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_RESOLVE_BENEATH) != 0)
125 if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
127 /* NOFOLLOW is pseudo flag */
128 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
129 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
132 if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
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 | AUDITVNODE1, pathseg, path, td);
340 mp = vfs_ref_from_vp(nd.ni_vp);
343 return (kern_do_statfs(td, mp, buf));
347 * Get filesystem statistics.
349 #ifndef _SYS_SYSPROTO_H_
350 struct fstatfs_args {
356 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
361 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
362 error = kern_fstatfs(td, uap->fd, sfp);
364 error = copyout(sfp, uap->buf, sizeof(struct statfs));
370 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
378 error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
383 if (AUDITING_TD(td)) {
384 vn_lock(vp, LK_SHARED | LK_RETRY);
385 AUDIT_ARG_VNODE1(vp);
389 mp = vfs_ref_from_vp(vp);
391 return (kern_do_statfs(td, mp, buf));
395 * Get statistics on all filesystems.
397 #ifndef _SYS_SYSPROTO_H_
398 struct getfsstat_args {
405 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
410 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
412 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
413 UIO_USERSPACE, uap->mode);
415 td->td_retval[0] = count;
420 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
421 * The caller is responsible for freeing memory which will be allocated
425 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
426 size_t *countp, enum uio_seg bufseg, int mode)
428 struct mount *mp, *nmp;
429 struct statfs *sfsp, *sp, *sptmp, *tofree;
430 size_t count, maxcount;
438 if (bufseg == UIO_SYSSPACE)
443 maxcount = bufsize / sizeof(struct statfs);
447 } else if (bufseg == UIO_USERSPACE) {
450 } else /* if (bufseg == UIO_SYSSPACE) */ {
452 mtx_lock(&mountlist_mtx);
453 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
456 mtx_unlock(&mountlist_mtx);
457 if (maxcount > count)
459 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
466 * If there is no target buffer they only want the count.
468 * This could be TAILQ_FOREACH but it is open-coded to match the original
472 mtx_lock(&mountlist_mtx);
473 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
474 if (prison_canseemount(td->td_ucred, mp) != 0) {
475 nmp = TAILQ_NEXT(mp, mnt_list);
479 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
480 nmp = TAILQ_NEXT(mp, mnt_list);
485 nmp = TAILQ_NEXT(mp, mnt_list);
487 mtx_unlock(&mountlist_mtx);
493 * They want the entire thing.
495 * Short-circuit the corner case of no room for anything, avoids
502 mtx_lock(&mountlist_mtx);
503 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
504 if (prison_canseemount(td->td_ucred, mp) != 0) {
505 nmp = TAILQ_NEXT(mp, mnt_list);
509 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
510 nmp = TAILQ_NEXT(mp, mnt_list);
514 if (mode == MNT_WAIT) {
515 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
517 * If vfs_busy() failed, and MBF_NOWAIT
518 * wasn't passed, then the mp is gone.
519 * Furthermore, because of MBF_MNTLSTLOCK,
520 * the mountlist_mtx was dropped. We have
521 * no other choice than to start over.
523 mtx_unlock(&mountlist_mtx);
524 free(tofree, M_STATFS);
528 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
529 nmp = TAILQ_NEXT(mp, mnt_list);
535 * If MNT_NOWAIT is specified, do not refresh
538 if (mode != MNT_NOWAIT) {
539 error = VFS_STATFS(mp, sp);
541 mtx_lock(&mountlist_mtx);
542 nmp = TAILQ_NEXT(mp, mnt_list);
547 if (priv_check_cred_vfs_generation(td->td_ucred)) {
548 sptmp = malloc(sizeof(struct statfs), M_STATFS,
551 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
552 prison_enforce_statfs(td->td_ucred, mp, sptmp);
556 if (bufseg == UIO_SYSSPACE) {
557 bcopy(sp, sfsp, sizeof(*sp));
558 free(sptmp, M_STATFS);
559 } else /* if (bufseg == UIO_USERSPACE) */ {
560 error = copyout(sp, sfsp, sizeof(*sp));
561 free(sptmp, M_STATFS);
570 if (count == maxcount) {
575 mtx_lock(&mountlist_mtx);
576 nmp = TAILQ_NEXT(mp, mnt_list);
579 mtx_unlock(&mountlist_mtx);
585 #ifdef COMPAT_FREEBSD4
587 * Get old format filesystem statistics.
589 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
591 #ifndef _SYS_SYSPROTO_H_
592 struct freebsd4_statfs_args {
598 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
604 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
605 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
607 freebsd4_cvtstatfs(sfp, &osb);
608 error = copyout(&osb, uap->buf, sizeof(osb));
615 * Get filesystem statistics.
617 #ifndef _SYS_SYSPROTO_H_
618 struct freebsd4_fstatfs_args {
624 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
630 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
631 error = kern_fstatfs(td, uap->fd, sfp);
633 freebsd4_cvtstatfs(sfp, &osb);
634 error = copyout(&osb, uap->buf, sizeof(osb));
641 * Get statistics on all filesystems.
643 #ifndef _SYS_SYSPROTO_H_
644 struct freebsd4_getfsstat_args {
651 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
653 struct statfs *buf, *sp;
658 if (uap->bufsize < 0)
660 count = uap->bufsize / sizeof(struct ostatfs);
661 if (count > SIZE_MAX / sizeof(struct statfs))
663 size = count * sizeof(struct statfs);
664 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
667 td->td_retval[0] = count;
670 while (count != 0 && error == 0) {
671 freebsd4_cvtstatfs(sp, &osb);
672 error = copyout(&osb, uap->buf, sizeof(osb));
683 * Implement fstatfs() for (NFS) file handles.
685 #ifndef _SYS_SYSPROTO_H_
686 struct freebsd4_fhstatfs_args {
687 struct fhandle *u_fhp;
692 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
699 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
702 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
703 error = kern_fhstatfs(td, fh, sfp);
705 freebsd4_cvtstatfs(sfp, &osb);
706 error = copyout(&osb, uap->buf, sizeof(osb));
713 * Convert a new format statfs structure to an old format statfs structure.
716 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
719 statfs_scale_blocks(nsp, LONG_MAX);
720 bzero(osp, sizeof(*osp));
721 osp->f_bsize = nsp->f_bsize;
722 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
723 osp->f_blocks = nsp->f_blocks;
724 osp->f_bfree = nsp->f_bfree;
725 osp->f_bavail = nsp->f_bavail;
726 osp->f_files = MIN(nsp->f_files, LONG_MAX);
727 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
728 osp->f_owner = nsp->f_owner;
729 osp->f_type = nsp->f_type;
730 osp->f_flags = nsp->f_flags;
731 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
732 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
733 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
734 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
735 strlcpy(osp->f_fstypename, nsp->f_fstypename,
736 MIN(MFSNAMELEN, OMFSNAMELEN));
737 strlcpy(osp->f_mntonname, nsp->f_mntonname,
738 MIN(MNAMELEN, OMNAMELEN));
739 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
740 MIN(MNAMELEN, OMNAMELEN));
741 osp->f_fsid = nsp->f_fsid;
743 #endif /* COMPAT_FREEBSD4 */
745 #if defined(COMPAT_FREEBSD11)
747 * Get old format filesystem statistics.
749 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
752 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
754 struct freebsd11_statfs osb;
758 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
759 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
761 freebsd11_cvtstatfs(sfp, &osb);
762 error = copyout(&osb, uap->buf, sizeof(osb));
769 * Get filesystem statistics.
772 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
774 struct freebsd11_statfs osb;
778 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
779 error = kern_fstatfs(td, uap->fd, sfp);
781 freebsd11_cvtstatfs(sfp, &osb);
782 error = copyout(&osb, uap->buf, sizeof(osb));
789 * Get statistics on all filesystems.
792 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
794 struct freebsd11_statfs osb;
795 struct statfs *buf, *sp;
799 count = uap->bufsize / sizeof(struct ostatfs);
800 size = count * sizeof(struct statfs);
801 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
804 td->td_retval[0] = count;
807 while (count > 0 && error == 0) {
808 freebsd11_cvtstatfs(sp, &osb);
809 error = copyout(&osb, uap->buf, sizeof(osb));
820 * Implement fstatfs() for (NFS) file handles.
823 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
825 struct freebsd11_statfs osb;
830 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
833 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
834 error = kern_fhstatfs(td, fh, sfp);
836 freebsd11_cvtstatfs(sfp, &osb);
837 error = copyout(&osb, uap->buf, sizeof(osb));
844 * Convert a new format statfs structure to an old format statfs structure.
