2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
49 #include <sys/capsicum.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
72 #include <sys/dirent.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
77 #include <sys/ktrace.h>
80 #include <machine/stdarg.h>
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
90 #include <ufs/ufs/quota.h>
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
94 SDT_PROVIDER_DEFINE(vfs);
95 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
96 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
98 static int kern_chflagsat(struct thread *td, int fd, const char *path,
99 enum uio_seg pathseg, u_long flags, int atflag);
100 static int setfflags(struct thread *td, struct vnode *, u_long);
101 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
102 static int getutimens(const struct timespec *, enum uio_seg,
103 struct timespec *, int *);
104 static int setutimes(struct thread *td, struct vnode *,
105 const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
110 * Sync each mounted filesystem.
112 #ifndef _SYS_SYSPROTO_H_
119 sys_sync(struct thread *td, struct sync_args *uap)
121 struct mount *mp, *nmp;
124 mtx_lock(&mountlist_mtx);
125 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
126 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
127 nmp = TAILQ_NEXT(mp, mnt_list);
130 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
131 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
132 save = curthread_pflags_set(TDP_SYNCIO);
133 vfs_msync(mp, MNT_NOWAIT);
134 VFS_SYNC(mp, MNT_NOWAIT);
135 curthread_pflags_restore(save);
136 vn_finished_write(mp);
138 mtx_lock(&mountlist_mtx);
139 nmp = TAILQ_NEXT(mp, mnt_list);
142 mtx_unlock(&mountlist_mtx);
147 * Change filesystem quotas.
149 #ifndef _SYS_SYSPROTO_H_
150 struct quotactl_args {
158 sys_quotactl(struct thread *td, struct quotactl_args *uap)
164 AUDIT_ARG_CMD(uap->cmd);
165 AUDIT_ARG_UID(uap->uid);
166 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
168 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
170 if ((error = namei(&nd)) != 0)
172 NDFREE(&nd, NDF_ONLY_PNBUF);
173 mp = nd.ni_vp->v_mount;
176 error = vfs_busy(mp, 0);
180 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
183 * Since quota on operation typically needs to open quota
184 * file, the Q_QUOTAON handler needs to unbusy the mount point
185 * before calling into namei. Otherwise, unmount might be
186 * started between two vfs_busy() invocations (first is our,
187 * second is from mount point cross-walk code in lookup()),
190 * Require that Q_QUOTAON handles the vfs_busy() reference on
191 * its own, always returning with ubusied mount point.
193 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
199 * Used by statfs conversion routines to scale the block size up if
200 * necessary so that all of the block counts are <= 'max_size'. Note
201 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
205 statfs_scale_blocks(struct statfs *sf, long max_size)
210 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
213 * Attempt to scale the block counts to give a more accurate
214 * overview to userland of the ratio of free space to used
215 * space. To do this, find the largest block count and compute
216 * a divisor that lets it fit into a signed integer <= max_size.
218 if (sf->f_bavail < 0)
219 count = -sf->f_bavail;
221 count = sf->f_bavail;
222 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
223 if (count <= max_size)
226 count >>= flsl(max_size);
233 sf->f_bsize <<= shift;
234 sf->f_blocks >>= shift;
235 sf->f_bfree >>= shift;
236 sf->f_bavail >>= shift;
240 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
247 error = vfs_busy(mp, 0);
252 error = mac_mount_check_stat(td->td_ucred, mp);
257 * Set these in case the underlying filesystem fails to do so.
260 sp->f_version = STATFS_VERSION;
261 sp->f_namemax = NAME_MAX;
262 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
263 error = VFS_STATFS(mp, sp);
267 if (priv_check(td, PRIV_VFS_GENERATION)) {
268 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
269 prison_enforce_statfs(td->td_ucred, mp, buf);
277 * Get filesystem statistics.
279 #ifndef _SYS_SYSPROTO_H_
286 sys_statfs(struct thread *td, struct statfs_args *uap)
291 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
292 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
294 error = copyout(sfp, uap->buf, sizeof(struct statfs));
300 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
307 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
312 mp = nd.ni_vp->v_mount;
314 NDFREE(&nd, NDF_ONLY_PNBUF);
316 return (kern_do_statfs(td, mp, buf));
320 * Get filesystem statistics.
322 #ifndef _SYS_SYSPROTO_H_
323 struct fstatfs_args {
329 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
334 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
335 error = kern_fstatfs(td, uap->fd, sfp);
337 error = copyout(sfp, uap->buf, sizeof(struct statfs));
343 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
352 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
356 vn_lock(vp, LK_SHARED | LK_RETRY);
358 AUDIT_ARG_VNODE1(vp);
365 return (kern_do_statfs(td, mp, buf));
369 * Get statistics on all filesystems.
371 #ifndef _SYS_SYSPROTO_H_
372 struct getfsstat_args {
379 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
384 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
386 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
387 UIO_USERSPACE, uap->mode);
389 td->td_retval[0] = count;
394 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
395 * The caller is responsible for freeing memory which will be allocated
399 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
400 size_t *countp, enum uio_seg bufseg, int mode)
402 struct mount *mp, *nmp;
403 struct statfs *sfsp, *sp, *sptmp, *tofree;
404 size_t count, maxcount;
412 if (bufseg == UIO_SYSSPACE)
417 maxcount = bufsize / sizeof(struct statfs);
421 } else if (bufseg == UIO_USERSPACE) {
424 } else /* if (bufseg == UIO_SYSSPACE) */ {
426 mtx_lock(&mountlist_mtx);
427 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
430 mtx_unlock(&mountlist_mtx);
431 if (maxcount > count)
433 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
437 mtx_lock(&mountlist_mtx);
438 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
439 if (prison_canseemount(td->td_ucred, mp) != 0) {
440 nmp = TAILQ_NEXT(mp, mnt_list);
444 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
445 nmp = TAILQ_NEXT(mp, mnt_list);
449 if (mode == MNT_WAIT) {
450 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
452 * If vfs_busy() failed, and MBF_NOWAIT
453 * wasn't passed, then the mp is gone.
454 * Furthermore, because of MBF_MNTLSTLOCK,
455 * the mountlist_mtx was dropped. We have
456 * no other choice than to start over.
458 mtx_unlock(&mountlist_mtx);
459 free(tofree, M_STATFS);
463 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
464 nmp = TAILQ_NEXT(mp, mnt_list);
468 if (sfsp != NULL && count < maxcount) {
471 * Set these in case the underlying filesystem
474 sp->f_version = STATFS_VERSION;
475 sp->f_namemax = NAME_MAX;
476 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
478 * If MNT_NOWAIT is specified, do not refresh
481 if (mode != MNT_NOWAIT) {
482 error = VFS_STATFS(mp, sp);
484 mtx_lock(&mountlist_mtx);
485 nmp = TAILQ_NEXT(mp, mnt_list);
490 if (priv_check(td, PRIV_VFS_GENERATION)) {
491 sptmp = malloc(sizeof(struct statfs), M_STATFS,
494 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
495 prison_enforce_statfs(td->td_ucred, mp, sptmp);
499 if (bufseg == UIO_SYSSPACE) {
500 bcopy(sp, sfsp, sizeof(*sp));
501 free(sptmp, M_STATFS);
502 } else /* if (bufseg == UIO_USERSPACE) */ {
503 error = copyout(sp, sfsp, sizeof(*sp));
504 free(sptmp, M_STATFS);
513 mtx_lock(&mountlist_mtx);
514 nmp = TAILQ_NEXT(mp, mnt_list);
517 mtx_unlock(&mountlist_mtx);
518 if (sfsp != NULL && count > maxcount)
525 #ifdef COMPAT_FREEBSD4
527 * Get old format filesystem statistics.
529 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
531 #ifndef _SYS_SYSPROTO_H_
532 struct freebsd4_statfs_args {
538 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
544 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
545 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
547 freebsd4_cvtstatfs(sfp, &osb);
548 error = copyout(&osb, uap->buf, sizeof(osb));
555 * Get filesystem statistics.
557 #ifndef _SYS_SYSPROTO_H_
558 struct freebsd4_fstatfs_args {
564 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
570 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
571 error = kern_fstatfs(td, uap->fd, sfp);
573 freebsd4_cvtstatfs(sfp, &osb);
574 error = copyout(&osb, uap->buf, sizeof(osb));
581 * Get statistics on all filesystems.
583 #ifndef _SYS_SYSPROTO_H_
584 struct freebsd4_getfsstat_args {
591 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
593 struct statfs *buf, *sp;
598 if (uap->bufsize < 0)
600 count = uap->bufsize / sizeof(struct ostatfs);
601 if (count > SIZE_MAX / sizeof(struct statfs))
603 size = count * sizeof(struct statfs);
604 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
607 td->td_retval[0] = count;
610 while (count != 0 && error == 0) {
611 freebsd4_cvtstatfs(sp, &osb);
612 error = copyout(&osb, uap->buf, sizeof(osb));
623 * Implement fstatfs() for (NFS) file handles.
