY_PNBUF);
1884 if (vp == nd.ni_dvp)
1888 VFS_UNLOCK_GIANT(vfslocked);
1893 * Reposition read/write file offset.
1895 #ifndef _SYS_SYSPROTO_H_
1906 register struct lseek_args /* {
1913 struct ucred *cred = td->td_ucred;
1921 AUDIT_ARG_FD(uap->fd);
1922 if ((error = fget(td, uap->fd, &fp)) != 0)
1924 if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1929 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1930 noneg = (vp->v_type != VCHR);
1931 offset = uap->offset;
1932 switch (uap->whence) {
1935 (fp->f_offset < 0 ||
1936 (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
1940 offset += fp->f_offset;
1943 vn_lock(vp, LK_SHARED | LK_RETRY);
1944 error = VOP_GETATTR(vp, &vattr, cred);
1949 (vattr.va_size > OFF_MAX ||
1950 (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1954 offset += vattr.va_size;
1959 error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1962 error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1967 if (error == 0 && noneg && offset < 0)
1971 fp->f_offset = offset;
1972 *(off_t *)(td->td_retval) = fp->f_offset;
1975 VFS_UNLOCK_GIANT(vfslocked);
1979 #if defined(COMPAT_43)
1981 * Reposition read/write file offset.
1983 #ifndef _SYS_SYSPROTO_H_
1984 struct olseek_args {
1993 register struct olseek_args /* {
1999 struct lseek_args /* {
2007 nuap.offset = uap->offset;
2008 nuap.whence = uap->whence;
2009 return (lseek(td, &nuap));
2011 #endif /* COMPAT_43 */
2013 /* Version with the 'pad' argument */
2015 freebsd6_lseek(td, uap)
2017 register struct freebsd6_lseek_args *uap;
2019 struct lseek_args ouap;
2022 ouap.offset = uap->offset;
2023 ouap.whence = uap->whence;
2024 return (lseek(td, &ouap));
2028 * Check access permissions using passed credentials.
2031 vn_access(vp, user_flags, cred, td)
2040 /* Flags == 0 means only check for existence. */
2044 if (user_flags & R_OK)
2046 if (user_flags & W_OK)
2048 if (user_flags & X_OK)
2051 error = mac_vnode_check_access(cred, vp, accmode);
2055 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2056 error = VOP_ACCESS(vp, accmode, cred, td);
2062 * Check access permissions using "real" credentials.
2064 #ifndef _SYS_SYSPROTO_H_
2065 struct access_args {
2073 register struct access_args /* {
2079 return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2082 #ifndef _SYS_SYSPROTO_H_
2083 struct faccessat_args {
2091 faccessat(struct thread *td, struct faccessat_args *uap)
2094 if (uap->flag & ~AT_EACCESS)
2096 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2101 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2104 return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2108 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2109 int flags, int mode)
2111 struct ucred *cred, *tmpcred;
2113 struct nameidata nd;
2118 * Create and modify a temporary credential instead of one that
2119 * is potentially shared. This could also mess up socket
2120 * buffer accounting which can run in an interrupt context.
2122 if (!(flags & AT_EACCESS)) {
2123 cred = td->td_ucred;
2124 tmpcred = crdup(cred);
2125 tmpcred->cr_uid = cred->cr_ruid;
2126 tmpcred->cr_groups[0] = cred->cr_rgid;
2127 td->td_ucred = tmpcred;
2129 cred = tmpcred = td->td_ucred;
2130 AUDIT_ARG_VALUE(mode);
2131 NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2132 AUDITVNODE1, pathseg, path, fd, td);
2133 if ((error = namei(&nd)) != 0)
2135 vfslocked = NDHASGIANT(&nd);
2138 error = vn_access(vp, mode, tmpcred, td);
2139 NDFREE(&nd, NDF_ONLY_PNBUF);
2141 VFS_UNLOCK_GIANT(vfslocked);
2143 if (!(flags & AT_EACCESS)) {
2144 td->td_ucred = cred;
2151 * Check access permissions using "effective" credentials.
2153 #ifndef _SYS_SYSPROTO_H_
2154 struct eaccess_args {
2162 register struct eaccess_args /* {
2168 return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2172 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2175 return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2178 #if defined(COMPAT_43)
2180 * Get file status; this version follows links.
2182 #ifndef _SYS_SYSPROTO_H_
2191 register struct ostat_args /* {
2200 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2204 error = copyout(&osb, uap->ub, sizeof (osb));
2209 * Get file status; this version does not follow links.
2211 #ifndef _SYS_SYSPROTO_H_
2212 struct olstat_args {
2220 register struct olstat_args /* {
2229 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2233 error = copyout(&osb, uap->ub, sizeof (osb));
2238 * Convert from an old to a new stat structure.
2246 ost->st_dev = st->st_dev;
2247 ost->st_ino = st->st_ino;
2248 ost->st_mode = st->st_mode;
2249 ost->st_nlink = st->st_nlink;
2250 ost->st_uid = st->st_uid;
2251 ost->st_gid = st->st_gid;
2252 ost->st_rdev = st->st_rdev;
2253 if (st->st_size < (quad_t)1 << 32)
2254 ost->st_size = st->st_size;
2257 ost->st_atime = st->st_atime;
2258 ost->st_mtime = st->st_mtime;
2259 ost->st_ctime = st->st_ctime;
2260 ost->st_blksize = st->st_blksize;
2261 ost->st_blocks = st->st_blocks;
2262 ost->st_flags = st->st_flags;
2263 ost->st_gen = st->st_gen;
2265 #endif /* COMPAT_43 */
2268 * Get file status; this version follows links.