847 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
850 bzero(osp, sizeof(*osp));
851 osp->f_version = FREEBSD11_STATFS_VERSION;
852 osp->f_type = nsp->f_type;
853 osp->f_flags = nsp->f_flags;
854 osp->f_bsize = nsp->f_bsize;
855 osp->f_iosize = nsp->f_iosize;
856 osp->f_blocks = nsp->f_blocks;
857 osp->f_bfree = nsp->f_bfree;
858 osp->f_bavail = nsp->f_bavail;
859 osp->f_files = nsp->f_files;
860 osp->f_ffree = nsp->f_ffree;
861 osp->f_syncwrites = nsp->f_syncwrites;
862 osp->f_asyncwrites = nsp->f_asyncwrites;
863 osp->f_syncreads = nsp->f_syncreads;
864 osp->f_asyncreads = nsp->f_asyncreads;
865 osp->f_namemax = nsp->f_namemax;
866 osp->f_owner = nsp->f_owner;
867 osp->f_fsid = nsp->f_fsid;
868 strlcpy(osp->f_fstypename, nsp->f_fstypename,
869 MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
870 strlcpy(osp->f_mntonname, nsp->f_mntonname,
871 MIN(MNAMELEN, sizeof(osp->f_mntonname)));
872 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
873 MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
875 #endif /* COMPAT_FREEBSD11 */
878 * Change current working directory to a given file descriptor.
880 #ifndef _SYS_SYSPROTO_H_
886 sys_fchdir(struct thread *td, struct fchdir_args *uap)
888 struct vnode *vp, *tdp;
893 AUDIT_ARG_FD(uap->fd);
894 error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
901 vn_lock(vp, LK_SHARED | LK_RETRY);
902 AUDIT_ARG_VNODE1(vp);
903 error = change_dir(vp, td);
904 while (!error && (mp = vp->v_mountedhere) != NULL) {
907 error = VFS_ROOT(mp, LK_SHARED, &tdp);
924 * Change current working directory (``.'').
926 #ifndef _SYS_SYSPROTO_H_
932 sys_chdir(struct thread *td, struct chdir_args *uap)
935 return (kern_chdir(td, uap->path, UIO_USERSPACE));
939 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
944 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
946 if ((error = namei(&nd)) != 0)
948 if ((error = change_dir(nd.ni_vp, td)) != 0) {
953 VOP_UNLOCK(nd.ni_vp);
955 pwd_chdir(td, nd.ni_vp);
960 * Change notion of root (``/'') directory.
962 #ifndef _SYS_SYSPROTO_H_
968 sys_chroot(struct thread *td, struct chroot_args *uap)
973 error = priv_check(td, PRIV_VFS_CHROOT);
976 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
977 UIO_USERSPACE, uap->path, td);
981 error = change_dir(nd.ni_vp, td);
985 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
989 VOP_UNLOCK(nd.ni_vp);
990 error = pwd_chroot(td, nd.ni_vp);
1002 * Common routine for chroot and chdir. Callers must provide a locked vnode
1006 change_dir(struct vnode *vp, struct thread *td)
1012 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1013 if (vp->v_type != VDIR)
1016 error = mac_vnode_check_chdir(td->td_ucred, vp);
1020 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1023 static __inline void
1024 flags_to_rights(int flags, cap_rights_t *rightsp)
1026 if (flags & O_EXEC) {
1027 cap_rights_set_one(rightsp, CAP_FEXECVE);
1031 switch ((flags & O_ACCMODE)) {
1033 cap_rights_set_one(rightsp, CAP_READ);
1036 cap_rights_set_one(rightsp, CAP_READ);
1039 cap_rights_set_one(rightsp, CAP_WRITE);
1040 if (!(flags & (O_APPEND | O_TRUNC)))
1041 cap_rights_set_one(rightsp, CAP_SEEK);
1046 if (flags & O_CREAT)
1047 cap_rights_set_one(rightsp, CAP_CREATE);
1049 if (flags & O_TRUNC)
1050 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1052 if (flags & (O_SYNC | O_FSYNC))
1053 cap_rights_set_one(rightsp, CAP_FSYNC);
1055 if (flags & (O_EXLOCK | O_SHLOCK))
1056 cap_rights_set_one(rightsp, CAP_FLOCK);
1060 * Check permissions, allocate an open file structure, and call the device
1061 * open routine if any.
1063 #ifndef _SYS_SYSPROTO_H_
1071 sys_open(struct thread *td, struct open_args *uap)
1074 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1075 uap->flags, uap->mode));
1078 #ifndef _SYS_SYSPROTO_H_
1079 struct openat_args {
1087 sys_openat(struct thread *td, struct openat_args *uap)
1090 AUDIT_ARG_FD(uap->fd);
1091 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1096 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1097 int flags, int mode)
1099 struct proc *p = td->td_proc;
1100 struct filedesc *fdp;
1101 struct pwddesc *pdp;
1104 struct nameidata nd;
1105 cap_rights_t rights;
1106 int cmode, error, indx;
1112 AUDIT_ARG_FFLAGS(flags);
1113 AUDIT_ARG_MODE(mode);
1114 cap_rights_init_one(&rights, CAP_LOOKUP);
1115 flags_to_rights(flags, &rights);
1118 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1119 * may be specified. On the other hand, for O_PATH any mode
1120 * except O_EXEC is ignored.
1122 if ((flags & O_PATH) != 0) {
1123 flags &= ~(O_CREAT | O_ACCMODE);
1124 } else if ((flags & O_EXEC) != 0) {
1125 if (flags & O_ACCMODE)
1127 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1130 flags = FFLAGS(flags);
1134 * Allocate a file structure. The descriptor to reference it
1135 * is allocated and used by finstall_refed() below.
1137 error = falloc_noinstall(td, &fp);
1140 /* Set the flags early so the finit in devfs can pick them up. */
1141 fp->f_flag = flags & FMASK;
1142 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1143 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1145 td->td_dupfd = -1; /* XXX check for fdopen */
1146 error = vn_open(&nd, &flags, cmode, fp);
1149 * If the vn_open replaced the method vector, something
1150 * wonderous happened deep below and we just pass it up
1151 * pretending we know what we do.
1153 if (error == ENXIO && fp->f_ops != &badfileops) {
1154 MPASS((flags & O_PATH) == 0);
1159 * Handle special fdopen() case. bleh.
1161 * Don't do this for relative (capability) lookups; we don't
1162 * understand exactly what would happen, and we don't think
1163 * that it ever should.
1165 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1166 (error == ENODEV || error == ENXIO) &&
1167 td->td_dupfd >= 0) {
1168 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1177 NDFREE(&nd, NDF_ONLY_PNBUF);
1181 * Store the vnode, for any f_type. Typically, the vnode use
1182 * count is decremented by direct call to vn_closefile() for
1183 * files that switched type in the cdevsw fdopen() method.
1188 * If the file wasn't claimed by devfs bind it to the normal
1189 * vnode operations here.
1191 if (fp->f_ops == &badfileops) {
1192 KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1193 ("Unexpected fifo fp %p vp %p", fp, vp));
1194 if ((flags & O_PATH) != 0) {
1195 finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1196 DTYPE_VNODE, NULL, &path_fileops);
1200 finit_vnode(fp, flags, NULL, &vnops);
1205 if (flags & O_TRUNC) {
1206 error = fo_truncate(fp, 0, td->td_ucred, td);
1212 * If we haven't already installed the FD (for dupfdopen), do so now.
1215 struct filecaps *fcaps;
1218 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1219 fcaps = &nd.ni_filecaps;
1223 error = finstall_refed(td, fp, &indx, flags, fcaps);
1224 /* On success finstall_refed() consumes fcaps. */
1226 filecaps_free(&nd.ni_filecaps);
1230 filecaps_free(&nd.ni_filecaps);
1231 falloc_abort(td, fp);
1234 td->td_retval[0] = indx;
1237 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1238 falloc_abort(td, fp);
1246 #ifndef _SYS_SYSPROTO_H_
1247 struct ocreat_args {
1253 ocreat(struct thread *td, struct ocreat_args *uap)
1256 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1257 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1259 #endif /* COMPAT_43 */
1262 * Create a special file.
1264 #ifndef _SYS_SYSPROTO_H_
1265 struct mknodat_args {
1273 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1276 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1280 #if defined(COMPAT_FREEBSD11)
1282 freebsd11_mknod(struct thread *td,
1283 struct freebsd11_mknod_args *uap)
1286 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1287 uap->mode, uap->dev));
1291 freebsd11_mknodat(struct thread *td,
1292 struct freebsd11_mknodat_args *uap)
1295 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1298 #endif /* COMPAT_FREEBSD11 */
1301 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1302 int mode, dev_t dev)
1307 struct nameidata nd;
1308 int error, whiteout = 0;
1310 AUDIT_ARG_MODE(mode);
1312 switch (mode & S_IFMT) {
1315 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1316 if (error == 0 && dev == VNOVAL)
1320 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1324 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1334 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1335 NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1337 if ((error = namei(&nd)) != 0)
1341 NDFREE(&nd, NDF_ONLY_PNBUF);
1342 if (vp == nd.ni_dvp)
1350 vattr.va_mode = (mode & ALLPERMS) &
1351 ~td->td_proc->p_pd->pd_cmask;
1352 vattr.va_rdev = dev;
1355 switch (mode & S_IFMT) {
1357 vattr.va_type = VCHR;
1360 vattr.va_type = VBLK;
1366 panic("kern_mknod: invalid mode");
1369 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1370 NDFREE(&nd, NDF_ONLY_PNBUF);
1372 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1377 if (error == 0 && !whiteout)
1378 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1379 &nd.ni_cnd, &vattr);
1383 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1385 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1386 &nd.ni_cnd, &vattr);
1389 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1391 vn_finished_write(mp);
1392 NDFREE(&nd, NDF_ONLY_PNBUF);
1393 if (error == ERELOOKUP)
1399 * Create a named pipe.
