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>
47 #ifdef COMPAT_FREEBSD11
48 #include <sys/abi_compat.h>
52 #include <sys/capsicum.h>
54 #include <sys/malloc.h>
55 #include <sys/mount.h>
56 #include <sys/mutex.h>
57 #include <sys/sysproto.h>
58 #include <sys/namei.h>
59 #include <sys/filedesc.h>
60 #include <sys/kernel.h>
61 #include <sys/fcntl.h>
63 #include <sys/filio.h>
64 #include <sys/limits.h>
65 #include <sys/linker.h>
66 #include <sys/rwlock.h>
70 #include <sys/unistd.h>
71 #include <sys/vnode.h>
74 #include <sys/dirent.h>
76 #include <sys/syscallsubr.h>
77 #include <sys/sysctl.h>
79 #include <sys/ktrace.h>
82 #include <machine/stdarg.h>
84 #include <security/audit/audit.h>
85 #include <security/mac/mac_framework.h>
88 #include <vm/vm_object.h>
89 #include <vm/vm_page.h>
92 #include <fs/devfs/devfs.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)
200 AUDIT_ARG_CMD(uap->cmd);
201 AUDIT_ARG_UID(uap->uid);
202 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
204 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
206 if ((error = namei(&nd)) != 0)
209 mp = nd.ni_vp->v_mount;
212 error = vfs_busy(mp, 0);
218 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
221 * Since quota on/off operations typically need to open quota
222 * files, the implementation may need to unbusy the mount point
223 * before calling into namei. Otherwise, unmount might be
224 * started between two vfs_busy() invocations (first is ours,
225 * second is from mount point cross-walk code in lookup()),
228 * Avoid unbusying mp if the implementation indicates it has
238 * Used by statfs conversion routines to scale the block size up if
239 * necessary so that all of the block counts are <= 'max_size'. Note
240 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
244 statfs_scale_blocks(struct statfs *sf, long max_size)
249 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
252 * Attempt to scale the block counts to give a more accurate
253 * overview to userland of the ratio of free space to used
254 * space. To do this, find the largest block count and compute
255 * a divisor that lets it fit into a signed integer <= max_size.
257 if (sf->f_bavail < 0)
258 count = -sf->f_bavail;
260 count = sf->f_bavail;
261 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
262 if (count <= max_size)
265 count >>= flsl(max_size);
272 sf->f_bsize <<= shift;
273 sf->f_blocks >>= shift;
274 sf->f_bfree >>= shift;
275 sf->f_bavail >>= shift;
279 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
285 error = vfs_busy(mp, 0);
290 error = mac_mount_check_stat(td->td_ucred, mp);
294 error = VFS_STATFS(mp, buf);
297 if (priv_check_cred_vfs_generation(td->td_ucred)) {
298 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
299 prison_enforce_statfs(td->td_ucred, mp, buf);
307 * Get filesystem statistics.
309 #ifndef _SYS_SYSPROTO_H_
316 sys_statfs(struct thread *td, struct statfs_args *uap)
321 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
322 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
324 error = copyout(sfp, uap->buf, sizeof(struct statfs));
330 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
337 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
341 mp = vfs_ref_from_vp(nd.ni_vp);
344 return (kern_do_statfs(td, mp, buf));
348 * Get filesystem statistics.
350 #ifndef _SYS_SYSPROTO_H_
351 struct fstatfs_args {
357 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
362 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
363 error = kern_fstatfs(td, uap->fd, sfp);
365 error = copyout(sfp, uap->buf, sizeof(struct statfs));
371 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
379 error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
384 if (AUDITING_TD(td)) {
385 vn_lock(vp, LK_SHARED | LK_RETRY);
386 AUDIT_ARG_VNODE1(vp);
390 mp = vfs_ref_from_vp(vp);
392 return (kern_do_statfs(td, mp, buf));
396 * Get statistics on all filesystems.
398 #ifndef _SYS_SYSPROTO_H_
399 struct getfsstat_args {
406 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
411 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
413 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
414 UIO_USERSPACE, uap->mode);
416 td->td_retval[0] = count;
421 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
422 * The caller is responsible for freeing memory which will be allocated
426 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
427 size_t *countp, enum uio_seg bufseg, int mode)
429 struct mount *mp, *nmp;
430 struct statfs *sfsp, *sp, *sptmp, *tofree;
431 size_t count, maxcount;
439 if (bufseg == UIO_SYSSPACE)
444 maxcount = bufsize / sizeof(struct statfs);
448 } else if (bufseg == UIO_USERSPACE) {
451 } else /* if (bufseg == UIO_SYSSPACE) */ {
453 mtx_lock(&mountlist_mtx);
454 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
457 mtx_unlock(&mountlist_mtx);
458 if (maxcount > count)
460 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
467 * If there is no target buffer they only want the count.
469 * This could be TAILQ_FOREACH but it is open-coded to match the original
473 mtx_lock(&mountlist_mtx);
474 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
475 if (prison_canseemount(td->td_ucred, mp) != 0) {
476 nmp = TAILQ_NEXT(mp, mnt_list);
480 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
481 nmp = TAILQ_NEXT(mp, mnt_list);
486 nmp = TAILQ_NEXT(mp, mnt_list);
488 mtx_unlock(&mountlist_mtx);
494 * They want the entire thing.
496 * Short-circuit the corner case of no room for anything, avoids
503 mtx_lock(&mountlist_mtx);
504 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
505 if (prison_canseemount(td->td_ucred, mp) != 0) {
506 nmp = TAILQ_NEXT(mp, mnt_list);
510 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
511 nmp = TAILQ_NEXT(mp, mnt_list);
515 if (mode == MNT_WAIT) {
516 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
518 * If vfs_busy() failed, and MBF_NOWAIT
519 * wasn't passed, then the mp is gone.
520 * Furthermore, because of MBF_MNTLSTLOCK,
521 * the mountlist_mtx was dropped. We have
522 * no other choice than to start over.
524 mtx_unlock(&mountlist_mtx);
525 free(tofree, M_STATFS);
529 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
530 nmp = TAILQ_NEXT(mp, mnt_list);
536 * If MNT_NOWAIT is specified, do not refresh
539 if (mode != MNT_NOWAIT) {
540 error = VFS_STATFS(mp, sp);
542 mtx_lock(&mountlist_mtx);
543 nmp = TAILQ_NEXT(mp, mnt_list);
548 if (priv_check_cred_vfs_generation(td->td_ucred)) {
549 sptmp = malloc(sizeof(struct statfs), M_STATFS,
552 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
553 prison_enforce_statfs(td->td_ucred, mp, sptmp);
557 if (bufseg == UIO_SYSSPACE) {
558 bcopy(sp, sfsp, sizeof(*sp));
559 free(sptmp, M_STATFS);
560 } else /* if (bufseg == UIO_USERSPACE) */ {
561 error = copyout(sp, sfsp, sizeof(*sp));
562 free(sptmp, M_STATFS);
571 if (count == maxcount) {
576 mtx_lock(&mountlist_mtx);
577 nmp = TAILQ_NEXT(mp, mnt_list);
580 mtx_unlock(&mountlist_mtx);
586 #ifdef COMPAT_FREEBSD4
588 * Get old format filesystem statistics.
590 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
592 #ifndef _SYS_SYSPROTO_H_
593 struct freebsd4_statfs_args {
599 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
605 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
606 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
608 freebsd4_cvtstatfs(sfp, &osb);
609 error = copyout(&osb, uap->buf, sizeof(osb));
616 * Get filesystem statistics.
618 #ifndef _SYS_SYSPROTO_H_
619 struct freebsd4_fstatfs_args {
625 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
631 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
632 error = kern_fstatfs(td, uap->fd, sfp);
634 freebsd4_cvtstatfs(sfp, &osb);
635 error = copyout(&osb, uap->buf, sizeof(osb));
642 * Get statistics on all filesystems.
644 #ifndef _SYS_SYSPROTO_H_
645 struct freebsd4_getfsstat_args {
652 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
654 struct statfs *buf, *sp;
659 if (uap->bufsize < 0)
661 count = uap->bufsize / sizeof(struct ostatfs);
662 if (count > SIZE_MAX / sizeof(struct statfs))
664 size = count * sizeof(struct statfs);
665 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
668 td->td_retval[0] = count;
671 while (count != 0 && error == 0) {
672 freebsd4_cvtstatfs(sp, &osb);
673 error = copyout(&osb, uap->buf, sizeof(osb));
684 * Implement fstatfs() for (NFS) file handles.
686 #ifndef _SYS_SYSPROTO_H_
687 struct freebsd4_fhstatfs_args {
688 struct fhandle *u_fhp;
693 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
700 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
703 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
704 error = kern_fhstatfs(td, fh, sfp);
706 freebsd4_cvtstatfs(sfp, &osb);
707 error = copyout(&osb, uap->buf, sizeof(osb));
714 * Convert a new format statfs structure to an old format statfs structure.