2270 #ifndef _SYS_SYSPROTO_H_
2279 register struct stat_args /* {
2287 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2289 error = copyout(&sb, uap->ub, sizeof (sb));
2293 #ifndef _SYS_SYSPROTO_H_
2294 struct fstatat_args {
2302 fstatat(struct thread *td, struct fstatat_args *uap)
2307 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2308 UIO_USERSPACE, &sb);
2310 error = copyout(&sb, uap->buf, sizeof (sb));
2315 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2318 return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2322 kern_statat(struct thread *td, int flag, int fd, char *path,
2323 enum uio_seg pathseg, struct stat *sbp)
2326 return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2330 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2331 enum uio_seg pathseg, struct stat *sbp,
2332 void (*hook)(struct vnode *vp, struct stat *sbp))
2334 struct nameidata nd;
2336 int error, vfslocked;
2338 if (flag & ~AT_SYMLINK_NOFOLLOW)
2341 NDINIT_AT(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2342 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2345 if ((error = namei(&nd)) != 0)
2347 vfslocked = NDHASGIANT(&nd);
2348 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2350 SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2351 if (S_ISREG(sb.st_mode))
2352 SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2353 if (__predict_false(hook != NULL))
2354 hook(nd.ni_vp, &sb);
2356 NDFREE(&nd, NDF_ONLY_PNBUF);
2358 VFS_UNLOCK_GIANT(vfslocked);
2363 if (KTRPOINT(td, KTR_STRUCT))
2370 * Get file status; this version does not follow links.
2372 #ifndef _SYS_SYSPROTO_H_
2381 register struct lstat_args /* {
2389 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2391 error = copyout(&sb, uap->ub, sizeof (sb));
2396 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2399 return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2404 * Implementation of the NetBSD [l]stat() functions.
2411 bzero(nsb, sizeof *nsb);
2412 nsb->st_dev = sb->st_dev;
2413 nsb->st_ino = sb->st_ino;
2414 nsb->st_mode = sb->st_mode;
2415 nsb->st_nlink = sb->st_nlink;
2416 nsb->st_uid = sb->st_uid;
2417 nsb->st_gid = sb->st_gid;
2418 nsb->st_rdev = sb->st_rdev;
2419 nsb->st_atimespec = sb->st_atimespec;
2420 nsb->st_mtimespec = sb->st_mtimespec;
2421 nsb->st_ctimespec = sb->st_ctimespec;
2422 nsb->st_size = sb->st_size;
2423 nsb->st_blocks = sb->st_blocks;
2424 nsb->st_blksize = sb->st_blksize;
2425 nsb->st_flags = sb->st_flags;
2426 nsb->st_gen = sb->st_gen;
2427 nsb->st_birthtimespec = sb->st_birthtimespec;
2430 #ifndef _SYS_SYSPROTO_H_
2439 register struct nstat_args /* {
2448 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2451 cvtnstat(&sb, &nsb);
2452 error = copyout(&nsb, uap->ub, sizeof (nsb));
2457 * NetBSD lstat. Get file status; this version does not follow links.
2459 #ifndef _SYS_SYSPROTO_H_
2468 register struct nlstat_args /* {
2477 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2480 cvtnstat(&sb, &nsb);
2481 error = copyout(&nsb, uap->ub, sizeof (nsb));
2486 * Get configurable pathname variables.
2488 #ifndef _SYS_SYSPROTO_H_
2489 struct pathconf_args {
2497 register struct pathconf_args /* {
2503 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2506 #ifndef _SYS_SYSPROTO_H_
2507 struct lpathconf_args {
2515 register struct lpathconf_args /* {
2521 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2525 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2528 struct nameidata nd;
2529 int error, vfslocked;
2531 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 |
2532 flags, pathseg, path, td);
2533 if ((error = namei(&nd)) != 0)
2535 vfslocked = NDHASGIANT(&nd);
2536 NDFREE(&nd, NDF_ONLY_PNBUF);
2538 /* If asynchronous I/O is available, it works for all files. */
2539 if (name == _PC_ASYNC_IO)
2540 td->td_retval[0] = async_io_version;
2542 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2544 VFS_UNLOCK_GIANT(vfslocked);
2549 * Return target name of a symbolic link.
