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>
49 #include <sys/sysent.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/mutex.h>
53 #include <sys/sysproto.h>
54 #include <sys/namei.h>
55 #include <sys/filedesc.h>
56 #include <sys/kernel.h>
57 #include <sys/fcntl.h>
59 #include <sys/filio.h>
60 #include <sys/limits.h>
61 #include <sys/linker.h>
65 #include <sys/unistd.h>
66 #include <sys/vnode.h>
69 #include <sys/dirent.h>
71 #include <sys/syscallsubr.h>
72 #include <sys/sysctl.h>
74 #include <sys/ktrace.h>
77 #include <machine/stdarg.h>
79 #include <security/audit/audit.h>
80 #include <security/mac/mac_framework.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
87 SDT_PROVIDER_DEFINE(vfs);
88 SDT_PROBE_DEFINE(vfs, , stat, mode);
89 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
90 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
91 SDT_PROBE_DEFINE(vfs, , stat, reg);
92 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
93 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
95 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
96 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
97 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
98 static int setfmode(struct thread *td, struct vnode *, int);
99 static int setfflags(struct thread *td, struct vnode *, int);
100 static int setutimes(struct thread *td, struct vnode *,
101 const struct timespec *, int, int);
102 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106 * The module initialization routine for POSIX asynchronous I/O will
107 * set this to the version of AIO that it implements. (Zero means
108 * that it is not implemented.) This value is used here by pathconf()
109 * and in kern_descrip.c by fpathconf().
111 int async_io_version;
114 static int syncprt = 0;
115 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
119 * Sync each mounted filesystem.
121 #ifndef _SYS_SYSPROTO_H_
130 struct sync_args *uap;
132 struct mount *mp, *nmp;
135 mtx_lock(&mountlist_mtx);
136 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
137 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
138 nmp = TAILQ_NEXT(mp, mnt_list);
141 vfslocked = VFS_LOCK_GIANT(mp);
142 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
143 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
146 mp->mnt_kern_flag &= ~MNTK_ASYNC;
148 vfs_msync(mp, MNT_NOWAIT);
149 VFS_SYNC(mp, MNT_NOWAIT);
152 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
153 mp->mnt_noasync == 0)
154 mp->mnt_kern_flag |= MNTK_ASYNC;
156 vn_finished_write(mp);
158 VFS_UNLOCK_GIANT(vfslocked);
159 mtx_lock(&mountlist_mtx);
160 nmp = TAILQ_NEXT(mp, mnt_list);
163 mtx_unlock(&mountlist_mtx);
168 * Change filesystem quotas.
170 #ifndef _SYS_SYSPROTO_H_
171 struct quotactl_args {
181 register struct quotactl_args /* {
193 AUDIT_ARG_CMD(uap->cmd);
194 AUDIT_ARG_UID(uap->uid);
195 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
197 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
198 UIO_USERSPACE, uap->path, td);
199 if ((error = namei(&nd)) != 0)
201 vfslocked = NDHASGIANT(&nd);
202 NDFREE(&nd, NDF_ONLY_PNBUF);
203 mp = nd.ni_vp->v_mount;
206 error = vfs_busy(mp, 0);
209 VFS_UNLOCK_GIANT(vfslocked);
212 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
214 VFS_UNLOCK_GIANT(vfslocked);
219 * Used by statfs conversion routines to scale the block size up if
220 * necessary so that all of the block counts are <= 'max_size'. Note
221 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
225 statfs_scale_blocks(struct statfs *sf, long max_size)
230 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
233 * Attempt to scale the block counts to give a more accurate
234 * overview to userland of the ratio of free space to used
235 * space. To do this, find the largest block count and compute
236 * a divisor that lets it fit into a signed integer <= max_size.
238 if (sf->f_bavail < 0)
239 count = -sf->f_bavail;
241 count = sf->f_bavail;
242 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
243 if (count <= max_size)
246 count >>= flsl(max_size);
253 sf->f_bsize <<= shift;
254 sf->f_blocks >>= shift;
255 sf->f_bfree >>= shift;
256 sf->f_bavail >>= shift;
260 * Get filesystem statistics.
262 #ifndef _SYS_SYSPROTO_H_
271 register struct statfs_args /* {
279 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
281 error = copyout(&sf, uap->buf, sizeof(sf));
286 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
290 struct statfs *sp, sb;
295 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
296 AUDITVNODE1, pathseg, path, td);
300 vfslocked = NDHASGIANT(&nd);
301 mp = nd.ni_vp->v_mount;
303 NDFREE(&nd, NDF_ONLY_PNBUF);
305 error = vfs_busy(mp, 0);
308 VFS_UNLOCK_GIANT(vfslocked);
312 error = mac_mount_check_stat(td->td_ucred, mp);
317 * Set these in case the underlying filesystem fails to do so.
320 sp->f_version = STATFS_VERSION;
321 sp->f_namemax = NAME_MAX;
322 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
323 error = VFS_STATFS(mp, sp);
326 if (priv_check(td, PRIV_VFS_GENERATION)) {
327 bcopy(sp, &sb, sizeof(sb));
328 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
329 prison_enforce_statfs(td->td_ucred, mp, &sb);
335 VFS_UNLOCK_GIANT(vfslocked);
340 * Get filesystem statistics.
342 #ifndef _SYS_SYSPROTO_H_
343 struct fstatfs_args {
351 register struct fstatfs_args /* {
359 error = kern_fstatfs(td, uap->fd, &sf);
361 error = copyout(&sf, uap->buf, sizeof(sf));
366 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
370 struct statfs *sp, sb;
376 error = getvnode(td->td_proc->p_fd, fd, &fp);
380 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
381 vn_lock(vp, LK_SHARED | LK_RETRY);
383 AUDIT_ARG_VNODE1(vp);
394 error = vfs_busy(mp, 0);
397 VFS_UNLOCK_GIANT(vfslocked);
401 error = mac_mount_check_stat(td->td_ucred, mp);
406 * Set these in case the underlying filesystem fails to do so.
409 sp->f_version = STATFS_VERSION;
410 sp->f_namemax = NAME_MAX;
411 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
412 error = VFS_STATFS(mp, sp);
415 if (priv_check(td, PRIV_VFS_GENERATION)) {
416 bcopy(sp, &sb, sizeof(sb));
417 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
418 prison_enforce_statfs(td->td_ucred, mp, &sb);
425 VFS_UNLOCK_GIANT(vfslocked);
430 * Get statistics on all filesystems.
432 #ifndef _SYS_SYSPROTO_H_
433 struct getfsstat_args {
442 register struct getfsstat_args /* {
449 return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
454 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
455 * The caller is responsible for freeing memory which will be allocated
459 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
460 enum uio_seg bufseg, int flags)
462 struct mount *mp, *nmp;
463 struct statfs *sfsp, *sp, sb;
464 size_t count, maxcount;
468 maxcount = bufsize / sizeof(struct statfs);
471 else if (bufseg == UIO_USERSPACE)
473 else /* if (bufseg == UIO_SYSSPACE) */ {
475 mtx_lock(&mountlist_mtx);
476 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
479 mtx_unlock(&mountlist_mtx);
480 if (maxcount > count)
482 sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
486 mtx_lock(&mountlist_mtx);
487 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
488 if (prison_canseemount(td->td_ucred, mp) != 0) {
489 nmp = TAILQ_NEXT(mp, mnt_list);
493 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
494 nmp = TAILQ_NEXT(mp, mnt_list);
498 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
499 nmp = TAILQ_NEXT(mp, mnt_list);
502 vfslocked = VFS_LOCK_GIANT(mp);
503 if (sfsp && count < maxcount) {
506 * Set these in case the underlying filesystem
509 sp->f_version = STATFS_VERSION;
510 sp->f_namemax = NAME_MAX;
511 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
513 * If MNT_NOWAIT or MNT_LAZY is specified, do not
514 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
515 * overrides MNT_WAIT.
517 if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
518 (flags & MNT_WAIT)) &&
519 (error = VFS_STATFS(mp, sp))) {
520 VFS_UNLOCK_GIANT(vfslocked);
521 mtx_lock(&mountlist_mtx);
522 nmp = TAILQ_NEXT(mp, mnt_list);
526 if (priv_check(td, PRIV_VFS_GENERATION)) {
527 bcopy(sp, &sb, sizeof(sb));
528 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
529 prison_enforce_statfs(td->td_ucred, mp, &sb);
532 if (bufseg == UIO_SYSSPACE)
533 bcopy(sp, sfsp, sizeof(*sp));
534 else /* if (bufseg == UIO_USERSPACE) */ {
535 error = copyout(sp, sfsp, sizeof(*sp));
538 VFS_UNLOCK_GIANT(vfslocked);
544 VFS_UNLOCK_GIANT(vfslocked);
546 mtx_lock(&mountlist_mtx);
547 nmp = TAILQ_NEXT(mp, mnt_list);
550 mtx_unlock(&mountlist_mtx);
551 if (sfsp && count > maxcount)
552 td->td_retval[0] = maxcount;
554 td->td_retval[0] = count;
558 #ifdef COMPAT_FREEBSD4
560 * Get old format filesystem statistics.
