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 <fs/devfs/devfs.h>
92 #include <ufs/ufs/quota.h>
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97 enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101 struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103 const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107 enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
109 size_t count, struct thread *td);
110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
111 const char *path, enum uio_seg segflag);
114 at2cnpflags(u_int at_flags, u_int mask)
118 MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
119 (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
123 if ((at_flags & AT_BENEATH) != 0)
125 if ((at_flags & AT_RESOLVE_BENEATH) != 0)
127 if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
129 /* NOFOLLOW is pseudo flag */
130 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
131 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
138 kern_sync(struct thread *td)
140 struct mount *mp, *nmp;
143 mtx_lock(&mountlist_mtx);
144 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
145 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
146 nmp = TAILQ_NEXT(mp, mnt_list);
149 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
150 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
151 save = curthread_pflags_set(TDP_SYNCIO);
152 vfs_periodic(mp, MNT_NOWAIT);
153 VFS_SYNC(mp, MNT_NOWAIT);
154 curthread_pflags_restore(save);
155 vn_finished_write(mp);
157 mtx_lock(&mountlist_mtx);
158 nmp = TAILQ_NEXT(mp, mnt_list);
161 mtx_unlock(&mountlist_mtx);
166 * Sync each mounted filesystem.
168 #ifndef _SYS_SYSPROTO_H_
175 sys_sync(struct thread *td, struct sync_args *uap)
178 return (kern_sync(td));
182 * Change filesystem quotas.
184 #ifndef _SYS_SYSPROTO_H_
185 struct quotactl_args {
193 sys_quotactl(struct thread *td, struct quotactl_args *uap)
199 AUDIT_ARG_CMD(uap->cmd);
200 AUDIT_ARG_UID(uap->uid);
201 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
203 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
205 if ((error = namei(&nd)) != 0)
207 NDFREE(&nd, NDF_ONLY_PNBUF);
208 mp = nd.ni_vp->v_mount;
211 error = vfs_busy(mp, 0);
216 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
219 * Since quota on operation typically needs to open quota
220 * file, the Q_QUOTAON handler needs to unbusy the mount point
221 * before calling into namei. Otherwise, unmount might be
222 * started between two vfs_busy() invocations (first is our,
223 * second is from mount point cross-walk code in lookup()),
226 * Require that Q_QUOTAON handles the vfs_busy() reference on
227 * its own, always returning with ubusied mount point.
229 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
230 (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
237 * Used by statfs conversion routines to scale the block size up if
238 * necessary so that all of the block counts are <= 'max_size'. Note
239 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
243 statfs_scale_blocks(struct statfs *sf, long max_size)
248 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
251 * Attempt to scale the block counts to give a more accurate
252 * overview to userland of the ratio of free space to used
253 * space. To do this, find the largest block count and compute
254 * a divisor that lets it fit into a signed integer <= max_size.
256 if (sf->f_bavail < 0)
257 count = -sf->f_bavail;
259 count = sf->f_bavail;
260 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
261 if (count <= max_size)
264 count >>= flsl(max_size);
271 sf->f_bsize <<= shift;
272 sf->f_blocks >>= shift;
273 sf->f_bfree >>= shift;
274 sf->f_bavail >>= shift;
278 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
284 error = vfs_busy(mp, 0);
289 error = mac_mount_check_stat(td->td_ucred, mp);
293 error = VFS_STATFS(mp, buf);
296 if (priv_check_cred_vfs_generation(td->td_ucred)) {
297 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
298 prison_enforce_statfs(td->td_ucred, mp, buf);
306 * Get filesystem statistics.
308 #ifndef _SYS_SYSPROTO_H_
315 sys_statfs(struct thread *td, struct statfs_args *uap)
320 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
321 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
323 error = copyout(sfp, uap->buf, sizeof(struct statfs));
329 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
336 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
341 mp = nd.ni_vp->v_mount;
345 return (kern_do_statfs(td, mp, buf));
349 * Get filesystem statistics.
351 #ifndef _SYS_SYSPROTO_H_
352 struct fstatfs_args {
358 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
363 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
364 error = kern_fstatfs(td, uap->fd, sfp);
366 error = copyout(sfp, uap->buf, sizeof(struct statfs));
372 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
380 error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
384 vn_lock(vp, LK_SHARED | LK_RETRY);
386 AUDIT_ARG_VNODE1(vp);
393 return (kern_do_statfs(td, mp, buf));
397 * Get statistics on all filesystems.
399 #ifndef _SYS_SYSPROTO_H_
400 struct getfsstat_args {
407 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
412 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
414 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
415 UIO_USERSPACE, uap->mode);
417 td->td_retval[0] = count;
422 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
423 * The caller is responsible for freeing memory which will be allocated
427 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
428 size_t *countp, enum uio_seg bufseg, int mode)
430 struct mount *mp, *nmp;
431 struct statfs *sfsp, *sp, *sptmp, *tofree;
432 size_t count, maxcount;
440 if (bufseg == UIO_SYSSPACE)
445 maxcount = bufsize / sizeof(struct statfs);
449 } else if (bufseg == UIO_USERSPACE) {
452 } else /* if (bufseg == UIO_SYSSPACE) */ {
454 mtx_lock(&mountlist_mtx);
455 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
458 mtx_unlock(&mountlist_mtx);
459 if (maxcount > count)
461 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
468 * If there is no target buffer they only want the count.
470 * This could be TAILQ_FOREACH but it is open-coded to match the original
474 mtx_lock(&mountlist_mtx);
475 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
476 if (prison_canseemount(td->td_ucred, mp) != 0) {
477 nmp = TAILQ_NEXT(mp, mnt_list);
481 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
482 nmp = TAILQ_NEXT(mp, mnt_list);
487 nmp = TAILQ_NEXT(mp, mnt_list);
489 mtx_unlock(&mountlist_mtx);
495 * They want the entire thing.
497 * Short-circuit the corner case of no room for anything, avoids
504 mtx_lock(&mountlist_mtx);
505 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
506 if (prison_canseemount(td->td_ucred, mp) != 0) {
507 nmp = TAILQ_NEXT(mp, mnt_list);
511 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
512 nmp = TAILQ_NEXT(mp, mnt_list);
516 if (mode == MNT_WAIT) {
517 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
519 * If vfs_busy() failed, and MBF_NOWAIT
520 * wasn't passed, then the mp is gone.
521 * Furthermore, because of MBF_MNTLSTLOCK,
522 * the mountlist_mtx was dropped. We have
523 * no other choice than to start over.
525 mtx_unlock(&mountlist_mtx);
526 free(tofree, M_STATFS);
530 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
531 nmp = TAILQ_NEXT(mp, mnt_list);
537 * If MNT_NOWAIT is specified, do not refresh
540 if (mode != MNT_NOWAIT) {
541 error = VFS_STATFS(mp, sp);
543 mtx_lock(&mountlist_mtx);
544 nmp = TAILQ_NEXT(mp, mnt_list);
549 if (priv_check_cred_vfs_generation(td->td_ucred)) {
550 sptmp = malloc(sizeof(struct statfs), M_STATFS,
553 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
554 prison_enforce_statfs(td->td_ucred, mp, sptmp);
558 if (bufseg == UIO_SYSSPACE) {
559 bcopy(sp, sfsp, sizeof(*sp));
560 free(sptmp, M_STATFS);
561 } else /* if (bufseg == UIO_USERSPACE) */ {
562 error = copyout(sp, sfsp, sizeof(*sp));
563 free(sptmp, M_STATFS);
572 if (count == maxcount) {
577 mtx_lock(&mountlist_mtx);
578 nmp = TAILQ_NEXT(mp, mnt_list);
581 mtx_unlock(&mountlist_mtx);
587 #ifdef COMPAT_FREEBSD4
589 * Get old format filesystem statistics.
591 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
593 #ifndef _SYS_SYSPROTO_H_
594 struct freebsd4_statfs_args {
600 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
606 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
607 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
609 freebsd4_cvtstatfs(sfp, &osb);
610 error = copyout(&osb, uap->buf, sizeof(osb));
617 * Get filesystem statistics.
619 #ifndef _SYS_SYSPROTO_H_
620 struct freebsd4_fstatfs_args {
626 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
632 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
633 error = kern_fstatfs(td, uap->fd, sfp);
635 freebsd4_cvtstatfs(sfp, &osb);
636 error = copyout(&osb, uap->buf, sizeof(osb));
643 * Get statistics on all filesystems.
645 #ifndef _SYS_SYSPROTO_H_
646 struct freebsd4_getfsstat_args {
653 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
655 struct statfs *buf, *sp;
660 if (uap->bufsize < 0)
662 count = uap->bufsize / sizeof(struct ostatfs);
663 if (count > SIZE_MAX / sizeof(struct statfs))
665 size = count * sizeof(struct statfs);
666 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
669 td->td_retval[0] = count;
672 while (count != 0 && error == 0) {
673 freebsd4_cvtstatfs(sp, &osb);
674 error = copyout(&osb, uap->buf, sizeof(osb));
685 * Implement fstatfs() for (NFS) file handles.
687 #ifndef _SYS_SYSPROTO_H_
688 struct freebsd4_fhstatfs_args {
689 struct fhandle *u_fhp;
694 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
701 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
704 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
705 error = kern_fhstatfs(td, fh, sfp);
707 freebsd4_cvtstatfs(sfp, &osb);
708 error = copyout(&osb, uap->buf, sizeof(osb));
715 * Convert a new format statfs structure to an old format statfs structure.