2551 #ifndef _SYS_SYSPROTO_H_
2552 struct readlink_args {
2561 register struct readlink_args /* {
2568 return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2569 UIO_USERSPACE, uap->count));
2571 #ifndef _SYS_SYSPROTO_H_
2572 struct readlinkat_args {
2580 readlinkat(struct thread *td, struct readlinkat_args *uap)
2583 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2584 uap->buf, UIO_USERSPACE, uap->bufsize));
2588 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2589 enum uio_seg bufseg, size_t count)
2592 return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2597 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2598 char *buf, enum uio_seg bufseg, size_t count)
2604 struct nameidata nd;
2607 if (count > INT_MAX)
2610 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2611 AUDITVNODE1, pathseg, path, fd, td);
2613 if ((error = namei(&nd)) != 0)
2615 NDFREE(&nd, NDF_ONLY_PNBUF);
2616 vfslocked = NDHASGIANT(&nd);
2619 error = mac_vnode_check_readlink(td->td_ucred, vp);
2622 VFS_UNLOCK_GIANT(vfslocked);
2626 if (vp->v_type != VLNK)
2629 aiov.iov_base = buf;
2630 aiov.iov_len = count;
2631 auio.uio_iov = &aiov;
2632 auio.uio_iovcnt = 1;
2633 auio.uio_offset = 0;
2634 auio.uio_rw = UIO_READ;
2635 auio.uio_segflg = bufseg;
2637 auio.uio_resid = count;
2638 error = VOP_READLINK(vp, &auio, td->td_ucred);
2641 VFS_UNLOCK_GIANT(vfslocked);
2642 td->td_retval[0] = count - auio.uio_resid;
2647 * Common implementation code for chflags() and fchflags().
2650 setfflags(td, vp, flags)
2660 * Prevent non-root users from setting flags on devices. When
2661 * a device is reused, users can retain ownership of the device
2662 * if they are allowed to set flags and programs assume that
2663 * chown can't fail when done as root.
2665 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2666 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2671 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2673 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2675 vattr.va_flags = flags;
2677 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2680 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2682 vn_finished_write(mp);
2687 * Change flags of a file given a path name.
2689 #ifndef _SYS_SYSPROTO_H_
2690 struct chflags_args {
2698 register struct chflags_args /* {
2704 struct nameidata nd;
2707 AUDIT_ARG_FFLAGS(uap->flags);
2708 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2710 if ((error = namei(&nd)) != 0)
2712 NDFREE(&nd, NDF_ONLY_PNBUF);
2713 vfslocked = NDHASGIANT(&nd);
2714 error = setfflags(td, nd.ni_vp, uap->flags);
2716 VFS_UNLOCK_GIANT(vfslocked);
2721 * Same as chflags() but doesn't follow symlinks.
2726 register struct lchflags_args /* {
2732 struct nameidata nd;
2735 AUDIT_ARG_FFLAGS(uap->flags);
2736 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2738 if ((error = namei(&nd)) != 0)
2740 vfslocked = NDHASGIANT(&nd);
2741 NDFREE(&nd, NDF_ONLY_PNBUF);
2742 error = setfflags(td, nd.ni_vp, uap->flags);
2744 VFS_UNLOCK_GIANT(vfslocked);
2749 * Change flags of a file given a file descriptor.
2751 #ifndef _SYS_SYSPROTO_H_
2752 struct fchflags_args {
2760 register struct fchflags_args /* {
2769 AUDIT_ARG_FD(uap->fd);
2770 AUDIT_ARG_FFLAGS(uap->flags);
2771 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2773 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2775 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2776 AUDIT_ARG_VNODE1(fp->f_vnode);
2777 VOP_UNLOCK(fp->f_vnode, 0);
2779 error = setfflags(td, fp->f_vnode, uap->flags);
2780 VFS_UNLOCK_GIANT(vfslocked);
2786 * Common implementation code for chmod(), lchmod() and fchmod().
2789 setfmode(td, vp, mode)
2798 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2800 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2802 vattr.va_mode = mode & ALLPERMS;
2804 error = mac_vnode_check_setmode(td->td_ucred, vp, vattr.va_mode);
2807 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2809 vn_finished_write(mp);
2814 * Change mode of a file given path name.
2816 #ifndef _SYS_SYSPROTO_H_
2825 register struct chmod_args /* {
2831 return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2834 #ifndef _SYS_SYSPROTO_H_
2835 struct fchmodat_args {
2843 fchmodat(struct thread *td, struct fchmodat_args *uap)
2845 int flag = uap->flag;
2847 char *path = uap->path;
2848 mode_t mode = uap->mode;
2850 if (flag & ~AT_SYMLINK_NOFOLLOW)
2853 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2857 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2860 return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2864 * Change mode of a file given path name (don't follow links.)
2866 #ifndef _SYS_SYSPROTO_H_
2867 struct lchmod_args {
2875 register struct lchmod_args /* {
2881 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2882 uap->mode, AT_SYMLINK_NOFOLLOW));
2887 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2888 mode_t mode, int flag)
2891 struct nameidata nd;
2895 AUDIT_ARG_MODE(mode);
2896 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2897 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
2899 if ((error = namei(&nd)) != 0)
2901 vfslocked = NDHASGIANT(&nd);
2902 NDFREE(&nd, NDF_ONLY_PNBUF);
2903 error = setfmode(td, nd.ni_vp, mode);
2905 VFS_UNLOCK_GIANT(vfslocked);
2910 * Change mode of a file given a file descriptor.
2912 #ifndef _SYS_SYSPROTO_H_
2913 struct fchmod_args {
2921 register struct fchmod_args /* {
2930 AUDIT_ARG_FD(uap->fd);
2931 AUDIT_ARG_MODE(uap->mode);
2932 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2934 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2936 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2937 AUDIT_ARG_VNODE1(fp->f_vnode);
2938 VOP_UNLOCK(fp->f_vnode, 0);
2940 error = setfmode(td, fp->f_vnode, uap->mode);
2941 VFS_UNLOCK_GIANT(vfslocked);
2947 * Common implementation for chown(), lchown(), and fchown()
2950 setfown(td, vp, uid, gid)
2960 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2962 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2967 error = mac_vnode_check_setowner(td->td_ucred, vp, vattr.va_uid,
2971 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2973 vn_finished_write(mp);
2978 * Set ownership given a path name.