625 #ifndef _SYS_SYSPROTO_H_
626 struct freebsd4_fhstatfs_args {
627 struct fhandle *u_fhp;
632 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
639 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
642 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
643 error = kern_fhstatfs(td, fh, sfp);
645 freebsd4_cvtstatfs(sfp, &osb);
646 error = copyout(&osb, uap->buf, sizeof(osb));
653 * Convert a new format statfs structure to an old format statfs structure.
656 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
659 statfs_scale_blocks(nsp, LONG_MAX);
660 bzero(osp, sizeof(*osp));
661 osp->f_bsize = nsp->f_bsize;
662 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
663 osp->f_blocks = nsp->f_blocks;
664 osp->f_bfree = nsp->f_bfree;
665 osp->f_bavail = nsp->f_bavail;
666 osp->f_files = MIN(nsp->f_files, LONG_MAX);
667 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
668 osp->f_owner = nsp->f_owner;
669 osp->f_type = nsp->f_type;
670 osp->f_flags = nsp->f_flags;
671 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
672 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
673 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
674 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
675 strlcpy(osp->f_fstypename, nsp->f_fstypename,
676 MIN(MFSNAMELEN, OMFSNAMELEN));
677 strlcpy(osp->f_mntonname, nsp->f_mntonname,
678 MIN(MNAMELEN, OMNAMELEN));
679 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
680 MIN(MNAMELEN, OMNAMELEN));
681 osp->f_fsid = nsp->f_fsid;
683 #endif /* COMPAT_FREEBSD4 */
685 #if defined(COMPAT_FREEBSD11)
687 * Get old format filesystem statistics.
689 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
692 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
694 struct freebsd11_statfs osb;
698 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
699 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
701 freebsd11_cvtstatfs(sfp, &osb);
702 error = copyout(&osb, uap->buf, sizeof(osb));
709 * Get filesystem statistics.
712 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
714 struct freebsd11_statfs osb;
718 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
719 error = kern_fstatfs(td, uap->fd, sfp);
721 freebsd11_cvtstatfs(sfp, &osb);
722 error = copyout(&osb, uap->buf, sizeof(osb));
729 * Get statistics on all filesystems.
732 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
734 struct freebsd11_statfs osb;
735 struct statfs *buf, *sp;
739 count = uap->bufsize / sizeof(struct ostatfs);
740 size = count * sizeof(struct statfs);
741 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
744 td->td_retval[0] = count;
747 while (count > 0 && error == 0) {
748 freebsd11_cvtstatfs(sp, &osb);
749 error = copyout(&osb, uap->buf, sizeof(osb));
760 * Implement fstatfs() for (NFS) file handles.
763 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
765 struct freebsd11_statfs osb;
770 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
773 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
774 error = kern_fhstatfs(td, fh, sfp);
776 freebsd11_cvtstatfs(sfp, &osb);
777 error = copyout(&osb, uap->buf, sizeof(osb));
784 * Convert a new format statfs structure to an old format statfs structure.
787 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
790 bzero(osp, sizeof(*osp));
791 osp->f_version = FREEBSD11_STATFS_VERSION;
792 osp->f_type = nsp->f_type;
793 osp->f_flags = nsp->f_flags;
794 osp->f_bsize = nsp->f_bsize;
795 osp->f_iosize = nsp->f_iosize;
796 osp->f_blocks = nsp->f_blocks;
797 osp->f_bfree = nsp->f_bfree;
798 osp->f_bavail = nsp->f_bavail;
799 osp->f_files = nsp->f_files;
800 osp->f_ffree = nsp->f_ffree;
801 osp->f_syncwrites = nsp->f_syncwrites;
802 osp->f_asyncwrites = nsp->f_asyncwrites;
803 osp->f_syncreads = nsp->f_syncreads;
804 osp->f_asyncreads = nsp->f_asyncreads;
805 osp->f_namemax = nsp->f_namemax;
806 osp->f_owner = nsp->f_owner;
807 osp->f_fsid = nsp->f_fsid;
808 strlcpy(osp->f_fstypename, nsp->f_fstypename,
809 MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
810 strlcpy(osp->f_mntonname, nsp->f_mntonname,
811 MIN(MNAMELEN, sizeof(osp->f_mntonname)));
812 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
813 MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
815 #endif /* COMPAT_FREEBSD11 */
818 * Change current working directory to a given file descriptor.
820 #ifndef _SYS_SYSPROTO_H_
826 sys_fchdir(struct thread *td, struct fchdir_args *uap)
828 struct vnode *vp, *tdp;
834 AUDIT_ARG_FD(uap->fd);
835 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
842 vn_lock(vp, LK_SHARED | LK_RETRY);
843 AUDIT_ARG_VNODE1(vp);
844 error = change_dir(vp, td);
845 while (!error && (mp = vp->v_mountedhere) != NULL) {
848 error = VFS_ROOT(mp, LK_SHARED, &tdp);
865 * Change current working directory (``.'').
867 #ifndef _SYS_SYSPROTO_H_
873 sys_chdir(struct thread *td, struct chdir_args *uap)
876 return (kern_chdir(td, uap->path, UIO_USERSPACE));
880 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
885 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
887 if ((error = namei(&nd)) != 0)
889 if ((error = change_dir(nd.ni_vp, td)) != 0) {
891 NDFREE(&nd, NDF_ONLY_PNBUF);
894 VOP_UNLOCK(nd.ni_vp, 0);
895 NDFREE(&nd, NDF_ONLY_PNBUF);
896 pwd_chdir(td, nd.ni_vp);
901 * Change notion of root (``/'') directory.
903 #ifndef _SYS_SYSPROTO_H_
909 sys_chroot(struct thread *td, struct chroot_args *uap)
914 error = priv_check(td, PRIV_VFS_CHROOT);
917 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
918 UIO_USERSPACE, uap->path, td);
922 error = change_dir(nd.ni_vp, td);
926 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
930 VOP_UNLOCK(nd.ni_vp, 0);
931 error = pwd_chroot(td, nd.ni_vp);
933 NDFREE(&nd, NDF_ONLY_PNBUF);
938 NDFREE(&nd, NDF_ONLY_PNBUF);
943 * Common routine for chroot and chdir. Callers must provide a locked vnode
947 change_dir(struct vnode *vp, struct thread *td)
953 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
954 if (vp->v_type != VDIR)
957 error = mac_vnode_check_chdir(td->td_ucred, vp);
961 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
965 flags_to_rights(int flags, cap_rights_t *rightsp)
968 if (flags & O_EXEC) {
969 cap_rights_set(rightsp, CAP_FEXECVE);
971 switch ((flags & O_ACCMODE)) {
973 cap_rights_set(rightsp, CAP_READ);
976 cap_rights_set(rightsp, CAP_READ);
979 cap_rights_set(rightsp, CAP_WRITE);
980 if (!(flags & (O_APPEND | O_TRUNC)))
981 cap_rights_set(rightsp, CAP_SEEK);
987 cap_rights_set(rightsp, CAP_CREATE);
990 cap_rights_set(rightsp, CAP_FTRUNCATE);
992 if (flags & (O_SYNC | O_FSYNC))
993 cap_rights_set(rightsp, CAP_FSYNC);
995 if (flags & (O_EXLOCK | O_SHLOCK))
996 cap_rights_set(rightsp, CAP_FLOCK);
1000 * Check permissions, allocate an open file structure, and call the device
1001 * open routine if any.
1003 #ifndef _SYS_SYSPROTO_H_
1011 sys_open(struct thread *td, struct open_args *uap)
1014 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1015 uap->flags, uap->mode));
1018 #ifndef _SYS_SYSPROTO_H_
1019 struct openat_args {
1027 sys_openat(struct thread *td, struct openat_args *uap)
1030 AUDIT_ARG_FD(uap->fd);
1031 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1036 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1037 int flags, int mode)
1039 struct proc *p = td->td_proc;
1040 struct filedesc *fdp = p->p_fd;
1043 struct nameidata nd;
1044 cap_rights_t rights;
1045 int cmode, error, indx;
1049 AUDIT_ARG_FFLAGS(flags);
1050 AUDIT_ARG_MODE(mode);
1051 cap_rights_init(&rights, CAP_LOOKUP);
1052 flags_to_rights(flags, &rights);
1054 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1057 if (flags & O_EXEC) {
1058 if (flags & O_ACCMODE)
1060 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1063 flags = FFLAGS(flags);
1067 * Allocate a file structure. The descriptor to reference it
1068 * is allocated and set by finstall() below.
1070 error = falloc_noinstall(td, &fp);
1074 * An extra reference on `fp' has been held for us by
1075 * falloc_noinstall().
1077 /* Set the flags early so the finit in devfs can pick them up. */
1078 fp->f_flag = flags & FMASK;
1079 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1080 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1082 td->td_dupfd = -1; /* XXX check for fdopen */
1083 error = vn_open(&nd, &flags, cmode, fp);
1086 * If the vn_open replaced the method vector, something
1087 * wonderous happened deep below and we just pass it up
1088 * pretending we know what we do.
