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 :
136 kern_sync(struct thread *td)
138 struct mount *mp, *nmp;
141 mtx_lock(&mountlist_mtx);
142 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
143 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
144 nmp = TAILQ_NEXT(mp, mnt_list);
147 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
148 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
149 save = curthread_pflags_set(TDP_SYNCIO);
150 vfs_periodic(mp, MNT_NOWAIT);
151 VFS_SYNC(mp, MNT_NOWAIT);
152 curthread_pflags_restore(save);
153 vn_finished_write(mp);
155 mtx_lock(&mountlist_mtx);
156 nmp = TAILQ_NEXT(mp, mnt_list);
159 mtx_unlock(&mountlist_mtx);
164 * Sync each mounted filesystem.
166 #ifndef _SYS_SYSPROTO_H_
173 sys_sync(struct thread *td, struct sync_args *uap)
176 return (kern_sync(td));
180 * Change filesystem quotas.
182 #ifndef _SYS_SYSPROTO_H_
183 struct quotactl_args {
191 sys_quotactl(struct thread *td, struct quotactl_args *uap)
197 AUDIT_ARG_CMD(uap->cmd);
198 AUDIT_ARG_UID(uap->uid);
199 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
201 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
203 if ((error = namei(&nd)) != 0)
205 NDFREE(&nd, NDF_ONLY_PNBUF);
206 mp = nd.ni_vp->v_mount;
209 error = vfs_busy(mp, 0);
214 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
217 * Since quota on operation typically needs to open quota
218 * file, the Q_QUOTAON handler needs to unbusy the mount point
219 * before calling into namei. Otherwise, unmount might be
220 * started between two vfs_busy() invocations (first is our,
221 * second is from mount point cross-walk code in lookup()),
224 * Require that Q_QUOTAON handles the vfs_busy() reference on
225 * its own, always returning with ubusied mount point.
227 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
228 (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
235 * Used by statfs conversion routines to scale the block size up if
236 * necessary so that all of the block counts are <= 'max_size'. Note
237 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
241 statfs_scale_blocks(struct statfs *sf, long max_size)
246 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
249 * Attempt to scale the block counts to give a more accurate
250 * overview to userland of the ratio of free space to used
251 * space. To do this, find the largest block count and compute
252 * a divisor that lets it fit into a signed integer <= max_size.
254 if (sf->f_bavail < 0)
255 count = -sf->f_bavail;
257 count = sf->f_bavail;
258 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
259 if (count <= max_size)
262 count >>= flsl(max_size);
269 sf->f_bsize <<= shift;
270 sf->f_blocks >>= shift;
271 sf->f_bfree >>= shift;
272 sf->f_bavail >>= shift;
276 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
282 error = vfs_busy(mp, 0);
287 error = mac_mount_check_stat(td->td_ucred, mp);
291 error = VFS_STATFS(mp, buf);
294 if (priv_check_cred_vfs_generation(td->td_ucred)) {
295 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
296 prison_enforce_statfs(td->td_ucred, mp, buf);
304 * Get filesystem statistics.
306 #ifndef _SYS_SYSPROTO_H_
313 sys_statfs(struct thread *td, struct statfs_args *uap)
318 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
319 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
321 error = copyout(sfp, uap->buf, sizeof(struct statfs));
327 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
334 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
339 mp = nd.ni_vp->v_mount;
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(td, fd, &cap_fstatfs_rights, &fp);
382 vn_lock(vp, LK_SHARED | LK_RETRY);
384 AUDIT_ARG_VNODE1(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(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)
1027 if (flags & O_EXEC) {
1028 cap_rights_set_one(rightsp, CAP_FEXECVE);
1030 switch ((flags & O_ACCMODE)) {
1032 cap_rights_set_one(rightsp, CAP_READ);
1035 cap_rights_set_one(rightsp, CAP_READ);
1038 cap_rights_set_one(rightsp, CAP_WRITE);
1039 if (!(flags & (O_APPEND | O_TRUNC)))
1040 cap_rights_set_one(rightsp, CAP_SEEK);
1045 if (flags & O_CREAT)
1046 cap_rights_set_one(rightsp, CAP_CREATE);
1048 if (flags & O_TRUNC)
1049 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1051 if (flags & (O_SYNC | O_FSYNC))
1052 cap_rights_set_one(rightsp, CAP_FSYNC);
1054 if (flags & (O_EXLOCK | O_SHLOCK))
1055 cap_rights_set_one(rightsp, CAP_FLOCK);
1059 * Check permissions, allocate an open file structure, and call the device
1060 * open routine if any.
1062 #ifndef _SYS_SYSPROTO_H_
1070 sys_open(struct thread *td, struct open_args *uap)
1073 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1074 uap->flags, uap->mode));
1077 #ifndef _SYS_SYSPROTO_H_
1078 struct openat_args {
1086 sys_openat(struct thread *td, struct openat_args *uap)
1089 AUDIT_ARG_FD(uap->fd);
1090 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1095 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1096 int flags, int mode)
1098 struct proc *p = td->td_proc;
1099 struct filedesc *fdp;
1100 struct pwddesc *pdp;
1103 struct nameidata nd;
1104 cap_rights_t rights;
1105 int cmode, error, indx;
1111 AUDIT_ARG_FFLAGS(flags);
1112 AUDIT_ARG_MODE(mode);
1113 cap_rights_init_one(&rights, CAP_LOOKUP);
1114 flags_to_rights(flags, &rights);
1116 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1119 if (flags & O_EXEC) {
1120 if (flags & O_ACCMODE)
1122 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1125 flags = FFLAGS(flags);
1129 * Allocate a file structure. The descriptor to reference it
1130 * is allocated and used by finstall_refed() below.
1132 error = falloc_noinstall(td, &fp);
1135 /* Set the flags early so the finit in devfs can pick them up. */
1136 fp->f_flag = flags & FMASK;
1137 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1138 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1140 td->td_dupfd = -1; /* XXX check for fdopen */
1141 error = vn_open(&nd, &flags, cmode, fp);
1144 * If the vn_open replaced the method vector, something
1145 * wonderous happened deep below and we just pass it up
1146 * pretending we know what we do.
1148 if (error == ENXIO && fp->f_ops != &badfileops)
1152 * Handle special fdopen() case. bleh.
1154 * Don't do this for relative (capability) lookups; we don't
1155 * understand exactly what would happen, and we don't think
1156 * that it ever should.
1158 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1159 (error == ENODEV || error == ENXIO) &&
1160 td->td_dupfd >= 0) {
1161 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1170 NDFREE(&nd, NDF_ONLY_PNBUF);
1174 * Store the vnode, for any f_type. Typically, the vnode use
1175 * count is decremented by direct call to vn_closefile() for
1176 * files that switched type in the cdevsw fdopen() method.
1180 * If the file wasn't claimed by devfs bind it to the normal
1181 * vnode operations here.
1183 if (fp->f_ops == &badfileops) {
1184 KASSERT(vp->v_type != VFIFO,
1185 ("Unexpected fifo fp %p vp %p", fp, vp));
1186 finit_vnode(fp, flags, NULL, &vnops);
1190 if (flags & O_TRUNC) {
1191 error = fo_truncate(fp, 0, td->td_ucred, td);
1197 * If we haven't already installed the FD (for dupfdopen), do so now.
1200 struct filecaps *fcaps;
1203 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1204 fcaps = &nd.ni_filecaps;
1208 error = finstall_refed(td, fp, &indx, flags, fcaps);
1209 /* On success finstall_refed() consumes fcaps. */
1211 filecaps_free(&nd.ni_filecaps);
1215 filecaps_free(&nd.ni_filecaps);
1216 falloc_abort(td, fp);
1219 td->td_retval[0] = indx;
1222 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1223 falloc_abort(td, fp);
1231 #ifndef _SYS_SYSPROTO_H_
1232 struct ocreat_args {
1238 ocreat(struct thread *td, struct ocreat_args *uap)
1241 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1242 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1244 #endif /* COMPAT_43 */
1247 * Create a special file.