717 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
720 statfs_scale_blocks(nsp, LONG_MAX);
721 bzero(osp, sizeof(*osp));
722 osp->f_bsize = nsp->f_bsize;
723 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
724 osp->f_blocks = nsp->f_blocks;
725 osp->f_bfree = nsp->f_bfree;
726 osp->f_bavail = nsp->f_bavail;
727 osp->f_files = MIN(nsp->f_files, LONG_MAX);
728 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
729 osp->f_owner = nsp->f_owner;
730 osp->f_type = nsp->f_type;
731 osp->f_flags = nsp->f_flags;
732 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
733 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
734 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
735 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
736 strlcpy(osp->f_fstypename, nsp->f_fstypename,
737 MIN(MFSNAMELEN, OMFSNAMELEN));
738 strlcpy(osp->f_mntonname, nsp->f_mntonname,
739 MIN(MNAMELEN, OMNAMELEN));
740 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
741 MIN(MNAMELEN, OMNAMELEN));
742 osp->f_fsid = nsp->f_fsid;
744 #endif /* COMPAT_FREEBSD4 */
746 #if defined(COMPAT_FREEBSD11)
748 * Get old format filesystem statistics.
750 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
753 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
755 struct freebsd11_statfs osb;
759 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
760 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
762 freebsd11_cvtstatfs(sfp, &osb);
763 error = copyout(&osb, uap->buf, sizeof(osb));
770 * Get filesystem statistics.
773 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
775 struct freebsd11_statfs osb;
779 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
780 error = kern_fstatfs(td, uap->fd, sfp);
782 freebsd11_cvtstatfs(sfp, &osb);
783 error = copyout(&osb, uap->buf, sizeof(osb));
790 * Get statistics on all filesystems.
793 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
795 return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode));
799 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf,
800 long bufsize, int mode)
802 struct freebsd11_statfs osb;
803 struct statfs *buf, *sp;
810 count = bufsize / sizeof(struct ostatfs);
811 size = count * sizeof(struct statfs);
812 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode);
814 td->td_retval[0] = count;
817 while (count > 0 && error == 0) {
818 freebsd11_cvtstatfs(sp, &osb);
819 error = copyout(&osb, ubuf, sizeof(osb));
830 * Implement fstatfs() for (NFS) file handles.
833 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
835 struct freebsd11_statfs osb;
840 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
843 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
844 error = kern_fhstatfs(td, fh, sfp);
846 freebsd11_cvtstatfs(sfp, &osb);
847 error = copyout(&osb, uap->buf, sizeof(osb));
854 * Convert a new format statfs structure to an old format statfs structure.
857 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
860 bzero(osp, sizeof(*osp));
861 osp->f_version = FREEBSD11_STATFS_VERSION;
862 osp->f_type = nsp->f_type;
863 osp->f_flags = nsp->f_flags;
864 osp->f_bsize = nsp->f_bsize;
865 osp->f_iosize = nsp->f_iosize;
866 osp->f_blocks = nsp->f_blocks;
867 osp->f_bfree = nsp->f_bfree;
868 osp->f_bavail = nsp->f_bavail;
869 osp->f_files = nsp->f_files;
870 osp->f_ffree = nsp->f_ffree;
871 osp->f_syncwrites = nsp->f_syncwrites;
872 osp->f_asyncwrites = nsp->f_asyncwrites;
873 osp->f_syncreads = nsp->f_syncreads;
874 osp->f_asyncreads = nsp->f_asyncreads;
875 osp->f_namemax = nsp->f_namemax;
876 osp->f_owner = nsp->f_owner;
877 osp->f_fsid = nsp->f_fsid;
878 strlcpy(osp->f_fstypename, nsp->f_fstypename,
879 MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
880 strlcpy(osp->f_mntonname, nsp->f_mntonname,
881 MIN(MNAMELEN, sizeof(osp->f_mntonname)));
882 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
883 MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
885 #endif /* COMPAT_FREEBSD11 */
888 * Change current working directory to a given file descriptor.
890 #ifndef _SYS_SYSPROTO_H_
896 sys_fchdir(struct thread *td, struct fchdir_args *uap)
898 struct vnode *vp, *tdp;
903 AUDIT_ARG_FD(uap->fd);
904 error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
911 vn_lock(vp, LK_SHARED | LK_RETRY);
912 AUDIT_ARG_VNODE1(vp);
913 error = change_dir(vp, td);
914 while (!error && (mp = vp->v_mountedhere) != NULL) {
917 error = VFS_ROOT(mp, LK_SHARED, &tdp);
934 * Change current working directory (``.'').
936 #ifndef _SYS_SYSPROTO_H_
942 sys_chdir(struct thread *td, struct chdir_args *uap)
945 return (kern_chdir(td, uap->path, UIO_USERSPACE));
949 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
954 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
956 if ((error = namei(&nd)) != 0)
958 if ((error = change_dir(nd.ni_vp, td)) != 0) {
963 VOP_UNLOCK(nd.ni_vp);
965 pwd_chdir(td, nd.ni_vp);
969 static int unprivileged_chroot = 0;
970 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
971 &unprivileged_chroot, 0,
972 "Unprivileged processes can use chroot(2)");
974 * Change notion of root (``/'') directory.
976 #ifndef _SYS_SYSPROTO_H_
982 sys_chroot(struct thread *td, struct chroot_args *uap)
988 error = priv_check(td, PRIV_VFS_CHROOT);
992 if (unprivileged_chroot == 0 ||
993 (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
999 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
1000 UIO_USERSPACE, uap->path);
1004 error = change_dir(nd.ni_vp, td);
1008 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
1012 VOP_UNLOCK(nd.ni_vp);
1013 error = pwd_chroot(td, nd.ni_vp);
1015 NDFREE_NOTHING(&nd);
1020 NDFREE_NOTHING(&nd);
1025 * Common routine for chroot and chdir. Callers must provide a locked vnode
1029 change_dir(struct vnode *vp, struct thread *td)
1035 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1036 if (vp->v_type != VDIR)
1039 error = mac_vnode_check_chdir(td->td_ucred, vp);
1043 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1046 static __inline void
1047 flags_to_rights(int flags, cap_rights_t *rightsp)
1049 if (flags & O_EXEC) {
1050 cap_rights_set_one(rightsp, CAP_FEXECVE);
1054 switch ((flags & O_ACCMODE)) {
1056 cap_rights_set_one(rightsp, CAP_READ);
1059 cap_rights_set_one(rightsp, CAP_READ);
1062 cap_rights_set_one(rightsp, CAP_WRITE);
1063 if (!(flags & (O_APPEND | O_TRUNC)))
1064 cap_rights_set_one(rightsp, CAP_SEEK);
1069 if (flags & O_CREAT)
1070 cap_rights_set_one(rightsp, CAP_CREATE);
1072 if (flags & O_TRUNC)
1073 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1075 if (flags & (O_SYNC | O_FSYNC))
1076 cap_rights_set_one(rightsp, CAP_FSYNC);
1078 if (flags & (O_EXLOCK | O_SHLOCK))
1079 cap_rights_set_one(rightsp, CAP_FLOCK);
1083 * Check permissions, allocate an open file structure, and call the device
1084 * open routine if any.
1086 #ifndef _SYS_SYSPROTO_H_
1094 sys_open(struct thread *td, struct open_args *uap)
1097 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1098 uap->flags, uap->mode));
1101 #ifndef _SYS_SYSPROTO_H_
1102 struct openat_args {
1110 sys_openat(struct thread *td, struct openat_args *uap)
1113 AUDIT_ARG_FD(uap->fd);
1114 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1119 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1120 int flags, int mode)
1122 struct proc *p = td->td_proc;
1123 struct filedesc *fdp;
1124 struct pwddesc *pdp;
1127 struct nameidata nd;
1128 cap_rights_t rights;
1129 int cmode, error, indx;
1135 AUDIT_ARG_FFLAGS(flags);
1136 AUDIT_ARG_MODE(mode);
1137 cap_rights_init_one(&rights, CAP_LOOKUP);
1138 flags_to_rights(flags, &rights);
1141 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1142 * may be specified. On the other hand, for O_PATH any mode
1143 * except O_EXEC is ignored.
1145 if ((flags & O_PATH) != 0) {
1146 flags &= ~(O_CREAT | O_ACCMODE);
1147 } else if ((flags & O_EXEC) != 0) {
1148 if (flags & O_ACCMODE)
1150 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1153 flags = FFLAGS(flags);
1157 * Allocate a file structure. The descriptor to reference it
1158 * is allocated and used by finstall_refed() below.
1160 error = falloc_noinstall(td, &fp);
1163 /* Set the flags early so the finit in devfs can pick them up. */
1164 fp->f_flag = flags & FMASK;
1165 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1166 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | WANTIOCTLCAPS,
1167 pathseg, path, fd, &rights);
1168 td->td_dupfd = -1; /* XXX check for fdopen */
1169 error = vn_open_cred(&nd, &flags, cmode, VN_OPEN_WANTIOCTLCAPS,
1173 * If the vn_open replaced the method vector, something
1174 * wonderous happened deep below and we just pass it up
1175 * pretending we know what we do.
1177 if (error == ENXIO && fp->f_ops != &badfileops) {
1178 MPASS((flags & O_PATH) == 0);
1183 * Handle special fdopen() case. bleh.
1185 * Don't do this for relative (capability) lookups; we don't
1186 * understand exactly what would happen, and we don't think
1187 * that it ever should.
1189 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1190 (error == ENODEV || error == ENXIO) &&
1191 td->td_dupfd >= 0) {
1192 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1205 * Store the vnode, for any f_type. Typically, the vnode use
1206 * count is decremented by direct call to vn_closefile() for
1207 * files that switched type in the cdevsw fdopen() method.
1212 * If the file wasn't claimed by devfs bind it to the normal
1213 * vnode operations here.