1401 #ifndef _SYS_SYSPROTO_H_
1402 struct mkfifo_args {
1408 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1411 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1415 #ifndef _SYS_SYSPROTO_H_
1416 struct mkfifoat_args {
1423 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1426 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1431 kern_mkfifoat(struct thread *td, int fd, const char *path,
1432 enum uio_seg pathseg, int mode)
1436 struct nameidata nd;
1439 AUDIT_ARG_MODE(mode);
1442 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1443 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1445 if ((error = namei(&nd)) != 0)
1447 if (nd.ni_vp != NULL) {
1448 NDFREE(&nd, NDF_ONLY_PNBUF);
1449 if (nd.ni_vp == nd.ni_dvp)
1456 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1457 NDFREE(&nd, NDF_ONLY_PNBUF);
1459 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1464 vattr.va_type = VFIFO;
1465 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1467 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1472 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1476 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1477 vn_finished_write(mp);
1478 NDFREE(&nd, NDF_ONLY_PNBUF);
1479 if (error == ERELOOKUP)
1485 * Make a hard file link.
1487 #ifndef _SYS_SYSPROTO_H_
1494 sys_link(struct thread *td, struct link_args *uap)
1497 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1498 UIO_USERSPACE, FOLLOW));
1501 #ifndef _SYS_SYSPROTO_H_
1502 struct linkat_args {
1511 sys_linkat(struct thread *td, struct linkat_args *uap)
1516 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1517 AT_EMPTY_PATH)) != 0)
1520 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1521 UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW |
1522 AT_RESOLVE_BENEATH | AT_EMPTY_PATH)));
1525 int hardlink_check_uid = 0;
1526 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1527 &hardlink_check_uid, 0,
1528 "Unprivileged processes cannot create hard links to files owned by other "
1530 static int hardlink_check_gid = 0;
1531 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1532 &hardlink_check_gid, 0,
1533 "Unprivileged processes cannot create hard links to files owned by other "
1537 can_hardlink(struct vnode *vp, struct ucred *cred)
1542 if (!hardlink_check_uid && !hardlink_check_gid)
1545 error = VOP_GETATTR(vp, &va, cred);
1549 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1550 error = priv_check_cred(cred, PRIV_VFS_LINK);
1555 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1556 error = priv_check_cred(cred, PRIV_VFS_LINK);
1565 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1566 const char *path2, enum uio_seg segflag, int follow)
1568 struct nameidata nd;
1573 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1574 path1, fd1, &cap_linkat_source_rights, td);
1575 if ((error = namei(&nd)) != 0)
1577 NDFREE(&nd, NDF_ONLY_PNBUF);
1578 if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1579 error = priv_check(td, PRIV_VFS_FHOPEN);
1585 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1586 } while (error == EAGAIN || error == ERELOOKUP);
1591 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1592 enum uio_seg segflag)
1594 struct nameidata nd;
1598 if (vp->v_type == VDIR) {
1600 return (EPERM); /* POSIX */
1602 NDINIT_ATRIGHTS(&nd, CREATE,
1603 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1604 &cap_linkat_target_rights, td);
1605 if ((error = namei(&nd)) == 0) {
1606 if (nd.ni_vp != NULL) {
1607 NDFREE(&nd, NDF_ONLY_PNBUF);
1608 if (nd.ni_dvp == nd.ni_vp)
1615 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1617 * Cross-device link. No need to recheck
1618 * vp->v_type, since it cannot change, except
1621 NDFREE(&nd, NDF_ONLY_PNBUF);
1625 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1626 error = can_hardlink(vp, td->td_ucred);
1629 error = mac_vnode_check_link(td->td_ucred,
1630 nd.ni_dvp, vp, &nd.ni_cnd);
1635 NDFREE(&nd, NDF_ONLY_PNBUF);
1638 error = vn_start_write(vp, &mp, V_NOWAIT);
1642 NDFREE(&nd, NDF_ONLY_PNBUF);
1643 error = vn_start_write(NULL, &mp,
1649 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1650 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1651 vn_finished_write(mp);
1652 NDFREE(&nd, NDF_ONLY_PNBUF);
1656 NDFREE(&nd, NDF_ONLY_PNBUF);
1667 * Make a symbolic link.
1669 #ifndef _SYS_SYSPROTO_H_
1670 struct symlink_args {
1676 sys_symlink(struct thread *td, struct symlink_args *uap)
1679 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1683 #ifndef _SYS_SYSPROTO_H_
1684 struct symlinkat_args {
1691 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1694 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1699 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1700 enum uio_seg segflg)
1704 const char *syspath;
1706 struct nameidata nd;
1709 if (segflg == UIO_SYSSPACE) {
1712 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1713 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1717 AUDIT_ARG_TEXT(syspath);
1720 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1721 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1723 if ((error = namei(&nd)) != 0)
1726 NDFREE(&nd, NDF_ONLY_PNBUF);
1727 if (nd.ni_vp == nd.ni_dvp)
1736 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1737 NDFREE(&nd, NDF_ONLY_PNBUF);
1739 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1744 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1746 vattr.va_type = VLNK;
1747 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1752 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1756 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1757 vn_finished_write(mp);
1758 NDFREE(&nd, NDF_ONLY_PNBUF);
1759 if (error == ERELOOKUP)
1762 if (segflg != UIO_SYSSPACE)
1763 uma_zfree(namei_zone, tmppath);
1768 * Delete a whiteout from the filesystem.
1770 #ifndef _SYS_SYSPROTO_H_
1771 struct undelete_args {
1776 sys_undelete(struct thread *td, struct undelete_args *uap)
1779 struct nameidata nd;
1784 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1785 UIO_USERSPACE, uap->path, td);
1790 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1791 NDFREE(&nd, NDF_ONLY_PNBUF);
1792 if (nd.ni_vp == nd.ni_dvp)
1800 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1801 NDFREE(&nd, NDF_ONLY_PNBUF);
1803 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1807 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1808 NDFREE(&nd, NDF_ONLY_PNBUF);
1810 vn_finished_write(mp);
1811 if (error == ERELOOKUP)
1817 * Delete a name from the filesystem.
1819 #ifndef _SYS_SYSPROTO_H_
1820 struct unlink_args {
1825 sys_unlink(struct thread *td, struct unlink_args *uap)
1828 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1833 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1834 int flag, enum uio_seg pathseg, ino_t oldinum)
1837 if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1840 if ((flag & AT_REMOVEDIR) != 0)
1841 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1843 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1846 #ifndef _SYS_SYSPROTO_H_
1847 struct unlinkat_args {
1854 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1857 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1861 #ifndef _SYS_SYSPROTO_H_
1862 struct funlinkat_args {
1870 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1873 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1878 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1879 enum uio_seg pathseg, int flag, ino_t oldinum)
1884 struct nameidata nd;
1889 if (fd != FD_NONE) {
1890 error = getvnode_path(td, fd, &cap_no_rights, &fp);
1897 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1898 at2cnpflags(flag, AT_RESOLVE_BENEATH),
1899 pathseg, path, dfd, &cap_unlinkat_rights, td);
1900 if ((error = namei(&nd)) != 0) {
1901 if (error == EINVAL)
1906 if (vp->v_type == VDIR && oldinum == 0) {
1907 error = EPERM; /* POSIX */
1908 } else if (oldinum != 0 &&
1909 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1910 sb.st_ino != oldinum) {
1911 error = EIDRM; /* Identifier removed */
1912 } else if (fp != NULL && fp->f_vnode != vp) {
1913 if (VN_IS_DOOMED(fp->f_vnode))
1919 * The root of a mounted filesystem cannot be deleted.
1921 * XXX: can this only be a VDIR case?
1923 if (vp->v_vflag & VV_ROOT)
1927 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1928 NDFREE(&nd, NDF_ONLY_PNBUF);
1930 if (vp == nd.ni_dvp)
1934 if ((error = vn_start_write(NULL, &mp,
1935 V_XSLEEP | PCATCH)) != 0) {
1941 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1946 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1947 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1951 vn_finished_write(mp);
1953 NDFREE(&nd, NDF_ONLY_PNBUF);
1955 if (vp == nd.ni_dvp)
1959 if (error == ERELOOKUP)
1968 * Reposition read/write file offset.
1970 #ifndef _SYS_SYSPROTO_H_
1979 sys_lseek(struct thread *td, struct lseek_args *uap)
1982 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1986 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1992 error = fget(td, fd, &cap_seek_rights, &fp);
1995 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1996 fo_seek(fp, offset, whence, td) : ESPIPE;
2001 #if defined(COMPAT_43)
2003 * Reposition read/write file offset.
2005 #ifndef _SYS_SYSPROTO_H_
2006 struct olseek_args {
2013 olseek(struct thread *td, struct olseek_args *uap)
2016 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2018 #endif /* COMPAT_43 */
2020 #if defined(COMPAT_FREEBSD6)
2021 /* Version with the 'pad' argument */
2023 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2026 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2031 * Check access permissions using passed credentials.