1249 #ifndef _SYS_SYSPROTO_H_
1250 struct mknodat_args {
1258 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1261 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1265 #if defined(COMPAT_FREEBSD11)
1267 freebsd11_mknod(struct thread *td,
1268 struct freebsd11_mknod_args *uap)
1271 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1272 uap->mode, uap->dev));
1276 freebsd11_mknodat(struct thread *td,
1277 struct freebsd11_mknodat_args *uap)
1280 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1283 #endif /* COMPAT_FREEBSD11 */
1286 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1287 int mode, dev_t dev)
1292 struct nameidata nd;
1293 int error, whiteout = 0;
1295 AUDIT_ARG_MODE(mode);
1297 switch (mode & S_IFMT) {
1300 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1301 if (error == 0 && dev == VNOVAL)
1305 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1309 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1319 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1320 NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1322 if ((error = namei(&nd)) != 0)
1326 NDFREE(&nd, NDF_ONLY_PNBUF);
1327 if (vp == nd.ni_dvp)
1335 vattr.va_mode = (mode & ALLPERMS) &
1336 ~td->td_proc->p_pd->pd_cmask;
1337 vattr.va_rdev = dev;
1340 switch (mode & S_IFMT) {
1342 vattr.va_type = VCHR;
1345 vattr.va_type = VBLK;
1351 panic("kern_mknod: invalid mode");
1354 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1355 NDFREE(&nd, NDF_ONLY_PNBUF);
1357 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1362 if (error == 0 && !whiteout)
1363 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1364 &nd.ni_cnd, &vattr);
1368 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1370 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1371 &nd.ni_cnd, &vattr);
1374 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1376 vn_finished_write(mp);
1377 NDFREE(&nd, NDF_ONLY_PNBUF);
1378 if (error == ERELOOKUP)
1384 * Create a named pipe.
1386 #ifndef _SYS_SYSPROTO_H_
1387 struct mkfifo_args {
1393 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1396 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1400 #ifndef _SYS_SYSPROTO_H_
1401 struct mkfifoat_args {
1408 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1411 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1416 kern_mkfifoat(struct thread *td, int fd, const char *path,
1417 enum uio_seg pathseg, int mode)
1421 struct nameidata nd;
1424 AUDIT_ARG_MODE(mode);
1427 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1428 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1430 if ((error = namei(&nd)) != 0)
1432 if (nd.ni_vp != NULL) {
1433 NDFREE(&nd, NDF_ONLY_PNBUF);
1434 if (nd.ni_vp == nd.ni_dvp)
1441 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1442 NDFREE(&nd, NDF_ONLY_PNBUF);
1444 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1449 vattr.va_type = VFIFO;
1450 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1452 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1457 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1461 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1462 vn_finished_write(mp);
1463 NDFREE(&nd, NDF_ONLY_PNBUF);
1464 if (error == ERELOOKUP)
1470 * Make a hard file link.
1472 #ifndef _SYS_SYSPROTO_H_
1479 sys_link(struct thread *td, struct link_args *uap)
1482 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1483 UIO_USERSPACE, FOLLOW));
1486 #ifndef _SYS_SYSPROTO_H_
1487 struct linkat_args {
1496 sys_linkat(struct thread *td, struct linkat_args *uap)
1501 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH)) != 0)
1504 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1505 UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW |
1506 AT_RESOLVE_BENEATH)));
1509 int hardlink_check_uid = 0;
1510 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1511 &hardlink_check_uid, 0,
1512 "Unprivileged processes cannot create hard links to files owned by other "
1514 static int hardlink_check_gid = 0;
1515 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1516 &hardlink_check_gid, 0,
1517 "Unprivileged processes cannot create hard links to files owned by other "
1521 can_hardlink(struct vnode *vp, struct ucred *cred)
1526 if (!hardlink_check_uid && !hardlink_check_gid)
1529 error = VOP_GETATTR(vp, &va, cred);
1533 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1534 error = priv_check_cred(cred, PRIV_VFS_LINK);
1539 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1540 error = priv_check_cred(cred, PRIV_VFS_LINK);
1549 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1550 const char *path2, enum uio_seg segflag, int follow)
1552 struct nameidata nd;
1557 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1558 path1, fd1, &cap_linkat_source_rights, td);
1559 if ((error = namei(&nd)) != 0)
1561 NDFREE(&nd, NDF_ONLY_PNBUF);
1562 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1563 } while (error == EAGAIN || error == ERELOOKUP);
1568 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1569 enum uio_seg segflag)
1571 struct nameidata nd;
1575 if (vp->v_type == VDIR) {
1577 return (EPERM); /* POSIX */
1579 NDINIT_ATRIGHTS(&nd, CREATE,
1580 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1581 &cap_linkat_target_rights, td);
1582 if ((error = namei(&nd)) == 0) {
1583 if (nd.ni_vp != NULL) {
1584 NDFREE(&nd, NDF_ONLY_PNBUF);
1585 if (nd.ni_dvp == nd.ni_vp)
1592 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1594 * Cross-device link. No need to recheck
1595 * vp->v_type, since it cannot change, except
1598 NDFREE(&nd, NDF_ONLY_PNBUF);
1602 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1603 error = can_hardlink(vp, td->td_ucred);
1606 error = mac_vnode_check_link(td->td_ucred,
1607 nd.ni_dvp, vp, &nd.ni_cnd);
1612 NDFREE(&nd, NDF_ONLY_PNBUF);
1615 error = vn_start_write(vp, &mp, V_NOWAIT);
1619 NDFREE(&nd, NDF_ONLY_PNBUF);
1620 error = vn_start_write(NULL, &mp,
1626 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1627 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1628 vn_finished_write(mp);
1629 NDFREE(&nd, NDF_ONLY_PNBUF);
1633 NDFREE(&nd, NDF_ONLY_PNBUF);
1644 * Make a symbolic link.
1646 #ifndef _SYS_SYSPROTO_H_
1647 struct symlink_args {
1653 sys_symlink(struct thread *td, struct symlink_args *uap)
1656 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1660 #ifndef _SYS_SYSPROTO_H_
1661 struct symlinkat_args {
1668 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1671 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1676 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1677 enum uio_seg segflg)
1681 const char *syspath;
1683 struct nameidata nd;
1686 if (segflg == UIO_SYSSPACE) {
1689 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1690 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1694 AUDIT_ARG_TEXT(syspath);
1697 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1698 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1700 if ((error = namei(&nd)) != 0)
1703 NDFREE(&nd, NDF_ONLY_PNBUF);
1704 if (nd.ni_vp == nd.ni_dvp)
1713 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1714 NDFREE(&nd, NDF_ONLY_PNBUF);
1716 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1721 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1723 vattr.va_type = VLNK;
1724 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1729 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1733 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1734 vn_finished_write(mp);
1735 NDFREE(&nd, NDF_ONLY_PNBUF);
1736 if (error == ERELOOKUP)
1739 if (segflg != UIO_SYSSPACE)
1740 uma_zfree(namei_zone, tmppath);
1745 * Delete a whiteout from the filesystem.
1747 #ifndef _SYS_SYSPROTO_H_
1748 struct undelete_args {
1753 sys_undelete(struct thread *td, struct undelete_args *uap)
1756 struct nameidata nd;
1761 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1762 UIO_USERSPACE, uap->path, td);
1767 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1768 NDFREE(&nd, NDF_ONLY_PNBUF);
1769 if (nd.ni_vp == nd.ni_dvp)
1777 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1778 NDFREE(&nd, NDF_ONLY_PNBUF);
1780 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1784 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1785 NDFREE(&nd, NDF_ONLY_PNBUF);
1787 vn_finished_write(mp);
1788 if (error == ERELOOKUP)
1794 * Delete a name from the filesystem.