718 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
721 statfs_scale_blocks(nsp, LONG_MAX);
722 bzero(osp, sizeof(*osp));
723 osp->f_bsize = nsp->f_bsize;
724 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
725 osp->f_blocks = nsp->f_blocks;
726 osp->f_bfree = nsp->f_bfree;
727 osp->f_bavail = nsp->f_bavail;
728 osp->f_files = MIN(nsp->f_files, LONG_MAX);
729 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
730 osp->f_owner = nsp->f_owner;
731 osp->f_type = nsp->f_type;
732 osp->f_flags = nsp->f_flags;
733 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
734 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
735 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
736 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
737 strlcpy(osp->f_fstypename, nsp->f_fstypename,
738 MIN(MFSNAMELEN, OMFSNAMELEN));
739 strlcpy(osp->f_mntonname, nsp->f_mntonname,
740 MIN(MNAMELEN, OMNAMELEN));
741 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
742 MIN(MNAMELEN, OMNAMELEN));
743 osp->f_fsid = nsp->f_fsid;
745 #endif /* COMPAT_FREEBSD4 */
747 #if defined(COMPAT_FREEBSD11)
749 * Get old format filesystem statistics.
751 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
754 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
756 struct freebsd11_statfs osb;
760 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
761 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
763 freebsd11_cvtstatfs(sfp, &osb);
764 error = copyout(&osb, uap->buf, sizeof(osb));
771 * Get filesystem statistics.
774 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
776 struct freebsd11_statfs osb;
780 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
781 error = kern_fstatfs(td, uap->fd, sfp);
783 freebsd11_cvtstatfs(sfp, &osb);
784 error = copyout(&osb, uap->buf, sizeof(osb));
791 * Get statistics on all filesystems.
794 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
796 struct freebsd11_statfs osb;
797 struct statfs *buf, *sp;
801 count = uap->bufsize / sizeof(struct ostatfs);
802 size = count * sizeof(struct statfs);
803 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
806 td->td_retval[0] = count;
809 while (count > 0 && error == 0) {
810 freebsd11_cvtstatfs(sp, &osb);
811 error = copyout(&osb, uap->buf, sizeof(osb));
822 * Implement fstatfs() for (NFS) file handles.
825 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
827 struct freebsd11_statfs osb;
832 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
835 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
836 error = kern_fhstatfs(td, fh, sfp);
838 freebsd11_cvtstatfs(sfp, &osb);
839 error = copyout(&osb, uap->buf, sizeof(osb));
846 * Convert a new format statfs structure to an old format statfs structure.
849 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
852 bzero(osp, sizeof(*osp));
853 osp->f_version = FREEBSD11_STATFS_VERSION;
854 osp->f_type = nsp->f_type;
855 osp->f_flags = nsp->f_flags;
856 osp->f_bsize = nsp->f_bsize;
857 osp->f_iosize = nsp->f_iosize;
858 osp->f_blocks = nsp->f_blocks;
859 osp->f_bfree = nsp->f_bfree;
860 osp->f_bavail = nsp->f_bavail;
861 osp->f_files = nsp->f_files;
862 osp->f_ffree = nsp->f_ffree;
863 osp->f_syncwrites = nsp->f_syncwrites;
864 osp->f_asyncwrites = nsp->f_asyncwrites;
865 osp->f_syncreads = nsp->f_syncreads;
866 osp->f_asyncreads = nsp->f_asyncreads;
867 osp->f_namemax = nsp->f_namemax;
868 osp->f_owner = nsp->f_owner;
869 osp->f_fsid = nsp->f_fsid;
870 strlcpy(osp->f_fstypename, nsp->f_fstypename,
871 MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
872 strlcpy(osp->f_mntonname, nsp->f_mntonname,
873 MIN(MNAMELEN, sizeof(osp->f_mntonname)));
874 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
875 MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
877 #endif /* COMPAT_FREEBSD11 */
880 * Change current working directory to a given file descriptor.
882 #ifndef _SYS_SYSPROTO_H_
888 sys_fchdir(struct thread *td, struct fchdir_args *uap)
890 struct vnode *vp, *tdp;
895 AUDIT_ARG_FD(uap->fd);
896 error = getvnode(td, uap->fd, &cap_fchdir_rights,
903 vn_lock(vp, LK_SHARED | LK_RETRY);
904 AUDIT_ARG_VNODE1(vp);
905 error = change_dir(vp, td);
906 while (!error && (mp = vp->v_mountedhere) != NULL) {
909 error = VFS_ROOT(mp, LK_SHARED, &tdp);
926 * Change current working directory (``.'').
928 #ifndef _SYS_SYSPROTO_H_
934 sys_chdir(struct thread *td, struct chdir_args *uap)
937 return (kern_chdir(td, uap->path, UIO_USERSPACE));
941 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
946 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
948 if ((error = namei(&nd)) != 0)
950 if ((error = change_dir(nd.ni_vp, td)) != 0) {
955 VOP_UNLOCK(nd.ni_vp);
957 pwd_chdir(td, nd.ni_vp);
962 * Change notion of root (``/'') directory.
964 #ifndef _SYS_SYSPROTO_H_
970 sys_chroot(struct thread *td, struct chroot_args *uap)
975 error = priv_check(td, PRIV_VFS_CHROOT);
978 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
979 UIO_USERSPACE, uap->path, td);
983 error = change_dir(nd.ni_vp, td);
987 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
991 VOP_UNLOCK(nd.ni_vp);
992 error = pwd_chroot(td, nd.ni_vp);
1004 * Common routine for chroot and chdir. Callers must provide a locked vnode
1008 change_dir(struct vnode *vp, struct thread *td)
1014 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1015 if (vp->v_type != VDIR)
1018 error = mac_vnode_check_chdir(td->td_ucred, vp);
1022 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1025 static __inline void
1026 flags_to_rights(int flags, cap_rights_t *rightsp)
1029 if (flags & O_EXEC) {
1030 cap_rights_set_one(rightsp, CAP_FEXECVE);
1032 switch ((flags & O_ACCMODE)) {
1034 cap_rights_set_one(rightsp, CAP_READ);
1037 cap_rights_set_one(rightsp, CAP_READ);
1040 cap_rights_set_one(rightsp, CAP_WRITE);
1041 if (!(flags & (O_APPEND | O_TRUNC)))
1042 cap_rights_set_one(rightsp, CAP_SEEK);
1047 if (flags & O_CREAT)
1048 cap_rights_set_one(rightsp, CAP_CREATE);
1050 if (flags & O_TRUNC)
1051 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1053 if (flags & (O_SYNC | O_FSYNC))
1054 cap_rights_set_one(rightsp, CAP_FSYNC);
1056 if (flags & (O_EXLOCK | O_SHLOCK))
1057 cap_rights_set_one(rightsp, CAP_FLOCK);
1061 * Check permissions, allocate an open file structure, and call the device
1062 * open routine if any.
1064 #ifndef _SYS_SYSPROTO_H_
1072 sys_open(struct thread *td, struct open_args *uap)
1075 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1076 uap->flags, uap->mode));
1079 #ifndef _SYS_SYSPROTO_H_
1080 struct openat_args {
1088 sys_openat(struct thread *td, struct openat_args *uap)
1091 AUDIT_ARG_FD(uap->fd);
1092 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1097 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1098 int flags, int mode)
1100 struct proc *p = td->td_proc;
1101 struct filedesc *fdp;
1102 struct pwddesc *pdp;
1105 struct nameidata nd;
1106 cap_rights_t rights;
1107 int cmode, error, indx;
1113 AUDIT_ARG_FFLAGS(flags);
1114 AUDIT_ARG_MODE(mode);
1115 cap_rights_init_one(&rights, CAP_LOOKUP);
1116 flags_to_rights(flags, &rights);
1118 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1121 if (flags & O_EXEC) {
1122 if (flags & O_ACCMODE)
1124 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1127 flags = FFLAGS(flags);
1131 * Allocate a file structure. The descriptor to reference it
1132 * is allocated and used by finstall_refed() below.
1134 error = falloc_noinstall(td, &fp);
1137 /* Set the flags early so the finit in devfs can pick them up. */
1138 fp->f_flag = flags & FMASK;
1139 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1140 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1142 td->td_dupfd = -1; /* XXX check for fdopen */
1143 error = vn_open(&nd, &flags, cmode, fp);
1146 * If the vn_open replaced the method vector, something
1147 * wonderous happened deep below and we just pass it up
1148 * pretending we know what we do.
1150 if (error == ENXIO && fp->f_ops != &badfileops)
1154 * Handle special fdopen() case. bleh.
1156 * Don't do this for relative (capability) lookups; we don't
1157 * understand exactly what would happen, and we don't think
1158 * that it ever should.
1160 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1161 (error == ENODEV || error == ENXIO) &&
1162 td->td_dupfd >= 0) {
1163 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1172 NDFREE(&nd, NDF_ONLY_PNBUF);
1176 * Store the vnode, for any f_type. Typically, the vnode use
1177 * count is decremented by direct call to vn_closefile() for
1178 * files that switched type in the cdevsw fdopen() method.
1182 * If the file wasn't claimed by devfs bind it to the normal
1183 * vnode operations here.
1185 if (fp->f_ops == &badfileops) {
1186 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1187 finit_vnode(fp, flags, NULL, &vnops);
1191 if (flags & O_TRUNC) {
1192 error = fo_truncate(fp, 0, td->td_ucred, td);
1198 * If we haven't already installed the FD (for dupfdopen), do so now.