562 static void cvtstatfs(struct statfs *, struct ostatfs *);
564 #ifndef _SYS_SYSPROTO_H_
565 struct freebsd4_statfs_args {
571 freebsd4_statfs(td, uap)
573 struct freebsd4_statfs_args /* {
582 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
585 cvtstatfs(&sf, &osb);
586 return (copyout(&osb, uap->buf, sizeof(osb)));
590 * Get filesystem statistics.
592 #ifndef _SYS_SYSPROTO_H_
593 struct freebsd4_fstatfs_args {
599 freebsd4_fstatfs(td, uap)
601 struct freebsd4_fstatfs_args /* {
610 error = kern_fstatfs(td, uap->fd, &sf);
613 cvtstatfs(&sf, &osb);
614 return (copyout(&osb, uap->buf, sizeof(osb)));
618 * Get statistics on all filesystems.
620 #ifndef _SYS_SYSPROTO_H_
621 struct freebsd4_getfsstat_args {
628 freebsd4_getfsstat(td, uap)
630 register struct freebsd4_getfsstat_args /* {
636 struct statfs *buf, *sp;
641 count = uap->bufsize / sizeof(struct ostatfs);
642 size = count * sizeof(struct statfs);
643 error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
645 count = td->td_retval[0];
647 while (count > 0 && error == 0) {
649 error = copyout(&osb, uap->buf, sizeof(osb));
660 * Implement fstatfs() for (NFS) file handles.
662 #ifndef _SYS_SYSPROTO_H_
663 struct freebsd4_fhstatfs_args {
664 struct fhandle *u_fhp;
669 freebsd4_fhstatfs(td, uap)
671 struct freebsd4_fhstatfs_args /* {
672 struct fhandle *u_fhp;
681 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
684 error = kern_fhstatfs(td, fh, &sf);
687 cvtstatfs(&sf, &osb);
688 return (copyout(&osb, uap->buf, sizeof(osb)));
692 * Convert a new format statfs structure to an old format statfs structure.
700 statfs_scale_blocks(nsp, LONG_MAX);
701 bzero(osp, sizeof(*osp));
702 osp->f_bsize = nsp->f_bsize;
703 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
704 osp->f_blocks = nsp->f_blocks;
705 osp->f_bfree = nsp->f_bfree;
706 osp->f_bavail = nsp->f_bavail;
707 osp->f_files = MIN(nsp->f_files, LONG_MAX);
708 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
709 osp->f_owner = nsp->f_owner;
710 osp->f_type = nsp->f_type;
711 osp->f_flags = nsp->f_flags;
712 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
713 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
714 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
715 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
716 strlcpy(osp->f_fstypename, nsp->f_fstypename,
717 MIN(MFSNAMELEN, OMFSNAMELEN));
718 strlcpy(osp->f_mntonname, nsp->f_mntonname,
719 MIN(MNAMELEN, OMNAMELEN));
720 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
721 MIN(MNAMELEN, OMNAMELEN));
722 osp->f_fsid = nsp->f_fsid;
724 #endif /* COMPAT_FREEBSD4 */
727 * Change current working directory to a given file descriptor.
729 #ifndef _SYS_SYSPROTO_H_
737 struct fchdir_args /* {
741 register struct filedesc *fdp = td->td_proc->p_fd;
742 struct vnode *vp, *tdp, *vpold;
748 AUDIT_ARG_FD(uap->fd);
749 if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
754 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
755 vn_lock(vp, LK_SHARED | LK_RETRY);
756 AUDIT_ARG_VNODE1(vp);
757 error = change_dir(vp, td);
758 while (!error && (mp = vp->v_mountedhere) != NULL) {
762 tvfslocked = VFS_LOCK_GIANT(mp);
763 error = VFS_ROOT(mp, LK_SHARED, &tdp);
766 VFS_UNLOCK_GIANT(tvfslocked);
770 VFS_UNLOCK_GIANT(vfslocked);
772 vfslocked = tvfslocked;
776 VFS_UNLOCK_GIANT(vfslocked);
780 VFS_UNLOCK_GIANT(vfslocked);
782 vpold = fdp->fd_cdir;
784 FILEDESC_XUNLOCK(fdp);
785 vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
787 VFS_UNLOCK_GIANT(vfslocked);
792 * Change current working directory (``.'').
794 #ifndef _SYS_SYSPROTO_H_
802 struct chdir_args /* {
807 return (kern_chdir(td, uap->path, UIO_USERSPACE));
811 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
813 register struct filedesc *fdp = td->td_proc->p_fd;
819 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
820 MPSAFE, pathseg, path, td);
821 if ((error = namei(&nd)) != 0)
823 vfslocked = NDHASGIANT(&nd);
824 if ((error = change_dir(nd.ni_vp, td)) != 0) {
826 VFS_UNLOCK_GIANT(vfslocked);
827 NDFREE(&nd, NDF_ONLY_PNBUF);
830 VOP_UNLOCK(nd.ni_vp, 0);
831 VFS_UNLOCK_GIANT(vfslocked);
832 NDFREE(&nd, NDF_ONLY_PNBUF);
835 fdp->fd_cdir = nd.ni_vp;
836 FILEDESC_XUNLOCK(fdp);
837 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
839 VFS_UNLOCK_GIANT(vfslocked);
844 * Helper function for raised chroot(2) security function: Refuse if
845 * any filedescriptors are open directories.
848 chroot_refuse_vdir_fds(fdp)
849 struct filedesc *fdp;
855 FILEDESC_LOCK_ASSERT(fdp);
857 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
858 fp = fget_locked(fdp, fd);
861 if (fp->f_type == DTYPE_VNODE) {
863 if (vp->v_type == VDIR)
871 * This sysctl determines if we will allow a process to chroot(2) if it
872 * has a directory open:
873 * 0: disallowed for all processes.
874 * 1: allowed for processes that were not already chroot(2)'ed.
875 * 2: allowed for all processes.
878 static int chroot_allow_open_directories = 1;
880 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
881 &chroot_allow_open_directories, 0, "");
884 * Change notion of root (``/'') directory.
886 #ifndef _SYS_SYSPROTO_H_
894 struct chroot_args /* {
902 error = priv_check(td, PRIV_VFS_CHROOT);
905 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
906 AUDITVNODE1, UIO_USERSPACE, uap->path, td);
910 vfslocked = NDHASGIANT(&nd);
911 if ((error = change_dir(nd.ni_vp, td)) != 0)
914 if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
917 VOP_UNLOCK(nd.ni_vp, 0);
918 error = change_root(nd.ni_vp, td);
920 VFS_UNLOCK_GIANT(vfslocked);
921 NDFREE(&nd, NDF_ONLY_PNBUF);
925 VFS_UNLOCK_GIANT(vfslocked);
927 NDFREE(&nd, NDF_ONLY_PNBUF);
932 * Common routine for chroot and chdir. Callers must provide a locked vnode
942 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
943 if (vp->v_type != VDIR)
946 error = mac_vnode_check_chdir(td->td_ucred, vp);
950 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
955 * Common routine for kern_chroot() and jail_attach(). The caller is
956 * responsible for invoking priv_check() and mac_vnode_check_chroot() to
957 * authorize this operation.
964 struct filedesc *fdp;
969 VFS_ASSERT_GIANT(vp->v_mount);
970 fdp = td->td_proc->p_fd;
972 if (chroot_allow_open_directories == 0 ||
973 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
974 error = chroot_refuse_vdir_fds(fdp);
976 FILEDESC_XUNLOCK(fdp);
980 oldvp = fdp->fd_rdir;
987 FILEDESC_XUNLOCK(fdp);
988 vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
990 VFS_UNLOCK_GIANT(vfslocked);
995 * Check permissions, allocate an open file structure, and call the device
996 * open routine if any.
998 #ifndef _SYS_SYSPROTO_H_
1008 register struct open_args /* {
1015 return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1018 #ifndef _SYS_SYSPROTO_H_
1019 struct openat_args {
1027 openat(struct thread *td, struct openat_args *uap)
1030 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1035 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1039 return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1043 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1044 int flags, int mode)
1046 struct proc *p = td->td_proc;
1047 struct filedesc *fdp = p->p_fd;
1054 int type, indx, error;
1056 struct nameidata nd;
1059 AUDIT_ARG_FFLAGS(flags);
1060 AUDIT_ARG_MODE(mode);
1061 /* XXX: audit dirfd */
1063 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR may
1066 if (flags & O_EXEC) {
1067 if (flags & O_ACCMODE)
1069 } else if ((flags & O_ACCMODE) == O_ACCMODE)
1072 flags = FFLAGS(flags);
1074 error = falloc(td, &nfp, &indx);
1077 /* An extra reference on `nfp' has been held for us by falloc(). */
1079 /* Set the flags early so the finit in devfs can pick them up. */
1080 fp->f_flag = flags & FMASK;
1081 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1082 NDINIT_AT(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, path, fd,
1084 td->td_dupfd = -1; /* XXX check for fdopen */
1085 error = vn_open(&nd, &flags, cmode, fp);
1088 * If the vn_open replaced the method vector, something
1089 * wonderous happened deep below and we just pass it up
1090 * pretending we know what we do.