1215 if (fp->f_ops == &badfileops) {
1216 KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1217 ("Unexpected fifo fp %p vp %p", fp, vp));
1218 if ((flags & O_PATH) != 0) {
1219 finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1220 DTYPE_VNODE, NULL, &path_fileops);
1222 finit_vnode(fp, flags, NULL, &vnops);
1227 if (flags & O_TRUNC) {
1228 error = fo_truncate(fp, 0, td->td_ucred, td);
1234 * If we haven't already installed the FD (for dupfdopen), do so now.
1237 struct filecaps *fcaps;
1240 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1241 fcaps = &nd.ni_filecaps;
1245 error = finstall_refed(td, fp, &indx, flags, fcaps);
1246 /* On success finstall_refed() consumes fcaps. */
1251 NDFREE_IOCTLCAPS(&nd);
1252 falloc_abort(td, fp);
1255 td->td_retval[0] = indx;
1258 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1259 NDFREE_IOCTLCAPS(&nd);
1260 falloc_abort(td, fp);
1268 #ifndef _SYS_SYSPROTO_H_
1269 struct ocreat_args {
1275 ocreat(struct thread *td, struct ocreat_args *uap)
1278 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1279 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1281 #endif /* COMPAT_43 */
1284 * Create a special file.
1286 #ifndef _SYS_SYSPROTO_H_
1287 struct mknodat_args {
1295 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1298 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1302 #if defined(COMPAT_FREEBSD11)
1304 freebsd11_mknod(struct thread *td,
1305 struct freebsd11_mknod_args *uap)
1308 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1309 uap->mode, uap->dev));
1313 freebsd11_mknodat(struct thread *td,
1314 struct freebsd11_mknodat_args *uap)
1317 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1320 #endif /* COMPAT_FREEBSD11 */
1323 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1324 int mode, dev_t dev)
1329 struct nameidata nd;
1330 int error, whiteout = 0;
1332 AUDIT_ARG_MODE(mode);
1334 switch (mode & S_IFMT) {
1337 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1338 if (error == 0 && dev == VNOVAL)
1342 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1346 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1357 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1358 NOCACHE, pathseg, path, fd, &cap_mknodat_rights);
1359 if ((error = namei(&nd)) != 0)
1364 if (vp == nd.ni_dvp)
1372 vattr.va_mode = (mode & ALLPERMS) &
1373 ~td->td_proc->p_pd->pd_cmask;
1374 vattr.va_rdev = dev;
1377 switch (mode & S_IFMT) {
1379 vattr.va_type = VCHR;
1382 vattr.va_type = VBLK;
1388 panic("kern_mknod: invalid mode");
1391 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1394 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1399 if (error == 0 && !whiteout)
1400 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1401 &nd.ni_cnd, &vattr);
1405 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1407 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1408 &nd.ni_cnd, &vattr);
1411 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1413 vn_finished_write(mp);
1415 if (error == ERELOOKUP)
1421 * Create a named pipe.
1423 #ifndef _SYS_SYSPROTO_H_
1424 struct mkfifo_args {
1430 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1433 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1437 #ifndef _SYS_SYSPROTO_H_
1438 struct mkfifoat_args {
1445 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1448 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1453 kern_mkfifoat(struct thread *td, int fd, const char *path,
1454 enum uio_seg pathseg, int mode)
1458 struct nameidata nd;
1461 AUDIT_ARG_MODE(mode);
1465 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1466 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights);
1467 if ((error = namei(&nd)) != 0)
1469 if (nd.ni_vp != NULL) {
1471 if (nd.ni_vp == nd.ni_dvp)
1478 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1481 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1486 vattr.va_type = VFIFO;
1487 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1489 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1494 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1498 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1499 vn_finished_write(mp);
1501 if (error == ERELOOKUP)
1507 * Make a hard file link.
1509 #ifndef _SYS_SYSPROTO_H_
1516 sys_link(struct thread *td, struct link_args *uap)
1519 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1520 UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1523 #ifndef _SYS_SYSPROTO_H_
1524 struct linkat_args {
1533 sys_linkat(struct thread *td, struct linkat_args *uap)
1536 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1537 UIO_USERSPACE, uap->flag));
1540 int hardlink_check_uid = 0;
1541 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1542 &hardlink_check_uid, 0,
1543 "Unprivileged processes cannot create hard links to files owned by other "
1545 static int hardlink_check_gid = 0;
1546 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1547 &hardlink_check_gid, 0,
1548 "Unprivileged processes cannot create hard links to files owned by other "
1552 can_hardlink(struct vnode *vp, struct ucred *cred)
1557 if (!hardlink_check_uid && !hardlink_check_gid)
1560 error = VOP_GETATTR(vp, &va, cred);
1564 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1565 error = priv_check_cred(cred, PRIV_VFS_LINK);
1570 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1571 error = priv_check_cred(cred, PRIV_VFS_LINK);
1580 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1581 const char *path2, enum uio_seg segflag, int flag)
1583 struct nameidata nd;
1586 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1587 AT_EMPTY_PATH)) != 0)
1593 NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1594 AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1595 segflag, path1, fd1, &cap_linkat_source_rights);
1596 if ((error = namei(&nd)) != 0)
1599 if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1600 error = priv_check(td, PRIV_VFS_FHOPEN);
1606 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1607 } while (error == EAGAIN || error == ERELOOKUP);
1612 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1613 enum uio_seg segflag)
1615 struct nameidata nd;
1619 if (vp->v_type == VDIR) {
1621 return (EPERM); /* POSIX */
1623 NDINIT_ATRIGHTS(&nd, CREATE,
1624 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1625 &cap_linkat_target_rights);
1626 if ((error = namei(&nd)) == 0) {
1627 if (nd.ni_vp != NULL) {
1629 if (nd.ni_dvp == nd.ni_vp)
1636 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1638 * Cross-device link. No need to recheck
1639 * vp->v_type, since it cannot change, except
1646 } else if (vn_lock(vp, LK_EXCLUSIVE) == 0) {
1647 error = can_hardlink(vp, td->td_ucred);
1650 error = mac_vnode_check_link(td->td_ucred,
1651 nd.ni_dvp, vp, &nd.ni_cnd);
1659 error = vn_start_write(vp, &mp, V_NOWAIT);
1664 error = vn_start_write(NULL, &mp,
1670 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1671 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1672 vn_finished_write(mp);
1688 * Make a symbolic link.
1690 #ifndef _SYS_SYSPROTO_H_
1691 struct symlink_args {
1697 sys_symlink(struct thread *td, struct symlink_args *uap)
1700 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1704 #ifndef _SYS_SYSPROTO_H_
1705 struct symlinkat_args {
1712 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1715 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1720 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1721 enum uio_seg segflg)
1725 const char *syspath;
1727 struct nameidata nd;
1730 if (segflg == UIO_SYSSPACE) {
1733 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1734 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1738 AUDIT_ARG_TEXT(syspath);
1742 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1743 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights);
1744 if ((error = namei(&nd)) != 0)
1748 if (nd.ni_vp == nd.ni_dvp)
1757 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1760 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1765 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1767 vattr.va_type = VLNK;
1768 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1773 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1777 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1778 vn_finished_write(mp);
1780 if (error == ERELOOKUP)
1783 if (segflg != UIO_SYSSPACE)
1784 uma_zfree(namei_zone, tmppath);
1789 * Delete a whiteout from the filesystem.
1791 #ifndef _SYS_SYSPROTO_H_
1792 struct undelete_args {
1797 sys_undelete(struct thread *td, struct undelete_args *uap)
1800 struct nameidata nd;
1806 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1807 UIO_USERSPACE, uap->path);
1812 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1814 if (nd.ni_vp == nd.ni_dvp)
1822 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1825 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1829 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1832 vn_finished_write(mp);
1833 if (error == ERELOOKUP)
1839 * Delete a name from the filesystem.
1841 #ifndef _SYS_SYSPROTO_H_
1842 struct unlink_args {
1847 sys_unlink(struct thread *td, struct unlink_args *uap)
1850 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1855 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1856 int flag, enum uio_seg pathseg, ino_t oldinum)
1859 if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1862 if ((flag & AT_REMOVEDIR) != 0)
1863 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1865 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1868 #ifndef _SYS_SYSPROTO_H_
1869 struct unlinkat_args {
1876 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1879 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1883 #ifndef _SYS_SYSPROTO_H_
1884 struct funlinkat_args {
1892 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1895 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1900 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1901 enum uio_seg pathseg, int flag, ino_t oldinum)
1906 struct nameidata nd;
1911 if (fd != FD_NONE) {
1912 error = getvnode_path(td, fd, &cap_no_rights, &fp);
1920 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1921 at2cnpflags(flag, AT_RESOLVE_BENEATH),
1922 pathseg, path, dfd, &cap_unlinkat_rights);
1923 if ((error = namei(&nd)) != 0) {
1924 if (error == EINVAL)
1929 if (vp->v_type == VDIR && oldinum == 0) {
1930 error = EPERM; /* POSIX */
1931 } else if (oldinum != 0 &&
1932 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
1933 sb.st_ino != oldinum) {
1934 error = EIDRM; /* Identifier removed */
1935 } else if (fp != NULL && fp->f_vnode != vp) {
1936 if (VN_IS_DOOMED(fp->f_vnode))
1942 * The root of a mounted filesystem cannot be deleted.
1944 * XXX: can this only be a VDIR case?
1946 if (vp->v_vflag & VV_ROOT)
1950 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1953 if (vp == nd.ni_dvp)
1957 if ((error = vn_start_write(NULL, &mp,
1958 V_XSLEEP | PCATCH)) != 0) {
1964 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1969 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1970 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1974 vn_finished_write(mp);
1978 if (vp == nd.ni_dvp)
1982 if (error == ERELOOKUP)
1991 * Reposition read/write file offset.