1201 struct filecaps *fcaps;
1204 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1205 fcaps = &nd.ni_filecaps;
1209 error = finstall_refed(td, fp, &indx, flags, fcaps);
1210 /* On success finstall_refed() consumes fcaps. */
1212 filecaps_free(&nd.ni_filecaps);
1216 filecaps_free(&nd.ni_filecaps);
1217 fdrop_close(fp, td);
1220 td->td_retval[0] = indx;
1223 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1224 fdrop_close(fp, td);
1232 #ifndef _SYS_SYSPROTO_H_
1233 struct ocreat_args {
1239 ocreat(struct thread *td, struct ocreat_args *uap)
1242 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1243 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1245 #endif /* COMPAT_43 */
1248 * Create a special file.
1250 #ifndef _SYS_SYSPROTO_H_
1251 struct mknodat_args {
1259 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1262 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1266 #if defined(COMPAT_FREEBSD11)
1268 freebsd11_mknod(struct thread *td,
1269 struct freebsd11_mknod_args *uap)
1272 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1273 uap->mode, uap->dev));
1277 freebsd11_mknodat(struct thread *td,
1278 struct freebsd11_mknodat_args *uap)
1281 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1284 #endif /* COMPAT_FREEBSD11 */
1287 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1288 int mode, dev_t dev)
1293 struct nameidata nd;
1294 int error, whiteout = 0;
1296 AUDIT_ARG_MODE(mode);
1298 switch (mode & S_IFMT) {
1301 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1302 if (error == 0 && dev == VNOVAL)
1306 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1310 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1320 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1321 NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1323 if ((error = namei(&nd)) != 0)
1327 NDFREE(&nd, NDF_ONLY_PNBUF);
1328 if (vp == nd.ni_dvp)
1336 vattr.va_mode = (mode & ALLPERMS) &
1337 ~td->td_proc->p_pd->pd_cmask;
1338 vattr.va_rdev = dev;
1341 switch (mode & S_IFMT) {
1343 vattr.va_type = VCHR;
1346 vattr.va_type = VBLK;
1352 panic("kern_mknod: invalid mode");
1355 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1356 NDFREE(&nd, NDF_ONLY_PNBUF);
1358 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1363 if (error == 0 && !whiteout)
1364 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1365 &nd.ni_cnd, &vattr);
1369 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1371 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1372 &nd.ni_cnd, &vattr);
1377 NDFREE(&nd, NDF_ONLY_PNBUF);
1379 vn_finished_write(mp);
1380 if (error == ERELOOKUP)
1386 * Create a named pipe.
1388 #ifndef _SYS_SYSPROTO_H_
1389 struct mkfifo_args {
1395 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1398 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1402 #ifndef _SYS_SYSPROTO_H_
1403 struct mkfifoat_args {
1410 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1413 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1418 kern_mkfifoat(struct thread *td, int fd, const char *path,
1419 enum uio_seg pathseg, int mode)
1423 struct nameidata nd;
1426 AUDIT_ARG_MODE(mode);
1429 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1430 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1432 if ((error = namei(&nd)) != 0)
1434 if (nd.ni_vp != NULL) {
1435 NDFREE(&nd, NDF_ONLY_PNBUF);
1436 if (nd.ni_vp == nd.ni_dvp)
1443 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1444 NDFREE(&nd, NDF_ONLY_PNBUF);
1446 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1451 vattr.va_type = VFIFO;
1452 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1454 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1459 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1466 vn_finished_write(mp);
1467 NDFREE(&nd, NDF_ONLY_PNBUF);
1468 if (error == ERELOOKUP)
1474 * Make a hard file link.
1476 #ifndef _SYS_SYSPROTO_H_
1483 sys_link(struct thread *td, struct link_args *uap)
1486 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1487 UIO_USERSPACE, FOLLOW));
1490 #ifndef _SYS_SYSPROTO_H_
1491 struct linkat_args {
1500 sys_linkat(struct thread *td, struct linkat_args *uap)
1505 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH |
1506 AT_RESOLVE_BENEATH)) != 0)
1509 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1510 UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW | AT_BENEATH |
1511 AT_RESOLVE_BENEATH)));
1514 int hardlink_check_uid = 0;
1515 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1516 &hardlink_check_uid, 0,
1517 "Unprivileged processes cannot create hard links to files owned by other "
1519 static int hardlink_check_gid = 0;
1520 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1521 &hardlink_check_gid, 0,
1522 "Unprivileged processes cannot create hard links to files owned by other "
1526 can_hardlink(struct vnode *vp, struct ucred *cred)
1531 if (!hardlink_check_uid && !hardlink_check_gid)
1534 error = VOP_GETATTR(vp, &va, cred);
1538 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1539 error = priv_check_cred(cred, PRIV_VFS_LINK);
1544 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1545 error = priv_check_cred(cred, PRIV_VFS_LINK);
1554 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1555 const char *path2, enum uio_seg segflag, int follow)
1557 struct nameidata nd;
1562 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1563 path1, fd1, &cap_linkat_source_rights, td);
1564 if ((error = namei(&nd)) != 0)
1566 NDFREE(&nd, NDF_ONLY_PNBUF);
1567 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1568 } while (error == EAGAIN || error == ERELOOKUP);
1573 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1574 enum uio_seg segflag)
1576 struct nameidata nd;
1580 if (vp->v_type == VDIR) {
1582 return (EPERM); /* POSIX */
1584 NDINIT_ATRIGHTS(&nd, CREATE,
1585 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1586 &cap_linkat_target_rights, td);
1587 if ((error = namei(&nd)) == 0) {
1588 if (nd.ni_vp != NULL) {
1589 NDFREE(&nd, NDF_ONLY_PNBUF);
1590 if (nd.ni_dvp == nd.ni_vp)
1597 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1599 * Cross-device link. No need to recheck
1600 * vp->v_type, since it cannot change, except
1603 NDFREE(&nd, NDF_ONLY_PNBUF);
1607 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1608 error = can_hardlink(vp, td->td_ucred);
1611 error = mac_vnode_check_link(td->td_ucred,
1612 nd.ni_dvp, vp, &nd.ni_cnd);
1617 NDFREE(&nd, NDF_ONLY_PNBUF);
1620 error = vn_start_write(vp, &mp, V_NOWAIT);
1624 NDFREE(&nd, NDF_ONLY_PNBUF);
1625 error = vn_start_write(NULL, &mp,
1631 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1634 vn_finished_write(mp);
1635 NDFREE(&nd, NDF_ONLY_PNBUF);
1638 NDFREE(&nd, NDF_ONLY_PNBUF);
1648 * Make a symbolic link.
1650 #ifndef _SYS_SYSPROTO_H_
1651 struct symlink_args {
1657 sys_symlink(struct thread *td, struct symlink_args *uap)
1660 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1664 #ifndef _SYS_SYSPROTO_H_
1665 struct symlinkat_args {
1672 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1675 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1680 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1681 enum uio_seg segflg)
1685 const char *syspath;
1687 struct nameidata nd;
1690 if (segflg == UIO_SYSSPACE) {
1693 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1694 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1698 AUDIT_ARG_TEXT(syspath);
1701 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1702 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1704 if ((error = namei(&nd)) != 0)
1707 NDFREE(&nd, NDF_ONLY_PNBUF);
1708 if (nd.ni_vp == nd.ni_dvp)
1716 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1717 NDFREE(&nd, NDF_ONLY_PNBUF);
1719 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1724 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1726 vattr.va_type = VLNK;
1727 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1732 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1738 NDFREE(&nd, NDF_ONLY_PNBUF);
1740 vn_finished_write(mp);
1741 if (error == ERELOOKUP)
1744 if (segflg != UIO_SYSSPACE)
1745 uma_zfree(namei_zone, tmppath);
1750 * Delete a whiteout from the filesystem.
1752 #ifndef _SYS_SYSPROTO_H_
1753 struct undelete_args {
1758 sys_undelete(struct thread *td, struct undelete_args *uap)
1761 struct nameidata nd;
1766 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1767 UIO_USERSPACE, uap->path, td);
1772 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1773 NDFREE(&nd, NDF_ONLY_PNBUF);
1774 if (nd.ni_vp == nd.ni_dvp)
1782 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1783 NDFREE(&nd, NDF_ONLY_PNBUF);
1785 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1789 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1790 NDFREE(&nd, NDF_ONLY_PNBUF);
1792 vn_finished_write(mp);
1793 if (error == ERELOOKUP)
1799 * Delete a name from the filesystem.
1801 #ifndef _SYS_SYSPROTO_H_
1802 struct unlink_args {
1807 sys_unlink(struct thread *td, struct unlink_args *uap)
1810 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1815 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1816 int flag, enum uio_seg pathseg, ino_t oldinum)
1819 if ((flag & ~AT_REMOVEDIR) != 0)
1822 if ((flag & AT_REMOVEDIR) != 0)
1823 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1825 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1828 #ifndef _SYS_SYSPROTO_H_
1829 struct unlinkat_args {
1836 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1839 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1843 #ifndef _SYS_SYSPROTO_H_
1844 struct funlinkat_args {
1852 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1855 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1860 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1861 enum uio_seg pathseg, int flag, ino_t oldinum)
1866 struct nameidata nd;
1871 if (fd != FD_NONE) {
1872 error = getvnode(td, fd, &cap_no_rights, &fp);
1879 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1880 at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
1881 pathseg, path, dfd, &cap_unlinkat_rights, td);
1882 if ((error = namei(&nd)) != 0) {
1883 if (error == EINVAL)
1888 if (vp->v_type == VDIR && oldinum == 0) {
1889 error = EPERM; /* POSIX */
1890 } else if (oldinum != 0 &&
1891 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1892 sb.st_ino != oldinum) {
1893 error = EIDRM; /* Identifier removed */
1894 } else if (fp != NULL && fp->f_vnode != vp) {
1895 if (VN_IS_DOOMED(fp->f_vnode))
1901 * The root of a mounted filesystem cannot be deleted.