1092 if (error == ENXIO && fp->f_ops != &badfileops) {
1094 td->td_retval[0] = indx;
1099 * handle special fdopen() case. bleh. dupfdopen() is
1100 * responsible for dropping the old contents of ofiles[indx]
1103 if ((error == ENODEV || error == ENXIO) &&
1104 td->td_dupfd >= 0 && /* XXX from fdopen */
1106 dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1107 td->td_retval[0] = indx;
1112 * Clean up the descriptor, but only if another thread hadn't
1113 * replaced or closed it.
1115 fdclose(fdp, fp, indx, td);
1118 if (error == ERESTART)
1123 vfslocked = NDHASGIANT(&nd);
1124 NDFREE(&nd, NDF_ONLY_PNBUF);
1127 fp->f_vnode = vp; /* XXX Does devfs need this? */
1129 * If the file wasn't claimed by devfs bind it to the normal
1130 * vnode operations here.
1132 if (fp->f_ops == &badfileops) {
1133 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1135 finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1139 if (flags & (O_EXLOCK | O_SHLOCK)) {
1140 lf.l_whence = SEEK_SET;
1143 if (flags & O_EXLOCK)
1144 lf.l_type = F_WRLCK;
1146 lf.l_type = F_RDLCK;
1148 if ((flags & FNONBLOCK) == 0)
1150 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1153 atomic_set_int(&fp->f_flag, FHASLOCK);
1155 if (flags & O_TRUNC) {
1156 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1160 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1162 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
1165 error = VOP_SETATTR(vp, &vat, td->td_ucred);
1167 vn_finished_write(mp);
1171 VFS_UNLOCK_GIANT(vfslocked);
1173 * Release our private reference, leaving the one associated with
1174 * the descriptor table intact.
1177 td->td_retval[0] = indx;
1180 VFS_UNLOCK_GIANT(vfslocked);
1181 fdclose(fdp, fp, indx, td);
1190 #ifndef _SYS_SYSPROTO_H_
1191 struct ocreat_args {
1199 register struct ocreat_args /* {
1205 return (kern_open(td, uap->path, UIO_USERSPACE,
1206 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1208 #endif /* COMPAT_43 */
1211 * Create a special file.
1213 #ifndef _SYS_SYSPROTO_H_
1223 register struct mknod_args /* {
1230 return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1233 #ifndef _SYS_SYSPROTO_H_
1234 struct mknodat_args {
1242 mknodat(struct thread *td, struct mknodat_args *uap)
1245 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1250 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1254 return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1258 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1266 struct nameidata nd;
1269 AUDIT_ARG_MODE(mode);
1271 switch (mode & S_IFMT) {
1274 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1277 error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1280 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1284 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1294 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1295 pathseg, path, fd, td);
1296 if ((error = namei(&nd)) != 0)
1298 vfslocked = NDHASGIANT(&nd);
1301 NDFREE(&nd, NDF_ONLY_PNBUF);
1302 if (vp == nd.ni_dvp)
1307 VFS_UNLOCK_GIANT(vfslocked);
1311 vattr.va_mode = (mode & ALLPERMS) &
1312 ~td->td_proc->p_fd->fd_cmask;
1313 vattr.va_rdev = dev;
1316 switch (mode & S_IFMT) {
1317 case S_IFMT: /* used by badsect to flag bad sectors */
1318 vattr.va_type = VBAD;
1321 vattr.va_type = VCHR;
1324 vattr.va_type = VBLK;
1330 panic("kern_mknod: invalid mode");
1333 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1334 NDFREE(&nd, NDF_ONLY_PNBUF);
1336 VFS_UNLOCK_GIANT(vfslocked);
1337 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1342 if (error == 0 && !whiteout)
1343 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1344 &nd.ni_cnd, &vattr);
1348 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1350 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1351 &nd.ni_cnd, &vattr);
1356 NDFREE(&nd, NDF_ONLY_PNBUF);
1358 vn_finished_write(mp);
1359 VFS_UNLOCK_GIANT(vfslocked);
1364 * Create a named pipe.
1366 #ifndef _SYS_SYSPROTO_H_
1367 struct mkfifo_args {
1375 register struct mkfifo_args /* {
1381 return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1384 #ifndef _SYS_SYSPROTO_H_
1385 struct mkfifoat_args {
1392 mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1395 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1400 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1403 return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1407 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1413 struct nameidata nd;
1416 AUDIT_ARG_MODE(mode);
1419 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1420 pathseg, path, fd, td);
1421 if ((error = namei(&nd)) != 0)
1423 vfslocked = NDHASGIANT(&nd);
1424 if (nd.ni_vp != NULL) {
1425 NDFREE(&nd, NDF_ONLY_PNBUF);
1426 if (nd.ni_vp == nd.ni_dvp)
1431 VFS_UNLOCK_GIANT(vfslocked);
1434 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1435 NDFREE(&nd, NDF_ONLY_PNBUF);
1437 VFS_UNLOCK_GIANT(vfslocked);
1438 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1443 vattr.va_type = VFIFO;
1444 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1446 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1451 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1458 vn_finished_write(mp);
1459 VFS_UNLOCK_GIANT(vfslocked);
1460 NDFREE(&nd, NDF_ONLY_PNBUF);
1465 * Make a hard file link.
1467 #ifndef _SYS_SYSPROTO_H_
1476 register struct link_args /* {
1482 return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1485 #ifndef _SYS_SYSPROTO_H_
1486 struct linkat_args {
1495 linkat(struct thread *td, struct linkat_args *uap)
1500 if (flag & ~AT_SYMLINK_FOLLOW)
1503 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1504 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1507 int hardlink_check_uid = 0;
1508 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1509 &hardlink_check_uid, 0,
1510 "Unprivileged processes cannot create hard links to files owned by other "
1512 static int hardlink_check_gid = 0;
1513 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1514 &hardlink_check_gid, 0,
1515 "Unprivileged processes cannot create hard links to files owned by other "
1519 can_hardlink(struct vnode *vp, struct ucred *cred)
1524 if (!hardlink_check_uid && !hardlink_check_gid)
1527 error = VOP_GETATTR(vp, &va, cred);
1531 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1532 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1537 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1538 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1547 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1550 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1554 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1555 enum uio_seg segflg, int follow)
1559 struct nameidata nd;
1565 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1568 if ((error = namei(&nd)) != 0)
1570 vfslocked = NDHASGIANT(&nd);
1571 NDFREE(&nd, NDF_ONLY_PNBUF);
1573 if (vp->v_type == VDIR) {
1575 VFS_UNLOCK_GIANT(vfslocked);
1576 return (EPERM); /* POSIX */
1578 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1580 VFS_UNLOCK_GIANT(vfslocked);
1583 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1584 segflg, path2, fd2, td);
1585 if ((error = namei(&nd)) == 0) {
1586 lvfslocked = NDHASGIANT(&nd);
1587 if (nd.ni_vp != NULL) {
1588 if (nd.ni_dvp == nd.ni_vp)
1594 } else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1596 error = can_hardlink(vp, td->td_ucred);
1599 error = mac_vnode_check_link(td->td_ucred,
1600 nd.ni_dvp, vp, &nd.ni_cnd);
1603 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1607 NDFREE(&nd, NDF_ONLY_PNBUF);
1608 VFS_UNLOCK_GIANT(lvfslocked);
1611 vn_finished_write(mp);
1612 VFS_UNLOCK_GIANT(vfslocked);
1617 * Make a symbolic link.
1619 #ifndef _SYS_SYSPROTO_H_
1620 struct symlink_args {
1628 register struct symlink_args /* {
1634 return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1637 #ifndef _SYS_SYSPROTO_H_
1638 struct symlinkat_args {
1645 symlinkat(struct thread *td, struct symlinkat_args *uap)
1648 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1653 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1656 return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1660 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1661 enum uio_seg segflg)
1667 struct nameidata nd;
1670 if (segflg == UIO_SYSSPACE) {
1673 syspath = uma_zalloc(namei_zone, M_WAITOK);
1674 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1677 AUDIT_ARG_TEXT(syspath);
1680 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1681 segflg, path2, fd, td);
1682 if ((error = namei(&nd)) != 0)
1684 vfslocked = NDHASGIANT(&nd);
1686 NDFREE(&nd, NDF_ONLY_PNBUF);
1687 if (nd.ni_vp == nd.ni_dvp)
1692 VFS_UNLOCK_GIANT(vfslocked);
1696 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1697 NDFREE(&nd, NDF_ONLY_PNBUF);
1699 VFS_UNLOCK_GIANT(vfslocked);
1700 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1705 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1707 vattr.va_type = VLNK;
1708 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1713 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1719 NDFREE(&nd, NDF_ONLY_PNBUF);
1721 vn_finished_write(mp);
1722 VFS_UNLOCK_GIANT(vfslocked);
1724 if (segflg != UIO_SYSSPACE)
1725 uma_zfree(namei_zone, syspath);
1730 * Delete a whiteout from the filesystem.