1090 if (error == ENXIO && fp->f_ops != &badfileops)
1094 * Handle special fdopen() case. bleh.
1096 * Don't do this for relative (capability) lookups; we don't
1097 * understand exactly what would happen, and we don't think
1098 * that it ever should.
1100 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1101 (error == ENODEV || error == ENXIO) &&
1102 td->td_dupfd >= 0) {
1103 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1112 NDFREE(&nd, NDF_ONLY_PNBUF);
1116 * Store the vnode, for any f_type. Typically, the vnode use
1117 * count is decremented by direct call to vn_closefile() for
1118 * files that switched type in the cdevsw fdopen() method.
1122 * If the file wasn't claimed by devfs bind it to the normal
1123 * vnode operations here.
1125 if (fp->f_ops == &badfileops) {
1126 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1128 finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1129 DTYPE_VNODE, vp, &vnops);
1133 if (flags & O_TRUNC) {
1134 error = fo_truncate(fp, 0, td->td_ucred, td);
1140 * If we haven't already installed the FD (for dupfdopen), do so now.
1143 struct filecaps *fcaps;
1146 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1147 fcaps = &nd.ni_filecaps;
1151 error = finstall(td, fp, &indx, flags, fcaps);
1152 /* On success finstall() consumes fcaps. */
1154 filecaps_free(&nd.ni_filecaps);
1158 filecaps_free(&nd.ni_filecaps);
1162 * Release our private reference, leaving the one associated with
1163 * the descriptor table intact.
1166 td->td_retval[0] = indx;
1169 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1178 #ifndef _SYS_SYSPROTO_H_
1179 struct ocreat_args {
1185 ocreat(struct thread *td, struct ocreat_args *uap)
1188 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1189 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1191 #endif /* COMPAT_43 */
1194 * Create a special file.
1196 #ifndef _SYS_SYSPROTO_H_
1197 struct mknodat_args {
1205 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1208 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1212 #if defined(COMPAT_FREEBSD11)
1214 freebsd11_mknod(struct thread *td,
1215 struct freebsd11_mknod_args *uap)
1218 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1219 uap->mode, uap->dev));
1223 freebsd11_mknodat(struct thread *td,
1224 struct freebsd11_mknodat_args *uap)
1227 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1230 #endif /* COMPAT_FREEBSD11 */
1233 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1234 int mode, dev_t dev)
1239 struct nameidata nd;
1240 cap_rights_t rights;
1241 int error, whiteout = 0;
1243 AUDIT_ARG_MODE(mode);
1245 switch (mode & S_IFMT) {
1248 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1249 if (error == 0 && dev == VNOVAL)
1253 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1257 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1267 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1268 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1270 if ((error = namei(&nd)) != 0)
1274 NDFREE(&nd, NDF_ONLY_PNBUF);
1275 if (vp == nd.ni_dvp)
1283 vattr.va_mode = (mode & ALLPERMS) &
1284 ~td->td_proc->p_fd->fd_cmask;
1285 vattr.va_rdev = dev;
1288 switch (mode & S_IFMT) {
1290 vattr.va_type = VCHR;
1293 vattr.va_type = VBLK;
1299 panic("kern_mknod: invalid mode");
1302 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1303 NDFREE(&nd, NDF_ONLY_PNBUF);
1305 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1310 if (error == 0 && !whiteout)
1311 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1312 &nd.ni_cnd, &vattr);
1316 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1318 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1319 &nd.ni_cnd, &vattr);
1324 NDFREE(&nd, NDF_ONLY_PNBUF);
1326 vn_finished_write(mp);
1331 * Create a named pipe.
1333 #ifndef _SYS_SYSPROTO_H_
1334 struct mkfifo_args {
1340 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1343 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1347 #ifndef _SYS_SYSPROTO_H_
1348 struct mkfifoat_args {
1355 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1358 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1363 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1368 struct nameidata nd;
1369 cap_rights_t rights;
1372 AUDIT_ARG_MODE(mode);
1375 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1376 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1378 if ((error = namei(&nd)) != 0)
1380 if (nd.ni_vp != NULL) {
1381 NDFREE(&nd, NDF_ONLY_PNBUF);
1382 if (nd.ni_vp == nd.ni_dvp)
1389 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1390 NDFREE(&nd, NDF_ONLY_PNBUF);
1392 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1397 vattr.va_type = VFIFO;
1398 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1400 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1405 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1412 vn_finished_write(mp);
1413 NDFREE(&nd, NDF_ONLY_PNBUF);
1418 * Make a hard file link.
1420 #ifndef _SYS_SYSPROTO_H_
1427 sys_link(struct thread *td, struct link_args *uap)
1430 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1431 UIO_USERSPACE, FOLLOW));
1434 #ifndef _SYS_SYSPROTO_H_
1435 struct linkat_args {
1444 sys_linkat(struct thread *td, struct linkat_args *uap)
1449 if (flag & ~AT_SYMLINK_FOLLOW)
1452 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1453 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1456 int hardlink_check_uid = 0;
1457 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1458 &hardlink_check_uid, 0,
1459 "Unprivileged processes cannot create hard links to files owned by other "
1461 static int hardlink_check_gid = 0;
1462 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1463 &hardlink_check_gid, 0,
1464 "Unprivileged processes cannot create hard links to files owned by other "
1468 can_hardlink(struct vnode *vp, struct ucred *cred)
1473 if (!hardlink_check_uid && !hardlink_check_gid)
1476 error = VOP_GETATTR(vp, &va, cred);
1480 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1481 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1486 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1487 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1496 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1497 enum uio_seg segflg, int follow)
1501 struct nameidata nd;
1502 cap_rights_t rights;
1507 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1508 cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1510 if ((error = namei(&nd)) != 0)
1512 NDFREE(&nd, NDF_ONLY_PNBUF);
1514 if (vp->v_type == VDIR) {
1516 return (EPERM); /* POSIX */
1518 NDINIT_ATRIGHTS(&nd, CREATE,
1519 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1520 cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1521 if ((error = namei(&nd)) == 0) {
1522 if (nd.ni_vp != NULL) {
1523 NDFREE(&nd, NDF_ONLY_PNBUF);
1524 if (nd.ni_dvp == nd.ni_vp)
1531 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1533 * Cross-device link. No need to recheck
1534 * vp->v_type, since it cannot change, except
1537 NDFREE(&nd, NDF_ONLY_PNBUF);
1541 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1542 error = can_hardlink(vp, td->td_ucred);
1545 error = mac_vnode_check_link(td->td_ucred,
1546 nd.ni_dvp, vp, &nd.ni_cnd);
1551 NDFREE(&nd, NDF_ONLY_PNBUF);
1554 error = vn_start_write(vp, &mp, V_NOWAIT);
1558 NDFREE(&nd, NDF_ONLY_PNBUF);
1559 error = vn_start_write(NULL, &mp,
1565 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1568 vn_finished_write(mp);
1569 NDFREE(&nd, NDF_ONLY_PNBUF);
1572 NDFREE(&nd, NDF_ONLY_PNBUF);
1582 * Make a symbolic link.
1584 #ifndef _SYS_SYSPROTO_H_
1585 struct symlink_args {
1591 sys_symlink(struct thread *td, struct symlink_args *uap)
1594 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1598 #ifndef _SYS_SYSPROTO_H_
1599 struct symlinkat_args {
1606 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1609 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1614 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1615 enum uio_seg segflg)
1620 struct nameidata nd;
1622 cap_rights_t rights;
1624 if (segflg == UIO_SYSSPACE) {
1627 syspath = uma_zalloc(namei_zone, M_WAITOK);
1628 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1631 AUDIT_ARG_TEXT(syspath);
1634 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1635 NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1637 if ((error = namei(&nd)) != 0)
1640 NDFREE(&nd, NDF_ONLY_PNBUF);
1641 if (nd.ni_vp == nd.ni_dvp)
1649 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1650 NDFREE(&nd, NDF_ONLY_PNBUF);
1652 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1657 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1659 vattr.va_type = VLNK;
1660 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1665 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1671 NDFREE(&nd, NDF_ONLY_PNBUF);
1673 vn_finished_write(mp);
1675 if (segflg != UIO_SYSSPACE)
1676 uma_zfree(namei_zone, syspath);
1681 * Delete a whiteout from the filesystem.
1683 #ifndef _SYS_SYSPROTO_H_
1684 struct undelete_args {
1689 sys_undelete(struct thread *td, struct undelete_args *uap)
1692 struct nameidata nd;
1697 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1698 UIO_USERSPACE, uap->path, td);
1703 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1704 NDFREE(&nd, NDF_ONLY_PNBUF);
1705 if (nd.ni_vp == nd.ni_dvp)
1713 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1714 NDFREE(&nd, NDF_ONLY_PNBUF);
1716 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1720 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1721 NDFREE(&nd, NDF_ONLY_PNBUF);
1723 vn_finished_write(mp);
1728 * Delete a name from the filesystem.