1993 #ifndef _SYS_SYSPROTO_H_
2002 sys_lseek(struct thread *td, struct lseek_args *uap)
2005 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2009 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2015 error = fget(td, fd, &cap_seek_rights, &fp);
2018 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2019 fo_seek(fp, offset, whence, td) : ESPIPE;
2024 #if defined(COMPAT_43)
2026 * Reposition read/write file offset.
2028 #ifndef _SYS_SYSPROTO_H_
2029 struct olseek_args {
2036 olseek(struct thread *td, struct olseek_args *uap)
2039 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2041 #endif /* COMPAT_43 */
2043 #if defined(COMPAT_FREEBSD6)
2044 /* Version with the 'pad' argument */
2046 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2049 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2054 * Check access permissions using passed credentials.
2057 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2063 /* Flags == 0 means only check for existence. */
2064 if (user_flags == 0)
2068 if (user_flags & R_OK)
2070 if (user_flags & W_OK)
2072 if (user_flags & X_OK)
2075 error = mac_vnode_check_access(cred, vp, accmode);
2079 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2080 error = VOP_ACCESS(vp, accmode, cred, td);
2085 * Check access permissions using "real" credentials.
2087 #ifndef _SYS_SYSPROTO_H_
2088 struct access_args {
2094 sys_access(struct thread *td, struct access_args *uap)
2097 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2101 #ifndef _SYS_SYSPROTO_H_
2102 struct faccessat_args {
2110 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2113 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2118 kern_accessat(struct thread *td, int fd, const char *path,
2119 enum uio_seg pathseg, int flag, int amode)
2121 struct ucred *cred, *usecred;
2123 struct nameidata nd;
2126 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2128 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2132 * Create and modify a temporary credential instead of one that
2133 * is potentially shared (if we need one).
2135 cred = td->td_ucred;
2136 if ((flag & AT_EACCESS) == 0 &&
2137 ((cred->cr_uid != cred->cr_ruid ||
2138 cred->cr_rgid != cred->cr_groups[0]))) {
2139 usecred = crdup(cred);
2140 usecred->cr_uid = cred->cr_ruid;
2141 usecred->cr_groups[0] = cred->cr_rgid;
2142 td->td_ucred = usecred;
2145 AUDIT_ARG_VALUE(amode);
2146 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2147 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2148 AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights);
2149 if ((error = namei(&nd)) != 0)
2153 error = vn_access(vp, amode, usecred, td);
2154 NDFREE_NOTHING(&nd);
2157 if (usecred != cred) {
2158 td->td_ucred = cred;
2165 * Check access permissions using "effective" credentials.
2167 #ifndef _SYS_SYSPROTO_H_
2168 struct eaccess_args {
2174 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2177 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2178 AT_EACCESS, uap->amode));
2181 #if defined(COMPAT_43)
2183 * Get file status; this version follows links.
2185 #ifndef _SYS_SYSPROTO_H_
2192 ostat(struct thread *td, struct ostat_args *uap)
2198 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2203 return (copyout(&osb, uap->ub, sizeof (osb)));
2207 * Get file status; this version does not follow links.
2209 #ifndef _SYS_SYSPROTO_H_
2210 struct olstat_args {
2216 olstat(struct thread *td, struct olstat_args *uap)
2222 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2223 UIO_USERSPACE, &sb, NULL);
2227 return (copyout(&osb, uap->ub, sizeof (osb)));
2231 * Convert from an old to a new stat structure.
2232 * XXX: many values are blindly truncated.
2235 cvtstat(struct stat *st, struct ostat *ost)
2238 bzero(ost, sizeof(*ost));
2239 ost->st_dev = st->st_dev;
2240 ost->st_ino = st->st_ino;
2241 ost->st_mode = st->st_mode;
2242 ost->st_nlink = st->st_nlink;
2243 ost->st_uid = st->st_uid;
2244 ost->st_gid = st->st_gid;
2245 ost->st_rdev = st->st_rdev;
2246 ost->st_size = MIN(st->st_size, INT32_MAX);
2247 ost->st_atim = st->st_atim;
2248 ost->st_mtim = st->st_mtim;
2249 ost->st_ctim = st->st_ctim;
2250 ost->st_blksize = st->st_blksize;
2251 ost->st_blocks = st->st_blocks;
2252 ost->st_flags = st->st_flags;
2253 ost->st_gen = st->st_gen;
2255 #endif /* COMPAT_43 */
2257 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2258 int ino64_trunc_error;
2259 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2260 &ino64_trunc_error, 0,
2261 "Error on truncation of device, file or inode number, or link count");
2264 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2267 ost->st_dev = st->st_dev;
2268 if (ost->st_dev != st->st_dev) {
2269 switch (ino64_trunc_error) {
2272 * Since dev_t is almost raw, don't clamp to the
2273 * maximum for case 2, but ignore the error.
2280 ost->st_ino = st->st_ino;
2281 if (ost->st_ino != st->st_ino) {
2282 switch (ino64_trunc_error) {
2289 ost->st_ino = UINT32_MAX;
2293 ost->st_mode = st->st_mode;
2294 ost->st_nlink = st->st_nlink;
2295 if (ost->st_nlink != st->st_nlink) {
2296 switch (ino64_trunc_error) {
2303 ost->st_nlink = UINT16_MAX;
2307 ost->st_uid = st->st_uid;
2308 ost->st_gid = st->st_gid;
2309 ost->st_rdev = st->st_rdev;
2310 if (ost->st_rdev != st->st_rdev) {
2311 switch (ino64_trunc_error) {
2318 ost->st_atim = st->st_atim;
2319 ost->st_mtim = st->st_mtim;
2320 ost->st_ctim = st->st_ctim;
2321 ost->st_size = st->st_size;
2322 ost->st_blocks = st->st_blocks;
2323 ost->st_blksize = st->st_blksize;
2324 ost->st_flags = st->st_flags;
2325 ost->st_gen = st->st_gen;
2327 ost->st_birthtim = st->st_birthtim;
2328 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2329 sizeof(*ost) - offsetof(struct freebsd11_stat,
2330 st_birthtim) - sizeof(ost->st_birthtim));
2335 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2338 struct freebsd11_stat osb;
2341 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2345 error = freebsd11_cvtstat(&sb, &osb);
2347 error = copyout(&osb, uap->ub, sizeof(osb));
2352 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2355 struct freebsd11_stat osb;
2358 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2359 UIO_USERSPACE, &sb, NULL);
2362 error = freebsd11_cvtstat(&sb, &osb);
2364 error = copyout(&osb, uap->ub, sizeof(osb));
2369 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2373 struct freebsd11_stat osb;
2376 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2379 error = kern_fhstat(td, fh, &sb);
2382 error = freebsd11_cvtstat(&sb, &osb);
2384 error = copyout(&osb, uap->sb, sizeof(osb));
2389 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2392 struct freebsd11_stat osb;
2395 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2396 UIO_USERSPACE, &sb, NULL);
2399 error = freebsd11_cvtstat(&sb, &osb);
2401 error = copyout(&osb, uap->buf, sizeof(osb));
2404 #endif /* COMPAT_FREEBSD11 */
2409 #ifndef _SYS_SYSPROTO_H_
2410 struct fstatat_args {
2418 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2423 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2424 UIO_USERSPACE, &sb, NULL);
2426 error = copyout(&sb, uap->buf, sizeof (sb));
2431 kern_statat(struct thread *td, int flag, int fd, const char *path,
2432 enum uio_seg pathseg, struct stat *sbp,
2433 void (*hook)(struct vnode *vp, struct stat *sbp))
2435 struct nameidata nd;
2438 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2439 AT_EMPTY_PATH)) != 0)
2442 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2443 AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2444 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights);
2446 if ((error = namei(&nd)) != 0) {
2447 if (error == ENOTDIR &&
2448 (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2449 error = kern_fstat(td, fd, sbp);
2452 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2454 if (__predict_false(hook != NULL))
2455 hook(nd.ni_vp, sbp);
2457 NDFREE_NOTHING(&nd);
2459 #ifdef __STAT_TIME_T_EXT
2460 sbp->st_atim_ext = 0;
2461 sbp->st_mtim_ext = 0;
2462 sbp->st_ctim_ext = 0;
2463 sbp->st_btim_ext = 0;
2466 if (KTRPOINT(td, KTR_STRUCT))
2467 ktrstat_error(sbp, error);
2472 #if defined(COMPAT_FREEBSD11)
2474 * Implementation of the NetBSD [l]stat() functions.
2477 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2479 struct freebsd11_stat sb11;
2482 error = freebsd11_cvtstat(sb, &sb11);
2486 bzero(nsb, sizeof(*nsb));
2487 CP(sb11, *nsb, st_dev);
2488 CP(sb11, *nsb, st_ino);
2489 CP(sb11, *nsb, st_mode);
2490 CP(sb11, *nsb, st_nlink);
2491 CP(sb11, *nsb, st_uid);
2492 CP(sb11, *nsb, st_gid);
2493 CP(sb11, *nsb, st_rdev);
2494 CP(sb11, *nsb, st_atim);
2495 CP(sb11, *nsb, st_mtim);
2496 CP(sb11, *nsb, st_ctim);
2497 CP(sb11, *nsb, st_size);
2498 CP(sb11, *nsb, st_blocks);
2499 CP(sb11, *nsb, st_blksize);
2500 CP(sb11, *nsb, st_flags);
2501 CP(sb11, *nsb, st_gen);
2502 CP(sb11, *nsb, st_birthtim);
2506 #ifndef _SYS_SYSPROTO_H_
2507 struct freebsd11_nstat_args {
2513 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2519 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2523 error = freebsd11_cvtnstat(&sb, &nsb);
2525 error = copyout(&nsb, uap->ub, sizeof (nsb));
2530 * NetBSD lstat. Get file status; this version does not follow links.