1735 register struct undelete_args /* {
1741 struct nameidata nd;
1746 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1747 UIO_USERSPACE, uap->path, td);
1751 vfslocked = NDHASGIANT(&nd);
1753 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1754 NDFREE(&nd, NDF_ONLY_PNBUF);
1755 if (nd.ni_vp == nd.ni_dvp)
1761 VFS_UNLOCK_GIANT(vfslocked);
1764 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1765 NDFREE(&nd, NDF_ONLY_PNBUF);
1767 VFS_UNLOCK_GIANT(vfslocked);
1768 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1772 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1773 NDFREE(&nd, NDF_ONLY_PNBUF);
1775 vn_finished_write(mp);
1776 VFS_UNLOCK_GIANT(vfslocked);
1781 * Delete a name from the filesystem.
1783 #ifndef _SYS_SYSPROTO_H_
1784 struct unlink_args {
1791 struct unlink_args /* {
1796 return (kern_unlink(td, uap->path, UIO_USERSPACE));
1799 #ifndef _SYS_SYSPROTO_H_
1800 struct unlinkat_args {
1807 unlinkat(struct thread *td, struct unlinkat_args *uap)
1809 int flag = uap->flag;
1811 char *path = uap->path;
1813 if (flag & ~AT_REMOVEDIR)
1816 if (flag & AT_REMOVEDIR)
1817 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1819 return (kern_unlinkat(td, fd, path, UIO_USERSPACE));
1823 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1826 return (kern_unlinkat(td, AT_FDCWD, path, pathseg));
1830 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
1835 struct nameidata nd;
1840 NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1841 pathseg, path, fd, td);
1842 if ((error = namei(&nd)) != 0)
1843 return (error == EINVAL ? EPERM : error);
1844 vfslocked = NDHASGIANT(&nd);
1846 if (vp->v_type == VDIR)
1847 error = EPERM; /* POSIX */
1850 * The root of a mounted filesystem cannot be deleted.
1852 * XXX: can this only be a VDIR case?
1854 if (vp->v_vflag & VV_ROOT)
1858 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1859 NDFREE(&nd, NDF_ONLY_PNBUF);
1861 if (vp == nd.ni_dvp)
1865 VFS_UNLOCK_GIANT(vfslocked);
1866 if ((error = vn_start_write(NULL, &mp,
1867 V_XSLEEP | PCATCH)) != 0)
1872 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1877 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1881 vn_finished_write(mp);
1883 NDFREE(&nd, NDF_ONLY_PNBUF);
1885 if (vp == nd.ni_dvp)
1889 VFS_UNLOCK_GIANT(vfslocked);
1894 * Reposition read/write file offset.
1896 #ifndef _SYS_SYSPROTO_H_
1907 register struct lseek_args /* {
1914 struct ucred *cred = td->td_ucred;
1922 AUDIT_ARG_FD(uap->fd);
1923 if ((error = fget(td, uap->fd, &fp)) != 0)
1925 if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1930 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1931 noneg = (vp->v_type != VCHR);
1932 offset = uap->offset;
1933 switch (uap->whence) {
1936 (fp->f_offset < 0 ||
1937 (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
1941 offset += fp->f_offset;
1944 vn_lock(vp, LK_SHARED | LK_RETRY);
1945 error = VOP_GETATTR(vp, &vattr, cred);
1951 * If the file references a disk device, then fetch
1952 * the media size and use that to determine the ending
1955 if (vattr.va_size == 0 && vp->v_type == VCHR &&
1956 fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
1957 vattr.va_size = size;
1959 (vattr.va_size > OFF_MAX ||
1960 (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1964 offset += vattr.va_size;
1969 error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1972 error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1977 if (error == 0 && noneg && offset < 0)
1981 fp->f_offset = offset;
1982 *(off_t *)(td->td_retval) = fp->f_offset;
1985 VFS_UNLOCK_GIANT(vfslocked);
1989 #if defined(COMPAT_43)
1991 * Reposition read/write file offset.
1993 #ifndef _SYS_SYSPROTO_H_
1994 struct olseek_args {
2003 register struct olseek_args /* {
2009 struct lseek_args /* {
2017 nuap.offset = uap->offset;
2018 nuap.whence = uap->whence;
2019 return (lseek(td, &nuap));
2021 #endif /* COMPAT_43 */
2023 /* Version with the 'pad' argument */
2025 freebsd6_lseek(td, uap)
2027 register struct freebsd6_lseek_args *uap;
2029 struct lseek_args ouap;
2032 ouap.offset = uap->offset;
2033 ouap.whence = uap->whence;
2034 return (lseek(td, &ouap));
2038 * Check access permissions using passed credentials.
2041 vn_access(vp, user_flags, cred, td)
2050 /* Flags == 0 means only check for existence. */
2054 if (user_flags & R_OK)
2056 if (user_flags & W_OK)
2058 if (user_flags & X_OK)
2061 error = mac_vnode_check_access(cred, vp, accmode);
2065 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2066 error = VOP_ACCESS(vp, accmode, cred, td);
2072 * Check access permissions using "real" credentials.
2074 #ifndef _SYS_SYSPROTO_H_
2075 struct access_args {
2083 register struct access_args /* {
2089 return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2092 #ifndef _SYS_SYSPROTO_H_
2093 struct faccessat_args {
2101 faccessat(struct thread *td, struct faccessat_args *uap)
2104 if (uap->flag & ~AT_EACCESS)
2106 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2111 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2114 return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2118 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2119 int flags, int mode)
2121 struct ucred *cred, *tmpcred;
2123 struct nameidata nd;
2128 * Create and modify a temporary credential instead of one that
2129 * is potentially shared. This could also mess up socket
2130 * buffer accounting which can run in an interrupt context.
2132 if (!(flags & AT_EACCESS)) {
2133 cred = td->td_ucred;
2134 tmpcred = crdup(cred);
2135 tmpcred->cr_uid = cred->cr_ruid;
2136 tmpcred->cr_groups[0] = cred->cr_rgid;
2137 td->td_ucred = tmpcred;
2139 cred = tmpcred = td->td_ucred;
2140 AUDIT_ARG_VALUE(mode);
2141 NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2142 AUDITVNODE1, pathseg, path, fd, td);
2143 if ((error = namei(&nd)) != 0)
2145 vfslocked = NDHASGIANT(&nd);
2148 error = vn_access(vp, mode, tmpcred, td);
2149 NDFREE(&nd, NDF_ONLY_PNBUF);
2151 VFS_UNLOCK_GIANT(vfslocked);
2153 if (!(flags & AT_EACCESS)) {
2154 td->td_ucred = cred;
2161 * Check access permissions using "effective" credentials.
2163 #ifndef _SYS_SYSPROTO_H_
2164 struct eaccess_args {
2172 register struct eaccess_args /* {
2178 return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2182 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2185 return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2188 #if defined(COMPAT_43)
2190 * Get file status; this version follows links.
2192 #ifndef _SYS_SYSPROTO_H_
2201 register struct ostat_args /* {
2210 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2214 error = copyout(&osb, uap->ub, sizeof (osb));
2219 * Get file status; this version does not follow links.
2221 #ifndef _SYS_SYSPROTO_H_
2222 struct olstat_args {
2230 register struct olstat_args /* {
2239 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2243 error = copyout(&osb, uap->ub, sizeof (osb));
2248 * Convert from an old to a new stat structure.
2256 ost->st_dev = st->st_dev;
2257 ost->st_ino = st->st_ino;
2258 ost->st_mode = st->st_mode;
2259 ost->st_nlink = st->st_nlink;
2260 ost->st_uid = st->st_uid;
2261 ost->st_gid = st->st_gid;
2262 ost->st_rdev = st->st_rdev;
2263 if (st->st_size < (quad_t)1 << 32)
2264 ost->st_size = st->st_size;
2267 ost->st_atime = st->st_atime;
2268 ost->st_mtime = st->st_mtime;
2269 ost->st_ctime = st->st_ctime;
2270 ost->st_blksize = st->st_blksize;
2271 ost->st_blocks = st->st_blocks;
2272 ost->st_flags = st->st_flags;
2273 ost->st_gen = st->st_gen;
2275 #endif /* COMPAT_43 */
2278 * Get file status; this version follows links.