1796 #ifndef _SYS_SYSPROTO_H_
1797 struct unlink_args {
1802 sys_unlink(struct thread *td, struct unlink_args *uap)
1805 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1810 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1811 int flag, enum uio_seg pathseg, ino_t oldinum)
1814 if ((flag & ~AT_REMOVEDIR) != 0)
1817 if ((flag & AT_REMOVEDIR) != 0)
1818 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1820 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1823 #ifndef _SYS_SYSPROTO_H_
1824 struct unlinkat_args {
1831 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1834 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1838 #ifndef _SYS_SYSPROTO_H_
1839 struct funlinkat_args {
1847 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1850 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1855 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1856 enum uio_seg pathseg, int flag, ino_t oldinum)
1861 struct nameidata nd;
1866 if (fd != FD_NONE) {
1867 error = getvnode(td, fd, &cap_no_rights, &fp);
1874 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1875 at2cnpflags(flag, AT_RESOLVE_BENEATH),
1876 pathseg, path, dfd, &cap_unlinkat_rights, td);
1877 if ((error = namei(&nd)) != 0) {
1878 if (error == EINVAL)
1883 if (vp->v_type == VDIR && oldinum == 0) {
1884 error = EPERM; /* POSIX */
1885 } else if (oldinum != 0 &&
1886 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1887 sb.st_ino != oldinum) {
1888 error = EIDRM; /* Identifier removed */
1889 } else if (fp != NULL && fp->f_vnode != vp) {
1890 if (VN_IS_DOOMED(fp->f_vnode))
1896 * The root of a mounted filesystem cannot be deleted.
1898 * XXX: can this only be a VDIR case?
1900 if (vp->v_vflag & VV_ROOT)
1904 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1905 NDFREE(&nd, NDF_ONLY_PNBUF);
1907 if (vp == nd.ni_dvp)
1911 if ((error = vn_start_write(NULL, &mp,
1912 V_XSLEEP | PCATCH)) != 0) {
1918 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1923 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1924 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1928 vn_finished_write(mp);
1930 NDFREE(&nd, NDF_ONLY_PNBUF);
1932 if (vp == nd.ni_dvp)
1936 if (error == ERELOOKUP)
1945 * Reposition read/write file offset.
1947 #ifndef _SYS_SYSPROTO_H_
1956 sys_lseek(struct thread *td, struct lseek_args *uap)
1959 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1963 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1969 error = fget(td, fd, &cap_seek_rights, &fp);
1972 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1973 fo_seek(fp, offset, whence, td) : ESPIPE;
1978 #if defined(COMPAT_43)
1980 * Reposition read/write file offset.
1982 #ifndef _SYS_SYSPROTO_H_
1983 struct olseek_args {
1990 olseek(struct thread *td, struct olseek_args *uap)
1993 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1995 #endif /* COMPAT_43 */
1997 #if defined(COMPAT_FREEBSD6)
1998 /* Version with the 'pad' argument */
2000 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2003 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2008 * Check access permissions using passed credentials.
2011 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2017 /* Flags == 0 means only check for existence. */
2018 if (user_flags == 0)
2022 if (user_flags & R_OK)
2024 if (user_flags & W_OK)
2026 if (user_flags & X_OK)
2029 error = mac_vnode_check_access(cred, vp, accmode);
2033 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2034 error = VOP_ACCESS(vp, accmode, cred, td);
2039 * Check access permissions using "real" credentials.
2041 #ifndef _SYS_SYSPROTO_H_
2042 struct access_args {
2048 sys_access(struct thread *td, struct access_args *uap)
2051 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2055 #ifndef _SYS_SYSPROTO_H_
2056 struct faccessat_args {
2064 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2067 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2072 kern_accessat(struct thread *td, int fd, const char *path,
2073 enum uio_seg pathseg, int flag, int amode)
2075 struct ucred *cred, *usecred;
2077 struct nameidata nd;
2080 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH)) != 0)
2082 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2086 * Create and modify a temporary credential instead of one that
2087 * is potentially shared (if we need one).
2089 cred = td->td_ucred;
2090 if ((flag & AT_EACCESS) == 0 &&
2091 ((cred->cr_uid != cred->cr_ruid ||
2092 cred->cr_rgid != cred->cr_groups[0]))) {
2093 usecred = crdup(cred);
2094 usecred->cr_uid = cred->cr_ruid;
2095 usecred->cr_groups[0] = cred->cr_rgid;
2096 td->td_ucred = usecred;
2099 AUDIT_ARG_VALUE(amode);
2100 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2101 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH),
2102 pathseg, path, fd, &cap_fstat_rights, td);
2103 if ((error = namei(&nd)) != 0)
2107 error = vn_access(vp, amode, usecred, td);
2108 NDFREE_NOTHING(&nd);
2111 if (usecred != cred) {
2112 td->td_ucred = cred;
2119 * Check access permissions using "effective" credentials.
2121 #ifndef _SYS_SYSPROTO_H_
2122 struct eaccess_args {
2128 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2131 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2132 AT_EACCESS, uap->amode));
2135 #if defined(COMPAT_43)
2137 * Get file status; this version follows links.
2139 #ifndef _SYS_SYSPROTO_H_
2146 ostat(struct thread *td, struct ostat_args *uap)
2152 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2157 return (copyout(&osb, uap->ub, sizeof (osb)));
2161 * Get file status; this version does not follow links.
2163 #ifndef _SYS_SYSPROTO_H_
2164 struct olstat_args {
2170 olstat(struct thread *td, struct olstat_args *uap)
2176 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2177 UIO_USERSPACE, &sb, NULL);
2181 return (copyout(&osb, uap->ub, sizeof (osb)));
2185 * Convert from an old to a new stat structure.
2186 * XXX: many values are blindly truncated.
2189 cvtstat(struct stat *st, struct ostat *ost)
2192 bzero(ost, sizeof(*ost));
2193 ost->st_dev = st->st_dev;
2194 ost->st_ino = st->st_ino;
2195 ost->st_mode = st->st_mode;
2196 ost->st_nlink = st->st_nlink;
2197 ost->st_uid = st->st_uid;
2198 ost->st_gid = st->st_gid;
2199 ost->st_rdev = st->st_rdev;
2200 ost->st_size = MIN(st->st_size, INT32_MAX);
2201 ost->st_atim = st->st_atim;
2202 ost->st_mtim = st->st_mtim;
2203 ost->st_ctim = st->st_ctim;
2204 ost->st_blksize = st->st_blksize;
2205 ost->st_blocks = st->st_blocks;
2206 ost->st_flags = st->st_flags;
2207 ost->st_gen = st->st_gen;
2209 #endif /* COMPAT_43 */
2211 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2212 int ino64_trunc_error;
2213 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2214 &ino64_trunc_error, 0,
2215 "Error on truncation of device, file or inode number, or link count");
2218 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2221 ost->st_dev = st->st_dev;
2222 if (ost->st_dev != st->st_dev) {
2223 switch (ino64_trunc_error) {
2226 * Since dev_t is almost raw, don't clamp to the
2227 * maximum for case 2, but ignore the error.