1903 * XXX: can this only be a VDIR case?
1905 if (vp->v_vflag & VV_ROOT)
1909 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1910 NDFREE(&nd, NDF_ONLY_PNBUF);
1912 if (vp == nd.ni_dvp)
1916 if ((error = vn_start_write(NULL, &mp,
1917 V_XSLEEP | PCATCH)) != 0) {
1923 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1928 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1929 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1933 vn_finished_write(mp);
1935 NDFREE(&nd, NDF_ONLY_PNBUF);
1937 if (vp == nd.ni_dvp)
1941 if (error == ERELOOKUP)
1950 * Reposition read/write file offset.
1952 #ifndef _SYS_SYSPROTO_H_
1961 sys_lseek(struct thread *td, struct lseek_args *uap)
1964 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1968 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1974 error = fget(td, fd, &cap_seek_rights, &fp);
1977 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1978 fo_seek(fp, offset, whence, td) : ESPIPE;
1983 #if defined(COMPAT_43)
1985 * Reposition read/write file offset.
1987 #ifndef _SYS_SYSPROTO_H_
1988 struct olseek_args {
1995 olseek(struct thread *td, struct olseek_args *uap)
1998 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2000 #endif /* COMPAT_43 */
2002 #if defined(COMPAT_FREEBSD6)
2003 /* Version with the 'pad' argument */
2005 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2008 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2013 * Check access permissions using passed credentials.
2016 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2022 /* Flags == 0 means only check for existence. */
2023 if (user_flags == 0)
2027 if (user_flags & R_OK)
2029 if (user_flags & W_OK)
2031 if (user_flags & X_OK)
2034 error = mac_vnode_check_access(cred, vp, accmode);
2038 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2039 error = VOP_ACCESS(vp, accmode, cred, td);
2044 * Check access permissions using "real" credentials.
2046 #ifndef _SYS_SYSPROTO_H_
2047 struct access_args {
2053 sys_access(struct thread *td, struct access_args *uap)
2056 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2060 #ifndef _SYS_SYSPROTO_H_
2061 struct faccessat_args {
2069 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2072 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2077 kern_accessat(struct thread *td, int fd, const char *path,
2078 enum uio_seg pathseg, int flag, int amode)
2080 struct ucred *cred, *usecred;
2082 struct nameidata nd;
2085 if ((flag & ~(AT_EACCESS | AT_BENEATH | AT_RESOLVE_BENEATH)) != 0)
2087 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2091 * Create and modify a temporary credential instead of one that
2092 * is potentially shared (if we need one).
2094 cred = td->td_ucred;
2095 if ((flag & AT_EACCESS) == 0 &&
2096 ((cred->cr_uid != cred->cr_ruid ||
2097 cred->cr_rgid != cred->cr_groups[0]))) {
2098 usecred = crdup(cred);
2099 usecred->cr_uid = cred->cr_ruid;
2100 usecred->cr_groups[0] = cred->cr_rgid;
2101 td->td_ucred = usecred;
2104 AUDIT_ARG_VALUE(amode);
2105 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2106 AUDITVNODE1 | at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
2107 pathseg, path, fd, &cap_fstat_rights, td);
2108 if ((error = namei(&nd)) != 0)
2112 error = vn_access(vp, amode, usecred, td);
2113 NDFREE_NOTHING(&nd);
2116 if (usecred != cred) {
2117 td->td_ucred = cred;
2124 * Check access permissions using "effective" credentials.
2126 #ifndef _SYS_SYSPROTO_H_
2127 struct eaccess_args {
2133 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2136 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2137 AT_EACCESS, uap->amode));
2140 #if defined(COMPAT_43)
2142 * Get file status; this version follows links.
2144 #ifndef _SYS_SYSPROTO_H_
2151 ostat(struct thread *td, struct ostat_args *uap)
2157 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2162 return (copyout(&osb, uap->ub, sizeof (osb)));
2166 * Get file status; this version does not follow links.
2168 #ifndef _SYS_SYSPROTO_H_
2169 struct olstat_args {
2175 olstat(struct thread *td, struct olstat_args *uap)
2181 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2182 UIO_USERSPACE, &sb, NULL);
2186 return (copyout(&osb, uap->ub, sizeof (osb)));
2190 * Convert from an old to a new stat structure.
2191 * XXX: many values are blindly truncated.
2194 cvtstat(struct stat *st, struct ostat *ost)
2197 bzero(ost, sizeof(*ost));
2198 ost->st_dev = st->st_dev;
2199 ost->st_ino = st->st_ino;
2200 ost->st_mode = st->st_mode;
2201 ost->st_nlink = st->st_nlink;
2202 ost->st_uid = st->st_uid;
2203 ost->st_gid = st->st_gid;
2204 ost->st_rdev = st->st_rdev;
2205 ost->st_size = MIN(st->st_size, INT32_MAX);
2206 ost->st_atim = st->st_atim;
2207 ost->st_mtim = st->st_mtim;
2208 ost->st_ctim = st->st_ctim;
2209 ost->st_blksize = st->st_blksize;
2210 ost->st_blocks = st->st_blocks;
2211 ost->st_flags = st->st_flags;
2212 ost->st_gen = st->st_gen;
2214 #endif /* COMPAT_43 */
2216 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2217 int ino64_trunc_error;
2218 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2219 &ino64_trunc_error, 0,
2220 "Error on truncation of device, file or inode number, or link count");
2223 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2226 ost->st_dev = st->st_dev;
2227 if (ost->st_dev != st->st_dev) {
2228 switch (ino64_trunc_error) {
2231 * Since dev_t is almost raw, don't clamp to the
2232 * maximum for case 2, but ignore the error.
2239 ost->st_ino = st->st_ino;
2240 if (ost->st_ino != st->st_ino) {
2241 switch (ino64_trunc_error) {
2248 ost->st_ino = UINT32_MAX;
2252 ost->st_mode = st->st_mode;
2253 ost->st_nlink = st->st_nlink;
2254 if (ost->st_nlink != st->st_nlink) {
2255 switch (ino64_trunc_error) {
2262 ost->st_nlink = UINT16_MAX;
2266 ost->st_uid = st->st_uid;
2267 ost->st_gid = st->st_gid;
2268 ost->st_rdev = st->st_rdev;
2269 if (ost->st_rdev != st->st_rdev) {
2270 switch (ino64_trunc_error) {
2277 ost->st_atim = st->st_atim;
2278 ost->st_mtim = st->st_mtim;
2279 ost->st_ctim = st->st_ctim;
2280 ost->st_size = st->st_size;
2281 ost->st_blocks = st->st_blocks;
2282 ost->st_blksize = st->st_blksize;
2283 ost->st_flags = st->st_flags;
2284 ost->st_gen = st->st_gen;
2286 ost->st_birthtim = st->st_birthtim;
2287 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2288 sizeof(*ost) - offsetof(struct freebsd11_stat,
2289 st_birthtim) - sizeof(ost->st_birthtim));
2294 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2297 struct freebsd11_stat osb;
2300 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2304 error = freebsd11_cvtstat(&sb, &osb);
2306 error = copyout(&osb, uap->ub, sizeof(osb));
2311 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2314 struct freebsd11_stat osb;
2317 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2318 UIO_USERSPACE, &sb, NULL);
2321 error = freebsd11_cvtstat(&sb, &osb);
2323 error = copyout(&osb, uap->ub, sizeof(osb));
2328 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2332 struct freebsd11_stat osb;
2335 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2338 error = kern_fhstat(td, fh, &sb);
2341 error = freebsd11_cvtstat(&sb, &osb);
2343 error = copyout(&osb, uap->sb, sizeof(osb));
2348 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2351 struct freebsd11_stat osb;
2354 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2355 UIO_USERSPACE, &sb, NULL);
2358 error = freebsd11_cvtstat(&sb, &osb);
2360 error = copyout(&osb, uap->buf, sizeof(osb));
2363 #endif /* COMPAT_FREEBSD11 */
2368 #ifndef _SYS_SYSPROTO_H_
2369 struct fstatat_args {
2377 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2382 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2383 UIO_USERSPACE, &sb, NULL);
2385 error = copyout(&sb, uap->buf, sizeof (sb));
2390 kern_statat(struct thread *td, int flag, int fd, const char *path,
2391 enum uio_seg pathseg, struct stat *sbp,
2392 void (*hook)(struct vnode *vp, struct stat *sbp))
2394 struct nameidata nd;
2397 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2398 AT_RESOLVE_BENEATH)) != 0)
2401 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_BENEATH |
2402 AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF |
2403 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2405 if ((error = namei(&nd)) != 0)
2407 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2409 if (__predict_false(hook != NULL))
2410 hook(nd.ni_vp, sbp);
2412 NDFREE_NOTHING(&nd);
2414 #ifdef __STAT_TIME_T_EXT
2415 sbp->st_atim_ext = 0;
2416 sbp->st_mtim_ext = 0;
2417 sbp->st_ctim_ext = 0;
2418 sbp->st_btim_ext = 0;
2421 if (KTRPOINT(td, KTR_STRUCT))
2422 ktrstat_error(sbp, error);
2427 #if defined(COMPAT_FREEBSD11)
2429 * Implementation of the NetBSD [l]stat() functions.