2280 #ifndef _SYS_SYSPROTO_H_
2289 register struct stat_args /* {
2297 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2299 error = copyout(&sb, uap->ub, sizeof (sb));
2303 #ifndef _SYS_SYSPROTO_H_
2304 struct fstatat_args {
2312 fstatat(struct thread *td, struct fstatat_args *uap)
2317 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2318 UIO_USERSPACE, &sb);
2320 error = copyout(&sb, uap->buf, sizeof (sb));
2325 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2328 return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2332 kern_statat(struct thread *td, int flag, int fd, char *path,
2333 enum uio_seg pathseg, struct stat *sbp)
2336 return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2340 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2341 enum uio_seg pathseg, struct stat *sbp,
2342 void (*hook)(struct vnode *vp, struct stat *sbp))
2344 struct nameidata nd;
2346 int error, vfslocked;
2348 if (flag & ~AT_SYMLINK_NOFOLLOW)
2351 NDINIT_AT(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2352 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2355 if ((error = namei(&nd)) != 0)
2357 vfslocked = NDHASGIANT(&nd);
2358 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2360 SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2361 if (S_ISREG(sb.st_mode))
2362 SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2363 if (__predict_false(hook != NULL))
2364 hook(nd.ni_vp, &sb);
2366 NDFREE(&nd, NDF_ONLY_PNBUF);
2368 VFS_UNLOCK_GIANT(vfslocked);
2373 if (KTRPOINT(td, KTR_STRUCT))
2380 * Get file status; this version does not follow links.
2382 #ifndef _SYS_SYSPROTO_H_
2391 register struct lstat_args /* {
2399 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2401 error = copyout(&sb, uap->ub, sizeof (sb));
2406 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2409 return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2414 * Implementation of the NetBSD [l]stat() functions.
2421 bzero(nsb, sizeof *nsb);
2422 nsb->st_dev = sb->st_dev;
2423 nsb->st_ino = sb->st_ino;
2424 nsb->st_mode = sb->st_mode;
2425 nsb->st_nlink = sb->st_nlink;
2426 nsb->st_uid = sb->st_uid;
2427 nsb->st_gid = sb->st_gid;
2428 nsb->st_rdev = sb->st_rdev;
2429 nsb->st_atimespec = sb->st_atimespec;
2430 nsb->st_mtimespec = sb->st_mtimespec;
2431 nsb->st_ctimespec = sb->st_ctimespec;
2432 nsb->st_size = sb->st_size;
2433 nsb->st_blocks = sb->st_blocks;
2434 nsb->st_blksize = sb->st_blksize;
2435 nsb->st_flags = sb->st_flags;
2436 nsb->st_gen = sb->st_gen;
2437 nsb->st_birthtimespec = sb->st_birthtimespec;
2440 #ifndef _SYS_SYSPROTO_H_
2449 register struct nstat_args /* {
2458 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2461 cvtnstat(&sb, &nsb);
2462 error = copyout(&nsb, uap->ub, sizeof (nsb));
2467 * NetBSD lstat. Get file status; this version does not follow links.
2469 #ifndef _SYS_SYSPROTO_H_
2478 register struct nlstat_args /* {
2487 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2490 cvtnstat(&sb, &nsb);
2491 error = copyout(&nsb, uap->ub, sizeof (nsb));
2496 * Get configurable pathname variables.
2498 #ifndef _SYS_SYSPROTO_H_
2499 struct pathconf_args {
2507 register struct pathconf_args /* {
2513 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2516 #ifndef _SYS_SYSPROTO_H_
2517 struct lpathconf_args {
2525 register struct lpathconf_args /* {
2531 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2535 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2538 struct nameidata nd;
2539 int error, vfslocked;
2541 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 |
2542 flags, pathseg, path, td);
2543 if ((error = namei(&nd)) != 0)
2545 vfslocked = NDHASGIANT(&nd);
2546 NDFREE(&nd, NDF_ONLY_PNBUF);
2548 /* If asynchronous I/O is available, it works for all files. */
2549 if (name == _PC_ASYNC_IO)
2550 td->td_retval[0] = async_io_version;
2552 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2554 VFS_UNLOCK_GIANT(vfslocked);
2559 * Return target name of a symbolic link.
2561 #ifndef _SYS_SYSPROTO_H_
2562 struct readlink_args {
2571 register struct readlink_args /* {
2578 return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2579 UIO_USERSPACE, uap->count));
2581 #ifndef _SYS_SYSPROTO_H_
2582 struct readlinkat_args {
2590 readlinkat(struct thread *td, struct readlinkat_args *uap)
2593 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2594 uap->buf, UIO_USERSPACE, uap->bufsize));
2598 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2599 enum uio_seg bufseg, size_t count)
2602 return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2607 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2608 char *buf, enum uio_seg bufseg, size_t count)
2614 struct nameidata nd;
2617 if (count > INT_MAX)
2620 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2621 AUDITVNODE1, pathseg, path, fd, td);
2623 if ((error = namei(&nd)) != 0)
2625 NDFREE(&nd, NDF_ONLY_PNBUF);
2626 vfslocked = NDHASGIANT(&nd);
2629 error = mac_vnode_check_readlink(td->td_ucred, vp);
2632 VFS_UNLOCK_GIANT(vfslocked);
2636 if (vp->v_type != VLNK)
2639 aiov.iov_base = buf;
2640 aiov.iov_len = count;
2641 auio.uio_iov = &aiov;
2642 auio.uio_iovcnt = 1;
2643 auio.uio_offset = 0;
2644 auio.uio_rw = UIO_READ;
2645 auio.uio_segflg = bufseg;
2647 auio.uio_resid = count;
2648 error = VOP_READLINK(vp, &auio, td->td_ucred);
2651 VFS_UNLOCK_GIANT(vfslocked);
2652 td->td_retval[0] = count - auio.uio_resid;
2657 * Common implementation code for chflags() and fchflags().
2660 setfflags(td, vp, flags)
2670 * Prevent non-root users from setting flags on devices. When
2671 * a device is reused, users can retain ownership of the device
2672 * if they are allowed to set flags and programs assume that
2673 * chown can't fail when done as root.
2675 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2676 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2681 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2683 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2685 vattr.va_flags = flags;
2687 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2690 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2692 vn_finished_write(mp);
2697 * Change flags of a file given a path name.
2699 #ifndef _SYS_SYSPROTO_H_
2700 struct chflags_args {
2708 register struct chflags_args /* {
2714 struct nameidata nd;
2717 AUDIT_ARG_FFLAGS(uap->flags);
2718 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2720 if ((error = namei(&nd)) != 0)
2722 NDFREE(&nd, NDF_ONLY_PNBUF);
2723 vfslocked = NDHASGIANT(&nd);
2724 error = setfflags(td, nd.ni_vp, uap->flags);
2726 VFS_UNLOCK_GIANT(vfslocked);
2731 * Same as chflags() but doesn't follow symlinks.
2736 register struct lchflags_args /* {
2742 struct nameidata nd;
2745 AUDIT_ARG_FFLAGS(uap->flags);
2746 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2748 if ((error = namei(&nd)) != 0)
2750 vfslocked = NDHASGIANT(&nd);
2751 NDFREE(&nd, NDF_ONLY_PNBUF);
2752 error = setfflags(td, nd.ni_vp, uap->flags);
2754 VFS_UNLOCK_GIANT(vfslocked);
2759 * Change flags of a file given a file descriptor.
2761 #ifndef _SYS_SYSPROTO_H_
2762 struct fchflags_args {
2770 register struct fchflags_args /* {
2779 AUDIT_ARG_FD(uap->fd);
2780 AUDIT_ARG_FFLAGS(uap->flags);
2781 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2783 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2785 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2786 AUDIT_ARG_VNODE1(fp->f_vnode);
2787 VOP_UNLOCK(fp->f_vnode, 0);
2789 error = setfflags(td, fp->f_vnode, uap->flags);
2790 VFS_UNLOCK_GIANT(vfslocked);
2796 * Common implementation code for chmod(), lchmod() and fchmod().
2799 setfmode(td, vp, mode)
2808 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2810 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2812 vattr.va_mode = mode & ALLPERMS;
2814 error = mac_vnode_check_setmode(td->td_ucred, vp, vattr.va_mode);
2817 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2819 vn_finished_write(mp);
2824 * Change mode of a file given path name.
2826 #ifndef _SYS_SYSPROTO_H_
2835 register struct chmod_args /* {
2841 return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2844 #ifndef _SYS_SYSPROTO_H_
2845 struct fchmodat_args {
2853 fchmodat(struct thread *td, struct fchmodat_args *uap)
2855 int flag = uap->flag;
2857 char *path = uap->path;
2858 mode_t mode = uap->mode;
2860 if (flag & ~AT_SYMLINK_NOFOLLOW)
2863 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2867 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2870 return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2874 * Change mode of a file given path name (don't follow links.)
2876 #ifndef _SYS_SYSPROTO_H_
2877 struct lchmod_args {
2885 register struct lchmod_args /* {
2891 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2892 uap->mode, AT_SYMLINK_NOFOLLOW));
2897 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2898 mode_t mode, int flag)
2901 struct nameidata nd;
2905 AUDIT_ARG_MODE(mode);
2906 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2907 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
2909 if ((error = namei(&nd)) != 0)
2911 vfslocked = NDHASGIANT(&nd);
2912 NDFREE(&nd, NDF_ONLY_PNBUF);
2913 error = setfmode(td, nd.ni_vp, mode);
2915 VFS_UNLOCK_GIANT(vfslocked);
2920 * Change mode of a file given a file descriptor.