2034 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2040 /* Flags == 0 means only check for existence. */
2041 if (user_flags == 0)
2045 if (user_flags & R_OK)
2047 if (user_flags & W_OK)
2049 if (user_flags & X_OK)
2052 error = mac_vnode_check_access(cred, vp, accmode);
2056 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2057 error = VOP_ACCESS(vp, accmode, cred, td);
2062 * Check access permissions using "real" credentials.
2064 #ifndef _SYS_SYSPROTO_H_
2065 struct access_args {
2071 sys_access(struct thread *td, struct access_args *uap)
2074 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2078 #ifndef _SYS_SYSPROTO_H_
2079 struct faccessat_args {
2087 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2090 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2095 kern_accessat(struct thread *td, int fd, const char *path,
2096 enum uio_seg pathseg, int flag, int amode)
2098 struct ucred *cred, *usecred;
2100 struct nameidata nd;
2103 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2105 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2109 * Create and modify a temporary credential instead of one that
2110 * is potentially shared (if we need one).
2112 cred = td->td_ucred;
2113 if ((flag & AT_EACCESS) == 0 &&
2114 ((cred->cr_uid != cred->cr_ruid ||
2115 cred->cr_rgid != cred->cr_groups[0]))) {
2116 usecred = crdup(cred);
2117 usecred->cr_uid = cred->cr_ruid;
2118 usecred->cr_groups[0] = cred->cr_rgid;
2119 td->td_ucred = usecred;
2122 AUDIT_ARG_VALUE(amode);
2123 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2124 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2125 AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights, td);
2126 if ((error = namei(&nd)) != 0)
2130 error = vn_access(vp, amode, usecred, td);
2131 NDFREE_NOTHING(&nd);
2134 if (usecred != cred) {
2135 td->td_ucred = cred;
2142 * Check access permissions using "effective" credentials.
2144 #ifndef _SYS_SYSPROTO_H_
2145 struct eaccess_args {
2151 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2154 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2155 AT_EACCESS, uap->amode));
2158 #if defined(COMPAT_43)
2160 * Get file status; this version follows links.
2162 #ifndef _SYS_SYSPROTO_H_
2169 ostat(struct thread *td, struct ostat_args *uap)
2175 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2180 return (copyout(&osb, uap->ub, sizeof (osb)));
2184 * Get file status; this version does not follow links.
2186 #ifndef _SYS_SYSPROTO_H_
2187 struct olstat_args {
2193 olstat(struct thread *td, struct olstat_args *uap)
2199 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2200 UIO_USERSPACE, &sb, NULL);
2204 return (copyout(&osb, uap->ub, sizeof (osb)));
2208 * Convert from an old to a new stat structure.
2209 * XXX: many values are blindly truncated.
2212 cvtstat(struct stat *st, struct ostat *ost)
2215 bzero(ost, sizeof(*ost));
2216 ost->st_dev = st->st_dev;
2217 ost->st_ino = st->st_ino;
2218 ost->st_mode = st->st_mode;
2219 ost->st_nlink = st->st_nlink;
2220 ost->st_uid = st->st_uid;
2221 ost->st_gid = st->st_gid;
2222 ost->st_rdev = st->st_rdev;
2223 ost->st_size = MIN(st->st_size, INT32_MAX);
2224 ost->st_atim = st->st_atim;
2225 ost->st_mtim = st->st_mtim;
2226 ost->st_ctim = st->st_ctim;
2227 ost->st_blksize = st->st_blksize;
2228 ost->st_blocks = st->st_blocks;
2229 ost->st_flags = st->st_flags;
2230 ost->st_gen = st->st_gen;
2232 #endif /* COMPAT_43 */
2234 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2235 int ino64_trunc_error;
2236 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2237 &ino64_trunc_error, 0,
2238 "Error on truncation of device, file or inode number, or link count");
2241 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2244 ost->st_dev = st->st_dev;
2245 if (ost->st_dev != st->st_dev) {
2246 switch (ino64_trunc_error) {
2249 * Since dev_t is almost raw, don't clamp to the
2250 * maximum for case 2, but ignore the error.
2257 ost->st_ino = st->st_ino;
2258 if (ost->st_ino != st->st_ino) {
2259 switch (ino64_trunc_error) {
2266 ost->st_ino = UINT32_MAX;
2270 ost->st_mode = st->st_mode;
2271 ost->st_nlink = st->st_nlink;
2272 if (ost->st_nlink != st->st_nlink) {
2273 switch (ino64_trunc_error) {
2280 ost->st_nlink = UINT16_MAX;
2284 ost->st_uid = st->st_uid;
2285 ost->st_gid = st->st_gid;
2286 ost->st_rdev = st->st_rdev;
2287 if (ost->st_rdev != st->st_rdev) {
2288 switch (ino64_trunc_error) {
2295 ost->st_atim = st->st_atim;
2296 ost->st_mtim = st->st_mtim;
2297 ost->st_ctim = st->st_ctim;
2298 ost->st_size = st->st_size;
2299 ost->st_blocks = st->st_blocks;
2300 ost->st_blksize = st->st_blksize;
2301 ost->st_flags = st->st_flags;
2302 ost->st_gen = st->st_gen;
2304 ost->st_birthtim = st->st_birthtim;
2305 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2306 sizeof(*ost) - offsetof(struct freebsd11_stat,
2307 st_birthtim) - sizeof(ost->st_birthtim));
2312 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2315 struct freebsd11_stat osb;
2318 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2322 error = freebsd11_cvtstat(&sb, &osb);
2324 error = copyout(&osb, uap->ub, sizeof(osb));
2329 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2332 struct freebsd11_stat osb;
2335 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2336 UIO_USERSPACE, &sb, NULL);
2339 error = freebsd11_cvtstat(&sb, &osb);
2341 error = copyout(&osb, uap->ub, sizeof(osb));
2346 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2350 struct freebsd11_stat osb;
2353 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2356 error = kern_fhstat(td, fh, &sb);
2359 error = freebsd11_cvtstat(&sb, &osb);
2361 error = copyout(&osb, uap->sb, sizeof(osb));
2366 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2369 struct freebsd11_stat osb;
2372 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2373 UIO_USERSPACE, &sb, NULL);
2376 error = freebsd11_cvtstat(&sb, &osb);
2378 error = copyout(&osb, uap->buf, sizeof(osb));
2381 #endif /* COMPAT_FREEBSD11 */
2386 #ifndef _SYS_SYSPROTO_H_
2387 struct fstatat_args {
2395 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2400 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2401 UIO_USERSPACE, &sb, NULL);
2403 error = copyout(&sb, uap->buf, sizeof (sb));
2408 kern_statat(struct thread *td, int flag, int fd, const char *path,
2409 enum uio_seg pathseg, struct stat *sbp,
2410 void (*hook)(struct vnode *vp, struct stat *sbp))
2412 struct nameidata nd;
2415 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2416 AT_EMPTY_PATH)) != 0)
2419 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2420 AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2421 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2423 if ((error = namei(&nd)) != 0)
2425 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2427 if (__predict_false(hook != NULL))
2428 hook(nd.ni_vp, sbp);
2430 NDFREE_NOTHING(&nd);
2432 #ifdef __STAT_TIME_T_EXT
2433 sbp->st_atim_ext = 0;
2434 sbp->st_mtim_ext = 0;
2435 sbp->st_ctim_ext = 0;
2436 sbp->st_btim_ext = 0;
2439 if (KTRPOINT(td, KTR_STRUCT))
2440 ktrstat_error(sbp, error);
2445 #if defined(COMPAT_FREEBSD11)
2447 * Implementation of the NetBSD [l]stat() functions.
2450 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2453 bzero(nsb, sizeof(*nsb));
2454 nsb->st_dev = sb->st_dev;
2455 nsb->st_ino = sb->st_ino;
2456 nsb->st_mode = sb->st_mode;
2457 nsb->st_nlink = sb->st_nlink;
2458 nsb->st_uid = sb->st_uid;
2459 nsb->st_gid = sb->st_gid;
2460 nsb->st_rdev = sb->st_rdev;
2461 nsb->st_atim = sb->st_atim;
2462 nsb->st_mtim = sb->st_mtim;
2463 nsb->st_ctim = sb->st_ctim;
2464 nsb->st_size = sb->st_size;
2465 nsb->st_blocks = sb->st_blocks;
2466 nsb->st_blksize = sb->st_blksize;
2467 nsb->st_flags = sb->st_flags;
2468 nsb->st_gen = sb->st_gen;
2469 nsb->st_birthtim = sb->st_birthtim;
2472 #ifndef _SYS_SYSPROTO_H_
2473 struct freebsd11_nstat_args {
2479 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2485 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2489 freebsd11_cvtnstat(&sb, &nsb);
2490 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2494 * NetBSD lstat. Get file status; this version does not follow links.
2496 #ifndef _SYS_SYSPROTO_H_
2497 struct freebsd11_nlstat_args {
2503 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2509 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2510 UIO_USERSPACE, &sb, NULL);
2513 freebsd11_cvtnstat(&sb, &nsb);
2514 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2516 #endif /* COMPAT_FREEBSD11 */
2519 * Get configurable pathname variables.