2532 #ifndef _SYS_SYSPROTO_H_
2533 struct freebsd11_nlstat_args {
2539 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2545 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2546 UIO_USERSPACE, &sb, NULL);
2549 error = freebsd11_cvtnstat(&sb, &nsb);
2551 error = copyout(&nsb, uap->ub, sizeof (nsb));
2554 #endif /* COMPAT_FREEBSD11 */
2557 * Get configurable pathname variables.
2559 #ifndef _SYS_SYSPROTO_H_
2560 struct pathconf_args {
2566 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2571 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2574 td->td_retval[0] = value;
2578 #ifndef _SYS_SYSPROTO_H_
2579 struct lpathconf_args {
2585 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2590 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2593 td->td_retval[0] = value;
2598 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2599 int name, u_long flags, long *valuep)
2601 struct nameidata nd;
2604 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2606 if ((error = namei(&nd)) != 0)
2608 NDFREE_NOTHING(&nd);
2610 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2616 * Return target name of a symbolic link.
2618 #ifndef _SYS_SYSPROTO_H_
2619 struct readlink_args {
2626 sys_readlink(struct thread *td, struct readlink_args *uap)
2629 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2630 uap->buf, UIO_USERSPACE, uap->count));
2632 #ifndef _SYS_SYSPROTO_H_
2633 struct readlinkat_args {
2641 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2644 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2645 uap->buf, UIO_USERSPACE, uap->bufsize));
2649 kern_readlinkat(struct thread *td, int fd, const char *path,
2650 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2653 struct nameidata nd;
2656 if (count > IOSIZE_MAX)
2659 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2660 EMPTYPATH, pathseg, path, fd);
2662 if ((error = namei(&nd)) != 0)
2664 NDFREE_NOTHING(&nd);
2667 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2674 * Helper function to readlink from a vnode
2677 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2684 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2686 error = mac_vnode_check_readlink(td->td_ucred, vp);
2690 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2693 aiov.iov_base = buf;
2694 aiov.iov_len = count;
2695 auio.uio_iov = &aiov;
2696 auio.uio_iovcnt = 1;
2697 auio.uio_offset = 0;
2698 auio.uio_rw = UIO_READ;
2699 auio.uio_segflg = bufseg;
2701 auio.uio_resid = count;
2702 error = VOP_READLINK(vp, &auio, td->td_ucred);
2703 td->td_retval[0] = count - auio.uio_resid;
2708 * Common implementation code for chflags() and fchflags().
2711 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2717 /* We can't support the value matching VNOVAL. */
2718 if (flags == VNOVAL)
2719 return (EOPNOTSUPP);
2722 * Prevent non-root users from setting flags on devices. When
2723 * a device is reused, users can retain ownership of the device
2724 * if they are allowed to set flags and programs assume that
2725 * chown can't fail when done as root.
2727 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2728 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2733 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2736 vattr.va_flags = flags;
2737 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2739 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2742 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2744 vn_finished_write(mp);
2749 * Change flags of a file given a path name.
2751 #ifndef _SYS_SYSPROTO_H_
2752 struct chflags_args {
2758 sys_chflags(struct thread *td, struct chflags_args *uap)
2761 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2765 #ifndef _SYS_SYSPROTO_H_
2766 struct chflagsat_args {
2774 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2777 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2778 uap->flags, uap->atflag));
2782 * Same as chflags() but doesn't follow symlinks.
2784 #ifndef _SYS_SYSPROTO_H_
2785 struct lchflags_args {
2791 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2794 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2795 uap->flags, AT_SYMLINK_NOFOLLOW));
2799 kern_chflagsat(struct thread *td, int fd, const char *path,
2800 enum uio_seg pathseg, u_long flags, int atflag)
2802 struct nameidata nd;
2805 if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2806 AT_EMPTY_PATH)) != 0)
2809 AUDIT_ARG_FFLAGS(flags);
2810 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2811 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2812 fd, &cap_fchflags_rights);
2813 if ((error = namei(&nd)) != 0)
2815 NDFREE_NOTHING(&nd);
2816 error = setfflags(td, nd.ni_vp, flags);
2822 * Change flags of a file given a file descriptor.
2824 #ifndef _SYS_SYSPROTO_H_
2825 struct fchflags_args {
2831 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2836 AUDIT_ARG_FD(uap->fd);
2837 AUDIT_ARG_FFLAGS(uap->flags);
2838 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2843 if (AUDITING_TD(td)) {
2844 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2845 AUDIT_ARG_VNODE1(fp->f_vnode);
2846 VOP_UNLOCK(fp->f_vnode);
2849 error = setfflags(td, fp->f_vnode, uap->flags);
2855 * Common implementation code for chmod(), lchmod() and fchmod().
2858 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2864 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2866 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2868 vattr.va_mode = mode & ALLPERMS;
2870 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2873 error = VOP_SETATTR(vp, &vattr, cred);
2875 vn_finished_write(mp);
2880 * Change mode of a file given path name.
2882 #ifndef _SYS_SYSPROTO_H_
2889 sys_chmod(struct thread *td, struct chmod_args *uap)
2892 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2896 #ifndef _SYS_SYSPROTO_H_
2897 struct fchmodat_args {
2905 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2908 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2909 uap->mode, uap->flag));
2913 * Change mode of a file given path name (don't follow links.)
2915 #ifndef _SYS_SYSPROTO_H_
2916 struct lchmod_args {
2922 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2925 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2926 uap->mode, AT_SYMLINK_NOFOLLOW));
2930 kern_fchmodat(struct thread *td, int fd, const char *path,
2931 enum uio_seg pathseg, mode_t mode, int flag)
2933 struct nameidata nd;
2936 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2937 AT_EMPTY_PATH)) != 0)
2940 AUDIT_ARG_MODE(mode);
2941 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2942 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2943 fd, &cap_fchmod_rights);
2944 if ((error = namei(&nd)) != 0)
2946 NDFREE_NOTHING(&nd);
2947 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2953 * Change mode of a file given a file descriptor.
2955 #ifndef _SYS_SYSPROTO_H_
2956 struct fchmod_args {
2962 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2967 AUDIT_ARG_FD(uap->fd);
2968 AUDIT_ARG_MODE(uap->mode);
2970 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2973 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2979 * Common implementation for chown(), lchown(), and fchown()
2982 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2989 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2991 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2996 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3000 error = VOP_SETATTR(vp, &vattr, cred);
3002 vn_finished_write(mp);
3007 * Set ownership given a path name.
3009 #ifndef _SYS_SYSPROTO_H_
3017 sys_chown(struct thread *td, struct chown_args *uap)
3020 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3024 #ifndef _SYS_SYSPROTO_H_
3025 struct fchownat_args {
3034 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3037 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3038 uap->gid, uap->flag));
3042 kern_fchownat(struct thread *td, int fd, const char *path,
3043 enum uio_seg pathseg, int uid, int gid, int flag)
3045 struct nameidata nd;
3048 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3049 AT_EMPTY_PATH)) != 0)
3052 AUDIT_ARG_OWNER(uid, gid);
3053 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3054 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3055 fd, &cap_fchown_rights);
3057 if ((error = namei(&nd)) != 0)
3059 NDFREE_NOTHING(&nd);
3060 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3066 * Set ownership given a path name, do not cross symlinks.
3068 #ifndef _SYS_SYSPROTO_H_
3069 struct lchown_args {
3076 sys_lchown(struct thread *td, struct lchown_args *uap)
3079 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3080 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3084 * Set ownership given a file descriptor.
3086 #ifndef _SYS_SYSPROTO_H_
3087 struct fchown_args {
3094 sys_fchown(struct thread *td, struct fchown_args *uap)
3099 AUDIT_ARG_FD(uap->fd);
3100 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3101 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3104 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3110 * Common implementation code for utimes(), lutimes(), and futimes().
3113 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3114 struct timespec *tsp)
3116 struct timeval tv[2];
3117 const struct timeval *tvp;
3120 if (usrtvp == NULL) {
3121 vfs_timestamp(&tsp[0]);
3124 if (tvpseg == UIO_SYSSPACE) {
3127 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3132 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3133 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3135 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3136 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3142 * Common implementation code for futimens(), utimensat().
3144 #define UTIMENS_NULL 0x1
3145 #define UTIMENS_EXIT 0x2
3147 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3148 struct timespec *tsp, int *retflags)
3150 struct timespec tsnow;
3153 vfs_timestamp(&tsnow);
3155 if (usrtsp == NULL) {
3158 *retflags |= UTIMENS_NULL;
3161 if (tspseg == UIO_SYSSPACE) {
3164 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3166 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3167 *retflags |= UTIMENS_EXIT;
3168 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3169 *retflags |= UTIMENS_NULL;
3170 if (tsp[0].tv_nsec == UTIME_OMIT)
3171 tsp[0].tv_sec = VNOVAL;
3172 else if (tsp[0].tv_nsec == UTIME_NOW)
3174 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3176 if (tsp[1].tv_nsec == UTIME_OMIT)
3177 tsp[1].tv_sec = VNOVAL;
3178 else if (tsp[1].tv_nsec == UTIME_NOW)
3180 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3187 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3191 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3192 int numtimes, int nullflag)
3199 setbirthtime = false;
3200 vattr.va_birthtime.tv_sec = VNOVAL;
3201 vattr.va_birthtime.tv_nsec = 0;
3203 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3205 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3206 if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3207 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3208 setbirthtime = true;
3210 vattr.va_atime = ts[0];
3211 vattr.va_mtime = ts[1];
3213 vattr.va_birthtime = ts[1];
3215 vattr.va_birthtime = ts[2];
3217 vattr.va_vaflags |= VA_UTIMES_NULL;
3219 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3223 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3225 vn_finished_write(mp);
3230 * Set the access and modification times of a file.