2922 #ifndef _SYS_SYSPROTO_H_
2923 struct fchmod_args {
2931 register struct fchmod_args /* {
2940 AUDIT_ARG_FD(uap->fd);
2941 AUDIT_ARG_MODE(uap->mode);
2942 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2944 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2946 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2947 AUDIT_ARG_VNODE1(fp->f_vnode);
2948 VOP_UNLOCK(fp->f_vnode, 0);
2950 error = setfmode(td, fp->f_vnode, uap->mode);
2951 VFS_UNLOCK_GIANT(vfslocked);
2957 * Common implementation for chown(), lchown(), and fchown()
2960 setfown(td, vp, uid, gid)
2970 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2972 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2977 error = mac_vnode_check_setowner(td->td_ucred, vp, vattr.va_uid,
2981 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2983 vn_finished_write(mp);
2988 * Set ownership given a path name.
2990 #ifndef _SYS_SYSPROTO_H_
3000 register struct chown_args /* {
3007 return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3010 #ifndef _SYS_SYSPROTO_H_
3011 struct fchownat_args {
3020 fchownat(struct thread *td, struct fchownat_args *uap)
3025 if (flag & ~AT_SYMLINK_NOFOLLOW)
3028 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3029 uap->gid, uap->flag));
3033 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3037 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3041 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3042 int uid, int gid, int flag)
3044 struct nameidata nd;
3045 int error, vfslocked, follow;
3047 AUDIT_ARG_OWNER(uid, gid);
3048 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3049 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
3052 if ((error = namei(&nd)) != 0)
3054 vfslocked = NDHASGIANT(&nd);
3055 NDFREE(&nd, NDF_ONLY_PNBUF);
3056 error = setfown(td, nd.ni_vp, uid, gid);
3058 VFS_UNLOCK_GIANT(vfslocked);
3063 * Set ownership given a path name, do not cross symlinks.
3065 #ifndef _SYS_SYSPROTO_H_
3066 struct lchown_args {
3075 register struct lchown_args /* {
3082 return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3086 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3090 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3091 AT_SYMLINK_NOFOLLOW));
3095 * Set ownership given a file descriptor.
3097 #ifndef _SYS_SYSPROTO_H_
3098 struct fchown_args {
3107 register struct fchown_args /* {
3117 AUDIT_ARG_FD(uap->fd);
3118 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3119 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3121 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3123 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3124 AUDIT_ARG_VNODE1(fp->f_vnode);
3125 VOP_UNLOCK(fp->f_vnode, 0);
3127 error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
3128 VFS_UNLOCK_GIANT(vfslocked);
3134 * Common implementation code for utimes(), lutimes(), and futimes().
3137 getutimes(usrtvp, tvpseg, tsp)
3138 const struct timeval *usrtvp;
3139 enum uio_seg tvpseg;
3140 struct timespec *tsp;
3142 struct timeval tv[2];
3143 const struct timeval *tvp;
3146 if (usrtvp == NULL) {
3147 vfs_timestamp(&tsp[0]);
3150 if (tvpseg == UIO_SYSSPACE) {
3153 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3158 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3159 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3161 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3162 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3168 * Common implementation code for utimes(), lutimes(), and futimes().
3171 setutimes(td, vp, ts, numtimes, nullflag)
3174 const struct timespec *ts;
3178 int error, setbirthtime;
3182 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3184 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3186 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3187 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3190 vattr.va_atime = ts[0];
3191 vattr.va_mtime = ts[1];
3193 vattr.va_birthtime = ts[1];
3195 vattr.va_birthtime = ts[2];
3197 vattr.va_vaflags |= VA_UTIMES_NULL;
3199 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3203 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3205 vn_finished_write(mp);
3210 * Set the access and modification times of a file.
3212 #ifndef _SYS_SYSPROTO_H_
3213 struct utimes_args {
3215 struct timeval *tptr;
3221 register struct utimes_args /* {
3223 struct timeval *tptr;
3227 return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3231 #ifndef _SYS_SYSPROTO_H_
3232 struct futimesat_args {
3235 const struct timeval * times;
3239 futimesat(struct thread *td, struct futimesat_args *uap)
3242 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3243 uap->times, UIO_USERSPACE));
3247 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3248 struct timeval *tptr, enum uio_seg tptrseg)
3251 return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3255 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3256 struct timeval *tptr, enum uio_seg tptrseg)
3258 struct nameidata nd;
3259 struct timespec ts[2];
3260 int error, vfslocked;
3262 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3264 NDINIT_AT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path,
3267 if ((error = namei(&nd)) != 0)
3269 vfslocked = NDHASGIANT(&nd);
3270 NDFREE(&nd, NDF_ONLY_PNBUF);
3271 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3273 VFS_UNLOCK_GIANT(vfslocked);
3278 * Set the access and modification times of a file.
3280 #ifndef _SYS_SYSPROTO_H_
3281 struct lutimes_args {
3283 struct timeval *tptr;
3289 register struct lutimes_args /* {
3291 struct timeval *tptr;
3295 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3300 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3301 struct timeval *tptr, enum uio_seg tptrseg)
3303 struct timespec ts[2];
3305 struct nameidata nd;
3308 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3310 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3311 if ((error = namei(&nd)) != 0)
3313 vfslocked = NDHASGIANT(&nd);
3314 NDFREE(&nd, NDF_ONLY_PNBUF);
3315 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3317 VFS_UNLOCK_GIANT(vfslocked);
3322 * Set the access and modification times of a file.
3324 #ifndef _SYS_SYSPROTO_H_
3325 struct futimes_args {
3327 struct timeval *tptr;
3333 register struct futimes_args /* {
3335 struct timeval *tptr;
3339 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3343 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3344 enum uio_seg tptrseg)
3346 struct timespec ts[2];
3352 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3354 if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
3356 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3358 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3359 AUDIT_ARG_VNODE1(fp->f_vnode);
3360 VOP_UNLOCK(fp->f_vnode, 0);
3362 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3363 VFS_UNLOCK_GIANT(vfslocked);
3369 * Truncate a file given its path name.
3371 #ifndef _SYS_SYSPROTO_H_
3372 struct truncate_args {
3381 register struct truncate_args /* {
3388 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3392 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3398 struct nameidata nd;
3403 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3404 if ((error = namei(&nd)) != 0)
3406 vfslocked = NDHASGIANT(&nd);
3408 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3410 VFS_UNLOCK_GIANT(vfslocked);
3413 NDFREE(&nd, NDF_ONLY_PNBUF);
3414 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3415 if (vp->v_type == VDIR)
3418 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3421 else if ((error = vn_writechk(vp)) == 0 &&
3422 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3424 vattr.va_size = length;
3425 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3428 vn_finished_write(mp);
3429 VFS_UNLOCK_GIANT(vfslocked);
3433 #if defined(COMPAT_43)
3435 * Truncate a file given its path name.
3437 #ifndef _SYS_SYSPROTO_H_
3438 struct otruncate_args {
3446 register struct otruncate_args /* {
3451 struct truncate_args /* {
3457 nuap.path = uap->path;
3458 nuap.length = uap->length;
3459 return (truncate(td, &nuap));
3461 #endif /* COMPAT_43 */
3463 /* Versions with the pad argument */
3465 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3467 struct truncate_args ouap;
3469 ouap.path = uap->path;
3470 ouap.length = uap->length;
3471 return (truncate(td, &ouap));
3475 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3477 struct ftruncate_args ouap;
3480 ouap.length = uap->length;
3481 return (ftruncate(td, &ouap));
3485 * Sync an open file.
3487 #ifndef _SYS_SYSPROTO_H_
3495 struct fsync_args /* {
3503 int error, lock_flags;
3505 AUDIT_ARG_FD(uap->fd);
3506 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3509 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3510 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3512 if (MNT_SHARED_WRITES(mp) ||
3513 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3514 lock_flags = LK_SHARED;
3516 lock_flags = LK_EXCLUSIVE;
3518 vn_lock(vp, lock_flags | LK_RETRY);
3519 AUDIT_ARG_VNODE1(vp);
3520 if (vp->v_object != NULL) {
3521 VM_OBJECT_LOCK(vp->v_object);
3522 vm_object_page_clean(vp->v_object, 0, 0, 0);
3523 VM_OBJECT_UNLOCK(vp->v_object);
3525 error = VOP_FSYNC(vp, MNT_WAIT, td);
3528 vn_finished_write(mp);
3530 VFS_UNLOCK_GIANT(vfslocked);
3536 * Rename files. Source and destination must either both be directories, or
3537 * both not be directories. If target is a directory, it must be empty.