2234 ost->st_ino = st->st_ino;
2235 if (ost->st_ino != st->st_ino) {
2236 switch (ino64_trunc_error) {
2243 ost->st_ino = UINT32_MAX;
2247 ost->st_mode = st->st_mode;
2248 ost->st_nlink = st->st_nlink;
2249 if (ost->st_nlink != st->st_nlink) {
2250 switch (ino64_trunc_error) {
2257 ost->st_nlink = UINT16_MAX;
2261 ost->st_uid = st->st_uid;
2262 ost->st_gid = st->st_gid;
2263 ost->st_rdev = st->st_rdev;
2264 if (ost->st_rdev != st->st_rdev) {
2265 switch (ino64_trunc_error) {
2272 ost->st_atim = st->st_atim;
2273 ost->st_mtim = st->st_mtim;
2274 ost->st_ctim = st->st_ctim;
2275 ost->st_size = st->st_size;
2276 ost->st_blocks = st->st_blocks;
2277 ost->st_blksize = st->st_blksize;
2278 ost->st_flags = st->st_flags;
2279 ost->st_gen = st->st_gen;
2281 ost->st_birthtim = st->st_birthtim;
2282 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2283 sizeof(*ost) - offsetof(struct freebsd11_stat,
2284 st_birthtim) - sizeof(ost->st_birthtim));
2289 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2292 struct freebsd11_stat osb;
2295 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2299 error = freebsd11_cvtstat(&sb, &osb);
2301 error = copyout(&osb, uap->ub, sizeof(osb));
2306 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2309 struct freebsd11_stat osb;
2312 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2313 UIO_USERSPACE, &sb, NULL);
2316 error = freebsd11_cvtstat(&sb, &osb);
2318 error = copyout(&osb, uap->ub, sizeof(osb));
2323 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2327 struct freebsd11_stat osb;
2330 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2333 error = kern_fhstat(td, fh, &sb);
2336 error = freebsd11_cvtstat(&sb, &osb);
2338 error = copyout(&osb, uap->sb, sizeof(osb));
2343 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2346 struct freebsd11_stat osb;
2349 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2350 UIO_USERSPACE, &sb, NULL);
2353 error = freebsd11_cvtstat(&sb, &osb);
2355 error = copyout(&osb, uap->buf, sizeof(osb));
2358 #endif /* COMPAT_FREEBSD11 */
2363 #ifndef _SYS_SYSPROTO_H_
2364 struct fstatat_args {
2372 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2377 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2378 UIO_USERSPACE, &sb, NULL);
2380 error = copyout(&sb, uap->buf, sizeof (sb));
2385 kern_statat(struct thread *td, int flag, int fd, const char *path,
2386 enum uio_seg pathseg, struct stat *sbp,
2387 void (*hook)(struct vnode *vp, struct stat *sbp))
2389 struct nameidata nd;
2392 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2395 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2396 AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2397 pathseg, path, fd, &cap_fstat_rights, td);
2399 if ((error = namei(&nd)) != 0)
2401 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2403 if (__predict_false(hook != NULL))
2404 hook(nd.ni_vp, sbp);
2406 NDFREE_NOTHING(&nd);
2408 #ifdef __STAT_TIME_T_EXT
2409 sbp->st_atim_ext = 0;
2410 sbp->st_mtim_ext = 0;
2411 sbp->st_ctim_ext = 0;
2412 sbp->st_btim_ext = 0;
2415 if (KTRPOINT(td, KTR_STRUCT))
2416 ktrstat_error(sbp, error);
2421 #if defined(COMPAT_FREEBSD11)
2423 * Implementation of the NetBSD [l]stat() functions.
2426 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2429 bzero(nsb, sizeof(*nsb));
2430 nsb->st_dev = sb->st_dev;
2431 nsb->st_ino = sb->st_ino;
2432 nsb->st_mode = sb->st_mode;
2433 nsb->st_nlink = sb->st_nlink;
2434 nsb->st_uid = sb->st_uid;
2435 nsb->st_gid = sb->st_gid;
2436 nsb->st_rdev = sb->st_rdev;
2437 nsb->st_atim = sb->st_atim;
2438 nsb->st_mtim = sb->st_mtim;
2439 nsb->st_ctim = sb->st_ctim;
2440 nsb->st_size = sb->st_size;
2441 nsb->st_blocks = sb->st_blocks;
2442 nsb->st_blksize = sb->st_blksize;
2443 nsb->st_flags = sb->st_flags;
2444 nsb->st_gen = sb->st_gen;
2445 nsb->st_birthtim = sb->st_birthtim;
2448 #ifndef _SYS_SYSPROTO_H_
2449 struct freebsd11_nstat_args {
2455 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2461 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2465 freebsd11_cvtnstat(&sb, &nsb);
2466 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2470 * NetBSD lstat. Get file status; this version does not follow links.
2472 #ifndef _SYS_SYSPROTO_H_
2473 struct freebsd11_nlstat_args {
2479 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2485 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2486 UIO_USERSPACE, &sb, NULL);
2489 freebsd11_cvtnstat(&sb, &nsb);
2490 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2492 #endif /* COMPAT_FREEBSD11 */
2495 * Get configurable pathname variables.
2497 #ifndef _SYS_SYSPROTO_H_
2498 struct pathconf_args {
2504 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2509 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2512 td->td_retval[0] = value;
2516 #ifndef _SYS_SYSPROTO_H_
2517 struct lpathconf_args {
2523 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2528 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2531 td->td_retval[0] = value;
2536 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2537 int name, u_long flags, long *valuep)
2539 struct nameidata nd;
2542 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2544 if ((error = namei(&nd)) != 0)
2546 NDFREE_NOTHING(&nd);
2548 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2554 * Return target name of a symbolic link.
2556 #ifndef _SYS_SYSPROTO_H_
2557 struct readlink_args {
2564 sys_readlink(struct thread *td, struct readlink_args *uap)
2567 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2568 uap->buf, UIO_USERSPACE, uap->count));
2570 #ifndef _SYS_SYSPROTO_H_
2571 struct readlinkat_args {
2579 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2582 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2583 uap->buf, UIO_USERSPACE, uap->bufsize));
2587 kern_readlinkat(struct thread *td, int fd, const char *path,
2588 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2591 struct nameidata nd;
2594 if (count > IOSIZE_MAX)
2597 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2598 pathseg, path, fd, td);
2600 if ((error = namei(&nd)) != 0)
2602 NDFREE_NOTHING(&nd);
2605 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2612 * Helper function to readlink from a vnode
2615 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2622 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2624 error = mac_vnode_check_readlink(td->td_ucred, vp);
2628 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2631 aiov.iov_base = buf;
2632 aiov.iov_len = count;
2633 auio.uio_iov = &aiov;
2634 auio.uio_iovcnt = 1;
2635 auio.uio_offset = 0;
2636 auio.uio_rw = UIO_READ;
2637 auio.uio_segflg = bufseg;
2639 auio.uio_resid = count;
2640 error = VOP_READLINK(vp, &auio, td->td_ucred);
2641 td->td_retval[0] = count - auio.uio_resid;
2646 * Common implementation code for chflags() and fchflags().
2649 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2655 /* We can't support the value matching VNOVAL. */
2656 if (flags == VNOVAL)
2657 return (EOPNOTSUPP);
2660 * Prevent non-root users from setting flags on devices. When
2661 * a device is reused, users can retain ownership of the device
2662 * if they are allowed to set flags and programs assume that
2663 * chown can't fail when done as root.
2665 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2666 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2671 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2674 vattr.va_flags = flags;
2675 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2677 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2680 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2682 vn_finished_write(mp);
2687 * Change flags of a file given a path name.
2689 #ifndef _SYS_SYSPROTO_H_
2690 struct chflags_args {
2696 sys_chflags(struct thread *td, struct chflags_args *uap)
2699 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2703 #ifndef _SYS_SYSPROTO_H_
2704 struct chflagsat_args {
2712 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2715 if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2718 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2719 uap->flags, uap->atflag));
2723 * Same as chflags() but doesn't follow symlinks.
2725 #ifndef _SYS_SYSPROTO_H_
2726 struct lchflags_args {
2732 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2735 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2736 uap->flags, AT_SYMLINK_NOFOLLOW));
2740 kern_chflagsat(struct thread *td, int fd, const char *path,
2741 enum uio_seg pathseg, u_long flags, int atflag)
2743 struct nameidata nd;
2746 AUDIT_ARG_FFLAGS(flags);
2747 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2748 AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2749 &cap_fchflags_rights, td);
2750 if ((error = namei(&nd)) != 0)
2752 NDFREE_NOTHING(&nd);
2753 error = setfflags(td, nd.ni_vp, flags);
2759 * Change flags of a file given a file descriptor.
2761 #ifndef _SYS_SYSPROTO_H_
2762 struct fchflags_args {
2768 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2773 AUDIT_ARG_FD(uap->fd);
2774 AUDIT_ARG_FFLAGS(uap->flags);
2775 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2780 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2781 AUDIT_ARG_VNODE1(fp->f_vnode);
2782 VOP_UNLOCK(fp->f_vnode);
2784 error = setfflags(td, fp->f_vnode, uap->flags);
2790 * Common implementation code for chmod(), lchmod() and fchmod().
2793 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2799 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2801 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2803 vattr.va_mode = mode & ALLPERMS;
2805 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2808 error = VOP_SETATTR(vp, &vattr, cred);
2810 vn_finished_write(mp);
2815 * Change mode of a file given path name.