1730 #ifndef _SYS_SYSPROTO_H_
1731 struct unlink_args {
1736 sys_unlink(struct thread *td, struct unlink_args *uap)
1739 return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1742 #ifndef _SYS_SYSPROTO_H_
1743 struct unlinkat_args {
1750 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1752 int flag = uap->flag;
1754 char *path = uap->path;
1756 if (flag & ~AT_REMOVEDIR)
1759 if (flag & AT_REMOVEDIR)
1760 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1762 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1766 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1771 struct nameidata nd;
1773 cap_rights_t rights;
1778 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1779 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1780 if ((error = namei(&nd)) != 0)
1781 return (error == EINVAL ? EPERM : error);
1783 if (vp->v_type == VDIR && oldinum == 0) {
1784 error = EPERM; /* POSIX */
1785 } else if (oldinum != 0 &&
1786 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1787 sb.st_ino != oldinum) {
1788 error = EIDRM; /* Identifier removed */
1791 * The root of a mounted filesystem cannot be deleted.
1793 * XXX: can this only be a VDIR case?
1795 if (vp->v_vflag & VV_ROOT)
1799 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1800 NDFREE(&nd, NDF_ONLY_PNBUF);
1802 if (vp == nd.ni_dvp)
1806 if ((error = vn_start_write(NULL, &mp,
1807 V_XSLEEP | PCATCH)) != 0)
1812 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1817 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1818 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1822 vn_finished_write(mp);
1824 NDFREE(&nd, NDF_ONLY_PNBUF);
1826 if (vp == nd.ni_dvp)
1834 * Reposition read/write file offset.
1836 #ifndef _SYS_SYSPROTO_H_
1845 sys_lseek(struct thread *td, struct lseek_args *uap)
1848 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1852 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1855 cap_rights_t rights;
1859 error = fget(td, fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1862 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1863 fo_seek(fp, offset, whence, td) : ESPIPE;
1868 #if defined(COMPAT_43)
1870 * Reposition read/write file offset.
1872 #ifndef _SYS_SYSPROTO_H_
1873 struct olseek_args {
1880 olseek(struct thread *td, struct olseek_args *uap)
1883 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1885 #endif /* COMPAT_43 */
1887 #if defined(COMPAT_FREEBSD6)
1888 /* Version with the 'pad' argument */
1890 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1893 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1898 * Check access permissions using passed credentials.
1901 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
1907 /* Flags == 0 means only check for existence. */
1908 if (user_flags == 0)
1912 if (user_flags & R_OK)
1914 if (user_flags & W_OK)
1916 if (user_flags & X_OK)
1919 error = mac_vnode_check_access(cred, vp, accmode);
1923 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1924 error = VOP_ACCESS(vp, accmode, cred, td);
1929 * Check access permissions using "real" credentials.
1931 #ifndef _SYS_SYSPROTO_H_
1932 struct access_args {
1938 sys_access(struct thread *td, struct access_args *uap)
1941 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1945 #ifndef _SYS_SYSPROTO_H_
1946 struct faccessat_args {
1954 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1957 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1962 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1963 int flag, int amode)
1965 struct ucred *cred, *usecred;
1967 struct nameidata nd;
1968 cap_rights_t rights;
1971 if (flag & ~AT_EACCESS)
1973 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1977 * Create and modify a temporary credential instead of one that
1978 * is potentially shared (if we need one).
1980 cred = td->td_ucred;
1981 if ((flag & AT_EACCESS) == 0 &&
1982 ((cred->cr_uid != cred->cr_ruid ||
1983 cred->cr_rgid != cred->cr_groups[0]))) {
1984 usecred = crdup(cred);
1985 usecred->cr_uid = cred->cr_ruid;
1986 usecred->cr_groups[0] = cred->cr_rgid;
1987 td->td_ucred = usecred;
1990 AUDIT_ARG_VALUE(amode);
1991 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1992 AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1994 if ((error = namei(&nd)) != 0)
1998 error = vn_access(vp, amode, usecred, td);
1999 NDFREE(&nd, NDF_ONLY_PNBUF);
2002 if (usecred != cred) {
2003 td->td_ucred = cred;
2010 * Check access permissions using "effective" credentials.
2012 #ifndef _SYS_SYSPROTO_H_
2013 struct eaccess_args {
2019 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2022 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2023 AT_EACCESS, uap->amode));
2026 #if defined(COMPAT_43)
2028 * Get file status; this version follows links.
2030 #ifndef _SYS_SYSPROTO_H_
2037 ostat(struct thread *td, struct ostat_args *uap)
2043 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2048 return (copyout(&osb, uap->ub, sizeof (osb)));
2052 * Get file status; this version does not follow links.
2054 #ifndef _SYS_SYSPROTO_H_
2055 struct olstat_args {
2061 olstat(struct thread *td, struct olstat_args *uap)
2067 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2068 UIO_USERSPACE, &sb, NULL);
2072 return (copyout(&osb, uap->ub, sizeof (osb)));
2076 * Convert from an old to a new stat structure.
2079 cvtstat(struct stat *st, struct ostat *ost)
2082 bzero(ost, sizeof(*ost));
2083 ost->st_dev = st->st_dev;
2084 ost->st_ino = st->st_ino;
2085 ost->st_mode = st->st_mode;
2086 ost->st_nlink = st->st_nlink;
2087 ost->st_uid = st->st_uid;
2088 ost->st_gid = st->st_gid;
2089 ost->st_rdev = st->st_rdev;
2090 if (st->st_size < (quad_t)1 << 32)
2091 ost->st_size = st->st_size;
2094 ost->st_atim = st->st_atim;
2095 ost->st_mtim = st->st_mtim;
2096 ost->st_ctim = st->st_ctim;
2097 ost->st_blksize = st->st_blksize;
2098 ost->st_blocks = st->st_blocks;
2099 ost->st_flags = st->st_flags;
2100 ost->st_gen = st->st_gen;
2102 #endif /* COMPAT_43 */
2104 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2105 int ino64_trunc_error;
2106 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2107 &ino64_trunc_error, 0,
2108 "Error on truncation of inode number, device id or link count");
2110 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2113 ost->st_dev = st->st_dev;
2114 ost->st_ino = st->st_ino;
2115 if (ost->st_ino != st->st_ino) {
2116 switch (ino64_trunc_error) {
2123 ost->st_ino = UINT32_MAX;
2127 ost->st_mode = st->st_mode;
2128 ost->st_nlink = st->st_nlink;
2129 if (ost->st_nlink != st->st_nlink) {
2130 switch (ino64_trunc_error) {
2137 ost->st_nlink = UINT16_MAX;
2141 ost->st_uid = st->st_uid;
2142 ost->st_gid = st->st_gid;
2143 ost->st_rdev = st->st_rdev;
2144 ost->st_atim = st->st_atim;
2145 ost->st_mtim = st->st_mtim;
2146 ost->st_ctim = st->st_ctim;
2147 ost->st_size = st->st_size;
2148 ost->st_blocks = st->st_blocks;
2149 ost->st_blksize = st->st_blksize;
2150 ost->st_flags = st->st_flags;
2151 ost->st_gen = st->st_gen;
2153 ost->st_birthtim = st->st_birthtim;
2154 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2155 sizeof(*ost) - offsetof(struct freebsd11_stat,
2156 st_birthtim) - sizeof(ost->st_birthtim));
2161 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2164 struct freebsd11_stat osb;
2167 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2171 error = freebsd11_cvtstat(&sb, &osb);
2173 error = copyout(&osb, uap->ub, sizeof(osb));
2178 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2181 struct freebsd11_stat osb;
2184 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2185 UIO_USERSPACE, &sb, NULL);
2188 error = freebsd11_cvtstat(&sb, &osb);
2190 error = copyout(&osb, uap->ub, sizeof(osb));
2195 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2199 struct freebsd11_stat osb;
2202 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2205 error = kern_fhstat(td, fh, &sb);
2208 error = freebsd11_cvtstat(&sb, &osb);
2210 error = copyout(&osb, uap->sb, sizeof(osb));
2215 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2218 struct freebsd11_stat osb;
2221 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2222 UIO_USERSPACE, &sb, NULL);
2225 error = freebsd11_cvtstat(&sb, &osb);
2227 error = copyout(&osb, uap->buf, sizeof(osb));
2230 #endif /* COMPAT_FREEBSD11 */
2235 #ifndef _SYS_SYSPROTO_H_
2236 struct fstatat_args {
2244 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2249 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2250 UIO_USERSPACE, &sb, NULL);
2252 error = copyout(&sb, uap->buf, sizeof (sb));
2257 kern_statat(struct thread *td, int flag, int fd, char *path,
2258 enum uio_seg pathseg, struct stat *sbp,
2259 void (*hook)(struct vnode *vp, struct stat *sbp))
2261 struct nameidata nd;
2263 cap_rights_t rights;
2266 if (flag & ~AT_SYMLINK_NOFOLLOW)
2269 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2270 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2271 cap_rights_init(&rights, CAP_FSTAT), td);
2273 if ((error = namei(&nd)) != 0)
2275 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2277 SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2278 if (S_ISREG(sb.st_mode))
2279 SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2280 if (__predict_false(hook != NULL))
2281 hook(nd.ni_vp, &sb);
2283 NDFREE(&nd, NDF_ONLY_PNBUF);
2287 #ifdef __STAT_TIME_T_EXT
2295 if (KTRPOINT(td, KTR_STRUCT))
2301 #if defined(COMPAT_FREEBSD11)
2303 * Implementation of the NetBSD [l]stat() functions.