2521 #ifndef _SYS_SYSPROTO_H_
2522 struct pathconf_args {
2528 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2533 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2536 td->td_retval[0] = value;
2540 #ifndef _SYS_SYSPROTO_H_
2541 struct lpathconf_args {
2547 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2552 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2555 td->td_retval[0] = value;
2560 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2561 int name, u_long flags, long *valuep)
2563 struct nameidata nd;
2566 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2568 if ((error = namei(&nd)) != 0)
2570 NDFREE_NOTHING(&nd);
2572 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2578 * Return target name of a symbolic link.
2580 #ifndef _SYS_SYSPROTO_H_
2581 struct readlink_args {
2588 sys_readlink(struct thread *td, struct readlink_args *uap)
2591 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2592 uap->buf, UIO_USERSPACE, uap->count));
2594 #ifndef _SYS_SYSPROTO_H_
2595 struct readlinkat_args {
2603 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2606 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2607 uap->buf, UIO_USERSPACE, uap->bufsize));
2611 kern_readlinkat(struct thread *td, int fd, const char *path,
2612 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2615 struct nameidata nd;
2618 if (count > IOSIZE_MAX)
2621 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2622 pathseg, path, fd, td);
2624 if ((error = namei(&nd)) != 0)
2626 NDFREE_NOTHING(&nd);
2629 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2636 * Helper function to readlink from a vnode
2639 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2646 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2648 error = mac_vnode_check_readlink(td->td_ucred, vp);
2652 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2655 aiov.iov_base = buf;
2656 aiov.iov_len = count;
2657 auio.uio_iov = &aiov;
2658 auio.uio_iovcnt = 1;
2659 auio.uio_offset = 0;
2660 auio.uio_rw = UIO_READ;
2661 auio.uio_segflg = bufseg;
2663 auio.uio_resid = count;
2664 error = VOP_READLINK(vp, &auio, td->td_ucred);
2665 td->td_retval[0] = count - auio.uio_resid;
2670 * Common implementation code for chflags() and fchflags().
2673 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2679 /* We can't support the value matching VNOVAL. */
2680 if (flags == VNOVAL)
2681 return (EOPNOTSUPP);
2684 * Prevent non-root users from setting flags on devices. When
2685 * a device is reused, users can retain ownership of the device
2686 * if they are allowed to set flags and programs assume that
2687 * chown can't fail when done as root.
2689 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2690 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2695 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2698 vattr.va_flags = flags;
2699 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2701 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2704 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2706 vn_finished_write(mp);
2711 * Change flags of a file given a path name.
2713 #ifndef _SYS_SYSPROTO_H_
2714 struct chflags_args {
2720 sys_chflags(struct thread *td, struct chflags_args *uap)
2723 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2727 #ifndef _SYS_SYSPROTO_H_
2728 struct chflagsat_args {
2736 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2739 if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2740 AT_EMPTY_PATH)) != 0)
2743 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2744 uap->flags, uap->atflag));
2748 * Same as chflags() but doesn't follow symlinks.
2750 #ifndef _SYS_SYSPROTO_H_
2751 struct lchflags_args {
2757 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2760 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2761 uap->flags, AT_SYMLINK_NOFOLLOW));
2765 kern_chflagsat(struct thread *td, int fd, const char *path,
2766 enum uio_seg pathseg, u_long flags, int atflag)
2768 struct nameidata nd;
2771 AUDIT_ARG_FFLAGS(flags);
2772 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2773 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2774 fd, &cap_fchflags_rights, td);
2775 if ((error = namei(&nd)) != 0)
2777 NDFREE_NOTHING(&nd);
2778 error = setfflags(td, nd.ni_vp, flags);
2784 * Change flags of a file given a file descriptor.
2786 #ifndef _SYS_SYSPROTO_H_
2787 struct fchflags_args {
2793 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2798 AUDIT_ARG_FD(uap->fd);
2799 AUDIT_ARG_FFLAGS(uap->flags);
2800 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2805 if (AUDITING_TD(td)) {
2806 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2807 AUDIT_ARG_VNODE1(fp->f_vnode);
2808 VOP_UNLOCK(fp->f_vnode);
2811 error = setfflags(td, fp->f_vnode, uap->flags);
2817 * Common implementation code for chmod(), lchmod() and fchmod().
2820 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2826 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2828 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2830 vattr.va_mode = mode & ALLPERMS;
2832 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2835 error = VOP_SETATTR(vp, &vattr, cred);
2837 vn_finished_write(mp);
2842 * Change mode of a file given path name.
2844 #ifndef _SYS_SYSPROTO_H_
2851 sys_chmod(struct thread *td, struct chmod_args *uap)
2854 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2858 #ifndef _SYS_SYSPROTO_H_
2859 struct fchmodat_args {
2867 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2870 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2871 AT_EMPTY_PATH)) != 0)
2874 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2875 uap->mode, uap->flag));
2879 * Change mode of a file given path name (don't follow links.)
2881 #ifndef _SYS_SYSPROTO_H_
2882 struct lchmod_args {
2888 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2891 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2892 uap->mode, AT_SYMLINK_NOFOLLOW));
2896 kern_fchmodat(struct thread *td, int fd, const char *path,
2897 enum uio_seg pathseg, mode_t mode, int flag)
2899 struct nameidata nd;
2902 AUDIT_ARG_MODE(mode);
2903 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2904 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2905 fd, &cap_fchmod_rights, td);
2906 if ((error = namei(&nd)) != 0)
2908 NDFREE_NOTHING(&nd);
2909 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2915 * Change mode of a file given a file descriptor.
2917 #ifndef _SYS_SYSPROTO_H_
2918 struct fchmod_args {
2924 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2929 AUDIT_ARG_FD(uap->fd);
2930 AUDIT_ARG_MODE(uap->mode);
2932 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2935 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2941 * Common implementation for chown(), lchown(), and fchown()
2944 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2951 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2953 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2958 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2962 error = VOP_SETATTR(vp, &vattr, cred);
2964 vn_finished_write(mp);
2969 * Set ownership given a path name.
2971 #ifndef _SYS_SYSPROTO_H_
2979 sys_chown(struct thread *td, struct chown_args *uap)
2982 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2986 #ifndef _SYS_SYSPROTO_H_
2987 struct fchownat_args {
2996 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2999 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3000 AT_EMPTY_PATH)) != 0)
3003 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3004 uap->gid, uap->flag));
3008 kern_fchownat(struct thread *td, int fd, const char *path,
3009 enum uio_seg pathseg, int uid, int gid, int flag)
3011 struct nameidata nd;
3014 AUDIT_ARG_OWNER(uid, gid);
3015 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3016 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3017 fd, &cap_fchown_rights, td);
3019 if ((error = namei(&nd)) != 0)
3021 NDFREE_NOTHING(&nd);
3022 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3028 * Set ownership given a path name, do not cross symlinks.
3030 #ifndef _SYS_SYSPROTO_H_
3031 struct lchown_args {
3038 sys_lchown(struct thread *td, struct lchown_args *uap)
3041 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3042 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3046 * Set ownership given a file descriptor.
3048 #ifndef _SYS_SYSPROTO_H_
3049 struct fchown_args {
3056 sys_fchown(struct thread *td, struct fchown_args *uap)
3061 AUDIT_ARG_FD(uap->fd);
3062 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3063 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3066 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3072 * Common implementation code for utimes(), lutimes(), and futimes().
3075 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3076 struct timespec *tsp)
3078 struct timeval tv[2];
3079 const struct timeval *tvp;
3082 if (usrtvp == NULL) {
3083 vfs_timestamp(&tsp[0]);
3086 if (tvpseg == UIO_SYSSPACE) {
3089 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3094 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3095 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3097 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3098 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3104 * Common implementation code for futimens(), utimensat().
3106 #define UTIMENS_NULL 0x1
3107 #define UTIMENS_EXIT 0x2
3109 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3110 struct timespec *tsp, int *retflags)
3112 struct timespec tsnow;
3115 vfs_timestamp(&tsnow);
3117 if (usrtsp == NULL) {
3120 *retflags |= UTIMENS_NULL;
3123 if (tspseg == UIO_SYSSPACE) {
3126 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3128 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3129 *retflags |= UTIMENS_EXIT;
3130 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3131 *retflags |= UTIMENS_NULL;
3132 if (tsp[0].tv_nsec == UTIME_OMIT)
3133 tsp[0].tv_sec = VNOVAL;
3134 else if (tsp[0].tv_nsec == UTIME_NOW)
3136 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3138 if (tsp[1].tv_nsec == UTIME_OMIT)
3139 tsp[1].tv_sec = VNOVAL;
3140 else if (tsp[1].tv_nsec == UTIME_NOW)
3142 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3149 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3153 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3154 int numtimes, int nullflag)
3158 int error, setbirthtime;
3160 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3162 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3164 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3165 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3168 vattr.va_atime = ts[0];
3169 vattr.va_mtime = ts[1];
3171 vattr.va_birthtime = ts[1];
3173 vattr.va_birthtime = ts[2];
3175 vattr.va_vaflags |= VA_UTIMES_NULL;
3177 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3181 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3183 vn_finished_write(mp);
3188 * Set the access and modification times of a file.