3539 #ifndef _SYS_SYSPROTO_H_
3540 struct rename_args {
3548 register struct rename_args /* {
3554 return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3557 #ifndef _SYS_SYSPROTO_H_
3558 struct renameat_args {
3566 renameat(struct thread *td, struct renameat_args *uap)
3569 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3574 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3577 return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3581 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3582 enum uio_seg pathseg)
3584 struct mount *mp = NULL;
3585 struct vnode *tvp, *fvp, *tdvp;
3586 struct nameidata fromnd, tond;
3593 NDINIT_AT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE |
3594 AUDITVNODE1, pathseg, old, oldfd, td);
3596 NDINIT_AT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3597 AUDITVNODE1, pathseg, old, oldfd, td);
3600 if ((error = namei(&fromnd)) != 0)
3602 fvfslocked = NDHASGIANT(&fromnd);
3605 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3606 fromnd.ni_vp, &fromnd.ni_cnd);
3607 VOP_UNLOCK(fromnd.ni_dvp, 0);
3608 if (fromnd.ni_dvp != fromnd.ni_vp)
3609 VOP_UNLOCK(fromnd.ni_vp, 0);
3613 error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3615 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3616 vrele(fromnd.ni_dvp);
3620 NDINIT_AT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3621 MPSAFE | AUDITVNODE2, pathseg, new, newfd, td);
3622 if (fromnd.ni_vp->v_type == VDIR)
3623 tond.ni_cnd.cn_flags |= WILLBEDIR;
3624 if ((error = namei(&tond)) != 0) {
3625 /* Translate error code for rename("dir1", "dir2/."). */
3626 if (error == EISDIR && fvp->v_type == VDIR)
3628 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3629 vrele(fromnd.ni_dvp);
3631 vn_finished_write(mp);
3634 tvfslocked = NDHASGIANT(&tond);
3638 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3641 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3651 * If the source is the same as the destination (that is, if they
3652 * are links to the same vnode), then there is nothing to do.
3658 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3659 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3663 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3664 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3665 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3666 NDFREE(&tond, NDF_ONLY_PNBUF);
3668 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3669 NDFREE(&tond, NDF_ONLY_PNBUF);
3676 vrele(fromnd.ni_dvp);
3679 vrele(tond.ni_startdir);
3680 vn_finished_write(mp);
3682 if (fromnd.ni_startdir)
3683 vrele(fromnd.ni_startdir);
3684 VFS_UNLOCK_GIANT(fvfslocked);
3685 VFS_UNLOCK_GIANT(tvfslocked);
3692 * Make a directory file.
3694 #ifndef _SYS_SYSPROTO_H_
3703 register struct mkdir_args /* {
3709 return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3712 #ifndef _SYS_SYSPROTO_H_
3713 struct mkdirat_args {
3720 mkdirat(struct thread *td, struct mkdirat_args *uap)
3723 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3727 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3730 return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3734 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3741 struct nameidata nd;
3744 AUDIT_ARG_MODE(mode);
3747 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
3748 segflg, path, fd, td);
3749 nd.ni_cnd.cn_flags |= WILLBEDIR;
3750 if ((error = namei(&nd)) != 0)
3752 vfslocked = NDHASGIANT(&nd);
3755 NDFREE(&nd, NDF_ONLY_PNBUF);
3757 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3758 * the strange behaviour of leaving the vnode unlocked
3759 * if the target is the same vnode as the parent.
3761 if (vp == nd.ni_dvp)
3766 VFS_UNLOCK_GIANT(vfslocked);
3769 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3770 NDFREE(&nd, NDF_ONLY_PNBUF);
3772 VFS_UNLOCK_GIANT(vfslocked);
3773 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3778 vattr.va_type = VDIR;
3779 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3781 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3786 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3790 NDFREE(&nd, NDF_ONLY_PNBUF);
3794 vn_finished_write(mp);
3795 VFS_UNLOCK_GIANT(vfslocked);
3800 * Remove a directory file.
3802 #ifndef _SYS_SYSPROTO_H_
3810 struct rmdir_args /* {
3815 return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3819 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3822 return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3826 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3831 struct nameidata nd;
3836 NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
3837 pathseg, path, fd, td);
3838 if ((error = namei(&nd)) != 0)
3840 vfslocked = NDHASGIANT(&nd);
3842 if (vp->v_type != VDIR) {
3847 * No rmdir "." please.
3849 if (nd.ni_dvp == vp) {
3854 * The root of a mounted filesystem cannot be deleted.
3856 if (vp->v_vflag & VV_ROOT) {
3861 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3866 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3867 NDFREE(&nd, NDF_ONLY_PNBUF);
3869 if (nd.ni_dvp == vp)
3873 VFS_UNLOCK_GIANT(vfslocked);
3874 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3878 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3879 vn_finished_write(mp);
3881 NDFREE(&nd, NDF_ONLY_PNBUF);
3883 if (nd.ni_dvp == vp)
3887 VFS_UNLOCK_GIANT(vfslocked);
3893 * Read a block of directory entries in a filesystem independent format.
3895 #ifndef _SYS_SYSPROTO_H_
3896 struct ogetdirentries_args {
3904 ogetdirentries(td, uap)
3906 register struct ogetdirentries_args /* {
3915 struct uio auio, kuio;
3916 struct iovec aiov, kiov;
3917 struct dirent *dp, *edp;
3919 int error, eofflag, readcnt, vfslocked;
3922 /* XXX arbitrary sanity limit on `count'. */
3923 if (uap->count > 64 * 1024)
3925 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3927 if ((fp->f_flag & FREAD) == 0) {
3933 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3934 if (vp->v_type != VDIR) {
3935 VFS_UNLOCK_GIANT(vfslocked);
3939 aiov.iov_base = uap->buf;
3940 aiov.iov_len = uap->count;
3941 auio.uio_iov = &aiov;
3942 auio.uio_iovcnt = 1;
3943 auio.uio_rw = UIO_READ;
3944 auio.uio_segflg = UIO_USERSPACE;
3946 auio.uio_resid = uap->count;
3947 vn_lock(vp, LK_SHARED | LK_RETRY);
3948 loff = auio.uio_offset = fp->f_offset;
3950 error = mac_vnode_check_readdir(td->td_ucred, vp);
3953 VFS_UNLOCK_GIANT(vfslocked);
3958 # if (BYTE_ORDER != LITTLE_ENDIAN)
3959 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3960 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3962 fp->f_offset = auio.uio_offset;
3967 kuio.uio_iov = &kiov;
3968 kuio.uio_segflg = UIO_SYSSPACE;
3969 kiov.iov_len = uap->count;
3970 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3971 kiov.iov_base = dirbuf;
3972 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3974 fp->f_offset = kuio.uio_offset;
3976 readcnt = uap->count - kuio.uio_resid;
3977 edp = (struct dirent *)&dirbuf[readcnt];
3978 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3979 # if (BYTE_ORDER == LITTLE_ENDIAN)
3981 * The expected low byte of
3982 * dp->d_namlen is our dp->d_type.
3983 * The high MBZ byte of dp->d_namlen
3984 * is our dp->d_namlen.
3986 dp->d_type = dp->d_namlen;
3990 * The dp->d_type is the high byte
3991 * of the expected dp->d_namlen,
3992 * so must be zero'ed.
3996 if (dp->d_reclen > 0) {
3997 dp = (struct dirent *)
3998 ((char *)dp + dp->d_reclen);
4005 error = uiomove(dirbuf, readcnt, &auio);
4007 free(dirbuf, M_TEMP);
4011 VFS_UNLOCK_GIANT(vfslocked);
4015 if (uap->count == auio.uio_resid &&
4016 (vp->v_vflag & VV_ROOT) &&
4017 (vp->v_mount->mnt_flag & MNT_UNION)) {
4018 struct vnode *tvp = vp;
4019 vp = vp->v_mount->mnt_vnodecovered;
4025 VFS_UNLOCK_GIANT(vfslocked);
4029 VFS_UNLOCK_GIANT(vfslocked);
4030 error = copyout(&loff, uap->basep, sizeof(long));
4032 td->td_retval[0] = uap->count - auio.uio_resid;
4035 #endif /* COMPAT_43 */
4038 * Read a block of directory entries in a filesystem independent format.