2432 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2435 bzero(nsb, sizeof(*nsb));
2436 nsb->st_dev = sb->st_dev;
2437 nsb->st_ino = sb->st_ino;
2438 nsb->st_mode = sb->st_mode;
2439 nsb->st_nlink = sb->st_nlink;
2440 nsb->st_uid = sb->st_uid;
2441 nsb->st_gid = sb->st_gid;
2442 nsb->st_rdev = sb->st_rdev;
2443 nsb->st_atim = sb->st_atim;
2444 nsb->st_mtim = sb->st_mtim;
2445 nsb->st_ctim = sb->st_ctim;
2446 nsb->st_size = sb->st_size;
2447 nsb->st_blocks = sb->st_blocks;
2448 nsb->st_blksize = sb->st_blksize;
2449 nsb->st_flags = sb->st_flags;
2450 nsb->st_gen = sb->st_gen;
2451 nsb->st_birthtim = sb->st_birthtim;
2454 #ifndef _SYS_SYSPROTO_H_
2455 struct freebsd11_nstat_args {
2461 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2467 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2471 freebsd11_cvtnstat(&sb, &nsb);
2472 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2476 * NetBSD lstat. Get file status; this version does not follow links.
2478 #ifndef _SYS_SYSPROTO_H_
2479 struct freebsd11_nlstat_args {
2485 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2491 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2492 UIO_USERSPACE, &sb, NULL);
2495 freebsd11_cvtnstat(&sb, &nsb);
2496 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2498 #endif /* COMPAT_FREEBSD11 */
2501 * Get configurable pathname variables.
2503 #ifndef _SYS_SYSPROTO_H_
2504 struct pathconf_args {
2510 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2515 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2518 td->td_retval[0] = value;
2522 #ifndef _SYS_SYSPROTO_H_
2523 struct lpathconf_args {
2529 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2534 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2537 td->td_retval[0] = value;
2542 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2543 int name, u_long flags, long *valuep)
2545 struct nameidata nd;
2548 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2550 if ((error = namei(&nd)) != 0)
2552 NDFREE_NOTHING(&nd);
2554 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2560 * Return target name of a symbolic link.
2562 #ifndef _SYS_SYSPROTO_H_
2563 struct readlink_args {
2570 sys_readlink(struct thread *td, struct readlink_args *uap)
2573 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2574 uap->buf, UIO_USERSPACE, uap->count));
2576 #ifndef _SYS_SYSPROTO_H_
2577 struct readlinkat_args {
2585 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2588 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2589 uap->buf, UIO_USERSPACE, uap->bufsize));
2593 kern_readlinkat(struct thread *td, int fd, const char *path,
2594 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2597 struct nameidata nd;
2600 if (count > IOSIZE_MAX)
2603 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2604 pathseg, path, fd, td);
2606 if ((error = namei(&nd)) != 0)
2608 NDFREE_NOTHING(&nd);
2611 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2618 * Helper function to readlink from a vnode
2621 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2628 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2630 error = mac_vnode_check_readlink(td->td_ucred, vp);
2634 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2637 aiov.iov_base = buf;
2638 aiov.iov_len = count;
2639 auio.uio_iov = &aiov;
2640 auio.uio_iovcnt = 1;
2641 auio.uio_offset = 0;
2642 auio.uio_rw = UIO_READ;
2643 auio.uio_segflg = bufseg;
2645 auio.uio_resid = count;
2646 error = VOP_READLINK(vp, &auio, td->td_ucred);
2647 td->td_retval[0] = count - auio.uio_resid;
2652 * Common implementation code for chflags() and fchflags().
2655 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2661 /* We can't support the value matching VNOVAL. */
2662 if (flags == VNOVAL)
2663 return (EOPNOTSUPP);
2666 * Prevent non-root users from setting flags on devices. When
2667 * a device is reused, users can retain ownership of the device
2668 * if they are allowed to set flags and programs assume that
2669 * chown can't fail when done as root.
2671 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2672 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2677 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2680 vattr.va_flags = flags;
2681 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2683 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2686 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2688 vn_finished_write(mp);
2693 * Change flags of a file given a path name.
2695 #ifndef _SYS_SYSPROTO_H_
2696 struct chflags_args {
2702 sys_chflags(struct thread *td, struct chflags_args *uap)
2705 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2709 #ifndef _SYS_SYSPROTO_H_
2710 struct chflagsat_args {
2718 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2721 if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2722 AT_RESOLVE_BENEATH)) != 0)
2725 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2726 uap->flags, uap->atflag));
2730 * Same as chflags() but doesn't follow symlinks.
2732 #ifndef _SYS_SYSPROTO_H_
2733 struct lchflags_args {
2739 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2742 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2743 uap->flags, AT_SYMLINK_NOFOLLOW));
2747 kern_chflagsat(struct thread *td, int fd, const char *path,
2748 enum uio_seg pathseg, u_long flags, int atflag)
2750 struct nameidata nd;
2753 AUDIT_ARG_FFLAGS(flags);
2754 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2755 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2756 &cap_fchflags_rights, td);
2757 if ((error = namei(&nd)) != 0)
2759 NDFREE_NOTHING(&nd);
2760 error = setfflags(td, nd.ni_vp, flags);
2766 * Change flags of a file given a file descriptor.
2768 #ifndef _SYS_SYSPROTO_H_
2769 struct fchflags_args {
2775 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2780 AUDIT_ARG_FD(uap->fd);
2781 AUDIT_ARG_FFLAGS(uap->flags);
2782 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2787 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2788 AUDIT_ARG_VNODE1(fp->f_vnode);
2789 VOP_UNLOCK(fp->f_vnode);
2791 error = setfflags(td, fp->f_vnode, uap->flags);
2797 * Common implementation code for chmod(), lchmod() and fchmod().
2800 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2806 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2808 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2810 vattr.va_mode = mode & ALLPERMS;
2812 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2815 error = VOP_SETATTR(vp, &vattr, cred);
2817 vn_finished_write(mp);
2822 * Change mode of a file given path name.
2824 #ifndef _SYS_SYSPROTO_H_
2831 sys_chmod(struct thread *td, struct chmod_args *uap)
2834 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2838 #ifndef _SYS_SYSPROTO_H_
2839 struct fchmodat_args {
2847 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2850 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2851 AT_RESOLVE_BENEATH)) != 0)
2854 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2855 uap->mode, uap->flag));
2859 * Change mode of a file given path name (don't follow links.)
2861 #ifndef _SYS_SYSPROTO_H_
2862 struct lchmod_args {
2868 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2871 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2872 uap->mode, AT_SYMLINK_NOFOLLOW));
2876 kern_fchmodat(struct thread *td, int fd, const char *path,
2877 enum uio_seg pathseg, mode_t mode, int flag)
2879 struct nameidata nd;
2882 AUDIT_ARG_MODE(mode);
2883 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2884 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2885 &cap_fchmod_rights, td);
2886 if ((error = namei(&nd)) != 0)
2888 NDFREE_NOTHING(&nd);
2889 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2895 * Change mode of a file given a file descriptor.
2897 #ifndef _SYS_SYSPROTO_H_
2898 struct fchmod_args {
2904 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2909 AUDIT_ARG_FD(uap->fd);
2910 AUDIT_ARG_MODE(uap->mode);
2912 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2915 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2921 * Common implementation for chown(), lchown(), and fchown()
2924 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2931 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2933 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2938 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2942 error = VOP_SETATTR(vp, &vattr, cred);
2944 vn_finished_write(mp);
2949 * Set ownership given a path name.
2951 #ifndef _SYS_SYSPROTO_H_
2959 sys_chown(struct thread *td, struct chown_args *uap)
2962 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2966 #ifndef _SYS_SYSPROTO_H_
2967 struct fchownat_args {
2976 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2979 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2980 AT_RESOLVE_BENEATH)) != 0)
2983 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2984 uap->gid, uap->flag));
2988 kern_fchownat(struct thread *td, int fd, const char *path,
2989 enum uio_seg pathseg, int uid, int gid, int flag)
2991 struct nameidata nd;
2994 AUDIT_ARG_OWNER(uid, gid);
2995 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2996 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2997 &cap_fchown_rights, td);
2999 if ((error = namei(&nd)) != 0)
3001 NDFREE_NOTHING(&nd);
3002 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3008 * Set ownership given a path name, do not cross symlinks.
3010 #ifndef _SYS_SYSPROTO_H_
3011 struct lchown_args {
3018 sys_lchown(struct thread *td, struct lchown_args *uap)
3021 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3022 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3026 * Set ownership given a file descriptor.
3028 #ifndef _SYS_SYSPROTO_H_
3029 struct fchown_args {
3036 sys_fchown(struct thread *td, struct fchown_args *uap)
3041 AUDIT_ARG_FD(uap->fd);
3042 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3043 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3046 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3052 * Common implementation code for utimes(), lutimes(), and futimes().
3055 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3056 struct timespec *tsp)
3058 struct timeval tv[2];
3059 const struct timeval *tvp;
3062 if (usrtvp == NULL) {
3063 vfs_timestamp(&tsp[0]);
3066 if (tvpseg == UIO_SYSSPACE) {
3069 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3074 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3075 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3077 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3078 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3084 * Common implementation code for futimens(), utimensat().