3232 #ifndef _SYS_SYSPROTO_H_
3233 struct utimes_args {
3235 struct timeval *tptr;
3239 sys_utimes(struct thread *td, struct utimes_args *uap)
3242 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3243 uap->tptr, UIO_USERSPACE));
3246 #ifndef _SYS_SYSPROTO_H_
3247 struct futimesat_args {
3250 const struct timeval * times;
3254 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3257 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3258 uap->times, UIO_USERSPACE));
3262 kern_utimesat(struct thread *td, int fd, const char *path,
3263 enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg)
3265 struct nameidata nd;
3266 struct timespec ts[2];
3269 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3271 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3272 &cap_futimes_rights);
3274 if ((error = namei(&nd)) != 0)
3276 NDFREE_NOTHING(&nd);
3277 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3283 * Set the access and modification times of a file.
3285 #ifndef _SYS_SYSPROTO_H_
3286 struct lutimes_args {
3288 struct timeval *tptr;
3292 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3295 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3300 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3301 const struct timeval *tptr, enum uio_seg tptrseg)
3303 struct timespec ts[2];
3304 struct nameidata nd;
3307 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3309 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path);
3310 if ((error = namei(&nd)) != 0)
3312 NDFREE_NOTHING(&nd);
3313 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3319 * Set the access and modification times of a file.
3321 #ifndef _SYS_SYSPROTO_H_
3322 struct futimes_args {
3324 struct timeval *tptr;
3328 sys_futimes(struct thread *td, struct futimes_args *uap)
3331 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3335 kern_futimes(struct thread *td, int fd, const struct timeval *tptr,
3336 enum uio_seg tptrseg)
3338 struct timespec ts[2];
3343 error = getutimes(tptr, tptrseg, ts);
3346 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3350 if (AUDITING_TD(td)) {
3351 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3352 AUDIT_ARG_VNODE1(fp->f_vnode);
3353 VOP_UNLOCK(fp->f_vnode);
3356 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3362 sys_futimens(struct thread *td, struct futimens_args *uap)
3365 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3369 kern_futimens(struct thread *td, int fd, const struct timespec *tptr,
3370 enum uio_seg tptrseg)
3372 struct timespec ts[2];
3377 error = getutimens(tptr, tptrseg, ts, &flags);
3380 if (flags & UTIMENS_EXIT)
3382 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3386 if (AUDITING_TD(td)) {
3387 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3388 AUDIT_ARG_VNODE1(fp->f_vnode);
3389 VOP_UNLOCK(fp->f_vnode);
3392 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3398 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3401 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3402 uap->times, UIO_USERSPACE, uap->flag));
3406 kern_utimensat(struct thread *td, int fd, const char *path,
3407 enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg,
3410 struct nameidata nd;
3411 struct timespec ts[2];
3414 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3415 AT_EMPTY_PATH)) != 0)
3418 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3420 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3421 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3422 pathseg, path, fd, &cap_futimes_rights);
3423 if ((error = namei(&nd)) != 0)
3426 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3428 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3429 * "Search permission is denied by a component of the path prefix."
3431 NDFREE_NOTHING(&nd);
3432 if ((flags & UTIMENS_EXIT) == 0)
3433 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3439 * Truncate a file given its path name.
3441 #ifndef _SYS_SYSPROTO_H_
3442 struct truncate_args {
3449 sys_truncate(struct thread *td, struct truncate_args *uap)
3452 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3456 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3462 struct nameidata nd;
3469 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
3470 if ((error = namei(&nd)) != 0)
3473 NDFREE_NOTHING(&nd);
3474 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3475 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3476 vn_rangelock_unlock(vp, rl_cookie);
3480 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3481 if (vp->v_type == VDIR) {
3486 error = mac_vnode_check_write(td->td_ucred, NOCRED, vp);
3490 error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td);
3494 error = vn_truncate_locked(vp, length, false, td->td_ucred);
3497 vn_finished_write(mp);
3498 vn_rangelock_unlock(vp, rl_cookie);
3500 if (error == ERELOOKUP)
3505 #if defined(COMPAT_43)
3507 * Truncate a file given its path name.
3509 #ifndef _SYS_SYSPROTO_H_
3510 struct otruncate_args {
3516 otruncate(struct thread *td, struct otruncate_args *uap)
3519 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3521 #endif /* COMPAT_43 */
3523 #if defined(COMPAT_FREEBSD6)
3524 /* Versions with the pad argument */
3526 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3529 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3533 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3536 return (kern_ftruncate(td, uap->fd, uap->length));
3541 kern_fsync(struct thread *td, int fd, bool fullsync)
3549 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3555 /* XXXKIB: compete outstanding aio writes */;
3558 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3561 vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3562 AUDIT_ARG_VNODE1(vp);
3563 if (vp->v_object != NULL) {
3564 VM_OBJECT_WLOCK(vp->v_object);
3565 vm_object_page_clean(vp->v_object, 0, 0, 0);
3566 VM_OBJECT_WUNLOCK(vp->v_object);
3568 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3570 vn_finished_write(mp);
3571 if (error == ERELOOKUP)
3579 * Sync an open file.
3581 #ifndef _SYS_SYSPROTO_H_
3587 sys_fsync(struct thread *td, struct fsync_args *uap)
3590 return (kern_fsync(td, uap->fd, true));
3594 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3597 return (kern_fsync(td, uap->fd, false));
3601 * Rename files. Source and destination must either both be directories, or
3602 * both not be directories. If target is a directory, it must be empty.
3604 #ifndef _SYS_SYSPROTO_H_
3605 struct rename_args {
3611 sys_rename(struct thread *td, struct rename_args *uap)
3614 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3615 uap->to, UIO_USERSPACE));
3618 #ifndef _SYS_SYSPROTO_H_
3619 struct renameat_args {
3627 sys_renameat(struct thread *td, struct renameat_args *uap)
3630 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3636 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3637 const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3641 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3642 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights);
3643 if ((error = namei(fromnd)) != 0)
3645 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3646 fromnd->ni_vp, &fromnd->ni_cnd);
3647 VOP_UNLOCK(fromnd->ni_dvp);
3648 if (fromnd->ni_dvp != fromnd->ni_vp)
3649 VOP_UNLOCK(fromnd->ni_vp);
3651 NDFREE_PNBUF(fromnd);
3652 vrele(fromnd->ni_dvp);
3653 vrele(fromnd->ni_vp);
3654 if (fromnd->ni_startdir)
3655 vrele(fromnd->ni_startdir);
3662 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3663 const char *new, enum uio_seg pathseg)
3665 struct mount *mp = NULL;
3666 struct vnode *tvp, *fvp, *tdvp;
3667 struct nameidata fromnd, tond;
3674 if (mac_vnode_check_rename_from_enabled()) {
3675 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3681 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3682 pathseg, old, oldfd, &cap_renameat_source_rights);
3683 if ((error = namei(&fromnd)) != 0)
3689 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3690 if (fromnd.ni_vp->v_type == VDIR)
3691 tondflags |= WILLBEDIR;
3692 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3693 &cap_renameat_target_rights);
3694 if ((error = namei(&tond)) != 0) {
3695 /* Translate error code for rename("dir1", "dir2/."). */
3696 if (error == EISDIR && fvp->v_type == VDIR)
3698 NDFREE_PNBUF(&fromnd);
3699 vrele(fromnd.ni_dvp);
3705 error = vn_start_write(fvp, &mp, V_NOWAIT);
3707 NDFREE_PNBUF(&fromnd);
3708 NDFREE_PNBUF(&tond);
3715 vrele(fromnd.ni_dvp);
3717 vrele(tond.ni_startdir);
3718 if (fromnd.ni_startdir != NULL)
3719 vrele(fromnd.ni_startdir);
3720 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3726 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3729 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3734 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3736 * If the target already exists we require CAP_UNLINKAT
3737 * from 'newfd', when newfd was used for the lookup.
3739 error = cap_check(&tond.ni_filecaps.fc_rights,
3740 &cap_unlinkat_rights);
3751 * If the source is the same as the destination (that is, if they
3752 * are links to the same vnode), then there is nothing to do.
3758 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3759 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3763 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3764 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3765 NDFREE_PNBUF(&fromnd);
3766 NDFREE_PNBUF(&tond);
3768 NDFREE_PNBUF(&fromnd);
3769 NDFREE_PNBUF(&tond);
3776 vrele(fromnd.ni_dvp);
3779 vrele(tond.ni_startdir);
3780 vn_finished_write(mp);
3782 if (fromnd.ni_startdir)
3783 vrele(fromnd.ni_startdir);
3784 if (error == ERESTART)
3786 if (error == ERELOOKUP)
3792 * Make a directory file.