4040 #ifndef _SYS_SYSPROTO_H_
4041 struct getdirentries_args {
4049 getdirentries(td, uap)
4051 register struct getdirentries_args /* {
4061 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4064 if (uap->basep != NULL)
4065 error = copyout(&base, uap->basep, sizeof(long));
4070 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4082 if (count > INT_MAX)
4084 if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
4086 if ((fp->f_flag & FREAD) == 0) {
4092 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4093 if (vp->v_type != VDIR) {
4094 VFS_UNLOCK_GIANT(vfslocked);
4098 aiov.iov_base = buf;
4099 aiov.iov_len = count;
4100 auio.uio_iov = &aiov;
4101 auio.uio_iovcnt = 1;
4102 auio.uio_rw = UIO_READ;
4103 auio.uio_segflg = UIO_USERSPACE;
4105 auio.uio_resid = count;
4106 vn_lock(vp, LK_SHARED | LK_RETRY);
4107 AUDIT_ARG_VNODE1(vp);
4108 loff = auio.uio_offset = fp->f_offset;
4110 error = mac_vnode_check_readdir(td->td_ucred, vp);
4113 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4115 fp->f_offset = auio.uio_offset;
4118 VFS_UNLOCK_GIANT(vfslocked);
4121 if (count == auio.uio_resid &&
4122 (vp->v_vflag & VV_ROOT) &&
4123 (vp->v_mount->mnt_flag & MNT_UNION)) {
4124 struct vnode *tvp = vp;
4125 vp = vp->v_mount->mnt_vnodecovered;
4131 VFS_UNLOCK_GIANT(vfslocked);
4135 VFS_UNLOCK_GIANT(vfslocked);
4137 td->td_retval[0] = count - auio.uio_resid;
4143 #ifndef _SYS_SYSPROTO_H_
4144 struct getdents_args {
4153 register struct getdents_args /* {
4159 struct getdirentries_args ap;
4162 ap.count = uap->count;
4164 return (getdirentries(td, &ap));
4168 * Set the mode mask for creation of filesystem nodes.
4170 #ifndef _SYS_SYSPROTO_H_
4178 struct umask_args /* {
4182 register struct filedesc *fdp;
4184 FILEDESC_XLOCK(td->td_proc->p_fd);
4185 fdp = td->td_proc->p_fd;
4186 td->td_retval[0] = fdp->fd_cmask;
4187 fdp->fd_cmask = uap->newmask & ALLPERMS;
4188 FILEDESC_XUNLOCK(td->td_proc->p_fd);
4193 * Void all references to file by ripping underlying filesystem away from
4196 #ifndef _SYS_SYSPROTO_H_
4197 struct revoke_args {
4204 register struct revoke_args /* {
4211 struct nameidata nd;
4214 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4215 UIO_USERSPACE, uap->path, td);
4216 if ((error = namei(&nd)) != 0)
4218 vfslocked = NDHASGIANT(&nd);
4220 NDFREE(&nd, NDF_ONLY_PNBUF);
4221 if (vp->v_type != VCHR) {
4226 error = mac_vnode_check_revoke(td->td_ucred, vp);
4230 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4233 if (td->td_ucred->cr_uid != vattr.va_uid) {
4234 error = priv_check(td, PRIV_VFS_ADMIN);
4239 VOP_REVOKE(vp, REVOKEALL);
4242 VFS_UNLOCK_GIANT(vfslocked);
4247 * Convert a user file descriptor to a kernel file entry.
4248 * A reference on the file entry is held upon returning.
4251 getvnode(fdp, fd, fpp)
4252 struct filedesc *fdp;
4261 if (fdp == NULL || (fp = fget_unlocked(fdp, fd)) == NULL)
4263 else if (fp->f_vnode == NULL) {
4265 fdrop(fp, curthread);
4272 * Get an (NFS) file handle.
4274 #ifndef _SYS_SYSPROTO_H_
4275 struct lgetfh_args {
4283 register struct lgetfh_args *uap;
4285 struct nameidata nd;
4287 register struct vnode *vp;
4291 error = priv_check(td, PRIV_VFS_GETFH);
4294 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4295 UIO_USERSPACE, uap->fname, td);
4299 vfslocked = NDHASGIANT(&nd);
4300 NDFREE(&nd, NDF_ONLY_PNBUF);
4302 bzero(&fh, sizeof(fh));
4303 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4304 error = VOP_VPTOFH(vp, &fh.fh_fid);
4306 VFS_UNLOCK_GIANT(vfslocked);
4309 error = copyout(&fh, uap->fhp, sizeof (fh));
4313 #ifndef _SYS_SYSPROTO_H_
4322 register struct getfh_args *uap;
4324 struct nameidata nd;
4326 register struct vnode *vp;
4330 error = priv_check(td, PRIV_VFS_GETFH);
4333 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4334 UIO_USERSPACE, uap->fname, td);
4338 vfslocked = NDHASGIANT(&nd);
4339 NDFREE(&nd, NDF_ONLY_PNBUF);
4341 bzero(&fh, sizeof(fh));
4342 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4343 error = VOP_VPTOFH(vp, &fh.fh_fid);
4345 VFS_UNLOCK_GIANT(vfslocked);
4348 error = copyout(&fh, uap->fhp, sizeof (fh));
4353 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4356 * warning: do not remove the priv_check() call or this becomes one giant
4359 #ifndef _SYS_SYSPROTO_H_
4360 struct fhopen_args {
4361 const struct fhandle *u_fhp;
4368 struct fhopen_args /* {
4369 const struct fhandle *u_fhp;
4373 struct proc *p = td->td_proc;
4378 struct vattr *vap = &vat;
4381 register struct filedesc *fdp = p->p_fd;
4382 int fmode, error, type;
4388 error = priv_check(td, PRIV_VFS_FHOPEN);
4391 fmode = FFLAGS(uap->flags);
4392 /* why not allow a non-read/write open for our lockd? */
4393 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4395 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4398 /* find the mount point */
4399 mp = vfs_busyfs(&fhp.fh_fsid);
4402 vfslocked = VFS_LOCK_GIANT(mp);
4403 /* now give me my vnode, it gets returned to me locked */
4404 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
4409 * from now on we have to make sure not
4410 * to forget about the vnode
4411 * any error that causes an abort must vput(vp)
4412 * just set error = err and 'goto bad;'.
4418 if (vp->v_type == VLNK) {
4422 if (vp->v_type == VSOCK) {
4427 if (fmode & (FWRITE | O_TRUNC)) {
4428 if (vp->v_type == VDIR) {
4432 error = vn_writechk(vp);
4439 if (fmode & O_APPEND)
4442 error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4447 error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4451 if (fmode & O_TRUNC) {
4453 VOP_UNLOCK(vp, 0); /* XXX */
4454 if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4459 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
4463 * We don't yet have fp->f_cred, so use td->td_ucred, which
4466 error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4471 error = VOP_SETATTR(vp, vap, td->td_ucred);
4475 vn_finished_write(mp);
4479 error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4487 * end of vn_open code
4490 if ((error = falloc(td, &nfp, &indx)) != 0) {
4495 /* An extra reference on `nfp' has been held for us by falloc(). */
4498 finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4499 if (fmode & (O_EXLOCK | O_SHLOCK)) {
4500 lf.l_whence = SEEK_SET;
4503 if (fmode & O_EXLOCK)
4504 lf.l_type = F_WRLCK;
4506 lf.l_type = F_RDLCK;
4508 if ((fmode & FNONBLOCK) == 0)
4511 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4514 * The lock request failed. Normally close the
4515 * descriptor but handle the case where someone might
4516 * have dup()d or close()d it when we weren't looking.
4518 fdclose(fdp, fp, indx, td);
4521 * release our private reference
4526 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4527 atomic_set_int(&fp->f_flag, FHASLOCK);
4532 VFS_UNLOCK_GIANT(vfslocked);
4533 td->td_retval[0] = indx;
4539 VFS_UNLOCK_GIANT(vfslocked);
4544 * Stat an (NFS) file handle.
4546 #ifndef _SYS_SYSPROTO_H_
4547 struct fhstat_args {
4548 struct fhandle *u_fhp;
4555 register struct fhstat_args /* {
4556 struct fhandle *u_fhp;
4567 error = priv_check(td, PRIV_VFS_FHSTAT);
4570 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4573 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4575 vfslocked = VFS_LOCK_GIANT(mp);
4576 error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4579 VFS_UNLOCK_GIANT(vfslocked);
4582 error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4584 VFS_UNLOCK_GIANT(vfslocked);
4587 error = copyout(&sb, uap->sb, sizeof(sb));
4592 * Implement fstatfs() for (NFS) file handles.
4594 #ifndef _SYS_SYSPROTO_H_
4595 struct fhstatfs_args {
4596 struct fhandle *u_fhp;
4603 struct fhstatfs_args /* {
4604 struct fhandle *u_fhp;
4612 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4615 error = kern_fhstatfs(td, fh, &sf);
4618 return (copyout(&sf, uap->buf, sizeof(sf)));
4622 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4630 error = priv_check(td, PRIV_VFS_FHSTATFS);
4633 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4635 vfslocked = VFS_LOCK_GIANT(mp);
4636 error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4639 VFS_UNLOCK_GIANT(vfslocked);
4643 error = prison_canseemount(td->td_ucred, mp);
4647 error = mac_mount_check_stat(td->td_ucred, mp);
4652 * Set these in case the underlying filesystem fails to do so.
4655 sp->f_version = STATFS_VERSION;
4656 sp->f_namemax = NAME_MAX;
4657 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4658 error = VFS_STATFS(mp, sp);
4663 VFS_UNLOCK_GIANT(vfslocked);