2980 #ifndef _SYS_SYSPROTO_H_
2990 register struct chown_args /* {
2997 return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3000 #ifndef _SYS_SYSPROTO_H_
3001 struct fchownat_args {
3010 fchownat(struct thread *td, struct fchownat_args *uap)
3015 if (flag & ~AT_SYMLINK_NOFOLLOW)
3018 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3019 uap->gid, uap->flag));
3023 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3027 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3031 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3032 int uid, int gid, int flag)
3034 struct nameidata nd;
3035 int error, vfslocked, follow;
3037 AUDIT_ARG_OWNER(uid, gid);
3038 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3039 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
3042 if ((error = namei(&nd)) != 0)
3044 vfslocked = NDHASGIANT(&nd);
3045 NDFREE(&nd, NDF_ONLY_PNBUF);
3046 error = setfown(td, nd.ni_vp, uid, gid);
3048 VFS_UNLOCK_GIANT(vfslocked);
3053 * Set ownership given a path name, do not cross symlinks.
3055 #ifndef _SYS_SYSPROTO_H_
3056 struct lchown_args {
3065 register struct lchown_args /* {
3072 return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3076 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3080 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3081 AT_SYMLINK_NOFOLLOW));
3085 * Set ownership given a file descriptor.
3087 #ifndef _SYS_SYSPROTO_H_
3088 struct fchown_args {
3097 register struct fchown_args /* {
3107 AUDIT_ARG_FD(uap->fd);
3108 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3109 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3111 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3113 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3114 AUDIT_ARG_VNODE1(fp->f_vnode);
3115 VOP_UNLOCK(fp->f_vnode, 0);
3117 error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
3118 VFS_UNLOCK_GIANT(vfslocked);
3124 * Common implementation code for utimes(), lutimes(), and futimes().
3127 getutimes(usrtvp, tvpseg, tsp)
3128 const struct timeval *usrtvp;
3129 enum uio_seg tvpseg;
3130 struct timespec *tsp;
3132 struct timeval tv[2];
3133 const struct timeval *tvp;
3136 if (usrtvp == NULL) {
3137 vfs_timestamp(&tsp[0]);
3140 if (tvpseg == UIO_SYSSPACE) {
3143 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3148 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3149 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3151 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3152 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3158 * Common implementation code for utimes(), lutimes(), and futimes().
3161 setutimes(td, vp, ts, numtimes, nullflag)
3164 const struct timespec *ts;
3168 int error, setbirthtime;
3172 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3174 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3176 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3177 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3180 vattr.va_atime = ts[0];
3181 vattr.va_mtime = ts[1];
3183 vattr.va_birthtime = ts[1];
3185 vattr.va_birthtime = ts[2];
3187 vattr.va_vaflags |= VA_UTIMES_NULL;
3189 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3193 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3195 vn_finished_write(mp);
3200 * Set the access and modification times of a file.
3202 #ifndef _SYS_SYSPROTO_H_
3203 struct utimes_args {
3205 struct timeval *tptr;
3211 register struct utimes_args /* {
3213 struct timeval *tptr;
3217 return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3221 #ifndef _SYS_SYSPROTO_H_
3222 struct futimesat_args {
3225 const struct timeval * times;
3229 futimesat(struct thread *td, struct futimesat_args *uap)
3232 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3233 uap->times, UIO_USERSPACE));
3237 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3238 struct timeval *tptr, enum uio_seg tptrseg)
3241 return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3245 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3246 struct timeval *tptr, enum uio_seg tptrseg)
3248 struct nameidata nd;
3249 struct timespec ts[2];
3250 int error, vfslocked;
3252 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3254 NDINIT_AT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path,
3257 if ((error = namei(&nd)) != 0)
3259 vfslocked = NDHASGIANT(&nd);
3260 NDFREE(&nd, NDF_ONLY_PNBUF);
3261 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3263 VFS_UNLOCK_GIANT(vfslocked);
3268 * Set the access and modification times of a file.
3270 #ifndef _SYS_SYSPROTO_H_
3271 struct lutimes_args {
3273 struct timeval *tptr;
3279 register struct lutimes_args /* {
3281 struct timeval *tptr;
3285 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3290 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3291 struct timeval *tptr, enum uio_seg tptrseg)
3293 struct timespec ts[2];
3295 struct nameidata nd;
3298 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3300 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3301 if ((error = namei(&nd)) != 0)
3303 vfslocked = NDHASGIANT(&nd);
3304 NDFREE(&nd, NDF_ONLY_PNBUF);
3305 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3307 VFS_UNLOCK_GIANT(vfslocked);
3312 * Set the access and modification times of a file.