2306 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2309 bzero(nsb, sizeof(*nsb));
2310 nsb->st_dev = sb->st_dev;
2311 nsb->st_ino = sb->st_ino;
2312 nsb->st_mode = sb->st_mode;
2313 nsb->st_nlink = sb->st_nlink;
2314 nsb->st_uid = sb->st_uid;
2315 nsb->st_gid = sb->st_gid;
2316 nsb->st_rdev = sb->st_rdev;
2317 nsb->st_atim = sb->st_atim;
2318 nsb->st_mtim = sb->st_mtim;
2319 nsb->st_ctim = sb->st_ctim;
2320 nsb->st_size = sb->st_size;
2321 nsb->st_blocks = sb->st_blocks;
2322 nsb->st_blksize = sb->st_blksize;
2323 nsb->st_flags = sb->st_flags;
2324 nsb->st_gen = sb->st_gen;
2325 nsb->st_birthtim = sb->st_birthtim;
2328 #ifndef _SYS_SYSPROTO_H_
2329 struct freebsd11_nstat_args {
2335 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2341 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2345 freebsd11_cvtnstat(&sb, &nsb);
2346 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2350 * NetBSD lstat. Get file status; this version does not follow links.
2352 #ifndef _SYS_SYSPROTO_H_
2353 struct freebsd11_nlstat_args {
2359 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2365 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2366 UIO_USERSPACE, &sb, NULL);
2369 freebsd11_cvtnstat(&sb, &nsb);
2370 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2372 #endif /* COMPAT_FREEBSD11 */
2375 * Get configurable pathname variables.
2377 #ifndef _SYS_SYSPROTO_H_
2378 struct pathconf_args {
2384 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2389 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2392 td->td_retval[0] = value;
2396 #ifndef _SYS_SYSPROTO_H_
2397 struct lpathconf_args {
2403 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2408 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2411 td->td_retval[0] = value;
2416 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2417 u_long flags, long *valuep)
2419 struct nameidata nd;
2422 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2424 if ((error = namei(&nd)) != 0)
2426 NDFREE(&nd, NDF_ONLY_PNBUF);
2428 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2434 * Return target name of a symbolic link.
2436 #ifndef _SYS_SYSPROTO_H_
2437 struct readlink_args {
2444 sys_readlink(struct thread *td, struct readlink_args *uap)
2447 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2448 uap->buf, UIO_USERSPACE, uap->count));
2450 #ifndef _SYS_SYSPROTO_H_
2451 struct readlinkat_args {
2459 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2462 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2463 uap->buf, UIO_USERSPACE, uap->bufsize));
2467 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2468 char *buf, enum uio_seg bufseg, size_t count)
2473 struct nameidata nd;
2476 if (count > IOSIZE_MAX)
2479 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2480 pathseg, path, fd, td);
2482 if ((error = namei(&nd)) != 0)
2484 NDFREE(&nd, NDF_ONLY_PNBUF);
2487 error = mac_vnode_check_readlink(td->td_ucred, vp);
2493 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2496 aiov.iov_base = buf;
2497 aiov.iov_len = count;
2498 auio.uio_iov = &aiov;
2499 auio.uio_iovcnt = 1;
2500 auio.uio_offset = 0;
2501 auio.uio_rw = UIO_READ;
2502 auio.uio_segflg = bufseg;
2504 auio.uio_resid = count;
2505 error = VOP_READLINK(vp, &auio, td->td_ucred);
2506 td->td_retval[0] = count - auio.uio_resid;
2513 * Common implementation code for chflags() and fchflags().
2516 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2522 /* We can't support the value matching VNOVAL. */
2523 if (flags == VNOVAL)
2524 return (EOPNOTSUPP);
2527 * Prevent non-root users from setting flags on devices. When
2528 * a device is reused, users can retain ownership of the device
2529 * if they are allowed to set flags and programs assume that
2530 * chown can't fail when done as root.
2532 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2533 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2538 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2541 vattr.va_flags = flags;
2542 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2544 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2547 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2549 vn_finished_write(mp);
2554 * Change flags of a file given a path name.
2556 #ifndef _SYS_SYSPROTO_H_
2557 struct chflags_args {
2563 sys_chflags(struct thread *td, struct chflags_args *uap)
2566 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2570 #ifndef _SYS_SYSPROTO_H_
2571 struct chflagsat_args {
2579 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2582 const char *path = uap->path;
2583 u_long flags = uap->flags;
2584 int atflag = uap->atflag;
2586 if (atflag & ~AT_SYMLINK_NOFOLLOW)
2589 return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2593 * Same as chflags() but doesn't follow symlinks.
2595 #ifndef _SYS_SYSPROTO_H_
2596 struct lchflags_args {
2602 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2605 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2606 uap->flags, AT_SYMLINK_NOFOLLOW));
2610 kern_chflagsat(struct thread *td, int fd, const char *path,
2611 enum uio_seg pathseg, u_long flags, int atflag)
2613 struct nameidata nd;
2614 cap_rights_t rights;
2617 AUDIT_ARG_FFLAGS(flags);
2618 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2619 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2620 cap_rights_init(&rights, CAP_FCHFLAGS), td);
2621 if ((error = namei(&nd)) != 0)
2623 NDFREE(&nd, NDF_ONLY_PNBUF);
2624 error = setfflags(td, nd.ni_vp, flags);
2630 * Change flags of a file given a file descriptor.
2632 #ifndef _SYS_SYSPROTO_H_
2633 struct fchflags_args {
2639 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2642 cap_rights_t rights;
2645 AUDIT_ARG_FD(uap->fd);
2646 AUDIT_ARG_FFLAGS(uap->flags);
2647 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2652 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2653 AUDIT_ARG_VNODE1(fp->f_vnode);
2654 VOP_UNLOCK(fp->f_vnode, 0);
2656 error = setfflags(td, fp->f_vnode, uap->flags);
2662 * Common implementation code for chmod(), lchmod() and fchmod().
2665 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2671 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2673 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2675 vattr.va_mode = mode & ALLPERMS;
2677 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2680 error = VOP_SETATTR(vp, &vattr, cred);
2682 vn_finished_write(mp);
2687 * Change mode of a file given path name.
2689 #ifndef _SYS_SYSPROTO_H_
2696 sys_chmod(struct thread *td, struct chmod_args *uap)
2699 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2703 #ifndef _SYS_SYSPROTO_H_
2704 struct fchmodat_args {
2712 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2714 int flag = uap->flag;
2716 char *path = uap->path;
2717 mode_t mode = uap->mode;
2719 if (flag & ~AT_SYMLINK_NOFOLLOW)
2722 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2726 * Change mode of a file given path name (don't follow links.)
2728 #ifndef _SYS_SYSPROTO_H_
2729 struct lchmod_args {
2735 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2738 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2739 uap->mode, AT_SYMLINK_NOFOLLOW));
2743 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2744 mode_t mode, int flag)
2746 struct nameidata nd;
2747 cap_rights_t rights;
2750 AUDIT_ARG_MODE(mode);
2751 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2752 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2753 cap_rights_init(&rights, CAP_FCHMOD), td);
2754 if ((error = namei(&nd)) != 0)
2756 NDFREE(&nd, NDF_ONLY_PNBUF);
2757 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2763 * Change mode of a file given a file descriptor.
2765 #ifndef _SYS_SYSPROTO_H_
2766 struct fchmod_args {
2772 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2775 cap_rights_t rights;
2778 AUDIT_ARG_FD(uap->fd);
2779 AUDIT_ARG_MODE(uap->mode);
2781 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2784 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2790 * Common implementation for chown(), lchown(), and fchown()
2793 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2800 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2802 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2807 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2811 error = VOP_SETATTR(vp, &vattr, cred);
2813 vn_finished_write(mp);
2818 * Set ownership given a path name.
2820 #ifndef _SYS_SYSPROTO_H_
2828 sys_chown(struct thread *td, struct chown_args *uap)
2831 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2835 #ifndef _SYS_SYSPROTO_H_
2836 struct fchownat_args {
2845 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2850 if (flag & ~AT_SYMLINK_NOFOLLOW)
2853 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2854 uap->gid, uap->flag));
2858 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2859 int uid, int gid, int flag)
2861 struct nameidata nd;
2862 cap_rights_t rights;
2865 AUDIT_ARG_OWNER(uid, gid);
2866 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2867 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2868 cap_rights_init(&rights, CAP_FCHOWN), td);
2870 if ((error = namei(&nd)) != 0)
2872 NDFREE(&nd, NDF_ONLY_PNBUF);
2873 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2879 * Set ownership given a path name, do not cross symlinks.