3794 #ifndef _SYS_SYSPROTO_H_
3801 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3804 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3808 #ifndef _SYS_SYSPROTO_H_
3809 struct mkdirat_args {
3816 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3819 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3823 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3828 struct nameidata nd;
3831 AUDIT_ARG_MODE(mode);
3835 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3836 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3837 segflg, path, fd, &cap_mkdirat_rights);
3838 if ((error = namei(&nd)) != 0)
3840 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3843 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3848 vattr.va_type = VDIR;
3849 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3851 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3856 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3861 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3862 vn_finished_write(mp);
3863 if (error == ERELOOKUP)
3869 * Remove a directory file.
3871 #ifndef _SYS_SYSPROTO_H_
3877 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3880 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3885 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3886 enum uio_seg pathseg, int flag)
3891 struct nameidata nd;
3892 cap_rights_t rights;
3896 if (fd != FD_NONE) {
3897 error = getvnode(td, fd, cap_rights_init_one(&rights,
3906 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3907 at2cnpflags(flag, AT_RESOLVE_BENEATH),
3908 pathseg, path, dfd, &cap_unlinkat_rights);
3909 if ((error = namei(&nd)) != 0)
3912 if (vp->v_type != VDIR) {
3917 * No rmdir "." please.
3919 if (nd.ni_dvp == vp) {
3924 * The root of a mounted filesystem cannot be deleted.
3926 if (vp->v_vflag & VV_ROOT) {
3931 if (fp != NULL && fp->f_vnode != vp) {
3932 if (VN_IS_DOOMED(fp->f_vnode))
3940 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3945 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3948 if (nd.ni_dvp == vp)
3952 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3956 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3957 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3958 vn_finished_write(mp);
3962 if (nd.ni_dvp == vp)
3966 if (error == ERELOOKUP)
3974 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3976 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3977 long *basep, void (*func)(struct freebsd11_dirent *))
3979 struct freebsd11_dirent dstdp;
3980 struct dirent *dp, *edp;
3983 ssize_t resid, ucount;
3986 /* XXX arbitrary sanity limit on `count'. */
3987 count = min(count, 64 * 1024);
3989 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3991 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3999 for (dp = (struct dirent *)dirbuf,
4000 edp = (struct dirent *)&dirbuf[count - resid];
4001 ucount < count && dp < edp; ) {
4002 if (dp->d_reclen == 0)
4004 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
4005 if (dp->d_namlen >= sizeof(dstdp.d_name))
4007 dstdp.d_type = dp->d_type;
4008 dstdp.d_namlen = dp->d_namlen;
4009 dstdp.d_fileno = dp->d_fileno; /* truncate */
4010 if (dstdp.d_fileno != dp->d_fileno) {
4011 switch (ino64_trunc_error) {
4019 dstdp.d_fileno = UINT32_MAX;
4023 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4024 ((dp->d_namlen + 1 + 3) &~ 3);
4025 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4026 bzero(dstdp.d_name + dstdp.d_namlen,
4027 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4029 MPASS(dstdp.d_reclen <= dp->d_reclen);
4030 MPASS(ucount + dstdp.d_reclen <= count);
4033 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4036 dp = (struct dirent *)((char *)dp + dp->d_reclen);
4037 ucount += dstdp.d_reclen;
4041 free(dirbuf, M_TEMP);
4043 td->td_retval[0] = ucount;
4050 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4052 #if (BYTE_ORDER == LITTLE_ENDIAN)
4054 * The expected low byte of dp->d_namlen is our dp->d_type.
4055 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4057 dp->d_type = dp->d_namlen;
4061 * The dp->d_type is the high byte of the expected dp->d_namlen,
4062 * so must be zero'ed.
4069 * Read a block of directory entries in a filesystem independent format.
4071 #ifndef _SYS_SYSPROTO_H_
4072 struct ogetdirentries_args {
4080 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4085 error = kern_ogetdirentries(td, uap, &loff);
4087 error = copyout(&loff, uap->basep, sizeof(long));
4092 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4098 /* XXX arbitrary sanity limit on `count'. */
4099 if (uap->count > 64 * 1024)
4102 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4103 &base, ogetdirentries_cvt);
4105 if (error == 0 && uap->basep != NULL)
4106 error = copyout(&base, uap->basep, sizeof(long));
4110 #endif /* COMPAT_43 */
4112 #if defined(COMPAT_FREEBSD11)
4113 #ifndef _SYS_SYSPROTO_H_
4114 struct freebsd11_getdirentries_args {
4122 freebsd11_getdirentries(struct thread *td,
4123 struct freebsd11_getdirentries_args *uap)
4128 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4131 if (error == 0 && uap->basep != NULL)
4132 error = copyout(&base, uap->basep, sizeof(long));
4137 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4139 struct freebsd11_getdirentries_args ap;
4143 ap.count = uap->count;
4145 return (freebsd11_getdirentries(td, &ap));
4147 #endif /* COMPAT_FREEBSD11 */
4150 * Read a block of directory entries in a filesystem independent format.
4153 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4158 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4159 NULL, UIO_USERSPACE);
4162 if (uap->basep != NULL)
4163 error = copyout(&base, uap->basep, sizeof(off_t));
4168 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4169 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4180 if (count > IOSIZE_MAX)
4182 auio.uio_resid = count;
4183 error = getvnode(td, fd, &cap_read_rights, &fp);
4186 if ((fp->f_flag & FREAD) == 0) {
4191 foffset = foffset_lock(fp, 0);
4193 if (vp->v_type != VDIR) {
4197 if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) {
4201 aiov.iov_base = buf;
4202 aiov.iov_len = count;
4203 auio.uio_iov = &aiov;
4204 auio.uio_iovcnt = 1;
4205 auio.uio_rw = UIO_READ;
4206 auio.uio_segflg = bufseg;
4208 vn_lock(vp, LK_SHARED | LK_RETRY);
4209 AUDIT_ARG_VNODE1(vp);
4210 loff = auio.uio_offset = foffset;
4212 error = mac_vnode_check_readdir(td->td_ucred, vp);
4215 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4217 foffset = auio.uio_offset;
4222 if (count == auio.uio_resid &&
4223 (vp->v_vflag & VV_ROOT) &&
4224 (vp->v_mount->mnt_flag & MNT_UNION)) {
4225 struct vnode *tvp = vp;
4227 vp = vp->v_mount->mnt_vnodecovered;
4237 *residp = auio.uio_resid;
4238 td->td_retval[0] = count - auio.uio_resid;
4240 foffset_unlock(fp, foffset, 0);
4246 * Set the mode mask for creation of filesystem nodes.
4248 #ifndef _SYS_SYSPROTO_H_
4254 sys_umask(struct thread *td, struct umask_args *uap)
4256 struct pwddesc *pdp;
4258 pdp = td->td_proc->p_pd;
4260 td->td_retval[0] = pdp->pd_cmask;
4261 pdp->pd_cmask = uap->newmask & ALLPERMS;
4262 PWDDESC_XUNLOCK(pdp);
4267 * Void all references to file by ripping underlying filesystem away from
4270 #ifndef _SYS_SYSPROTO_H_
4271 struct revoke_args {
4276 sys_revoke(struct thread *td, struct revoke_args *uap)
4280 struct nameidata nd;
4283 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4285 if ((error = namei(&nd)) != 0)
4288 NDFREE_NOTHING(&nd);
4289 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4294 error = mac_vnode_check_revoke(td->td_ucred, vp);
4298 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4301 if (td->td_ucred->cr_uid != vattr.va_uid) {
4302 error = priv_check(td, PRIV_VFS_ADMIN);
4306 if (devfs_usecount(vp) > 0)
4307 VOP_REVOKE(vp, REVOKEALL);
4314 * This variant of getvnode() allows O_PATH files. Caller should
4315 * ensure that returned file and vnode are only used for compatible
4319 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4325 error = fget_unlocked(td, fd, rightsp, &fp);
4330 * The file could be not of the vnode type, or it may be not
4331 * yet fully initialized, in which case the f_vnode pointer
4332 * may be set, but f_ops is still badfileops. E.g.,
4333 * devfs_open() transiently create such situation to
4334 * facilitate csw d_fdopen().
4336 * Dupfdopen() handling in kern_openat() installs the
4337 * half-baked file into the process descriptor table, allowing
4338 * other thread to dereference it. Guard against the race by
4341 if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4352 * Convert a user file descriptor to a kernel file entry and check
4353 * that, if it is a capability, the correct rights are present.
4354 * A reference on the file entry is held upon returning.
4357 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4361 error = getvnode_path(td, fd, rightsp, fpp);
4362 if (__predict_false(error != 0))
4366 * Filter out O_PATH file descriptors, most getvnode() callers
4367 * do not call fo_ methods.
4369 if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4379 * Get an (NFS) file handle.
4381 #ifndef _SYS_SYSPROTO_H_
4382 struct lgetfh_args {
4388 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4391 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4392 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4395 #ifndef _SYS_SYSPROTO_H_
4402 sys_getfh(struct thread *td, struct getfh_args *uap)
4405 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4406 uap->fhp, UIO_USERSPACE));
4410 * syscall for the rpc.lockd to use to translate an open descriptor into
4411 * a NFS file handle.