2817 #ifndef _SYS_SYSPROTO_H_
2824 sys_chmod(struct thread *td, struct chmod_args *uap)
2827 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2831 #ifndef _SYS_SYSPROTO_H_
2832 struct fchmodat_args {
2840 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2843 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2846 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2847 uap->mode, uap->flag));
2851 * Change mode of a file given path name (don't follow links.)
2853 #ifndef _SYS_SYSPROTO_H_
2854 struct lchmod_args {
2860 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2863 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2864 uap->mode, AT_SYMLINK_NOFOLLOW));
2868 kern_fchmodat(struct thread *td, int fd, const char *path,
2869 enum uio_seg pathseg, mode_t mode, int flag)
2871 struct nameidata nd;
2874 AUDIT_ARG_MODE(mode);
2875 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2876 AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2877 &cap_fchmod_rights, td);
2878 if ((error = namei(&nd)) != 0)
2880 NDFREE_NOTHING(&nd);
2881 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2887 * Change mode of a file given a file descriptor.
2889 #ifndef _SYS_SYSPROTO_H_
2890 struct fchmod_args {
2896 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2901 AUDIT_ARG_FD(uap->fd);
2902 AUDIT_ARG_MODE(uap->mode);
2904 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2907 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2913 * Common implementation for chown(), lchown(), and fchown()
2916 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2923 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2925 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2930 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2934 error = VOP_SETATTR(vp, &vattr, cred);
2936 vn_finished_write(mp);
2941 * Set ownership given a path name.
2943 #ifndef _SYS_SYSPROTO_H_
2951 sys_chown(struct thread *td, struct chown_args *uap)
2954 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2958 #ifndef _SYS_SYSPROTO_H_
2959 struct fchownat_args {
2968 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2971 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2974 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2975 uap->gid, uap->flag));
2979 kern_fchownat(struct thread *td, int fd, const char *path,
2980 enum uio_seg pathseg, int uid, int gid, int flag)
2982 struct nameidata nd;
2985 AUDIT_ARG_OWNER(uid, gid);
2986 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2987 AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2988 &cap_fchown_rights, td);
2990 if ((error = namei(&nd)) != 0)
2992 NDFREE_NOTHING(&nd);
2993 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2999 * Set ownership given a path name, do not cross symlinks.
3001 #ifndef _SYS_SYSPROTO_H_
3002 struct lchown_args {
3009 sys_lchown(struct thread *td, struct lchown_args *uap)
3012 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3013 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3017 * Set ownership given a file descriptor.
3019 #ifndef _SYS_SYSPROTO_H_
3020 struct fchown_args {
3027 sys_fchown(struct thread *td, struct fchown_args *uap)
3032 AUDIT_ARG_FD(uap->fd);
3033 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3034 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3037 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3043 * Common implementation code for utimes(), lutimes(), and futimes().
3046 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3047 struct timespec *tsp)
3049 struct timeval tv[2];
3050 const struct timeval *tvp;
3053 if (usrtvp == NULL) {
3054 vfs_timestamp(&tsp[0]);
3057 if (tvpseg == UIO_SYSSPACE) {
3060 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3065 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3066 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3068 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3069 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3075 * Common implementation code for futimens(), utimensat().
3077 #define UTIMENS_NULL 0x1
3078 #define UTIMENS_EXIT 0x2
3080 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3081 struct timespec *tsp, int *retflags)
3083 struct timespec tsnow;
3086 vfs_timestamp(&tsnow);
3088 if (usrtsp == NULL) {
3091 *retflags |= UTIMENS_NULL;
3094 if (tspseg == UIO_SYSSPACE) {
3097 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3099 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3100 *retflags |= UTIMENS_EXIT;
3101 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3102 *retflags |= UTIMENS_NULL;
3103 if (tsp[0].tv_nsec == UTIME_OMIT)
3104 tsp[0].tv_sec = VNOVAL;
3105 else if (tsp[0].tv_nsec == UTIME_NOW)
3107 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3109 if (tsp[1].tv_nsec == UTIME_OMIT)
3110 tsp[1].tv_sec = VNOVAL;
3111 else if (tsp[1].tv_nsec == UTIME_NOW)
3113 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3120 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3124 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3125 int numtimes, int nullflag)
3129 int error, setbirthtime;
3131 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3133 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3135 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3136 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3139 vattr.va_atime = ts[0];
3140 vattr.va_mtime = ts[1];
3142 vattr.va_birthtime = ts[1];
3144 vattr.va_birthtime = ts[2];
3146 vattr.va_vaflags |= VA_UTIMES_NULL;
3148 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3152 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3154 vn_finished_write(mp);
3159 * Set the access and modification times of a file.
3161 #ifndef _SYS_SYSPROTO_H_
3162 struct utimes_args {
3164 struct timeval *tptr;
3168 sys_utimes(struct thread *td, struct utimes_args *uap)
3171 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3172 uap->tptr, UIO_USERSPACE));
3175 #ifndef _SYS_SYSPROTO_H_
3176 struct futimesat_args {
3179 const struct timeval * times;
3183 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3186 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3187 uap->times, UIO_USERSPACE));
3191 kern_utimesat(struct thread *td, int fd, const char *path,
3192 enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3194 struct nameidata nd;
3195 struct timespec ts[2];
3198 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3200 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3201 &cap_futimes_rights, td);
3203 if ((error = namei(&nd)) != 0)
3205 NDFREE_NOTHING(&nd);
3206 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3212 * Set the access and modification times of a file.
3214 #ifndef _SYS_SYSPROTO_H_
3215 struct lutimes_args {
3217 struct timeval *tptr;
3221 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3224 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3229 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3230 struct timeval *tptr, enum uio_seg tptrseg)
3232 struct timespec ts[2];
3233 struct nameidata nd;
3236 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3238 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3239 if ((error = namei(&nd)) != 0)
3241 NDFREE_NOTHING(&nd);
3242 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3248 * Set the access and modification times of a file.
3250 #ifndef _SYS_SYSPROTO_H_
3251 struct futimes_args {
3253 struct timeval *tptr;
3257 sys_futimes(struct thread *td, struct futimes_args *uap)
3260 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3264 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3265 enum uio_seg tptrseg)
3267 struct timespec ts[2];
3272 error = getutimes(tptr, tptrseg, ts);
3275 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3279 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3280 AUDIT_ARG_VNODE1(fp->f_vnode);
3281 VOP_UNLOCK(fp->f_vnode);
3283 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3289 sys_futimens(struct thread *td, struct futimens_args *uap)
3292 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3296 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3297 enum uio_seg tptrseg)
3299 struct timespec ts[2];
3304 error = getutimens(tptr, tptrseg, ts, &flags);
3307 if (flags & UTIMENS_EXIT)
3309 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3313 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3314 AUDIT_ARG_VNODE1(fp->f_vnode);
3315 VOP_UNLOCK(fp->f_vnode);
3317 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3323 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3326 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3327 uap->times, UIO_USERSPACE, uap->flag));
3331 kern_utimensat(struct thread *td, int fd, const char *path,
3332 enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3335 struct nameidata nd;
3336 struct timespec ts[2];
3339 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
3342 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3344 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3345 AT_RESOLVE_BENEATH) | AUDITVNODE1,
3346 pathseg, path, fd, &cap_futimes_rights, td);
3347 if ((error = namei(&nd)) != 0)
3350 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3352 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3353 * "Search permission is denied by a component of the path prefix."
3355 NDFREE_NOTHING(&nd);
3356 if ((flags & UTIMENS_EXIT) == 0)
3357 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3363 * Truncate a file given its path name.