2881 #ifndef _SYS_SYSPROTO_H_
2882 struct lchown_args {
2889 sys_lchown(struct thread *td, struct lchown_args *uap)
2892 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2893 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2897 * Set ownership given a file descriptor.
2899 #ifndef _SYS_SYSPROTO_H_
2900 struct fchown_args {
2907 sys_fchown(struct thread *td, struct fchown_args *uap)
2910 cap_rights_t rights;
2913 AUDIT_ARG_FD(uap->fd);
2914 AUDIT_ARG_OWNER(uap->uid, uap->gid);
2915 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2918 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2924 * Common implementation code for utimes(), lutimes(), and futimes().
2927 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
2928 struct timespec *tsp)
2930 struct timeval tv[2];
2931 const struct timeval *tvp;
2934 if (usrtvp == NULL) {
2935 vfs_timestamp(&tsp[0]);
2938 if (tvpseg == UIO_SYSSPACE) {
2941 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2946 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2947 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2949 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2950 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2956 * Common implementation code for futimens(), utimensat().
2958 #define UTIMENS_NULL 0x1
2959 #define UTIMENS_EXIT 0x2
2961 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2962 struct timespec *tsp, int *retflags)
2964 struct timespec tsnow;
2967 vfs_timestamp(&tsnow);
2969 if (usrtsp == NULL) {
2972 *retflags |= UTIMENS_NULL;
2975 if (tspseg == UIO_SYSSPACE) {
2978 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2980 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2981 *retflags |= UTIMENS_EXIT;
2982 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2983 *retflags |= UTIMENS_NULL;
2984 if (tsp[0].tv_nsec == UTIME_OMIT)
2985 tsp[0].tv_sec = VNOVAL;
2986 else if (tsp[0].tv_nsec == UTIME_NOW)
2988 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2990 if (tsp[1].tv_nsec == UTIME_OMIT)
2991 tsp[1].tv_sec = VNOVAL;
2992 else if (tsp[1].tv_nsec == UTIME_NOW)
2994 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3001 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3005 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3006 int numtimes, int nullflag)
3010 int error, setbirthtime;
3012 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3014 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3016 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3017 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3020 vattr.va_atime = ts[0];
3021 vattr.va_mtime = ts[1];
3023 vattr.va_birthtime = ts[1];
3025 vattr.va_birthtime = ts[2];
3027 vattr.va_vaflags |= VA_UTIMES_NULL;
3029 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3033 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3035 vn_finished_write(mp);
3040 * Set the access and modification times of a file.
3042 #ifndef _SYS_SYSPROTO_H_
3043 struct utimes_args {
3045 struct timeval *tptr;
3049 sys_utimes(struct thread *td, struct utimes_args *uap)
3052 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3053 uap->tptr, UIO_USERSPACE));
3056 #ifndef _SYS_SYSPROTO_H_
3057 struct futimesat_args {
3060 const struct timeval * times;
3064 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3067 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3068 uap->times, UIO_USERSPACE));
3072 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3073 struct timeval *tptr, enum uio_seg tptrseg)
3075 struct nameidata nd;
3076 struct timespec ts[2];
3077 cap_rights_t rights;
3080 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3082 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3083 cap_rights_init(&rights, CAP_FUTIMES), td);
3085 if ((error = namei(&nd)) != 0)
3087 NDFREE(&nd, NDF_ONLY_PNBUF);
3088 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3094 * Set the access and modification times of a file.
3096 #ifndef _SYS_SYSPROTO_H_
3097 struct lutimes_args {
3099 struct timeval *tptr;
3103 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3106 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3111 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3112 struct timeval *tptr, enum uio_seg tptrseg)
3114 struct timespec ts[2];
3115 struct nameidata nd;
3118 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3120 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3121 if ((error = namei(&nd)) != 0)
3123 NDFREE(&nd, NDF_ONLY_PNBUF);
3124 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3130 * Set the access and modification times of a file.
3132 #ifndef _SYS_SYSPROTO_H_
3133 struct futimes_args {
3135 struct timeval *tptr;
3139 sys_futimes(struct thread *td, struct futimes_args *uap)
3142 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3146 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3147 enum uio_seg tptrseg)
3149 struct timespec ts[2];
3151 cap_rights_t rights;
3155 error = getutimes(tptr, tptrseg, ts);
3158 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3162 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3163 AUDIT_ARG_VNODE1(fp->f_vnode);
3164 VOP_UNLOCK(fp->f_vnode, 0);
3166 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3172 sys_futimens(struct thread *td, struct futimens_args *uap)
3175 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3179 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3180 enum uio_seg tptrseg)
3182 struct timespec ts[2];
3184 cap_rights_t rights;
3188 error = getutimens(tptr, tptrseg, ts, &flags);
3191 if (flags & UTIMENS_EXIT)
3193 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3197 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3198 AUDIT_ARG_VNODE1(fp->f_vnode);
3199 VOP_UNLOCK(fp->f_vnode, 0);
3201 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3207 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3210 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3211 uap->times, UIO_USERSPACE, uap->flag));
3215 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3216 struct timespec *tptr, enum uio_seg tptrseg, int flag)
3218 struct nameidata nd;
3219 struct timespec ts[2];
3220 cap_rights_t rights;
3223 if (flag & ~AT_SYMLINK_NOFOLLOW)
3226 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3228 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3229 FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3230 cap_rights_init(&rights, CAP_FUTIMES), td);
3231 if ((error = namei(&nd)) != 0)
3234 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3236 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3237 * "Search permission is denied by a component of the path prefix."
3239 NDFREE(&nd, NDF_ONLY_PNBUF);
3240 if ((flags & UTIMENS_EXIT) == 0)
3241 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3247 * Truncate a file given its path name.
3249 #ifndef _SYS_SYSPROTO_H_
3250 struct truncate_args {
3257 sys_truncate(struct thread *td, struct truncate_args *uap)
3260 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3264 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3270 struct nameidata nd;
3275 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3276 if ((error = namei(&nd)) != 0)
3279 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3280 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3281 vn_rangelock_unlock(vp, rl_cookie);
3285 NDFREE(&nd, NDF_ONLY_PNBUF);
3286 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3287 if (vp->v_type == VDIR)
3290 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3293 else if ((error = vn_writechk(vp)) == 0 &&
3294 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3296 vattr.va_size = length;
3297 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3300 vn_finished_write(mp);
3301 vn_rangelock_unlock(vp, rl_cookie);
3306 #if defined(COMPAT_43)
3308 * Truncate a file given its path name.
3310 #ifndef _SYS_SYSPROTO_H_
3311 struct otruncate_args {
3317 otruncate(struct thread *td, struct otruncate_args *uap)
3320 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3322 #endif /* COMPAT_43 */
3324 #if defined(COMPAT_FREEBSD6)
3325 /* Versions with the pad argument */
3327 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3330 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3334 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3337 return (kern_ftruncate(td, uap->fd, uap->length));
3342 kern_fsync(struct thread *td, int fd, bool fullsync)
3347 cap_rights_t rights;
3348 int error, lock_flags;
3351 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3357 /* XXXKIB: compete outstanding aio writes */;
3359 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3362 if (MNT_SHARED_WRITES(mp) ||
3363 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3364 lock_flags = LK_SHARED;
3366 lock_flags = LK_EXCLUSIVE;
3368 vn_lock(vp, lock_flags | LK_RETRY);
3369 AUDIT_ARG_VNODE1(vp);
3370 if (vp->v_object != NULL) {
3371 VM_OBJECT_WLOCK(vp->v_object);
3372 vm_object_page_clean(vp->v_object, 0, 0, 0);
3373 VM_OBJECT_WUNLOCK(vp->v_object);
3375 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3377 vn_finished_write(mp);
3384 * Sync an open file.
3386 #ifndef _SYS_SYSPROTO_H_
3392 sys_fsync(struct thread *td, struct fsync_args *uap)
3395 return (kern_fsync(td, uap->fd, true));
3399 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3402 return (kern_fsync(td, uap->fd, false));
3406 * Rename files. Source and destination must either both be directories, or
3407 * both not be directories. If target is a directory, it must be empty.