3314 #ifndef _SYS_SYSPROTO_H_
3315 struct futimes_args {
3317 struct timeval *tptr;
3323 register struct futimes_args /* {
3325 struct timeval *tptr;
3329 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3333 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3334 enum uio_seg tptrseg)
3336 struct timespec ts[2];
3342 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3344 if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
3346 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3348 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3349 AUDIT_ARG_VNODE1(fp->f_vnode);
3350 VOP_UNLOCK(fp->f_vnode, 0);
3352 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3353 VFS_UNLOCK_GIANT(vfslocked);
3359 * Truncate a file given its path name.
3361 #ifndef _SYS_SYSPROTO_H_
3362 struct truncate_args {
3371 register struct truncate_args /* {
3378 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3382 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3388 struct nameidata nd;
3393 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3394 if ((error = namei(&nd)) != 0)
3396 vfslocked = NDHASGIANT(&nd);
3398 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3400 VFS_UNLOCK_GIANT(vfslocked);
3403 NDFREE(&nd, NDF_ONLY_PNBUF);
3404 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3405 if (vp->v_type == VDIR)
3408 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3411 else if ((error = vn_writechk(vp)) == 0 &&
3412 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3414 vattr.va_size = length;
3415 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3418 vn_finished_write(mp);
3419 VFS_UNLOCK_GIANT(vfslocked);
3423 #if defined(COMPAT_43)
3425 * Truncate a file given its path name.
3427 #ifndef _SYS_SYSPROTO_H_
3428 struct otruncate_args {
3436 register struct otruncate_args /* {
3441 struct truncate_args /* {
3447 nuap.path = uap->path;
3448 nuap.length = uap->length;
3449 return (truncate(td, &nuap));
3451 #endif /* COMPAT_43 */
3453 /* Versions with the pad argument */
3455 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3457 struct truncate_args ouap;
3459 ouap.path = uap->path;
3460 ouap.length = uap->length;
3461 return (truncate(td, &ouap));
3465 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3467 struct ftruncate_args ouap;
3470 ouap.length = uap->length;
3471 return (ftruncate(td, &ouap));
3475 * Sync an open file.
3477 #ifndef _SYS_SYSPROTO_H_
3485 struct fsync_args /* {
3493 int error, lock_flags;
3495 AUDIT_ARG_FD(uap->fd);
3496 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3499 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3500 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3502 if (MNT_SHARED_WRITES(mp) ||
3503 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3504 lock_flags = LK_SHARED;
3506 lock_flags = LK_EXCLUSIVE;
3508 vn_lock(vp, lock_flags | LK_RETRY);
3509 AUDIT_ARG_VNODE1(vp);
3510 if (vp->v_object != NULL) {
3511 VM_OBJECT_LOCK(vp->v_object);
3512 vm_object_page_clean(vp->v_object, 0, 0, 0);
3513 VM_OBJECT_UNLOCK(vp->v_object);
3515 error = VOP_FSYNC(vp, MNT_WAIT, td);
3518 vn_finished_write(mp);
3520 VFS_UNLOCK_GIANT(vfslocked);
3526 * Rename files. Source and destination must either both be directories, or
3527 * both not be directories. If target is a directory, it must be empty.
3529 #ifndef _SYS_SYSPROTO_H_
3530 struct rename_args {
3538 register struct rename_args /* {
3544 return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3547 #ifndef _SYS_SYSPROTO_H_
3548 struct renameat_args {
3556 renameat(struct thread *td, struct renameat_args *uap)
3559 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3564 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3567 return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3571 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3572 enum uio_seg pathseg)
3574 struct mount *mp = NULL;
3575 struct vnode *tvp, *fvp, *tdvp;
3576 struct nameidata fromnd, tond;
3583 NDINIT_AT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE |
3584 AUDITVNODE1, pathseg, old, oldfd, td);
3586 NDINIT_AT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3587 AUDITVNODE1, pathseg, old, oldfd, td);
3590 if ((error = namei(&fromnd)) != 0)
3592 fvfslocked = NDHASGIANT(&fromnd);
3595 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3596 fromnd.ni_vp, &fromnd.ni_cnd);
3597 VOP_UNLOCK(fromnd.ni_dvp, 0);
3598 if (fromnd.ni_dvp != fromnd.ni_vp)
3599 VOP_UNLOCK(fromnd.ni_vp, 0);
3603 error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3605 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3606 vrele(fromnd.ni_dvp);
3610 NDINIT_AT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3611 MPSAFE | AUDITVNODE2, pathseg, new, newfd, td);
3612 if (fromnd.ni_vp->v_type == VDIR)
3613 tond.ni_cnd.cn_flags |= WILLBEDIR;
3614 if ((error = namei(&tond)) != 0) {
3615 /* Translate error code for rename("dir1", "dir2/."). */
3616 if (error == EISDIR && fvp->v_type == VDIR)
3618 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3619 vrele(fromnd.ni_dvp);
3621 vn_finished_write(mp);
3624 tvfslocked = NDHASGIANT(&tond);
3628 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3631 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3641 * If the source is the same as the destination (that is, if they
3642 * are links to the same vnode), then there is nothing to do.
3648 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3649 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3653 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3654 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3655 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3656 NDFREE(&tond, NDF_ONLY_PNBUF);
3658 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3659 NDFREE(&tond, NDF_ONLY_PNBUF);
3666 vrele(fromnd.ni_dvp);
3669 vrele(tond.ni_startdir);
3670 vn_finished_write(mp);
3672 if (fromnd.ni_startdir)
3673 vrele(fromnd.ni_startdir);
3674 VFS_UNLOCK_GIANT(fvfslocked);
3675 VFS_UNLOCK_GIANT(tvfslocked);
3682 * Make a directory file.