3365 #ifndef _SYS_SYSPROTO_H_
3366 struct truncate_args {
3373 sys_truncate(struct thread *td, struct truncate_args *uap)
3376 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3380 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3387 struct nameidata nd;
3393 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3394 if ((error = namei(&nd)) != 0)
3397 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3398 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3399 vn_rangelock_unlock(vp, rl_cookie);
3403 NDFREE(&nd, NDF_ONLY_PNBUF);
3404 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3405 if (vp->v_type == VDIR)
3408 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3411 else if ((error = vn_writechk(vp)) == 0 &&
3412 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3414 vattr.va_size = length;
3415 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3418 vn_finished_write(mp);
3419 vn_rangelock_unlock(vp, rl_cookie);
3421 if (error == ERELOOKUP)
3426 #if defined(COMPAT_43)
3428 * Truncate a file given its path name.
3430 #ifndef _SYS_SYSPROTO_H_
3431 struct otruncate_args {
3437 otruncate(struct thread *td, struct otruncate_args *uap)
3440 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3442 #endif /* COMPAT_43 */
3444 #if defined(COMPAT_FREEBSD6)
3445 /* Versions with the pad argument */
3447 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3450 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3454 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3457 return (kern_ftruncate(td, uap->fd, uap->length));
3462 kern_fsync(struct thread *td, int fd, bool fullsync)
3467 int error, lock_flags;
3470 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3476 /* XXXKIB: compete outstanding aio writes */;
3479 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3482 if (MNT_SHARED_WRITES(mp) ||
3483 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3484 lock_flags = LK_SHARED;
3486 lock_flags = LK_EXCLUSIVE;
3488 vn_lock(vp, lock_flags | LK_RETRY);
3489 AUDIT_ARG_VNODE1(vp);
3490 if (vp->v_object != NULL) {
3491 VM_OBJECT_WLOCK(vp->v_object);
3492 vm_object_page_clean(vp->v_object, 0, 0, 0);
3493 VM_OBJECT_WUNLOCK(vp->v_object);
3495 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3497 vn_finished_write(mp);
3498 if (error == ERELOOKUP)
3506 * Sync an open file.
3508 #ifndef _SYS_SYSPROTO_H_
3514 sys_fsync(struct thread *td, struct fsync_args *uap)
3517 return (kern_fsync(td, uap->fd, true));
3521 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3524 return (kern_fsync(td, uap->fd, false));
3528 * Rename files. Source and destination must either both be directories, or
3529 * both not be directories. If target is a directory, it must be empty.
3531 #ifndef _SYS_SYSPROTO_H_
3532 struct rename_args {
3538 sys_rename(struct thread *td, struct rename_args *uap)
3541 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3542 uap->to, UIO_USERSPACE));
3545 #ifndef _SYS_SYSPROTO_H_
3546 struct renameat_args {
3554 sys_renameat(struct thread *td, struct renameat_args *uap)
3557 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3563 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3564 const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3568 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3569 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3570 if ((error = namei(fromnd)) != 0)
3572 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3573 fromnd->ni_vp, &fromnd->ni_cnd);
3574 VOP_UNLOCK(fromnd->ni_dvp);
3575 if (fromnd->ni_dvp != fromnd->ni_vp)
3576 VOP_UNLOCK(fromnd->ni_vp);
3578 NDFREE(fromnd, NDF_ONLY_PNBUF);
3579 vrele(fromnd->ni_dvp);
3580 vrele(fromnd->ni_vp);
3581 if (fromnd->ni_startdir)
3582 vrele(fromnd->ni_startdir);
3589 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3590 const char *new, enum uio_seg pathseg)
3592 struct mount *mp = NULL;
3593 struct vnode *tvp, *fvp, *tdvp;
3594 struct nameidata fromnd, tond;
3595 u_int64_t tondflags;
3601 if (mac_vnode_check_rename_from_enabled()) {
3602 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3608 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3609 pathseg, old, oldfd, &cap_renameat_source_rights, td);
3610 if ((error = namei(&fromnd)) != 0)
3616 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3617 if (fromnd.ni_vp->v_type == VDIR)
3618 tondflags |= WILLBEDIR;
3619 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3620 &cap_renameat_target_rights, td);
3621 if ((error = namei(&tond)) != 0) {
3622 /* Translate error code for rename("dir1", "dir2/."). */
3623 if (error == EISDIR && fvp->v_type == VDIR)
3625 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3626 vrele(fromnd.ni_dvp);
3632 error = vn_start_write(fvp, &mp, V_NOWAIT);
3634 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3635 NDFREE(&tond, NDF_ONLY_PNBUF);
3642 vrele(fromnd.ni_dvp);
3644 vrele(tond.ni_startdir);
3645 if (fromnd.ni_startdir != NULL)
3646 vrele(fromnd.ni_startdir);
3647 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3653 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3656 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3661 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3663 * If the target already exists we require CAP_UNLINKAT
3664 * from 'newfd', when newfd was used for the lookup.
3666 error = cap_check(&tond.ni_filecaps.fc_rights,
3667 &cap_unlinkat_rights);
3678 * If the source is the same as the destination (that is, if they
3679 * are links to the same vnode), then there is nothing to do.
3685 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3686 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3690 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3691 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3692 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3693 NDFREE(&tond, NDF_ONLY_PNBUF);
3695 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3696 NDFREE(&tond, NDF_ONLY_PNBUF);
3703 vrele(fromnd.ni_dvp);
3706 vrele(tond.ni_startdir);
3707 vn_finished_write(mp);
3709 if (fromnd.ni_startdir)
3710 vrele(fromnd.ni_startdir);
3711 if (error == ERESTART)
3713 if (error == ERELOOKUP)
3719 * Make a directory file.
3721 #ifndef _SYS_SYSPROTO_H_
3728 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3731 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3735 #ifndef _SYS_SYSPROTO_H_
3736 struct mkdirat_args {
3743 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3746 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3750 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3755 struct nameidata nd;
3758 AUDIT_ARG_MODE(mode);
3761 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3762 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3763 segflg, path, fd, &cap_mkdirat_rights, td);
3764 if ((error = namei(&nd)) != 0)
3766 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3767 NDFREE(&nd, NDF_ONLY_PNBUF);
3769 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3774 vattr.va_type = VDIR;
3775 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3777 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3782 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3786 NDFREE(&nd, NDF_ONLY_PNBUF);
3787 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3788 vn_finished_write(mp);
3789 if (error == ERELOOKUP)
3795 * Remove a directory file.
3797 #ifndef _SYS_SYSPROTO_H_
3803 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3806 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3811 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3812 enum uio_seg pathseg, int flag)
3817 struct nameidata nd;
3818 cap_rights_t rights;
3822 if (fd != FD_NONE) {
3823 error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3831 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3832 at2cnpflags(flag, AT_RESOLVE_BENEATH),
3833 pathseg, path, dfd, &cap_unlinkat_rights, td);
3834 if ((error = namei(&nd)) != 0)
3837 if (vp->v_type != VDIR) {
3842 * No rmdir "." please.
3844 if (nd.ni_dvp == vp) {
3849 * The root of a mounted filesystem cannot be deleted.
3851 if (vp->v_vflag & VV_ROOT) {
3856 if (fp != NULL && fp->f_vnode != vp) {
3857 if (VN_IS_DOOMED(fp->f_vnode))
3865 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3870 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3871 NDFREE(&nd, NDF_ONLY_PNBUF);
3873 if (nd.ni_dvp == vp)
3877 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3881 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3882 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3883 vn_finished_write(mp);
3885 NDFREE(&nd, NDF_ONLY_PNBUF);
3887 if (nd.ni_dvp == vp)
3891 if (error == ERELOOKUP)
3899 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3901 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3902 long *basep, void (*func)(struct freebsd11_dirent *))
3904 struct freebsd11_dirent dstdp;
3905 struct dirent *dp, *edp;
3908 ssize_t resid, ucount;
3911 /* XXX arbitrary sanity limit on `count'. */
3912 count = min(count, 64 * 1024);
3914 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3916 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3924 for (dp = (struct dirent *)dirbuf,
3925 edp = (struct dirent *)&dirbuf[count - resid];
3926 ucount < count && dp < edp; ) {
3927 if (dp->d_reclen == 0)
3929 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3930 if (dp->d_namlen >= sizeof(dstdp.d_name))
3932 dstdp.d_type = dp->d_type;
3933 dstdp.d_namlen = dp->d_namlen;
3934 dstdp.d_fileno = dp->d_fileno; /* truncate */
3935 if (dstdp.d_fileno != dp->d_fileno) {
3936 switch (ino64_trunc_error) {
3944 dstdp.d_fileno = UINT32_MAX;
3948 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3949 ((dp->d_namlen + 1 + 3) &~ 3);
3950 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3951 bzero(dstdp.d_name + dstdp.d_namlen,
3952 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3954 MPASS(dstdp.d_reclen <= dp->d_reclen);
3955 MPASS(ucount + dstdp.d_reclen <= count);
3958 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3961 dp = (struct dirent *)((char *)dp + dp->d_reclen);
3962 ucount += dstdp.d_reclen;
3966 free(dirbuf, M_TEMP);
3968 td->td_retval[0] = ucount;
3975 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3977 #if (BYTE_ORDER == LITTLE_ENDIAN)
3979 * The expected low byte of dp->d_namlen is our dp->d_type.