3190 #ifndef _SYS_SYSPROTO_H_
3191 struct utimes_args {
3193 struct timeval *tptr;
3197 sys_utimes(struct thread *td, struct utimes_args *uap)
3200 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3201 uap->tptr, UIO_USERSPACE));
3204 #ifndef _SYS_SYSPROTO_H_
3205 struct futimesat_args {
3208 const struct timeval * times;
3212 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3215 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3216 uap->times, UIO_USERSPACE));
3220 kern_utimesat(struct thread *td, int fd, const char *path,
3221 enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3223 struct nameidata nd;
3224 struct timespec ts[2];
3227 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3229 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3230 &cap_futimes_rights, td);
3232 if ((error = namei(&nd)) != 0)
3234 NDFREE_NOTHING(&nd);
3235 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3241 * Set the access and modification times of a file.
3243 #ifndef _SYS_SYSPROTO_H_
3244 struct lutimes_args {
3246 struct timeval *tptr;
3250 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3253 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3258 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3259 struct timeval *tptr, enum uio_seg tptrseg)
3261 struct timespec ts[2];
3262 struct nameidata nd;
3265 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3267 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3268 if ((error = namei(&nd)) != 0)
3270 NDFREE_NOTHING(&nd);
3271 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3277 * Set the access and modification times of a file.
3279 #ifndef _SYS_SYSPROTO_H_
3280 struct futimes_args {
3282 struct timeval *tptr;
3286 sys_futimes(struct thread *td, struct futimes_args *uap)
3289 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3293 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3294 enum uio_seg tptrseg)
3296 struct timespec ts[2];
3301 error = getutimes(tptr, tptrseg, ts);
3304 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3308 if (AUDITING_TD(td)) {
3309 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3310 AUDIT_ARG_VNODE1(fp->f_vnode);
3311 VOP_UNLOCK(fp->f_vnode);
3314 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3320 sys_futimens(struct thread *td, struct futimens_args *uap)
3323 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3327 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3328 enum uio_seg tptrseg)
3330 struct timespec ts[2];
3335 error = getutimens(tptr, tptrseg, ts, &flags);
3338 if (flags & UTIMENS_EXIT)
3340 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3344 if (AUDITING_TD(td)) {
3345 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3346 AUDIT_ARG_VNODE1(fp->f_vnode);
3347 VOP_UNLOCK(fp->f_vnode);
3350 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3356 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3359 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3360 uap->times, UIO_USERSPACE, uap->flag));
3364 kern_utimensat(struct thread *td, int fd, const char *path,
3365 enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3368 struct nameidata nd;
3369 struct timespec ts[2];
3372 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3373 AT_EMPTY_PATH)) != 0)
3376 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3378 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3379 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3380 pathseg, path, fd, &cap_futimes_rights, td);
3381 if ((error = namei(&nd)) != 0)
3384 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3386 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3387 * "Search permission is denied by a component of the path prefix."
3389 NDFREE_NOTHING(&nd);
3390 if ((flags & UTIMENS_EXIT) == 0)
3391 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3397 * Truncate a file given its path name.
3399 #ifndef _SYS_SYSPROTO_H_
3400 struct truncate_args {
3407 sys_truncate(struct thread *td, struct truncate_args *uap)
3410 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3414 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3421 struct nameidata nd;
3427 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3428 if ((error = namei(&nd)) != 0)
3431 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3432 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3433 vn_rangelock_unlock(vp, rl_cookie);
3437 NDFREE(&nd, NDF_ONLY_PNBUF);
3438 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3439 if (vp->v_type == VDIR)
3442 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3445 else if ((error = vn_writechk(vp)) == 0 &&
3446 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3448 vattr.va_size = length;
3449 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3452 vn_finished_write(mp);
3453 vn_rangelock_unlock(vp, rl_cookie);
3455 if (error == ERELOOKUP)
3460 #if defined(COMPAT_43)
3462 * Truncate a file given its path name.
3464 #ifndef _SYS_SYSPROTO_H_
3465 struct otruncate_args {
3471 otruncate(struct thread *td, struct otruncate_args *uap)
3474 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3476 #endif /* COMPAT_43 */
3478 #if defined(COMPAT_FREEBSD6)
3479 /* Versions with the pad argument */
3481 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3484 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3488 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3491 return (kern_ftruncate(td, uap->fd, uap->length));
3496 kern_fsync(struct thread *td, int fd, bool fullsync)
3501 int error, lock_flags;
3504 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3510 /* XXXKIB: compete outstanding aio writes */;
3513 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3516 if (MNT_SHARED_WRITES(mp) ||
3517 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3518 lock_flags = LK_SHARED;
3520 lock_flags = LK_EXCLUSIVE;
3522 vn_lock(vp, lock_flags | LK_RETRY);
3523 AUDIT_ARG_VNODE1(vp);
3524 if (vp->v_object != NULL) {
3525 VM_OBJECT_WLOCK(vp->v_object);
3526 vm_object_page_clean(vp->v_object, 0, 0, 0);
3527 VM_OBJECT_WUNLOCK(vp->v_object);
3529 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3531 vn_finished_write(mp);
3532 if (error == ERELOOKUP)
3540 * Sync an open file.
3542 #ifndef _SYS_SYSPROTO_H_
3548 sys_fsync(struct thread *td, struct fsync_args *uap)
3551 return (kern_fsync(td, uap->fd, true));
3555 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3558 return (kern_fsync(td, uap->fd, false));
3562 * Rename files. Source and destination must either both be directories, or
3563 * both not be directories. If target is a directory, it must be empty.
3565 #ifndef _SYS_SYSPROTO_H_
3566 struct rename_args {
3572 sys_rename(struct thread *td, struct rename_args *uap)
3575 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3576 uap->to, UIO_USERSPACE));
3579 #ifndef _SYS_SYSPROTO_H_
3580 struct renameat_args {
3588 sys_renameat(struct thread *td, struct renameat_args *uap)
3591 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3597 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3598 const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3602 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3603 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3604 if ((error = namei(fromnd)) != 0)
3606 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3607 fromnd->ni_vp, &fromnd->ni_cnd);
3608 VOP_UNLOCK(fromnd->ni_dvp);
3609 if (fromnd->ni_dvp != fromnd->ni_vp)
3610 VOP_UNLOCK(fromnd->ni_vp);
3612 NDFREE(fromnd, NDF_ONLY_PNBUF);
3613 vrele(fromnd->ni_dvp);
3614 vrele(fromnd->ni_vp);
3615 if (fromnd->ni_startdir)
3616 vrele(fromnd->ni_startdir);
3623 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3624 const char *new, enum uio_seg pathseg)
3626 struct mount *mp = NULL;
3627 struct vnode *tvp, *fvp, *tdvp;
3628 struct nameidata fromnd, tond;
3629 u_int64_t tondflags;
3635 if (mac_vnode_check_rename_from_enabled()) {
3636 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3642 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3643 pathseg, old, oldfd, &cap_renameat_source_rights, td);
3644 if ((error = namei(&fromnd)) != 0)
3650 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3651 if (fromnd.ni_vp->v_type == VDIR)
3652 tondflags |= WILLBEDIR;
3653 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3654 &cap_renameat_target_rights, td);
3655 if ((error = namei(&tond)) != 0) {
3656 /* Translate error code for rename("dir1", "dir2/."). */
3657 if (error == EISDIR && fvp->v_type == VDIR)
3659 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3660 vrele(fromnd.ni_dvp);
3666 error = vn_start_write(fvp, &mp, V_NOWAIT);
3668 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3669 NDFREE(&tond, NDF_ONLY_PNBUF);
3676 vrele(fromnd.ni_dvp);
3678 vrele(tond.ni_startdir);
3679 if (fromnd.ni_startdir != NULL)
3680 vrele(fromnd.ni_startdir);
3681 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3687 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3690 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3695 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3697 * If the target already exists we require CAP_UNLINKAT
3698 * from 'newfd', when newfd was used for the lookup.
3700 error = cap_check(&tond.ni_filecaps.fc_rights,
3701 &cap_unlinkat_rights);
3712 * If the source is the same as the destination (that is, if they
3713 * are links to the same vnode), then there is nothing to do.
3719 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3720 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3724 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3725 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3726 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3727 NDFREE(&tond, NDF_ONLY_PNBUF);
3729 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3730 NDFREE(&tond, NDF_ONLY_PNBUF);
3737 vrele(fromnd.ni_dvp);
3740 vrele(tond.ni_startdir);
3741 vn_finished_write(mp);
3743 if (fromnd.ni_startdir)
3744 vrele(fromnd.ni_startdir);
3745 if (error == ERESTART)
3747 if (error == ERELOOKUP)
3753 * Make a directory file.
3755 #ifndef _SYS_SYSPROTO_H_
3762 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3765 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3769 #ifndef _SYS_SYSPROTO_H_
3770 struct mkdirat_args {
3777 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3780 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3784 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3789 struct nameidata nd;
3792 AUDIT_ARG_MODE(mode);
3795 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3796 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3797 segflg, path, fd, &cap_mkdirat_rights, td);
3798 if ((error = namei(&nd)) != 0)
3800 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3801 NDFREE(&nd, NDF_ONLY_PNBUF);
3803 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3808 vattr.va_type = VDIR;
3809 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3811 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3816 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3820 NDFREE(&nd, NDF_ONLY_PNBUF);
3821 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3822 vn_finished_write(mp);
3823 if (error == ERELOOKUP)
3829 * Remove a directory file.
