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/sysent.h>
55 #include <sys/malloc.h>
56 #include <sys/mount.h>
57 #include <sys/mutex.h>
58 #include <sys/sysproto.h>
59 #include <sys/namei.h>
60 #include <sys/filedesc.h>
61 #include <sys/kernel.h>
62 #include <sys/fcntl.h>
64 #include <sys/filio.h>
65 #include <sys/limits.h>
66 #include <sys/linker.h>
67 #include <sys/rwlock.h>
71 #include <sys/unistd.h>
72 #include <sys/vnode.h>
75 #include <sys/dirent.h>
77 #include <sys/syscallsubr.h>
78 #include <sys/sysctl.h>
80 #include <sys/ktrace.h>
83 #include <machine/stdarg.h>
85 #include <security/audit/audit.h>
86 #include <security/mac/mac_framework.h>
89 #include <vm/vm_object.h>
90 #include <vm/vm_page.h>
93 #include <fs/devfs/devfs.h>
95 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98 enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102 struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104 const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
108 enum uio_seg pathseg, fhandle_t *fhp);
109 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
110 size_t count, struct thread *td);
111 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
112 const char *path, enum uio_seg segflag);
115 at2cnpflags(u_int at_flags, u_int mask)
119 MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
120 (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
124 if ((at_flags & AT_RESOLVE_BENEATH) != 0)
126 if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
128 /* NOFOLLOW is pseudo flag */
129 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
130 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
133 if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
139 kern_sync(struct thread *td)
141 struct mount *mp, *nmp;
144 mtx_lock(&mountlist_mtx);
145 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
146 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
147 nmp = TAILQ_NEXT(mp, mnt_list);
150 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
151 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
152 save = curthread_pflags_set(TDP_SYNCIO);
153 vfs_periodic(mp, MNT_NOWAIT);
154 VFS_SYNC(mp, MNT_NOWAIT);
155 curthread_pflags_restore(save);
156 vn_finished_write(mp);
158 mtx_lock(&mountlist_mtx);
159 nmp = TAILQ_NEXT(mp, mnt_list);
162 mtx_unlock(&mountlist_mtx);
167 * Sync each mounted filesystem.
169 #ifndef _SYS_SYSPROTO_H_
176 sys_sync(struct thread *td, struct sync_args *uap)
179 return (kern_sync(td));
183 * Change filesystem quotas.
185 #ifndef _SYS_SYSPROTO_H_
186 struct quotactl_args {
194 sys_quotactl(struct thread *td, struct quotactl_args *uap)
201 AUDIT_ARG_CMD(uap->cmd);
202 AUDIT_ARG_UID(uap->uid);
203 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
205 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
207 if ((error = namei(&nd)) != 0)
210 mp = nd.ni_vp->v_mount;
213 error = vfs_busy(mp, 0);
219 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
222 * Since quota on/off operations typically need to open quota
223 * files, the implementation may need to unbusy the mount point
224 * before calling into namei. Otherwise, unmount might be
225 * started between two vfs_busy() invocations (first is ours,
226 * second is from mount point cross-walk code in lookup()),
229 * Avoid unbusying mp if the implementation indicates it has
239 * Used by statfs conversion routines to scale the block size up if
240 * necessary so that all of the block counts are <= 'max_size'. Note
241 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
245 statfs_scale_blocks(struct statfs *sf, long max_size)
250 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
253 * Attempt to scale the block counts to give a more accurate
254 * overview to userland of the ratio of free space to used
255 * space. To do this, find the largest block count and compute
256 * a divisor that lets it fit into a signed integer <= max_size.
258 if (sf->f_bavail < 0)
259 count = -sf->f_bavail;
261 count = sf->f_bavail;
262 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
263 if (count <= max_size)
266 count >>= flsl(max_size);
273 sf->f_bsize <<= shift;
274 sf->f_blocks >>= shift;
275 sf->f_bfree >>= shift;
276 sf->f_bavail >>= shift;
280 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
286 error = vfs_busy(mp, 0);
291 error = mac_mount_check_stat(td->td_ucred, mp);
295 error = VFS_STATFS(mp, buf);
298 if (priv_check_cred_vfs_generation(td->td_ucred)) {
299 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
300 prison_enforce_statfs(td->td_ucred, mp, buf);
308 * Get filesystem statistics.