3086 #define UTIMENS_NULL 0x1
3087 #define UTIMENS_EXIT 0x2
3089 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3090 struct timespec *tsp, int *retflags)
3092 struct timespec tsnow;
3095 vfs_timestamp(&tsnow);
3097 if (usrtsp == NULL) {
3100 *retflags |= UTIMENS_NULL;
3103 if (tspseg == UIO_SYSSPACE) {
3106 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3108 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3109 *retflags |= UTIMENS_EXIT;
3110 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3111 *retflags |= UTIMENS_NULL;
3112 if (tsp[0].tv_nsec == UTIME_OMIT)
3113 tsp[0].tv_sec = VNOVAL;
3114 else if (tsp[0].tv_nsec == UTIME_NOW)
3116 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3118 if (tsp[1].tv_nsec == UTIME_OMIT)
3119 tsp[1].tv_sec = VNOVAL;
3120 else if (tsp[1].tv_nsec == UTIME_NOW)
3122 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3129 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3133 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3134 int numtimes, int nullflag)
3138 int error, setbirthtime;
3140 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3142 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3144 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3145 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3148 vattr.va_atime = ts[0];
3149 vattr.va_mtime = ts[1];
3151 vattr.va_birthtime = ts[1];
3153 vattr.va_birthtime = ts[2];
3155 vattr.va_vaflags |= VA_UTIMES_NULL;
3157 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3161 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3163 vn_finished_write(mp);
3168 * Set the access and modification times of a file.
3170 #ifndef _SYS_SYSPROTO_H_
3171 struct utimes_args {
3173 struct timeval *tptr;
3177 sys_utimes(struct thread *td, struct utimes_args *uap)
3180 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3181 uap->tptr, UIO_USERSPACE));
3184 #ifndef _SYS_SYSPROTO_H_
3185 struct futimesat_args {
3188 const struct timeval * times;
3192 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3195 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3196 uap->times, UIO_USERSPACE));
3200 kern_utimesat(struct thread *td, int fd, const char *path,
3201 enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3203 struct nameidata nd;
3204 struct timespec ts[2];
3207 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3209 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3210 &cap_futimes_rights, td);
3212 if ((error = namei(&nd)) != 0)
3214 NDFREE_NOTHING(&nd);
3215 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3221 * Set the access and modification times of a file.
3223 #ifndef _SYS_SYSPROTO_H_
3224 struct lutimes_args {
3226 struct timeval *tptr;
3230 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3233 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3238 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3239 struct timeval *tptr, enum uio_seg tptrseg)
3241 struct timespec ts[2];
3242 struct nameidata nd;
3245 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3247 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3248 if ((error = namei(&nd)) != 0)
3250 NDFREE_NOTHING(&nd);
3251 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3257 * Set the access and modification times of a file.
3259 #ifndef _SYS_SYSPROTO_H_
3260 struct futimes_args {
3262 struct timeval *tptr;
3266 sys_futimes(struct thread *td, struct futimes_args *uap)
3269 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3273 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3274 enum uio_seg tptrseg)
3276 struct timespec ts[2];
3281 error = getutimes(tptr, tptrseg, ts);
3284 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3288 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3289 AUDIT_ARG_VNODE1(fp->f_vnode);
3290 VOP_UNLOCK(fp->f_vnode);
3292 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3298 sys_futimens(struct thread *td, struct futimens_args *uap)
3301 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3305 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3306 enum uio_seg tptrseg)
3308 struct timespec ts[2];
3313 error = getutimens(tptr, tptrseg, ts, &flags);
3316 if (flags & UTIMENS_EXIT)
3318 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3322 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3323 AUDIT_ARG_VNODE1(fp->f_vnode);
3324 VOP_UNLOCK(fp->f_vnode);
3326 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3332 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3335 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3336 uap->times, UIO_USERSPACE, uap->flag));
3340 kern_utimensat(struct thread *td, int fd, const char *path,
3341 enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3344 struct nameidata nd;
3345 struct timespec ts[2];
3348 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
3349 AT_RESOLVE_BENEATH)) != 0)
3352 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3354 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3355 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1,
3356 pathseg, path, fd, &cap_futimes_rights, td);
3357 if ((error = namei(&nd)) != 0)
3360 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3362 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3363 * "Search permission is denied by a component of the path prefix."
3365 NDFREE_NOTHING(&nd);
3366 if ((flags & UTIMENS_EXIT) == 0)
3367 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3373 * Truncate a file given its path name.
3375 #ifndef _SYS_SYSPROTO_H_
3376 struct truncate_args {
3383 sys_truncate(struct thread *td, struct truncate_args *uap)
3386 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3390 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3397 struct nameidata nd;
3403 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3404 if ((error = namei(&nd)) != 0)
3407 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3408 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3409 vn_rangelock_unlock(vp, rl_cookie);
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 vn_rangelock_unlock(vp, rl_cookie);
3431 if (error == ERELOOKUP)
3436 #if defined(COMPAT_43)
3438 * Truncate a file given its path name.
3440 #ifndef _SYS_SYSPROTO_H_
3441 struct otruncate_args {
3447 otruncate(struct thread *td, struct otruncate_args *uap)
3450 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3452 #endif /* COMPAT_43 */
3454 #if defined(COMPAT_FREEBSD6)
3455 /* Versions with the pad argument */
3457 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3460 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3464 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3467 return (kern_ftruncate(td, uap->fd, uap->length));
3472 kern_fsync(struct thread *td, int fd, bool fullsync)
3477 int error, lock_flags;
3480 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3486 /* XXXKIB: compete outstanding aio writes */;
3489 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3492 if (MNT_SHARED_WRITES(mp) ||
3493 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3494 lock_flags = LK_SHARED;
3496 lock_flags = LK_EXCLUSIVE;
3498 vn_lock(vp, lock_flags | LK_RETRY);
3499 AUDIT_ARG_VNODE1(vp);
3500 if (vp->v_object != NULL) {
3501 VM_OBJECT_WLOCK(vp->v_object);
3502 vm_object_page_clean(vp->v_object, 0, 0, 0);
3503 VM_OBJECT_WUNLOCK(vp->v_object);
3505 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3507 vn_finished_write(mp);
3508 if (error == ERELOOKUP)
3516 * Sync an open file.
3518 #ifndef _SYS_SYSPROTO_H_
3524 sys_fsync(struct thread *td, struct fsync_args *uap)
3527 return (kern_fsync(td, uap->fd, true));
3531 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3534 return (kern_fsync(td, uap->fd, false));
3538 * Rename files. Source and destination must either both be directories, or
3539 * both not be directories. If target is a directory, it must be empty.
3541 #ifndef _SYS_SYSPROTO_H_
3542 struct rename_args {
3548 sys_rename(struct thread *td, struct rename_args *uap)
3551 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3552 uap->to, UIO_USERSPACE));
3555 #ifndef _SYS_SYSPROTO_H_
3556 struct renameat_args {
3564 sys_renameat(struct thread *td, struct renameat_args *uap)
3567 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3573 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3574 const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3578 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3579 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3580 if ((error = namei(fromnd)) != 0)
3582 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3583 fromnd->ni_vp, &fromnd->ni_cnd);
3584 VOP_UNLOCK(fromnd->ni_dvp);
3585 if (fromnd->ni_dvp != fromnd->ni_vp)
3586 VOP_UNLOCK(fromnd->ni_vp);
3588 NDFREE(fromnd, NDF_ONLY_PNBUF);
3589 vrele(fromnd->ni_dvp);
3590 vrele(fromnd->ni_vp);
3591 if (fromnd->ni_startdir)
3592 vrele(fromnd->ni_startdir);
3599 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3600 const char *new, enum uio_seg pathseg)
3602 struct mount *mp = NULL;
3603 struct vnode *tvp, *fvp, *tdvp;
3604 struct nameidata fromnd, tond;
3605 u_int64_t tondflags;
3611 if (mac_vnode_check_rename_from_enabled()) {
3612 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3618 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3619 pathseg, old, oldfd, &cap_renameat_source_rights, td);
3620 if ((error = namei(&fromnd)) != 0)
3626 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3627 if (fromnd.ni_vp->v_type == VDIR)
3628 tondflags |= WILLBEDIR;
3629 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3630 &cap_renameat_target_rights, td);
3631 if ((error = namei(&tond)) != 0) {
3632 /* Translate error code for rename("dir1", "dir2/."). */
3633 if (error == EISDIR && fvp->v_type == VDIR)
3635 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3636 vrele(fromnd.ni_dvp);
3642 error = vn_start_write(fvp, &mp, V_NOWAIT);
3644 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3645 NDFREE(&tond, NDF_ONLY_PNBUF);
3652 vrele(fromnd.ni_dvp);
3654 vrele(tond.ni_startdir);
3655 if (fromnd.ni_startdir != NULL)
3656 vrele(fromnd.ni_startdir);
3657 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3663 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3666 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3671 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3673 * If the target already exists we require CAP_UNLINKAT
3674 * from 'newfd', when newfd was used for the lookup.
3676 error = cap_check(&tond.ni_filecaps.fc_rights,
3677 &cap_unlinkat_rights);
3688 * If the source is the same as the destination (that is, if they
3689 * are links to the same vnode), then there is nothing to do.
3695 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3696 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3700 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3701 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3702 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3703 NDFREE(&tond, NDF_ONLY_PNBUF);
3705 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3706 NDFREE(&tond, NDF_ONLY_PNBUF);
3713 vrele(fromnd.ni_dvp);
3716 vrele(tond.ni_startdir);
3717 vn_finished_write(mp);
3719 if (fromnd.ni_startdir)
3720 vrele(fromnd.ni_startdir);
3721 if (error == ERESTART)
3723 if (error == ERELOOKUP)
3729 * Make a directory file.