3684 #ifndef _SYS_SYSPROTO_H_
3693 register struct mkdir_args /* {
3699 return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3702 #ifndef _SYS_SYSPROTO_H_
3703 struct mkdirat_args {
3710 mkdirat(struct thread *td, struct mkdirat_args *uap)
3713 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3717 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3720 return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3724 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3731 struct nameidata nd;
3734 AUDIT_ARG_MODE(mode);
3737 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
3738 segflg, path, fd, td);
3739 nd.ni_cnd.cn_flags |= WILLBEDIR;
3740 if ((error = namei(&nd)) != 0)
3742 vfslocked = NDHASGIANT(&nd);
3745 NDFREE(&nd, NDF_ONLY_PNBUF);
3747 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3748 * the strange behaviour of leaving the vnode unlocked
3749 * if the target is the same vnode as the parent.
3751 if (vp == nd.ni_dvp)
3756 VFS_UNLOCK_GIANT(vfslocked);
3759 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3760 NDFREE(&nd, NDF_ONLY_PNBUF);
3762 VFS_UNLOCK_GIANT(vfslocked);
3763 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3768 vattr.va_type = VDIR;
3769 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3771 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3776 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3780 NDFREE(&nd, NDF_ONLY_PNBUF);
3784 vn_finished_write(mp);
3785 VFS_UNLOCK_GIANT(vfslocked);
3790 * Remove a directory file.
3792 #ifndef _SYS_SYSPROTO_H_
3800 struct rmdir_args /* {
3805 return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3809 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3812 return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3816 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3821 struct nameidata nd;
3826 NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
3827 pathseg, path, fd, td);
3828 if ((error = namei(&nd)) != 0)
3830 vfslocked = NDHASGIANT(&nd);
3832 if (vp->v_type != VDIR) {
3837 * No rmdir "." please.
3839 if (nd.ni_dvp == vp) {
3844 * The root of a mounted filesystem cannot be deleted.
3846 if (vp->v_vflag & VV_ROOT) {
3851 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3856 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3857 NDFREE(&nd, NDF_ONLY_PNBUF);
3859 if (nd.ni_dvp == vp)
3863 VFS_UNLOCK_GIANT(vfslocked);
3864 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3868 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3869 vn_finished_write(mp);
3871 NDFREE(&nd, NDF_ONLY_PNBUF);
3873 if (nd.ni_dvp == vp)
3877 VFS_UNLOCK_GIANT(vfslocked);
3883 * Read a block of directory entries in a filesystem independent format.
3885 #ifndef _SYS_SYSPROTO_H_
3886 struct ogetdirentries_args {
3894 ogetdirentries(td, uap)
3896 register struct ogetdirentries_args /* {
3905 struct uio auio, kuio;
3906 struct iovec aiov, kiov;
3907 struct dirent *dp, *edp;
3909 int error, eofflag, readcnt, vfslocked;
3912 /* XXX arbitrary sanity limit on `count'. */
3913 if (uap->count > 64 * 1024)
3915 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3917 if ((fp->f_flag & FREAD) == 0) {
3923 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3924 if (vp->v_type != VDIR) {
3925 VFS_UNLOCK_GIANT(vfslocked);
3929 aiov.iov_base = uap->buf;
3930 aiov.iov_len = uap->count;
3931 auio.uio_iov = &aiov;
3932 auio.uio_iovcnt = 1;
3933 auio.uio_rw = UIO_READ;
3934 auio.uio_segflg = UIO_USERSPACE;
3936 auio.uio_resid = uap->count;
3937 vn_lock(vp, LK_SHARED | LK_RETRY);
3938 loff = auio.uio_offset = fp->f_offset;
3940 error = mac_vnode_check_readdir(td->td_ucred, vp);
3943 VFS_UNLOCK_GIANT(vfslocked);
3948 # if (BYTE_ORDER != LITTLE_ENDIAN)
3949 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3950 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3952 fp->f_offset = auio.uio_offset;
3957 kuio.uio_iov = &kiov;
3958 kuio.uio_segflg = UIO_SYSSPACE;
3959 kiov.iov_len = uap->count;
3960 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3961 kiov.iov_base = dirbuf;
3962 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3964 fp->f_offset = kuio.uio_offset;
3966 readcnt = uap->count - kuio.uio_resid;
3967 edp = (struct dirent *)&dirbuf[readcnt];
3968 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3969 # if (BYTE_ORDER == LITTLE_ENDIAN)
3971 * The expected low byte of
3972 * dp->d_namlen is our dp->d_type.
3973 * The high MBZ byte of dp->d_namlen
3974 * is our dp->d_namlen.
3976 dp->d_type = dp->d_namlen;
3980 * The dp->d_type is the high byte
3981 * of the expected dp->d_namlen,
3982 * so must be zero'ed.