4413 * warning: do not remove the priv_check() call or this becomes one giant
4416 #ifndef _SYS_SYSPROTO_H_
4417 struct getfhat_args {
4425 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4428 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4429 uap->fhp, UIO_USERSPACE));
4433 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4434 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4436 struct nameidata nd;
4441 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4443 error = priv_check(td, PRIV_VFS_GETFH);
4446 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4447 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4452 NDFREE_NOTHING(&nd);
4454 bzero(&fh, sizeof(fh));
4455 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4456 error = VOP_VPTOFH(vp, &fh.fh_fid);
4459 if (fhseg == UIO_USERSPACE)
4460 error = copyout(&fh, fhp, sizeof (fh));
4462 memcpy(fhp, &fh, sizeof(fh));
4467 #ifndef _SYS_SYSPROTO_H_
4468 struct fhlink_args {
4474 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4477 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4480 #ifndef _SYS_SYSPROTO_H_
4481 struct fhlinkat_args {
4488 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4491 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4495 kern_fhlinkat(struct thread *td, int fd, const char *path,
4496 enum uio_seg pathseg, fhandle_t *fhp)
4503 error = priv_check(td, PRIV_VFS_GETFH);
4506 error = copyin(fhp, &fh, sizeof(fh));
4511 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4513 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4518 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4519 } while (error == EAGAIN || error == ERELOOKUP);
4523 #ifndef _SYS_SYSPROTO_H_
4524 struct fhreadlink_args {
4531 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4538 error = priv_check(td, PRIV_VFS_GETFH);
4541 if (uap->bufsize > IOSIZE_MAX)
4543 error = copyin(uap->fhp, &fh, sizeof(fh));
4546 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4548 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4552 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4558 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4561 * warning: do not remove the priv_check() call or this becomes one giant
4564 #ifndef _SYS_SYSPROTO_H_
4565 struct fhopen_args {
4566 const struct fhandle *u_fhp;
4571 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4573 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4577 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4586 error = priv_check(td, PRIV_VFS_FHOPEN);
4590 fmode = FFLAGS(flags);
4591 /* why not allow a non-read/write open for our lockd? */
4592 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4594 error = copyin(u_fhp, &fhp, sizeof(fhp));
4597 /* find the mount point */
4598 mp = vfs_busyfs(&fhp.fh_fsid);
4601 /* now give me my vnode, it gets returned to me locked */
4602 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4607 error = falloc_noinstall(td, &fp);
4613 * An extra reference on `fp' has been held for us by
4614 * falloc_noinstall().
4620 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4622 KASSERT(fp->f_ops == &badfileops,
4623 ("VOP_OPEN in fhopen() set f_ops"));
4624 KASSERT(td->td_dupfd < 0,
4625 ("fhopen() encountered fdopen()"));
4634 finit_vnode(fp, fmode, NULL, &vnops);
4636 if ((fmode & O_TRUNC) != 0) {
4637 error = fo_truncate(fp, 0, td->td_ucred, td);
4642 error = finstall(td, fp, &indx, fmode, NULL);
4645 td->td_retval[0] = indx;
4650 * Stat an (NFS) file handle.
4652 #ifndef _SYS_SYSPROTO_H_
4653 struct fhstat_args {
4654 struct fhandle *u_fhp;
4659 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4665 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4668 error = kern_fhstat(td, fh, &sb);
4670 error = copyout(&sb, uap->sb, sizeof(sb));
4675 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4681 error = priv_check(td, PRIV_VFS_FHSTAT);
4684 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4686 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4690 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4696 * Implement fstatfs() for (NFS) file handles.
4698 #ifndef _SYS_SYSPROTO_H_
4699 struct fhstatfs_args {
4700 struct fhandle *u_fhp;
4705 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4711 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4714 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4715 error = kern_fhstatfs(td, fh, sfp);
4717 error = copyout(sfp, uap->buf, sizeof(*sfp));
4718 free(sfp, M_STATFS);
4723 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4729 error = priv_check(td, PRIV_VFS_FHSTATFS);
4732 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4734 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4740 error = prison_canseemount(td->td_ucred, mp);
4744 error = mac_mount_check_stat(td->td_ucred, mp);
4748 error = VFS_STATFS(mp, buf);
4755 * Unlike madvise(2), we do not make a best effort to remember every
4756 * possible caching hint. Instead, we remember the last setting with
4757 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4758 * region of any current setting.
4761 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4764 struct fadvise_info *fa, *new;
4770 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4772 AUDIT_ARG_VALUE(advice);
4774 case POSIX_FADV_SEQUENTIAL:
4775 case POSIX_FADV_RANDOM:
4776 case POSIX_FADV_NOREUSE:
4777 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4779 case POSIX_FADV_NORMAL:
4780 case POSIX_FADV_WILLNEED:
4781 case POSIX_FADV_DONTNEED:
4787 /* XXX: CAP_POSIX_FADVISE? */
4789 error = fget(td, fd, &cap_no_rights, &fp);
4792 AUDIT_ARG_FILE(td->td_proc, fp);
4793 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4797 if (fp->f_type != DTYPE_VNODE) {
4802 if (vp->v_type != VREG) {
4809 end = offset + len - 1;
4811 case POSIX_FADV_SEQUENTIAL:
4812 case POSIX_FADV_RANDOM:
4813 case POSIX_FADV_NOREUSE:
4815 * Try to merge any existing non-standard region with
4816 * this new region if possible, otherwise create a new
4817 * non-standard region for this request.
4819 mtx_pool_lock(mtxpool_sleep, fp);
4821 if (fa != NULL && fa->fa_advice == advice &&
4822 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4823 (end != OFF_MAX && fa->fa_start == end + 1) ||
4824 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4825 if (offset < fa->fa_start)
4826 fa->fa_start = offset;
4827 if (end > fa->fa_end)
4830 new->fa_advice = advice;
4831 new->fa_start = offset;
4836 mtx_pool_unlock(mtxpool_sleep, fp);
4838 case POSIX_FADV_NORMAL:
4840 * If a the "normal" region overlaps with an existing
4841 * non-standard region, trim or remove the
4842 * non-standard region.
4844 mtx_pool_lock(mtxpool_sleep, fp);
4847 if (offset <= fa->fa_start && end >= fa->fa_end) {
4849 fp->f_advice = NULL;
4850 } else if (offset <= fa->fa_start &&
4851 end >= fa->fa_start)
4852 fa->fa_start = end + 1;
4853 else if (offset <= fa->fa_end && end >= fa->fa_end)
4854 fa->fa_end = offset - 1;
4855 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4857 * If the "normal" region is a middle
4858 * portion of the existing
4859 * non-standard region, just remove
4860 * the whole thing rather than picking
4861 * one side or the other to
4865 fp->f_advice = NULL;
4868 mtx_pool_unlock(mtxpool_sleep, fp);
4870 case POSIX_FADV_WILLNEED:
4871 case POSIX_FADV_DONTNEED:
4872 error = VOP_ADVISE(vp, offset, end, advice);
4878 free(new, M_FADVISE);
4883 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4887 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4889 return (kern_posix_error(td, error));
4893 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4894 off_t *outoffp, size_t len, unsigned int flags)
4896 struct file *infp, *outfp;
4897 struct vnode *invp, *outvp;
4900 void *rl_rcookie, *rl_wcookie;
4901 off_t savinoff, savoutoff;
4903 infp = outfp = NULL;
4904 rl_rcookie = rl_wcookie = NULL;
4913 if (len > SSIZE_MAX)
4915 * Although the len argument is size_t, the return argument
4916 * is ssize_t (which is signed). Therefore a size that won't
4917 * fit in ssize_t can't be returned.
4921 /* Get the file structures for the file descriptors. */
4922 error = fget_read(td, infd, &cap_read_rights, &infp);
4925 if (infp->f_ops == &badfileops) {
4929 if (infp->f_vnode == NULL) {
4933 error = fget_write(td, outfd, &cap_write_rights, &outfp);
4936 if (outfp->f_ops == &badfileops) {
4940 if (outfp->f_vnode == NULL) {
4945 /* Set the offset pointers to the correct place. */
4947 inoffp = &infp->f_offset;
4948 if (outoffp == NULL)
4949 outoffp = &outfp->f_offset;
4951 savoutoff = *outoffp;
4953 invp = infp->f_vnode;
4954 outvp = outfp->f_vnode;
4955 /* Sanity check the f_flag bits. */
4956 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4957 (infp->f_flag & FREAD) == 0) {
4962 /* If len == 0, just return 0. */
4967 * If infp and outfp refer to the same file, the byte ranges cannot
4970 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4971 savoutoff) || (savinoff > savoutoff && savoutoff + len >
4977 /* Range lock the byte ranges for both invp and outvp. */
4979 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4981 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4983 if (rl_rcookie != NULL)
4985 vn_rangelock_unlock(outvp, rl_wcookie);
4986 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4987 vn_rangelock_unlock(invp, rl_rcookie);
4991 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4992 flags, infp->f_cred, outfp->f_cred, td);
4994 if (rl_rcookie != NULL)
4995 vn_rangelock_unlock(invp, rl_rcookie);
4996 if (rl_wcookie != NULL)
4997 vn_rangelock_unlock(outvp, rl_wcookie);
4998 if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
5000 *outoffp = savoutoff;
5006 td->td_retval[0] = retlen;
5011 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
5013 off_t inoff, outoff, *inoffp, *outoffp;
5016 inoffp = outoffp = NULL;
5017 if (uap->inoffp != NULL) {
5018 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
5023 if (uap->outoffp != NULL) {
5024 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
5029 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5030 outoffp, uap->len, uap->flags);
5031 if (error == 0 && uap->inoffp != NULL)
5032 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5033 if (error == 0 && uap->outoffp != NULL)
5034 error = copyout(outoffp, uap->outoffp, sizeof(off_t));