3731 #ifndef _SYS_SYSPROTO_H_
3738 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3741 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3745 #ifndef _SYS_SYSPROTO_H_
3746 struct mkdirat_args {
3753 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3756 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3760 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3765 struct nameidata nd;
3768 AUDIT_ARG_MODE(mode);
3771 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3772 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3773 segflg, path, fd, &cap_mkdirat_rights, td);
3774 if ((error = namei(&nd)) != 0)
3776 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3777 NDFREE(&nd, NDF_ONLY_PNBUF);
3779 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3784 vattr.va_type = VDIR;
3785 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3787 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3792 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3796 NDFREE(&nd, NDF_ONLY_PNBUF);
3800 vn_finished_write(mp);
3801 if (error == ERELOOKUP)
3807 * Remove a directory file.
3809 #ifndef _SYS_SYSPROTO_H_
3815 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3818 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3823 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3824 enum uio_seg pathseg, int flag)
3829 struct nameidata nd;
3830 cap_rights_t rights;
3834 if (fd != FD_NONE) {
3835 error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3843 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3844 at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
3845 pathseg, path, dfd, &cap_unlinkat_rights, td);
3846 if ((error = namei(&nd)) != 0)
3849 if (vp->v_type != VDIR) {
3854 * No rmdir "." please.
3856 if (nd.ni_dvp == vp) {
3861 * The root of a mounted filesystem cannot be deleted.
3863 if (vp->v_vflag & VV_ROOT) {
3868 if (fp != NULL && fp->f_vnode != vp) {
3869 if (VN_IS_DOOMED(fp->f_vnode))
3877 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3882 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3883 NDFREE(&nd, NDF_ONLY_PNBUF);
3885 if (nd.ni_dvp == vp)
3889 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3893 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3894 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3895 vn_finished_write(mp);
3897 NDFREE(&nd, NDF_ONLY_PNBUF);
3899 if (nd.ni_dvp == vp)
3903 if (error == ERELOOKUP)
3911 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3913 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3914 long *basep, void (*func)(struct freebsd11_dirent *))
3916 struct freebsd11_dirent dstdp;
3917 struct dirent *dp, *edp;
3920 ssize_t resid, ucount;
3923 /* XXX arbitrary sanity limit on `count'. */
3924 count = min(count, 64 * 1024);
3926 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3928 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3936 for (dp = (struct dirent *)dirbuf,
3937 edp = (struct dirent *)&dirbuf[count - resid];
3938 ucount < count && dp < edp; ) {
3939 if (dp->d_reclen == 0)
3941 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3942 if (dp->d_namlen >= sizeof(dstdp.d_name))
3944 dstdp.d_type = dp->d_type;
3945 dstdp.d_namlen = dp->d_namlen;
3946 dstdp.d_fileno = dp->d_fileno; /* truncate */
3947 if (dstdp.d_fileno != dp->d_fileno) {
3948 switch (ino64_trunc_error) {
3956 dstdp.d_fileno = UINT32_MAX;
3960 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3961 ((dp->d_namlen + 1 + 3) &~ 3);
3962 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3963 bzero(dstdp.d_name + dstdp.d_namlen,
3964 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3966 MPASS(dstdp.d_reclen <= dp->d_reclen);
3967 MPASS(ucount + dstdp.d_reclen <= count);
3970 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3973 dp = (struct dirent *)((char *)dp + dp->d_reclen);
3974 ucount += dstdp.d_reclen;
3978 free(dirbuf, M_TEMP);
3980 td->td_retval[0] = ucount;
3987 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3989 #if (BYTE_ORDER == LITTLE_ENDIAN)
3991 * The expected low byte of dp->d_namlen is our dp->d_type.
3992 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3994 dp->d_type = dp->d_namlen;
3998 * The dp->d_type is the high byte of the expected dp->d_namlen,
3999 * so must be zero'ed.
4006 * Read a block of directory entries in a filesystem independent format.
4008 #ifndef _SYS_SYSPROTO_H_
4009 struct ogetdirentries_args {
4017 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4022 error = kern_ogetdirentries(td, uap, &loff);
4024 error = copyout(&loff, uap->basep, sizeof(long));
4029 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4035 /* XXX arbitrary sanity limit on `count'. */
4036 if (uap->count > 64 * 1024)
4039 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4040 &base, ogetdirentries_cvt);
4042 if (error == 0 && uap->basep != NULL)
4043 error = copyout(&base, uap->basep, sizeof(long));
4047 #endif /* COMPAT_43 */
4049 #if defined(COMPAT_FREEBSD11)
4050 #ifndef _SYS_SYSPROTO_H_
4051 struct freebsd11_getdirentries_args {
4059 freebsd11_getdirentries(struct thread *td,
4060 struct freebsd11_getdirentries_args *uap)
4065 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4068 if (error == 0 && uap->basep != NULL)
4069 error = copyout(&base, uap->basep, sizeof(long));
4074 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4076 struct freebsd11_getdirentries_args ap;
4080 ap.count = uap->count;
4082 return (freebsd11_getdirentries(td, &ap));
4084 #endif /* COMPAT_FREEBSD11 */
4087 * Read a block of directory entries in a filesystem independent format.
4090 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4095 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4096 NULL, UIO_USERSPACE);
4099 if (uap->basep != NULL)
4100 error = copyout(&base, uap->basep, sizeof(off_t));
4105 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4106 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4117 if (count > IOSIZE_MAX)
4119 auio.uio_resid = count;
4120 error = getvnode(td, fd, &cap_read_rights, &fp);
4123 if ((fp->f_flag & FREAD) == 0) {
4128 foffset = foffset_lock(fp, 0);
4130 if (vp->v_type != VDIR) {
4134 aiov.iov_base = buf;
4135 aiov.iov_len = count;
4136 auio.uio_iov = &aiov;
4137 auio.uio_iovcnt = 1;
4138 auio.uio_rw = UIO_READ;
4139 auio.uio_segflg = bufseg;
4141 vn_lock(vp, LK_SHARED | LK_RETRY);
4142 AUDIT_ARG_VNODE1(vp);
4143 loff = auio.uio_offset = foffset;
4145 error = mac_vnode_check_readdir(td->td_ucred, vp);
4148 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4150 foffset = auio.uio_offset;
4155 if (count == auio.uio_resid &&
4156 (vp->v_vflag & VV_ROOT) &&
4157 (vp->v_mount->mnt_flag & MNT_UNION)) {
4158 struct vnode *tvp = vp;
4160 vp = vp->v_mount->mnt_vnodecovered;
4170 *residp = auio.uio_resid;
4171 td->td_retval[0] = count - auio.uio_resid;
4173 foffset_unlock(fp, foffset, 0);
4179 * Set the mode mask for creation of filesystem nodes.
4181 #ifndef _SYS_SYSPROTO_H_
4187 sys_umask(struct thread *td, struct umask_args *uap)
4189 struct pwddesc *pdp;
4191 pdp = td->td_proc->p_pd;
4193 td->td_retval[0] = pdp->pd_cmask;
4194 pdp->pd_cmask = uap->newmask & ALLPERMS;
4195 PWDDESC_XUNLOCK(pdp);
4200 * Void all references to file by ripping underlying filesystem away from
4203 #ifndef _SYS_SYSPROTO_H_
4204 struct revoke_args {
4209 sys_revoke(struct thread *td, struct revoke_args *uap)
4213 struct nameidata nd;
4216 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4218 if ((error = namei(&nd)) != 0)
4221 NDFREE_NOTHING(&nd);
4222 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4227 error = mac_vnode_check_revoke(td->td_ucred, vp);
4231 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4234 if (td->td_ucred->cr_uid != vattr.va_uid) {
4235 error = priv_check(td, PRIV_VFS_ADMIN);
4239 if (devfs_usecount(vp) > 0)
4240 VOP_REVOKE(vp, REVOKEALL);
4247 * Convert a user file descriptor to a kernel file entry and check that, if it
4248 * is a capability, the correct rights are present. A reference on the file
4249 * entry is held upon returning.
4252 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4257 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4262 * The file could be not of the vnode type, or it may be not
4263 * yet fully initialized, in which case the f_vnode pointer
4264 * may be set, but f_ops is still badfileops. E.g.,
4265 * devfs_open() transiently create such situation to
4266 * facilitate csw d_fdopen().
4268 * Dupfdopen() handling in kern_openat() installs the
4269 * half-baked file into the process descriptor table, allowing
4270 * other thread to dereference it. Guard against the race by
4273 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4282 * Get an (NFS) file handle.
4284 #ifndef _SYS_SYSPROTO_H_
4285 struct lgetfh_args {
4291 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4294 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4295 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4298 #ifndef _SYS_SYSPROTO_H_
4305 sys_getfh(struct thread *td, struct getfh_args *uap)
4308 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4309 uap->fhp, UIO_USERSPACE));
4313 * syscall for the rpc.lockd to use to translate an open descriptor into
4314 * a NFS file handle.