3986 if (dp->d_reclen > 0) {
3987 dp = (struct dirent *)
3988 ((char *)dp + dp->d_reclen);
3995 error = uiomove(dirbuf, readcnt, &auio);
3997 free(dirbuf, M_TEMP);
4001 VFS_UNLOCK_GIANT(vfslocked);
4005 if (uap->count == auio.uio_resid &&
4006 (vp->v_vflag & VV_ROOT) &&
4007 (vp->v_mount->mnt_flag & MNT_UNION)) {
4008 struct vnode *tvp = vp;
4009 vp = vp->v_mount->mnt_vnodecovered;
4015 VFS_UNLOCK_GIANT(vfslocked);
4019 VFS_UNLOCK_GIANT(vfslocked);
4020 error = copyout(&loff, uap->basep, sizeof(long));
4022 td->td_retval[0] = uap->count - auio.uio_resid;
4025 #endif /* COMPAT_43 */
4028 * Read a block of directory entries in a filesystem independent format.
4030 #ifndef _SYS_SYSPROTO_H_
4031 struct getdirentries_args {
4039 getdirentries(td, uap)
4041 register struct getdirentries_args /* {
4051 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4054 if (uap->basep != NULL)
4055 error = copyout(&base, uap->basep, sizeof(long));
4060 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4072 if (count > INT_MAX)
4074 if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
4076 if ((fp->f_flag & FREAD) == 0) {
4082 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4083 if (vp->v_type != VDIR) {
4084 VFS_UNLOCK_GIANT(vfslocked);
4088 aiov.iov_base = buf;
4089 aiov.iov_len = count;
4090 auio.uio_iov = &aiov;
4091 auio.uio_iovcnt = 1;
4092 auio.uio_rw = UIO_READ;
4093 auio.uio_segflg = UIO_USERSPACE;
4095 auio.uio_resid = count;
4096 vn_lock(vp, LK_SHARED | LK_RETRY);
4097 AUDIT_ARG_VNODE1(vp);
4098 loff = auio.uio_offset = fp->f_offset;
4100 error = mac_vnode_check_readdir(td->td_ucred, vp);
4103 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4105 fp->f_offset = auio.uio_offset;
4108 VFS_UNLOCK_GIANT(vfslocked);
4111 if (count == auio.uio_resid &&
4112 (vp->v_vflag & VV_ROOT) &&
4113 (vp->v_mount->mnt_flag & MNT_UNION)) {
4114 struct vnode *tvp = vp;
4115 vp = vp->v_mount->mnt_vnodecovered;
4121 VFS_UNLOCK_GIANT(vfslocked);
4125 VFS_UNLOCK_GIANT(vfslocked);
4127 td->td_retval[0] = count - auio.uio_resid;
4133 #ifndef _SYS_SYSPROTO_H_
4134 struct getdents_args {
4143 register struct getdents_args /* {
4149 struct getdirentries_args ap;
4152 ap.count = uap->count;
4154 return (getdirentries(td, &ap));
4158 * Set the mode mask for creation of filesystem nodes.
4160 #ifndef _SYS_SYSPROTO_H_
4168 struct umask_args /* {
4172 register struct filedesc *fdp;
4174 FILEDESC_XLOCK(td->td_proc->p_fd);
4175 fdp = td->td_proc->p_fd;
4176 td->td_retval[0] = fdp->fd_cmask;
4177 fdp->fd_cmask = uap->newmask & ALLPERMS;
4178 FILEDESC_XUNLOCK(td->td_proc->p_fd);
4183 * Void all references to file by ripping underlying filesystem away from
4186 #ifndef _SYS_SYSPROTO_H_
4187 struct revoke_args {
4194 register struct revoke_args /* {
4201 struct nameidata nd;
4204 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4205 UIO_USERSPACE, uap->path, td);
4206 if ((error = namei(&nd)) != 0)
4208 vfslocked = NDHASGIANT(&nd);
4210 NDFREE(&nd, NDF_ONLY_PNBUF);
4211 if (vp->v_type != VCHR) {
4216 error = mac_vnode_check_revoke(td->td_ucred, vp);
4220 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4223 if (td->td_ucred->cr_uid != vattr.va_uid) {
4224 error = priv_check(td, PRIV_VFS_ADMIN);
4229 VOP_REVOKE(vp, REVOKEALL);
4232 VFS_UNLOCK_GIANT(vfslocked);
4237 * Convert a user file descriptor to a kernel file entry.
4238 * A reference on the file entry is held upon returning.
4241 getvnode(fdp, fd, fpp)
4242 struct filedesc *fdp;
4251 if (fdp == NULL || (fp = fget_unlocked(fdp, fd)) == NULL)
4253 else if (fp->f_vnode == NULL) {
4255 fdrop(fp, curthread);
4262 * Get an (NFS) file handle.