3831 #ifndef _SYS_SYSPROTO_H_
3837 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3840 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3845 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3846 enum uio_seg pathseg, int flag)
3851 struct nameidata nd;
3852 cap_rights_t rights;
3856 if (fd != FD_NONE) {
3857 error = getvnode(td, fd, cap_rights_init_one(&rights,
3865 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3866 at2cnpflags(flag, AT_RESOLVE_BENEATH),
3867 pathseg, path, dfd, &cap_unlinkat_rights, td);
3868 if ((error = namei(&nd)) != 0)
3871 if (vp->v_type != VDIR) {
3876 * No rmdir "." please.
3878 if (nd.ni_dvp == vp) {
3883 * The root of a mounted filesystem cannot be deleted.
3885 if (vp->v_vflag & VV_ROOT) {
3890 if (fp != NULL && fp->f_vnode != vp) {
3891 if (VN_IS_DOOMED(fp->f_vnode))
3899 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3904 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3905 NDFREE(&nd, NDF_ONLY_PNBUF);
3907 if (nd.ni_dvp == vp)
3911 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3915 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3916 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3917 vn_finished_write(mp);
3919 NDFREE(&nd, NDF_ONLY_PNBUF);
3921 if (nd.ni_dvp == vp)
3925 if (error == ERELOOKUP)
3933 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3935 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3936 long *basep, void (*func)(struct freebsd11_dirent *))
3938 struct freebsd11_dirent dstdp;
3939 struct dirent *dp, *edp;
3942 ssize_t resid, ucount;
3945 /* XXX arbitrary sanity limit on `count'. */
3946 count = min(count, 64 * 1024);
3948 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3950 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3958 for (dp = (struct dirent *)dirbuf,
3959 edp = (struct dirent *)&dirbuf[count - resid];
3960 ucount < count && dp < edp; ) {
3961 if (dp->d_reclen == 0)
3963 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3964 if (dp->d_namlen >= sizeof(dstdp.d_name))
3966 dstdp.d_type = dp->d_type;
3967 dstdp.d_namlen = dp->d_namlen;
3968 dstdp.d_fileno = dp->d_fileno; /* truncate */
3969 if (dstdp.d_fileno != dp->d_fileno) {
3970 switch (ino64_trunc_error) {
3978 dstdp.d_fileno = UINT32_MAX;
3982 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3983 ((dp->d_namlen + 1 + 3) &~ 3);
3984 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3985 bzero(dstdp.d_name + dstdp.d_namlen,
3986 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3988 MPASS(dstdp.d_reclen <= dp->d_reclen);
3989 MPASS(ucount + dstdp.d_reclen <= count);
3992 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3995 dp = (struct dirent *)((char *)dp + dp->d_reclen);
3996 ucount += dstdp.d_reclen;
4000 free(dirbuf, M_TEMP);
4002 td->td_retval[0] = ucount;
4009 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4011 #if (BYTE_ORDER == LITTLE_ENDIAN)
4013 * The expected low byte of dp->d_namlen is our dp->d_type.
4014 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4016 dp->d_type = dp->d_namlen;
4020 * The dp->d_type is the high byte of the expected dp->d_namlen,
4021 * so must be zero'ed.
4028 * Read a block of directory entries in a filesystem independent format.
4030 #ifndef _SYS_SYSPROTO_H_
4031 struct ogetdirentries_args {
4039 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4044 error = kern_ogetdirentries(td, uap, &loff);
4046 error = copyout(&loff, uap->basep, sizeof(long));
4051 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4057 /* XXX arbitrary sanity limit on `count'. */
4058 if (uap->count > 64 * 1024)
4061 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4062 &base, ogetdirentries_cvt);
4064 if (error == 0 && uap->basep != NULL)
4065 error = copyout(&base, uap->basep, sizeof(long));
4069 #endif /* COMPAT_43 */
4071 #if defined(COMPAT_FREEBSD11)
4072 #ifndef _SYS_SYSPROTO_H_
4073 struct freebsd11_getdirentries_args {
4081 freebsd11_getdirentries(struct thread *td,
4082 struct freebsd11_getdirentries_args *uap)
4087 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4090 if (error == 0 && uap->basep != NULL)
4091 error = copyout(&base, uap->basep, sizeof(long));
4096 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4098 struct freebsd11_getdirentries_args ap;
4102 ap.count = uap->count;
4104 return (freebsd11_getdirentries(td, &ap));
4106 #endif /* COMPAT_FREEBSD11 */
4109 * Read a block of directory entries in a filesystem independent format.
4112 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4117 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4118 NULL, UIO_USERSPACE);
4121 if (uap->basep != NULL)
4122 error = copyout(&base, uap->basep, sizeof(off_t));
4127 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4128 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4139 if (count > IOSIZE_MAX)
4141 auio.uio_resid = count;
4142 error = getvnode(td, fd, &cap_read_rights, &fp);
4145 if ((fp->f_flag & FREAD) == 0) {
4150 foffset = foffset_lock(fp, 0);
4152 if (vp->v_type != VDIR) {
4156 aiov.iov_base = buf;
4157 aiov.iov_len = count;
4158 auio.uio_iov = &aiov;
4159 auio.uio_iovcnt = 1;
4160 auio.uio_rw = UIO_READ;
4161 auio.uio_segflg = bufseg;
4163 vn_lock(vp, LK_SHARED | LK_RETRY);
4164 AUDIT_ARG_VNODE1(vp);
4165 loff = auio.uio_offset = foffset;
4167 error = mac_vnode_check_readdir(td->td_ucred, vp);
4170 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4172 foffset = auio.uio_offset;
4177 if (count == auio.uio_resid &&
4178 (vp->v_vflag & VV_ROOT) &&
4179 (vp->v_mount->mnt_flag & MNT_UNION)) {
4180 struct vnode *tvp = vp;
4182 vp = vp->v_mount->mnt_vnodecovered;
4192 *residp = auio.uio_resid;
4193 td->td_retval[0] = count - auio.uio_resid;
4195 foffset_unlock(fp, foffset, 0);
4201 * Set the mode mask for creation of filesystem nodes.
4203 #ifndef _SYS_SYSPROTO_H_
4209 sys_umask(struct thread *td, struct umask_args *uap)
4211 struct pwddesc *pdp;
4213 pdp = td->td_proc->p_pd;
4215 td->td_retval[0] = pdp->pd_cmask;
4216 pdp->pd_cmask = uap->newmask & ALLPERMS;
4217 PWDDESC_XUNLOCK(pdp);
4222 * Void all references to file by ripping underlying filesystem away from
4225 #ifndef _SYS_SYSPROTO_H_
4226 struct revoke_args {
4231 sys_revoke(struct thread *td, struct revoke_args *uap)
4235 struct nameidata nd;
4238 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4240 if ((error = namei(&nd)) != 0)
4243 NDFREE_NOTHING(&nd);
4244 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4249 error = mac_vnode_check_revoke(td->td_ucred, vp);
4253 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4256 if (td->td_ucred->cr_uid != vattr.va_uid) {
4257 error = priv_check(td, PRIV_VFS_ADMIN);
4261 if (devfs_usecount(vp) > 0)
4262 VOP_REVOKE(vp, REVOKEALL);
4269 * This variant of getvnode() allows O_PATH files. Caller should
4270 * ensure that returned file and vnode are only used for compatible
4274 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4280 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4285 * The file could be not of the vnode type, or it may be not
4286 * yet fully initialized, in which case the f_vnode pointer
4287 * may be set, but f_ops is still badfileops. E.g.,
4288 * devfs_open() transiently create such situation to
4289 * facilitate csw d_fdopen().
4291 * Dupfdopen() handling in kern_openat() installs the
4292 * half-baked file into the process descriptor table, allowing
4293 * other thread to dereference it. Guard against the race by
4296 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4306 * Convert a user file descriptor to a kernel file entry and check
4307 * that, if it is a capability, the correct rights are present.
4308 * A reference on the file entry is held upon returning.
4311 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4315 error = getvnode_path(td, fd, rightsp, fpp);
4318 * Filter out O_PATH file descriptors, most getvnode() callers
4319 * do not call fo_ methods.
4321 if (error == 0 && (*fpp)->f_ops == &path_fileops) {
4330 * Get an (NFS) file handle.
4332 #ifndef _SYS_SYSPROTO_H_
4333 struct lgetfh_args {
4339 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4342 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4343 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4346 #ifndef _SYS_SYSPROTO_H_
4353 sys_getfh(struct thread *td, struct getfh_args *uap)
4356 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4357 uap->fhp, UIO_USERSPACE));
4361 * syscall for the rpc.lockd to use to translate an open descriptor into
4362 * a NFS file handle.