3980 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3982 dp->d_type = dp->d_namlen;
3986 * The dp->d_type is the high byte of the expected dp->d_namlen,
3987 * so must be zero'ed.
3994 * Read a block of directory entries in a filesystem independent format.
3996 #ifndef _SYS_SYSPROTO_H_
3997 struct ogetdirentries_args {
4005 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4010 error = kern_ogetdirentries(td, uap, &loff);
4012 error = copyout(&loff, uap->basep, sizeof(long));
4017 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4023 /* XXX arbitrary sanity limit on `count'. */
4024 if (uap->count > 64 * 1024)
4027 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4028 &base, ogetdirentries_cvt);
4030 if (error == 0 && uap->basep != NULL)
4031 error = copyout(&base, uap->basep, sizeof(long));
4035 #endif /* COMPAT_43 */
4037 #if defined(COMPAT_FREEBSD11)
4038 #ifndef _SYS_SYSPROTO_H_
4039 struct freebsd11_getdirentries_args {
4047 freebsd11_getdirentries(struct thread *td,
4048 struct freebsd11_getdirentries_args *uap)
4053 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4056 if (error == 0 && uap->basep != NULL)
4057 error = copyout(&base, uap->basep, sizeof(long));
4062 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4064 struct freebsd11_getdirentries_args ap;
4068 ap.count = uap->count;
4070 return (freebsd11_getdirentries(td, &ap));
4072 #endif /* COMPAT_FREEBSD11 */
4075 * Read a block of directory entries in a filesystem independent format.
4078 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4083 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4084 NULL, UIO_USERSPACE);
4087 if (uap->basep != NULL)
4088 error = copyout(&base, uap->basep, sizeof(off_t));
4093 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4094 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4105 if (count > IOSIZE_MAX)
4107 auio.uio_resid = count;
4108 error = getvnode(td, fd, &cap_read_rights, &fp);
4111 if ((fp->f_flag & FREAD) == 0) {
4116 foffset = foffset_lock(fp, 0);
4118 if (vp->v_type != VDIR) {
4122 aiov.iov_base = buf;
4123 aiov.iov_len = count;
4124 auio.uio_iov = &aiov;
4125 auio.uio_iovcnt = 1;
4126 auio.uio_rw = UIO_READ;
4127 auio.uio_segflg = bufseg;
4129 vn_lock(vp, LK_SHARED | LK_RETRY);
4130 AUDIT_ARG_VNODE1(vp);
4131 loff = auio.uio_offset = foffset;
4133 error = mac_vnode_check_readdir(td->td_ucred, vp);
4136 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4138 foffset = auio.uio_offset;
4143 if (count == auio.uio_resid &&
4144 (vp->v_vflag & VV_ROOT) &&
4145 (vp->v_mount->mnt_flag & MNT_UNION)) {
4146 struct vnode *tvp = vp;
4148 vp = vp->v_mount->mnt_vnodecovered;
4158 *residp = auio.uio_resid;
4159 td->td_retval[0] = count - auio.uio_resid;
4161 foffset_unlock(fp, foffset, 0);
4167 * Set the mode mask for creation of filesystem nodes.
4169 #ifndef _SYS_SYSPROTO_H_
4175 sys_umask(struct thread *td, struct umask_args *uap)
4177 struct pwddesc *pdp;
4179 pdp = td->td_proc->p_pd;
4181 td->td_retval[0] = pdp->pd_cmask;
4182 pdp->pd_cmask = uap->newmask & ALLPERMS;
4183 PWDDESC_XUNLOCK(pdp);
4188 * Void all references to file by ripping underlying filesystem away from
4191 #ifndef _SYS_SYSPROTO_H_
4192 struct revoke_args {
4197 sys_revoke(struct thread *td, struct revoke_args *uap)
4201 struct nameidata nd;
4204 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4206 if ((error = namei(&nd)) != 0)
4209 NDFREE_NOTHING(&nd);
4210 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4215 error = mac_vnode_check_revoke(td->td_ucred, vp);
4219 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4222 if (td->td_ucred->cr_uid != vattr.va_uid) {
4223 error = priv_check(td, PRIV_VFS_ADMIN);
4227 if (devfs_usecount(vp) > 0)
4228 VOP_REVOKE(vp, REVOKEALL);
4235 * Convert a user file descriptor to a kernel file entry and check that, if it
4236 * is a capability, the correct rights are present. A reference on the file
4237 * entry is held upon returning.
4240 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4245 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4250 * The file could be not of the vnode type, or it may be not
4251 * yet fully initialized, in which case the f_vnode pointer
4252 * may be set, but f_ops is still badfileops. E.g.,
4253 * devfs_open() transiently create such situation to
4254 * facilitate csw d_fdopen().
4256 * Dupfdopen() handling in kern_openat() installs the
4257 * half-baked file into the process descriptor table, allowing
4258 * other thread to dereference it. Guard against the race by
4261 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4270 * Get an (NFS) file handle.
4272 #ifndef _SYS_SYSPROTO_H_
4273 struct lgetfh_args {
4279 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4282 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4283 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4286 #ifndef _SYS_SYSPROTO_H_
4293 sys_getfh(struct thread *td, struct getfh_args *uap)
4296 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4297 uap->fhp, UIO_USERSPACE));
4301 * syscall for the rpc.lockd to use to translate an open descriptor into
4302 * a NFS file handle.
4304 * warning: do not remove the priv_check() call or this becomes one giant
4307 #ifndef _SYS_SYSPROTO_H_
4308 struct getfhat_args {
4316 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4319 if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4321 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4322 uap->fhp, UIO_USERSPACE));
4326 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4327 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4329 struct nameidata nd;
4334 error = priv_check(td, PRIV_VFS_GETFH);
4337 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4338 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4343 NDFREE_NOTHING(&nd);
4345 bzero(&fh, sizeof(fh));
4346 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4347 error = VOP_VPTOFH(vp, &fh.fh_fid);
4350 if (fhseg == UIO_USERSPACE)
4351 error = copyout(&fh, fhp, sizeof (fh));
4353 memcpy(fhp, &fh, sizeof(fh));
4358 #ifndef _SYS_SYSPROTO_H_
4359 struct fhlink_args {
4365 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4368 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4371 #ifndef _SYS_SYSPROTO_H_
4372 struct fhlinkat_args {
4379 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4382 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4386 kern_fhlinkat(struct thread *td, int fd, const char *path,
4387 enum uio_seg pathseg, fhandle_t *fhp)
4394 error = priv_check(td, PRIV_VFS_GETFH);
4397 error = copyin(fhp, &fh, sizeof(fh));
4402 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4404 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4409 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4410 } while (error == EAGAIN || error == ERELOOKUP);
4414 #ifndef _SYS_SYSPROTO_H_
4415 struct fhreadlink_args {
4422 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4429 error = priv_check(td, PRIV_VFS_GETFH);
4432 if (uap->bufsize > IOSIZE_MAX)
4434 error = copyin(uap->fhp, &fh, sizeof(fh));
4437 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4439 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4443 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4449 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4452 * warning: do not remove the priv_check() call or this becomes one giant
4455 #ifndef _SYS_SYSPROTO_H_
4456 struct fhopen_args {
4457 const struct fhandle *u_fhp;
4462 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4464 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4468 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4477 error = priv_check(td, PRIV_VFS_FHOPEN);
4481 fmode = FFLAGS(flags);
4482 /* why not allow a non-read/write open for our lockd? */
4483 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4485 error = copyin(u_fhp, &fhp, sizeof(fhp));
4488 /* find the mount point */
4489 mp = vfs_busyfs(&fhp.fh_fsid);
4492 /* now give me my vnode, it gets returned to me locked */
4493 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4498 error = falloc_noinstall(td, &fp);
4504 * An extra reference on `fp' has been held for us by
4505 * falloc_noinstall().