4264 #ifndef _SYS_SYSPROTO_H_
4265 struct lgetfh_args {
4273 register struct lgetfh_args *uap;
4275 struct nameidata nd;
4277 register struct vnode *vp;
4281 error = priv_check(td, PRIV_VFS_GETFH);
4284 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4285 UIO_USERSPACE, uap->fname, td);
4289 vfslocked = NDHASGIANT(&nd);
4290 NDFREE(&nd, NDF_ONLY_PNBUF);
4292 bzero(&fh, sizeof(fh));
4293 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4294 error = VOP_VPTOFH(vp, &fh.fh_fid);
4296 VFS_UNLOCK_GIANT(vfslocked);
4299 error = copyout(&fh, uap->fhp, sizeof (fh));
4303 #ifndef _SYS_SYSPROTO_H_
4312 register struct getfh_args *uap;
4314 struct nameidata nd;
4316 register struct vnode *vp;
4320 error = priv_check(td, PRIV_VFS_GETFH);
4323 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4324 UIO_USERSPACE, uap->fname, td);
4328 vfslocked = NDHASGIANT(&nd);
4329 NDFREE(&nd, NDF_ONLY_PNBUF);
4331 bzero(&fh, sizeof(fh));
4332 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4333 error = VOP_VPTOFH(vp, &fh.fh_fid);
4335 VFS_UNLOCK_GIANT(vfslocked);
4338 error = copyout(&fh, uap->fhp, sizeof (fh));
4343 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4346 * warning: do not remove the priv_check() call or this becomes one giant
4349 #ifndef _SYS_SYSPROTO_H_
4350 struct fhopen_args {
4351 const struct fhandle *u_fhp;
4358 struct fhopen_args /* {
4359 const struct fhandle *u_fhp;
4363 struct proc *p = td->td_proc;
4368 struct vattr *vap = &vat;
4371 register struct filedesc *fdp = p->p_fd;
4372 int fmode, error, type;
4378 error = priv_check(td, PRIV_VFS_FHOPEN);
4381 fmode = FFLAGS(uap->flags);
4382 /* why not allow a non-read/write open for our lockd? */
4383 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4385 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4388 /* find the mount point */
4389 mp = vfs_busyfs(&fhp.fh_fsid);
4392 vfslocked = VFS_LOCK_GIANT(mp);
4393 /* now give me my vnode, it gets returned to me locked */
4394 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
4399 * from now on we have to make sure not
4400 * to forget about the vnode
4401 * any error that causes an abort must vput(vp)
4402 * just set error = err and 'goto bad;'.
4408 if (vp->v_type == VLNK) {
4412 if (vp->v_type == VSOCK) {
4417 if (fmode & (FWRITE | O_TRUNC)) {
4418 if (vp->v_type == VDIR) {
4422 error = vn_writechk(vp);
4429 if (fmode & O_APPEND)
4432 error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4437 error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4441 if (fmode & O_TRUNC) {
4443 VOP_UNLOCK(vp, 0); /* XXX */
4444 if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4449 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
4453 * We don't yet have fp->f_cred, so use td->td_ucred, which
4456 error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4461 error = VOP_SETATTR(vp, vap, td->td_ucred);
4465 vn_finished_write(mp);
4469 error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4477 * end of vn_open code
4480 if ((error = falloc(td, &nfp, &indx)) != 0) {
4485 /* An extra reference on `nfp' has been held for us by falloc(). */
4488 finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4489 if (fmode & (O_EXLOCK | O_SHLOCK)) {
4490 lf.l_whence = SEEK_SET;
4493 if (fmode & O_EXLOCK)
4494 lf.l_type = F_WRLCK;
4496 lf.l_type = F_RDLCK;
4498 if ((fmode & FNONBLOCK) == 0)
4501 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4504 * The lock request failed. Normally close the
4505 * descriptor but handle the case where someone might
4506 * have dup()d or close()d it when we weren't looking.
4508 fdclose(fdp, fp, indx, td);
4511 * release our private reference
4516 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4517 atomic_set_int(&fp->f_flag, FHASLOCK);
4522 VFS_UNLOCK_GIANT(vfslocked);
4523 td->td_retval[0] = indx;
4529 VFS_UNLOCK_GIANT(vfslocked);
4534 * Stat an (NFS) file handle.
4536 #ifndef _SYS_SYSPROTO_H_
4537 struct fhstat_args {
4538 struct fhandle *u_fhp;
4545 register struct fhstat_args /* {
4546 struct fhandle *u_fhp;
4557 error = priv_check(td, PRIV_VFS_FHSTAT);
4560 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4563 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4565 vfslocked = VFS_LOCK_GIANT(mp);
4566 error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4569 VFS_UNLOCK_GIANT(vfslocked);
4572 error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4574 VFS_UNLOCK_GIANT(vfslocked);
4577 error = copyout(&sb, uap->sb, sizeof(sb));
4582 * Implement fstatfs() for (NFS) file handles.
4584 #ifndef _SYS_SYSPROTO_H_
4585 struct fhstatfs_args {
4586 struct fhandle *u_fhp;
4593 struct fhstatfs_args /* {
4594 struct fhandle *u_fhp;
4602 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4605 error = kern_fhstatfs(td, fh, &sf);
4608 return (copyout(&sf, uap->buf, sizeof(sf)));
4612 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4620 error = priv_check(td, PRIV_VFS_FHSTATFS);
4623 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4625 vfslocked = VFS_LOCK_GIANT(mp);
4626 error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4629 VFS_UNLOCK_GIANT(vfslocked);
4633 error = prison_canseemount(td->td_ucred, mp);
4637 error = mac_mount_check_stat(td->td_ucred, mp);
4642 * Set these in case the underlying filesystem fails to do so.
4645 sp->f_version = STATFS_VERSION;
4646 sp->f_namemax = NAME_MAX;
4647 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4648 error = VFS_STATFS(mp, sp);
4653 VFS_UNLOCK_GIANT(vfslocked);