310 #ifndef _SYS_SYSPROTO_H_
317 sys_statfs(struct thread *td, struct statfs_args *uap)
322 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
323 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
325 error = copyout(sfp, uap->buf, sizeof(struct statfs));
331 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
338 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
342 mp = vfs_ref_from_vp(nd.ni_vp);
345 return (kern_do_statfs(td, mp, buf));
349 * Get filesystem statistics.
351 #ifndef _SYS_SYSPROTO_H_
352 struct fstatfs_args {
358 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
363 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
364 error = kern_fstatfs(td, uap->fd, sfp);
366 error = copyout(sfp, uap->buf, sizeof(struct statfs));
372 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
380 error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
385 if (AUDITING_TD(td)) {
386 vn_lock(vp, LK_SHARED | LK_RETRY);
387 AUDIT_ARG_VNODE1(vp);
391 mp = vfs_ref_from_vp(vp);
393 return (kern_do_statfs(td, mp, buf));
397 * Get statistics on all filesystems.
399 #ifndef _SYS_SYSPROTO_H_
400 struct getfsstat_args {
407 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
412 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
414 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
415 UIO_USERSPACE, uap->mode);
417 td->td_retval[0] = count;
422 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
423 * The caller is responsible for freeing memory which will be allocated
427 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
428 size_t *countp, enum uio_seg bufseg, int mode)
430 struct mount *mp, *nmp;
431 struct statfs *sfsp, *sp, *sptmp, *tofree;
432 size_t count, maxcount;
440 if (bufseg == UIO_SYSSPACE)
445 maxcount = bufsize / sizeof(struct statfs);
449 } else if (bufseg == UIO_USERSPACE) {
452 } else /* if (bufseg == UIO_SYSSPACE) */ {
454 mtx_lock(&mountlist_mtx);
455 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
458 mtx_unlock(&mountlist_mtx);
459 if (maxcount > count)
461 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
468 * If there is no target buffer they only want the count.
470 * This could be TAILQ_FOREACH but it is open-coded to match the original
474 mtx_lock(&mountlist_mtx);
475 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
476 if (prison_canseemount(td->td_ucred, mp) != 0) {
477 nmp = TAILQ_NEXT(mp, mnt_list);
481 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
482 nmp = TAILQ_NEXT(mp, mnt_list);
487 nmp = TAILQ_NEXT(mp, mnt_list);
489 mtx_unlock(&mountlist_mtx);
495 * They want the entire thing.
497 * Short-circuit the corner case of no room for anything, avoids
504 mtx_lock(&mountlist_mtx);
505 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
506 if (prison_canseemount(td->td_ucred, mp) != 0) {
507 nmp = TAILQ_NEXT(mp, mnt_list);
511 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
512 nmp = TAILQ_NEXT(mp, mnt_list);
516 if (mode == MNT_WAIT) {
517 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
519 * If vfs_busy() failed, and MBF_NOWAIT
520 * wasn't passed, then the mp is gone.
521 * Furthermore, because of MBF_MNTLSTLOCK,
522 * the mountlist_mtx was dropped. We have
523 * no other choice than to start over.
525 mtx_unlock(&mountlist_mtx);
526 free(tofree, M_STATFS);
530 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
531 nmp = TAILQ_NEXT(mp, mnt_list);
537 * If MNT_NOWAIT is specified, do not refresh
540 if (mode != MNT_NOWAIT) {
541 error = VFS_STATFS(mp, sp);
543 mtx_lock(&mountlist_mtx);
544 nmp = TAILQ_NEXT(mp, mnt_list);
549 if (priv_check_cred_vfs_generation(td->td_ucred)) {
550 sptmp = malloc(sizeof(struct statfs), M_STATFS,
553 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
554 prison_enforce_statfs(td->td_ucred, mp, sptmp);
558 if (bufseg == UIO_SYSSPACE) {
559 bcopy(sp, sfsp, sizeof(*sp));
560 free(sptmp, M_STATFS);
561 } else /* if (bufseg == UIO_USERSPACE) */ {
562 error = copyout(sp, sfsp, sizeof(*sp));
563 free(sptmp, M_STATFS);
572 if (count == maxcount) {
577 mtx_lock(&mountlist_mtx);
578 nmp = TAILQ_NEXT(mp, mnt_list);
581 mtx_unlock(&mountlist_mtx);
587 #ifdef COMPAT_FREEBSD4
589 * Get old format filesystem statistics.