4316 * warning: do not remove the priv_check() call or this becomes one giant
4319 #ifndef _SYS_SYSPROTO_H_
4320 struct getfhat_args {
4328 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4331 if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
4332 AT_RESOLVE_BENEATH)) != 0)
4334 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4335 uap->fhp, UIO_USERSPACE));
4339 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4340 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4342 struct nameidata nd;
4347 error = priv_check(td, PRIV_VFS_GETFH);
4350 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4351 AT_BENEATH | AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1,
4352 pathseg, path, fd, td);
4356 NDFREE_NOTHING(&nd);
4358 bzero(&fh, sizeof(fh));
4359 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4360 error = VOP_VPTOFH(vp, &fh.fh_fid);
4363 if (fhseg == UIO_USERSPACE)
4364 error = copyout(&fh, fhp, sizeof (fh));
4366 memcpy(fhp, &fh, sizeof(fh));
4371 #ifndef _SYS_SYSPROTO_H_
4372 struct fhlink_args {
4378 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4381 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4384 #ifndef _SYS_SYSPROTO_H_
4385 struct fhlinkat_args {
4392 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4395 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4399 kern_fhlinkat(struct thread *td, int fd, const char *path,
4400 enum uio_seg pathseg, fhandle_t *fhp)
4407 error = priv_check(td, PRIV_VFS_GETFH);
4410 error = copyin(fhp, &fh, sizeof(fh));
4415 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4417 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4422 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4423 } while (error == EAGAIN || error == ERELOOKUP);
4427 #ifndef _SYS_SYSPROTO_H_
4428 struct fhreadlink_args {
4435 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4442 error = priv_check(td, PRIV_VFS_GETFH);
4445 if (uap->bufsize > IOSIZE_MAX)
4447 error = copyin(uap->fhp, &fh, sizeof(fh));
4450 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4452 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4456 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4462 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4465 * warning: do not remove the priv_check() call or this becomes one giant
4468 #ifndef _SYS_SYSPROTO_H_
4469 struct fhopen_args {
4470 const struct fhandle *u_fhp;
4475 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4477 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4481 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4490 error = priv_check(td, PRIV_VFS_FHOPEN);
4494 fmode = FFLAGS(flags);
4495 /* why not allow a non-read/write open for our lockd? */
4496 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4498 error = copyin(u_fhp, &fhp, sizeof(fhp));
4501 /* find the mount point */
4502 mp = vfs_busyfs(&fhp.fh_fsid);
4505 /* now give me my vnode, it gets returned to me locked */
4506 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4511 error = falloc_noinstall(td, &fp);
4517 * An extra reference on `fp' has been held for us by
4518 * falloc_noinstall().
4524 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4526 KASSERT(fp->f_ops == &badfileops,
4527 ("VOP_OPEN in fhopen() set f_ops"));
4528 KASSERT(td->td_dupfd < 0,
4529 ("fhopen() encountered fdopen()"));
4538 finit_vnode(fp, fmode, NULL, &vnops);
4540 if ((fmode & O_TRUNC) != 0) {
4541 error = fo_truncate(fp, 0, td->td_ucred, td);
4546 error = finstall(td, fp, &indx, fmode, NULL);
4549 td->td_retval[0] = indx;
4554 * Stat an (NFS) file handle.
4556 #ifndef _SYS_SYSPROTO_H_
4557 struct fhstat_args {
4558 struct fhandle *u_fhp;
4563 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4569 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4572 error = kern_fhstat(td, fh, &sb);
4574 error = copyout(&sb, uap->sb, sizeof(sb));
4579 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4585 error = priv_check(td, PRIV_VFS_FHSTAT);
4588 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4590 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4594 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4600 * Implement fstatfs() for (NFS) file handles.
4602 #ifndef _SYS_SYSPROTO_H_
4603 struct fhstatfs_args {
4604 struct fhandle *u_fhp;
4609 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4615 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4618 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4619 error = kern_fhstatfs(td, fh, sfp);
4621 error = copyout(sfp, uap->buf, sizeof(*sfp));
4622 free(sfp, M_STATFS);
4627 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4633 error = priv_check(td, PRIV_VFS_FHSTATFS);
4636 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4638 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4644 error = prison_canseemount(td->td_ucred, mp);
4648 error = mac_mount_check_stat(td->td_ucred, mp);
4652 error = VFS_STATFS(mp, buf);
4659 * Unlike madvise(2), we do not make a best effort to remember every
4660 * possible caching hint. Instead, we remember the last setting with
4661 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4662 * region of any current setting.
4665 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4668 struct fadvise_info *fa, *new;
4674 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4676 AUDIT_ARG_VALUE(advice);
4678 case POSIX_FADV_SEQUENTIAL:
4679 case POSIX_FADV_RANDOM:
4680 case POSIX_FADV_NOREUSE:
4681 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4683 case POSIX_FADV_NORMAL:
4684 case POSIX_FADV_WILLNEED:
4685 case POSIX_FADV_DONTNEED:
4691 /* XXX: CAP_POSIX_FADVISE? */
4693 error = fget(td, fd, &cap_no_rights, &fp);
4696 AUDIT_ARG_FILE(td->td_proc, fp);
4697 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4701 if (fp->f_type != DTYPE_VNODE) {
4706 if (vp->v_type != VREG) {
4713 end = offset + len - 1;
4715 case POSIX_FADV_SEQUENTIAL:
4716 case POSIX_FADV_RANDOM:
4717 case POSIX_FADV_NOREUSE:
4719 * Try to merge any existing non-standard region with
4720 * this new region if possible, otherwise create a new
4721 * non-standard region for this request.
4723 mtx_pool_lock(mtxpool_sleep, fp);
4725 if (fa != NULL && fa->fa_advice == advice &&
4726 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4727 (end != OFF_MAX && fa->fa_start == end + 1) ||
4728 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4729 if (offset < fa->fa_start)
4730 fa->fa_start = offset;
4731 if (end > fa->fa_end)
4734 new->fa_advice = advice;
4735 new->fa_start = offset;
4740 mtx_pool_unlock(mtxpool_sleep, fp);
4742 case POSIX_FADV_NORMAL:
4744 * If a the "normal" region overlaps with an existing
4745 * non-standard region, trim or remove the
4746 * non-standard region.
4748 mtx_pool_lock(mtxpool_sleep, fp);
4751 if (offset <= fa->fa_start && end >= fa->fa_end) {
4753 fp->f_advice = NULL;
4754 } else if (offset <= fa->fa_start &&
4755 end >= fa->fa_start)
4756 fa->fa_start = end + 1;
4757 else if (offset <= fa->fa_end && end >= fa->fa_end)
4758 fa->fa_end = offset - 1;
4759 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4761 * If the "normal" region is a middle
4762 * portion of the existing
4763 * non-standard region, just remove
4764 * the whole thing rather than picking
4765 * one side or the other to
4769 fp->f_advice = NULL;
4772 mtx_pool_unlock(mtxpool_sleep, fp);
4774 case POSIX_FADV_WILLNEED:
4775 case POSIX_FADV_DONTNEED:
4776 error = VOP_ADVISE(vp, offset, end, advice);
4782 free(new, M_FADVISE);
4787 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4791 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4793 return (kern_posix_error(td, error));
4797 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4798 off_t *outoffp, size_t len, unsigned int flags)
4800 struct file *infp, *outfp;
4801 struct vnode *invp, *outvp;
4804 void *rl_rcookie, *rl_wcookie;
4805 off_t savinoff, savoutoff;
4807 infp = outfp = NULL;
4808 rl_rcookie = rl_wcookie = NULL;
4817 if (len > SSIZE_MAX)
4819 * Although the len argument is size_t, the return argument
4820 * is ssize_t (which is signed). Therefore a size that won't
4821 * fit in ssize_t can't be returned.
4825 /* Get the file structures for the file descriptors. */
4826 error = fget_read(td, infd, &cap_read_rights, &infp);
4829 if (infp->f_ops == &badfileops) {
4833 if (infp->f_vnode == NULL) {
4837 error = fget_write(td, outfd, &cap_write_rights, &outfp);
4840 if (outfp->f_ops == &badfileops) {
4844 if (outfp->f_vnode == NULL) {
4849 /* Set the offset pointers to the correct place. */
4851 inoffp = &infp->f_offset;
4852 if (outoffp == NULL)
4853 outoffp = &outfp->f_offset;
4855 savoutoff = *outoffp;
4857 invp = infp->f_vnode;
4858 outvp = outfp->f_vnode;
4859 /* Sanity check the f_flag bits. */
4860 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4861 (infp->f_flag & FREAD) == 0) {
4866 /* If len == 0, just return 0. */
4871 * If infp and outfp refer to the same file, the byte ranges cannot
4874 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4875 savoutoff) || (savinoff > savoutoff && savoutoff + len >
4881 /* Range lock the byte ranges for both invp and outvp. */
4883 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4885 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4887 if (rl_rcookie != NULL)
4889 vn_rangelock_unlock(outvp, rl_wcookie);
4890 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4891 vn_rangelock_unlock(invp, rl_rcookie);
4895 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4896 flags, infp->f_cred, outfp->f_cred, td);
4898 if (rl_rcookie != NULL)
4899 vn_rangelock_unlock(invp, rl_rcookie);
4900 if (rl_wcookie != NULL)
4901 vn_rangelock_unlock(outvp, rl_wcookie);
4902 if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4904 *outoffp = savoutoff;
4910 td->td_retval[0] = retlen;
4915 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4917 off_t inoff, outoff, *inoffp, *outoffp;
4920 inoffp = outoffp = NULL;
4921 if (uap->inoffp != NULL) {
4922 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4927 if (uap->outoffp != NULL) {
4928 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4933 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4934 outoffp, uap->len, uap->flags);
4935 if (error == 0 && uap->inoffp != NULL)
4936 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4937 if (error == 0 && uap->outoffp != NULL)
4938 error = copyout(outoffp, uap->outoffp, sizeof(off_t));