4511 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4513 KASSERT(fp->f_ops == &badfileops,
4514 ("VOP_OPEN in fhopen() set f_ops"));
4515 KASSERT(td->td_dupfd < 0,
4516 ("fhopen() encountered fdopen()"));
4525 finit_vnode(fp, fmode, NULL, &vnops);
4527 if ((fmode & O_TRUNC) != 0) {
4528 error = fo_truncate(fp, 0, td->td_ucred, td);
4533 error = finstall(td, fp, &indx, fmode, NULL);
4536 td->td_retval[0] = indx;
4541 * Stat an (NFS) file handle.
4543 #ifndef _SYS_SYSPROTO_H_
4544 struct fhstat_args {
4545 struct fhandle *u_fhp;
4550 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4556 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4559 error = kern_fhstat(td, fh, &sb);
4561 error = copyout(&sb, uap->sb, sizeof(sb));
4566 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4572 error = priv_check(td, PRIV_VFS_FHSTAT);
4575 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4577 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4581 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4587 * Implement fstatfs() for (NFS) file handles.
4589 #ifndef _SYS_SYSPROTO_H_
4590 struct fhstatfs_args {
4591 struct fhandle *u_fhp;
4596 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4602 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4605 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4606 error = kern_fhstatfs(td, fh, sfp);
4608 error = copyout(sfp, uap->buf, sizeof(*sfp));
4609 free(sfp, M_STATFS);
4614 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4620 error = priv_check(td, PRIV_VFS_FHSTATFS);
4623 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4625 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4631 error = prison_canseemount(td->td_ucred, mp);
4635 error = mac_mount_check_stat(td->td_ucred, mp);
4639 error = VFS_STATFS(mp, buf);
4646 * Unlike madvise(2), we do not make a best effort to remember every
4647 * possible caching hint. Instead, we remember the last setting with
4648 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4649 * region of any current setting.
4652 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4655 struct fadvise_info *fa, *new;
4661 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4663 AUDIT_ARG_VALUE(advice);
4665 case POSIX_FADV_SEQUENTIAL:
4666 case POSIX_FADV_RANDOM:
4667 case POSIX_FADV_NOREUSE:
4668 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4670 case POSIX_FADV_NORMAL:
4671 case POSIX_FADV_WILLNEED:
4672 case POSIX_FADV_DONTNEED:
4678 /* XXX: CAP_POSIX_FADVISE? */
4680 error = fget(td, fd, &cap_no_rights, &fp);
4683 AUDIT_ARG_FILE(td->td_proc, fp);
4684 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4688 if (fp->f_type != DTYPE_VNODE) {
4693 if (vp->v_type != VREG) {
4700 end = offset + len - 1;
4702 case POSIX_FADV_SEQUENTIAL:
4703 case POSIX_FADV_RANDOM:
4704 case POSIX_FADV_NOREUSE:
4706 * Try to merge any existing non-standard region with
4707 * this new region if possible, otherwise create a new
4708 * non-standard region for this request.
4710 mtx_pool_lock(mtxpool_sleep, fp);
4712 if (fa != NULL && fa->fa_advice == advice &&
4713 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4714 (end != OFF_MAX && fa->fa_start == end + 1) ||
4715 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4716 if (offset < fa->fa_start)
4717 fa->fa_start = offset;
4718 if (end > fa->fa_end)
4721 new->fa_advice = advice;
4722 new->fa_start = offset;
4727 mtx_pool_unlock(mtxpool_sleep, fp);
4729 case POSIX_FADV_NORMAL:
4731 * If a the "normal" region overlaps with an existing
4732 * non-standard region, trim or remove the
4733 * non-standard region.
4735 mtx_pool_lock(mtxpool_sleep, fp);
4738 if (offset <= fa->fa_start && end >= fa->fa_end) {
4740 fp->f_advice = NULL;
4741 } else if (offset <= fa->fa_start &&
4742 end >= fa->fa_start)
4743 fa->fa_start = end + 1;
4744 else if (offset <= fa->fa_end && end >= fa->fa_end)
4745 fa->fa_end = offset - 1;
4746 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4748 * If the "normal" region is a middle
4749 * portion of the existing
4750 * non-standard region, just remove
4751 * the whole thing rather than picking
4752 * one side or the other to
4756 fp->f_advice = NULL;
4759 mtx_pool_unlock(mtxpool_sleep, fp);
4761 case POSIX_FADV_WILLNEED:
4762 case POSIX_FADV_DONTNEED:
4763 error = VOP_ADVISE(vp, offset, end, advice);
4769 free(new, M_FADVISE);
4774 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4778 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4780 return (kern_posix_error(td, error));
4784 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4785 off_t *outoffp, size_t len, unsigned int flags)
4787 struct file *infp, *outfp;
4788 struct vnode *invp, *outvp;
4791 void *rl_rcookie, *rl_wcookie;
4792 off_t savinoff, savoutoff;
4794 infp = outfp = NULL;
4795 rl_rcookie = rl_wcookie = NULL;
4804 if (len > SSIZE_MAX)
4806 * Although the len argument is size_t, the return argument
4807 * is ssize_t (which is signed). Therefore a size that won't
4808 * fit in ssize_t can't be returned.
4812 /* Get the file structures for the file descriptors. */
4813 error = fget_read(td, infd, &cap_read_rights, &infp);
4816 if (infp->f_ops == &badfileops) {
4820 if (infp->f_vnode == NULL) {
4824 error = fget_write(td, outfd, &cap_write_rights, &outfp);
4827 if (outfp->f_ops == &badfileops) {
4831 if (outfp->f_vnode == NULL) {
4836 /* Set the offset pointers to the correct place. */
4838 inoffp = &infp->f_offset;
4839 if (outoffp == NULL)
4840 outoffp = &outfp->f_offset;
4842 savoutoff = *outoffp;
4844 invp = infp->f_vnode;
4845 outvp = outfp->f_vnode;
4846 /* Sanity check the f_flag bits. */
4847 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4848 (infp->f_flag & FREAD) == 0) {
4853 /* If len == 0, just return 0. */
4858 * If infp and outfp refer to the same file, the byte ranges cannot
4861 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4862 savoutoff) || (savinoff > savoutoff && savoutoff + len >
4868 /* Range lock the byte ranges for both invp and outvp. */
4870 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4872 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4874 if (rl_rcookie != NULL)
4876 vn_rangelock_unlock(outvp, rl_wcookie);
4877 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4878 vn_rangelock_unlock(invp, rl_rcookie);
4882 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4883 flags, infp->f_cred, outfp->f_cred, td);
4885 if (rl_rcookie != NULL)
4886 vn_rangelock_unlock(invp, rl_rcookie);
4887 if (rl_wcookie != NULL)
4888 vn_rangelock_unlock(outvp, rl_wcookie);
4889 if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4891 *outoffp = savoutoff;
4897 td->td_retval[0] = retlen;
4902 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4904 off_t inoff, outoff, *inoffp, *outoffp;
4907 inoffp = outoffp = NULL;
4908 if (uap->inoffp != NULL) {
4909 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4914 if (uap->outoffp != NULL) {
4915 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4920 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4921 outoffp, uap->len, uap->flags);
4922 if (error == 0 && uap->inoffp != NULL)
4923 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4924 if (error == 0 && uap->outoffp != NULL)
4925 error = copyout(outoffp, uap->outoffp, sizeof(off_t));