4364 * warning: do not remove the priv_check() call or this becomes one giant
4367 #ifndef _SYS_SYSPROTO_H_
4368 struct getfhat_args {
4376 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4379 if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4381 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4382 uap->fhp, UIO_USERSPACE));
4386 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4387 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4389 struct nameidata nd;
4394 error = priv_check(td, PRIV_VFS_GETFH);
4397 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4398 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4403 NDFREE_NOTHING(&nd);
4405 bzero(&fh, sizeof(fh));
4406 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4407 error = VOP_VPTOFH(vp, &fh.fh_fid);
4410 if (fhseg == UIO_USERSPACE)
4411 error = copyout(&fh, fhp, sizeof (fh));
4413 memcpy(fhp, &fh, sizeof(fh));
4418 #ifndef _SYS_SYSPROTO_H_
4419 struct fhlink_args {
4425 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4428 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4431 #ifndef _SYS_SYSPROTO_H_
4432 struct fhlinkat_args {
4439 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4442 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4446 kern_fhlinkat(struct thread *td, int fd, const char *path,
4447 enum uio_seg pathseg, fhandle_t *fhp)
4454 error = priv_check(td, PRIV_VFS_GETFH);
4457 error = copyin(fhp, &fh, sizeof(fh));
4462 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4464 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4469 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4470 } while (error == EAGAIN || error == ERELOOKUP);
4474 #ifndef _SYS_SYSPROTO_H_
4475 struct fhreadlink_args {
4482 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4489 error = priv_check(td, PRIV_VFS_GETFH);
4492 if (uap->bufsize > IOSIZE_MAX)
4494 error = copyin(uap->fhp, &fh, sizeof(fh));
4497 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4499 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4503 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4509 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4512 * warning: do not remove the priv_check() call or this becomes one giant
4515 #ifndef _SYS_SYSPROTO_H_
4516 struct fhopen_args {
4517 const struct fhandle *u_fhp;
4522 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4524 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4528 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4537 error = priv_check(td, PRIV_VFS_FHOPEN);
4541 fmode = FFLAGS(flags);
4542 /* why not allow a non-read/write open for our lockd? */
4543 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4545 error = copyin(u_fhp, &fhp, sizeof(fhp));
4548 /* find the mount point */
4549 mp = vfs_busyfs(&fhp.fh_fsid);
4552 /* now give me my vnode, it gets returned to me locked */
4553 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4558 error = falloc_noinstall(td, &fp);
4564 * An extra reference on `fp' has been held for us by
4565 * falloc_noinstall().
4571 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4573 KASSERT(fp->f_ops == &badfileops,
4574 ("VOP_OPEN in fhopen() set f_ops"));
4575 KASSERT(td->td_dupfd < 0,
4576 ("fhopen() encountered fdopen()"));
4585 finit_vnode(fp, fmode, NULL, &vnops);
4587 if ((fmode & O_TRUNC) != 0) {
4588 error = fo_truncate(fp, 0, td->td_ucred, td);
4593 error = finstall(td, fp, &indx, fmode, NULL);
4596 td->td_retval[0] = indx;
4601 * Stat an (NFS) file handle.
4603 #ifndef _SYS_SYSPROTO_H_
4604 struct fhstat_args {
4605 struct fhandle *u_fhp;
4610 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4616 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4619 error = kern_fhstat(td, fh, &sb);
4621 error = copyout(&sb, uap->sb, sizeof(sb));
4626 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4632 error = priv_check(td, PRIV_VFS_FHSTAT);
4635 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4637 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4641 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4647 * Implement fstatfs() for (NFS) file handles.
4649 #ifndef _SYS_SYSPROTO_H_
4650 struct fhstatfs_args {
4651 struct fhandle *u_fhp;
4656 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4662 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4665 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4666 error = kern_fhstatfs(td, fh, sfp);
4668 error = copyout(sfp, uap->buf, sizeof(*sfp));
4669 free(sfp, M_STATFS);
4674 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4680 error = priv_check(td, PRIV_VFS_FHSTATFS);
4683 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4685 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4691 error = prison_canseemount(td->td_ucred, mp);
4695 error = mac_mount_check_stat(td->td_ucred, mp);
4699 error = VFS_STATFS(mp, buf);
4706 * Unlike madvise(2), we do not make a best effort to remember every
4707 * possible caching hint. Instead, we remember the last setting with
4708 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4709 * region of any current setting.
4712 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4715 struct fadvise_info *fa, *new;
4721 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4723 AUDIT_ARG_VALUE(advice);
4725 case POSIX_FADV_SEQUENTIAL:
4726 case POSIX_FADV_RANDOM:
4727 case POSIX_FADV_NOREUSE:
4728 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4730 case POSIX_FADV_NORMAL:
4731 case POSIX_FADV_WILLNEED:
4732 case POSIX_FADV_DONTNEED:
4738 /* XXX: CAP_POSIX_FADVISE? */
4740 error = fget(td, fd, &cap_no_rights, &fp);
4743 AUDIT_ARG_FILE(td->td_proc, fp);
4744 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4748 if (fp->f_type != DTYPE_VNODE) {
4753 if (vp->v_type != VREG) {
4760 end = offset + len - 1;
4762 case POSIX_FADV_SEQUENTIAL:
4763 case POSIX_FADV_RANDOM:
4764 case POSIX_FADV_NOREUSE:
4766 * Try to merge any existing non-standard region with
4767 * this new region if possible, otherwise create a new
4768 * non-standard region for this request.
4770 mtx_pool_lock(mtxpool_sleep, fp);
4772 if (fa != NULL && fa->fa_advice == advice &&
4773 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4774 (end != OFF_MAX && fa->fa_start == end + 1) ||
4775 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4776 if (offset < fa->fa_start)
4777 fa->fa_start = offset;
4778 if (end > fa->fa_end)
4781 new->fa_advice = advice;
4782 new->fa_start = offset;
4787 mtx_pool_unlock(mtxpool_sleep, fp);
4789 case POSIX_FADV_NORMAL:
4791 * If a the "normal" region overlaps with an existing
4792 * non-standard region, trim or remove the
4793 * non-standard region.
4795 mtx_pool_lock(mtxpool_sleep, fp);
4798 if (offset <= fa->fa_start && end >= fa->fa_end) {
4800 fp->f_advice = NULL;
4801 } else if (offset <= fa->fa_start &&
4802 end >= fa->fa_start)
4803 fa->fa_start = end + 1;
4804 else if (offset <= fa->fa_end && end >= fa->fa_end)
4805 fa->fa_end = offset - 1;
4806 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4808 * If the "normal" region is a middle
4809 * portion of the existing
4810 * non-standard region, just remove
4811 * the whole thing rather than picking
4812 * one side or the other to
4816 fp->f_advice = NULL;
4819 mtx_pool_unlock(mtxpool_sleep, fp);
4821 case POSIX_FADV_WILLNEED:
4822 case POSIX_FADV_DONTNEED:
4823 error = VOP_ADVISE(vp, offset, end, advice);
4829 free(new, M_FADVISE);
4834 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4838 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4840 return (kern_posix_error(td, error));
4844 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4845 off_t *outoffp, size_t len, unsigned int flags)
4847 struct file *infp, *outfp;
4848 struct vnode *invp, *outvp;
4851 void *rl_rcookie, *rl_wcookie;
4852 off_t savinoff, savoutoff;
4854 infp = outfp = NULL;
4855 rl_rcookie = rl_wcookie = NULL;
4864 if (len > SSIZE_MAX)
4866 * Although the len argument is size_t, the return argument
4867 * is ssize_t (which is signed). Therefore a size that won't
4868 * fit in ssize_t can't be returned.
4872 /* Get the file structures for the file descriptors. */
4873 error = fget_read(td, infd, &cap_read_rights, &infp);
4876 if (infp->f_ops == &badfileops) {
4880 if (infp->f_vnode == NULL) {
4884 error = fget_write(td, outfd, &cap_write_rights, &outfp);
4887 if (outfp->f_ops == &badfileops) {
4891 if (outfp->f_vnode == NULL) {
4896 /* Set the offset pointers to the correct place. */
4898 inoffp = &infp->f_offset;
4899 if (outoffp == NULL)
4900 outoffp = &outfp->f_offset;
4902 savoutoff = *outoffp;
4904 invp = infp->f_vnode;
4905 outvp = outfp->f_vnode;
4906 /* Sanity check the f_flag bits. */
4907 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4908 (infp->f_flag & FREAD) == 0) {
4913 /* If len == 0, just return 0. */
4918 * If infp and outfp refer to the same file, the byte ranges cannot
4921 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4922 savoutoff) || (savinoff > savoutoff && savoutoff + len >
4928 /* Range lock the byte ranges for both invp and outvp. */
4930 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4932 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4934 if (rl_rcookie != NULL)
4936 vn_rangelock_unlock(outvp, rl_wcookie);
4937 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4938 vn_rangelock_unlock(invp, rl_rcookie);
4942 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4943 flags, infp->f_cred, outfp->f_cred, td);
4945 if (rl_rcookie != NULL)
4946 vn_rangelock_unlock(invp, rl_rcookie);
4947 if (rl_wcookie != NULL)
4948 vn_rangelock_unlock(outvp, rl_wcookie);
4949 if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4951 *outoffp = savoutoff;
4957 td->td_retval[0] = retlen;
4962 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4964 off_t inoff, outoff, *inoffp, *outoffp;
4967 inoffp = outoffp = NULL;
4968 if (uap->inoffp != NULL) {
4969 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4974 if (uap->outoffp != NULL) {
4975 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4980 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4981 outoffp, uap->len, uap->flags);
4982 if (error == 0 && uap->inoffp != NULL)
4983 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4984 if (error == 0 && uap->outoffp != NULL)
4985 error = copyout(outoffp, uap->outoffp, sizeof(off_t));