3409 #ifndef _SYS_SYSPROTO_H_
3410 struct rename_args {
3416 sys_rename(struct thread *td, struct rename_args *uap)
3419 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3420 uap->to, UIO_USERSPACE));
3423 #ifndef _SYS_SYSPROTO_H_
3424 struct renameat_args {
3432 sys_renameat(struct thread *td, struct renameat_args *uap)
3435 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3440 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3441 enum uio_seg pathseg)
3443 struct mount *mp = NULL;
3444 struct vnode *tvp, *fvp, *tdvp;
3445 struct nameidata fromnd, tond;
3446 cap_rights_t rights;
3452 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3453 AUDITVNODE1, pathseg, old, oldfd,
3454 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3456 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3457 pathseg, old, oldfd,
3458 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3461 if ((error = namei(&fromnd)) != 0)
3464 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3465 fromnd.ni_vp, &fromnd.ni_cnd);
3466 VOP_UNLOCK(fromnd.ni_dvp, 0);
3467 if (fromnd.ni_dvp != fromnd.ni_vp)
3468 VOP_UNLOCK(fromnd.ni_vp, 0);
3471 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3472 SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3473 cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3474 if (fromnd.ni_vp->v_type == VDIR)
3475 tond.ni_cnd.cn_flags |= WILLBEDIR;
3476 if ((error = namei(&tond)) != 0) {
3477 /* Translate error code for rename("dir1", "dir2/."). */
3478 if (error == EISDIR && fvp->v_type == VDIR)
3480 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3481 vrele(fromnd.ni_dvp);
3487 error = vn_start_write(fvp, &mp, V_NOWAIT);
3489 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3490 NDFREE(&tond, NDF_ONLY_PNBUF);
3497 vrele(fromnd.ni_dvp);
3499 vrele(tond.ni_startdir);
3500 if (fromnd.ni_startdir != NULL)
3501 vrele(fromnd.ni_startdir);
3502 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3508 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3511 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3516 if (newfd != AT_FDCWD) {
3518 * If the target already exists we require CAP_UNLINKAT
3521 error = cap_check(&tond.ni_filecaps.fc_rights,
3522 cap_rights_init(&rights, CAP_UNLINKAT));
3533 * If the source is the same as the destination (that is, if they
3534 * are links to the same vnode), then there is nothing to do.
3540 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3541 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3545 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3546 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3547 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3548 NDFREE(&tond, NDF_ONLY_PNBUF);
3550 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3551 NDFREE(&tond, NDF_ONLY_PNBUF);
3558 vrele(fromnd.ni_dvp);
3561 vrele(tond.ni_startdir);
3562 vn_finished_write(mp);
3564 if (fromnd.ni_startdir)
3565 vrele(fromnd.ni_startdir);
3572 * Make a directory file.
3574 #ifndef _SYS_SYSPROTO_H_
3581 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3584 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3588 #ifndef _SYS_SYSPROTO_H_
3589 struct mkdirat_args {
3596 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3599 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3603 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3609 struct nameidata nd;
3610 cap_rights_t rights;
3613 AUDIT_ARG_MODE(mode);
3616 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3617 NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3619 nd.ni_cnd.cn_flags |= WILLBEDIR;
3620 if ((error = namei(&nd)) != 0)
3624 NDFREE(&nd, NDF_ONLY_PNBUF);
3626 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3627 * the strange behaviour of leaving the vnode unlocked
3628 * if the target is the same vnode as the parent.
3630 if (vp == nd.ni_dvp)
3637 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3638 NDFREE(&nd, NDF_ONLY_PNBUF);
3640 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3645 vattr.va_type = VDIR;
3646 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3648 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3653 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3657 NDFREE(&nd, NDF_ONLY_PNBUF);
3661 vn_finished_write(mp);
3666 * Remove a directory file.
3668 #ifndef _SYS_SYSPROTO_H_
3674 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3677 return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3681 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3685 struct nameidata nd;
3686 cap_rights_t rights;
3691 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3692 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3693 if ((error = namei(&nd)) != 0)
3696 if (vp->v_type != VDIR) {
3701 * No rmdir "." please.
3703 if (nd.ni_dvp == vp) {
3708 * The root of a mounted filesystem cannot be deleted.
3710 if (vp->v_vflag & VV_ROOT) {
3715 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3720 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3721 NDFREE(&nd, NDF_ONLY_PNBUF);
3723 if (nd.ni_dvp == vp)
3727 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3731 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3732 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3733 vn_finished_write(mp);
3735 NDFREE(&nd, NDF_ONLY_PNBUF);
3737 if (nd.ni_dvp == vp)
3744 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3746 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3747 long *basep, void (*func)(struct freebsd11_dirent *))
3749 struct freebsd11_dirent dstdp;
3750 struct dirent *dp, *edp;
3753 ssize_t resid, ucount;
3756 /* XXX arbitrary sanity limit on `count'. */
3757 count = min(count, 64 * 1024);
3759 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3761 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3769 for (dp = (struct dirent *)dirbuf,
3770 edp = (struct dirent *)&dirbuf[count - resid];
3771 ucount < count && dp < edp; ) {
3772 if (dp->d_reclen == 0)
3774 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3775 if (dp->d_namlen >= sizeof(dstdp.d_name))
3777 dstdp.d_type = dp->d_type;
3778 dstdp.d_namlen = dp->d_namlen;
3779 dstdp.d_fileno = dp->d_fileno; /* truncate */
3780 if (dstdp.d_fileno != dp->d_fileno) {
3781 switch (ino64_trunc_error) {
3789 dstdp.d_fileno = UINT32_MAX;
3793 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3794 ((dp->d_namlen + 1 + 3) &~ 3);
3795 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3796 bzero(dstdp.d_name + dstdp.d_namlen,
3797 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3799 MPASS(dstdp.d_reclen <= dp->d_reclen);
3800 MPASS(ucount + dstdp.d_reclen <= count);
3803 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3806 dp = (struct dirent *)((char *)dp + dp->d_reclen);
3807 ucount += dstdp.d_reclen;
3811 free(dirbuf, M_TEMP);
3813 td->td_retval[0] = ucount;
3820 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3822 #if (BYTE_ORDER == LITTLE_ENDIAN)
3824 * The expected low byte of dp->d_namlen is our dp->d_type.
3825 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3827 dp->d_type = dp->d_namlen;
3831 * The dp->d_type is the high byte of the expected dp->d_namlen,
3832 * so must be zero'ed.
3839 * Read a block of directory entries in a filesystem independent format.
3841 #ifndef _SYS_SYSPROTO_H_
3842 struct ogetdirentries_args {
3850 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3855 error = kern_ogetdirentries(td, uap, &loff);
3857 error = copyout(&loff, uap->basep, sizeof(long));
3862 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3868 /* XXX arbitrary sanity limit on `count'. */
3869 if (uap->count > 64 * 1024)
3872 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3873 &base, ogetdirentries_cvt);
3875 if (error == 0 && uap->basep != NULL)
3876 error = copyout(&base, uap->basep, sizeof(long));
3880 #endif /* COMPAT_43 */
3882 #if defined(COMPAT_FREEBSD11)
3883 #ifndef _SYS_SYSPROTO_H_
3884 struct freebsd11_getdirentries_args {
3892 freebsd11_getdirentries(struct thread *td,
3893 struct freebsd11_getdirentries_args *uap)
3898 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3901 if (error == 0 && uap->basep != NULL)
3902 error = copyout(&base, uap->basep, sizeof(long));
3907 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
3909 struct freebsd11_getdirentries_args ap;
3913 ap.count = uap->count;
3915 return (freebsd11_getdirentries(td, &ap));
3917 #endif /* COMPAT_FREEBSD11 */
3920 * Read a block of directory entries in a filesystem independent format.
3923 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
3928 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3929 NULL, UIO_USERSPACE);
3932 if (uap->basep != NULL)
3933 error = copyout(&base, uap->basep, sizeof(off_t));
3938 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
3939 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
3945 cap_rights_t rights;
3951 if (count > IOSIZE_MAX)
3953 auio.uio_resid = count;
3954 error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3957 if ((fp->f_flag & FREAD) == 0) {
3962 foffset = foffset_lock(fp, 0);
3964 if (vp->v_type != VDIR) {
3968 aiov.iov_base = buf;
3969 aiov.iov_len = count;
3970 auio.uio_iov = &aiov;
3971 auio.uio_iovcnt = 1;
3972 auio.uio_rw = UIO_READ;
3973 auio.uio_segflg = bufseg;
3975 vn_lock(vp, LK_SHARED | LK_RETRY);
3976 AUDIT_ARG_VNODE1(vp);
3977 loff = auio.uio_offset = foffset;
3979 error = mac_vnode_check_readdir(td->td_ucred, vp);
3982 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3984 foffset = auio.uio_offset;
3989 if (count == auio.uio_resid &&
3990 (vp->v_vflag & VV_ROOT) &&
3991 (vp->v_mount->mnt_flag & MNT_UNION)) {
3992 struct vnode *tvp = vp;
3994 vp = vp->v_mount->mnt_vnodecovered;
4005 *residp = auio.uio_resid;
4006 td->td_retval[0] = count - auio.uio_resid;
4008 foffset_unlock(fp, foffset, 0);
4014 * Set the mode mask for creation of filesystem nodes.
4016 #ifndef _SYS_SYSPROTO_H_
4022 sys_umask(struct thread *td, struct umask_args *uap)
4024 struct filedesc *fdp;
4026 fdp = td->td_proc->p_fd;
4027 FILEDESC_XLOCK(fdp);
4028 td->td_retval[0] = fdp->fd_cmask;
4029 fdp->fd_cmask = uap->newmask & ALLPERMS;
4030 FILEDESC_XUNLOCK(fdp);
4035 * Void all references to file by ripping underlying filesystem away from
4038 #ifndef _SYS_SYSPROTO_H_
4039 struct revoke_args {
4044 sys_revoke(struct thread *td, struct revoke_args *uap)
4048 struct nameidata nd;
4051 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4053 if ((error = namei(&nd)) != 0)
4056 NDFREE(&nd, NDF_ONLY_PNBUF);
4057 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4062 error = mac_vnode_check_revoke(td->td_ucred, vp);
4066 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4069 if (td->td_ucred->cr_uid != vattr.va_uid) {
4070 error = priv_check(td, PRIV_VFS_ADMIN);
4075 VOP_REVOKE(vp, REVOKEALL);
4082 * Convert a user file descriptor to a kernel file entry and check that, if it
4083 * is a capability, the correct rights are present. A reference on the file
4084 * entry is held upon returning.
4087 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4092 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4097 * The file could be not of the vnode type, or it may be not
4098 * yet fully initialized, in which case the f_vnode pointer
4099 * may be set, but f_ops is still badfileops. E.g.,
4100 * devfs_open() transiently create such situation to
4101 * facilitate csw d_fdopen().
4103 * Dupfdopen() handling in kern_openat() installs the
4104 * half-baked file into the process descriptor table, allowing
4105 * other thread to dereference it. Guard against the race by
4108 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4118 * Get an (NFS) file handle.
4120 #ifndef _SYS_SYSPROTO_H_
4121 struct lgetfh_args {
4127 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4129 struct nameidata nd;
4134 error = priv_check(td, PRIV_VFS_GETFH);
4137 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4142 NDFREE(&nd, NDF_ONLY_PNBUF);
4144 bzero(&fh, sizeof(fh));
4145 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4146 error = VOP_VPTOFH(vp, &fh.fh_fid);
4149 error = copyout(&fh, uap->fhp, sizeof (fh));
4153 #ifndef _SYS_SYSPROTO_H_
4160 sys_getfh(struct thread *td, struct getfh_args *uap)
4162 struct nameidata nd;
4167 error = priv_check(td, PRIV_VFS_GETFH);
4170 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4175 NDFREE(&nd, NDF_ONLY_PNBUF);
4177 bzero(&fh, sizeof(fh));
4178 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4179 error = VOP_VPTOFH(vp, &fh.fh_fid);
4182 error = copyout(&fh, uap->fhp, sizeof (fh));
4187 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4190 * warning: do not remove the priv_check() call or this becomes one giant
4193 #ifndef _SYS_SYSPROTO_H_
4194 struct fhopen_args {
4195 const struct fhandle *u_fhp;
4200 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4209 error = priv_check(td, PRIV_VFS_FHOPEN);
4213 fmode = FFLAGS(uap->flags);
4214 /* why not allow a non-read/write open for our lockd? */
4215 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4217 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4220 /* find the mount point */
4221 mp = vfs_busyfs(&fhp.fh_fsid);
4224 /* now give me my vnode, it gets returned to me locked */
4225 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4230 error = falloc_noinstall(td, &fp);
4236 * An extra reference on `fp' has been held for us by
4237 * falloc_noinstall().
4243 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4245 KASSERT(fp->f_ops == &badfileops,
4246 ("VOP_OPEN in fhopen() set f_ops"));
4247 KASSERT(td->td_dupfd < 0,
4248 ("fhopen() encountered fdopen()"));
4258 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4261 if ((fmode & O_TRUNC) != 0) {
4262 error = fo_truncate(fp, 0, td->td_ucred, td);
4267 error = finstall(td, fp, &indx, fmode, NULL);
4270 td->td_retval[0] = indx;
4275 * Stat an (NFS) file handle.
4277 #ifndef _SYS_SYSPROTO_H_
4278 struct fhstat_args {
4279 struct fhandle *u_fhp;
4284 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4290 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4293 error = kern_fhstat(td, fh, &sb);
4295 error = copyout(&sb, uap->sb, sizeof(sb));
4300 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4306 error = priv_check(td, PRIV_VFS_FHSTAT);
4309 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4311 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4315 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4321 * Implement fstatfs() for (NFS) file handles.
4323 #ifndef _SYS_SYSPROTO_H_
4324 struct fhstatfs_args {
4325 struct fhandle *u_fhp;
4330 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4336 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4339 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4340 error = kern_fhstatfs(td, fh, sfp);
4342 error = copyout(sfp, uap->buf, sizeof(*sfp));
4343 free(sfp, M_STATFS);
4348 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4355 error = priv_check(td, PRIV_VFS_FHSTATFS);
4358 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4360 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4366 error = prison_canseemount(td->td_ucred, mp);
4370 error = mac_mount_check_stat(td->td_ucred, mp);
4375 * Set these in case the underlying filesystem fails to do so.
4378 sp->f_version = STATFS_VERSION;
4379 sp->f_namemax = NAME_MAX;
4380 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4381 error = VFS_STATFS(mp, sp);
4390 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4395 cap_rights_t rights;
4396 off_t olen, ooffset;
4399 int audited_vnode1 = 0;
4403 if (offset < 0 || len <= 0)
4405 /* Check for wrap. */
4406 if (offset > OFF_MAX - len)
4409 error = fget(td, fd, cap_rights_init(&rights, CAP_PWRITE), &fp);
4412 AUDIT_ARG_FILE(td->td_proc, fp);
4413 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4417 if ((fp->f_flag & FWRITE) == 0) {
4421 if (fp->f_type != DTYPE_VNODE) {
4426 if (vp->v_type != VREG) {
4431 /* Allocating blocks may take a long time, so iterate. */
4438 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4441 error = vn_lock(vp, LK_EXCLUSIVE);
4443 vn_finished_write(mp);
4447 if (!audited_vnode1) {
4448 AUDIT_ARG_VNODE1(vp);
4453 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4456 error = VOP_ALLOCATE(vp, &offset, &len);
4458 vn_finished_write(mp);
4460 if (olen + ooffset != offset + len) {
4461 panic("offset + len changed from %jx/%jx to %jx/%jx",
4462 ooffset, olen, offset, len);
4464 if (error != 0 || len == 0)
4466 KASSERT(olen > len, ("Iteration did not make progress?"));
4475 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4479 error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4480 return (kern_posix_error(td, error));
4484 * Unlike madvise(2), we do not make a best effort to remember every
4485 * possible caching hint. Instead, we remember the last setting with
4486 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4487 * region of any current setting.
4490 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4493 struct fadvise_info *fa, *new;
4496 cap_rights_t rights;
4500 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4502 AUDIT_ARG_VALUE(advice);
4504 case POSIX_FADV_SEQUENTIAL:
4505 case POSIX_FADV_RANDOM:
4506 case POSIX_FADV_NOREUSE:
4507 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4509 case POSIX_FADV_NORMAL:
4510 case POSIX_FADV_WILLNEED:
4511 case POSIX_FADV_DONTNEED:
4517 /* XXX: CAP_POSIX_FADVISE? */
4519 error = fget(td, fd, cap_rights_init(&rights), &fp);
4522 AUDIT_ARG_FILE(td->td_proc, fp);
4523 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4527 if (fp->f_type != DTYPE_VNODE) {
4532 if (vp->v_type != VREG) {
4539 end = offset + len - 1;
4541 case POSIX_FADV_SEQUENTIAL:
4542 case POSIX_FADV_RANDOM:
4543 case POSIX_FADV_NOREUSE:
4545 * Try to merge any existing non-standard region with
4546 * this new region if possible, otherwise create a new
4547 * non-standard region for this request.
4549 mtx_pool_lock(mtxpool_sleep, fp);
4551 if (fa != NULL && fa->fa_advice == advice &&
4552 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4553 (end != OFF_MAX && fa->fa_start == end + 1) ||
4554 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4555 if (offset < fa->fa_start)
4556 fa->fa_start = offset;
4557 if (end > fa->fa_end)
4560 new->fa_advice = advice;
4561 new->fa_start = offset;
4566 mtx_pool_unlock(mtxpool_sleep, fp);
4568 case POSIX_FADV_NORMAL:
4570 * If a the "normal" region overlaps with an existing
4571 * non-standard region, trim or remove the
4572 * non-standard region.
4574 mtx_pool_lock(mtxpool_sleep, fp);
4577 if (offset <= fa->fa_start && end >= fa->fa_end) {
4579 fp->f_advice = NULL;
4580 } else if (offset <= fa->fa_start &&
4581 end >= fa->fa_start)
4582 fa->fa_start = end + 1;
4583 else if (offset <= fa->fa_end && end >= fa->fa_end)
4584 fa->fa_end = offset - 1;
4585 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4587 * If the "normal" region is a middle
4588 * portion of the existing
4589 * non-standard region, just remove
4590 * the whole thing rather than picking
4591 * one side or the other to
4595 fp->f_advice = NULL;
4598 mtx_pool_unlock(mtxpool_sleep, fp);
4600 case POSIX_FADV_WILLNEED:
4601 case POSIX_FADV_DONTNEED:
4602 error = VOP_ADVISE(vp, offset, end, advice);
4608 free(new, M_FADVISE);
4613 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4617 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4619 return (kern_posix_error(td, error));