591 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
593 #ifndef _SYS_SYSPROTO_H_
594 struct freebsd4_statfs_args {
600 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
606 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
607 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
609 freebsd4_cvtstatfs(sfp, &osb);
610 error = copyout(&osb, uap->buf, sizeof(osb));
617 * Get filesystem statistics.
619 #ifndef _SYS_SYSPROTO_H_
620 struct freebsd4_fstatfs_args {
626 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
632 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
633 error = kern_fstatfs(td, uap->fd, sfp);
635 freebsd4_cvtstatfs(sfp, &osb);
636 error = copyout(&osb, uap->buf, sizeof(osb));
643 * Get statistics on all filesystems.
645 #ifndef _SYS_SYSPROTO_H_
646 struct freebsd4_getfsstat_args {
653 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
655 struct statfs *buf, *sp;
660 if (uap->bufsize < 0)
662 count = uap->bufsize / sizeof(struct ostatfs);
663 if (count > SIZE_MAX / sizeof(struct statfs))
665 size = count * sizeof(struct statfs);
666 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
669 td->td_retval[0] = count;
672 while (count != 0 && error == 0) {
673 freebsd4_cvtstatfs(sp, &osb);
674 error = copyout(&osb, uap->buf, sizeof(osb));
685 * Implement fstatfs() for (NFS) file handles.
687 #ifndef _SYS_SYSPROTO_H_
688 struct freebsd4_fhstatfs_args {
689 struct fhandle *u_fhp;
694 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
701 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
704 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
705 error = kern_fhstatfs(td, fh, sfp);
707 freebsd4_cvtstatfs(sfp, &osb);
708 error = copyout(&osb, uap->buf, sizeof(osb));
715 * Convert a new format statfs structure to an old format statfs structure.
718 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
721 statfs_scale_blocks(nsp, LONG_MAX);
722 bzero(osp, sizeof(*osp));
723 osp->f_bsize = nsp->f_bsize;
724 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
725 osp->f_blocks = nsp->f_blocks;
726 osp->f_bfree = nsp->f_bfree;
727 osp->f_bavail = nsp->f_bavail;
728 osp->f_files = MIN(nsp->f_files, LONG_MAX);
729 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
730 osp->f_owner = nsp->f_owner;
731 osp->f_type = nsp->f_type;
732 osp->f_flags = nsp->f_flags;
733 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
734 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
735 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
736 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
737 strlcpy(osp->f_fstypename, nsp->f_fstypename,
738 MIN(MFSNAMELEN, OMFSNAMELEN));
739 strlcpy(osp->f_mntonname, nsp->f_mntonname,
740 MIN(MNAMELEN, OMNAMELEN));
741 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
742 MIN(MNAMELEN, OMNAMELEN));
743 osp->f_fsid = nsp->f_fsid;
745 #endif /* COMPAT_FREEBSD4 */
747 #if defined(COMPAT_FREEBSD11)
749 * Get old format filesystem statistics.
751 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
754 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
756 struct freebsd11_statfs osb;
760 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
761 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
763 freebsd11_cvtstatfs(sfp, &osb);
764 error = copyout(&osb, uap->buf, sizeof(osb));
771 * Get filesystem statistics.
774 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
776 struct freebsd11_statfs osb;
780 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
781 error = kern_fstatfs(td, uap->fd, sfp);
783 freebsd11_cvtstatfs(sfp, &osb);
784 error = copyout(&osb, uap->buf, sizeof(osb));
791 * Get statistics on all filesystems.
794 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
796 return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode));
800 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf,
801 long bufsize, int mode)
803 struct freebsd11_statfs osb;
804 struct statfs *buf, *sp;
811 count = bufsize / sizeof(struct ostatfs);
812 size = count * sizeof(struct statfs);
813 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode);
815 td->td_retval[0] = count;
818 while (count > 0 && error == 0) {
819 freebsd11_cvtstatfs(sp, &osb);
820 error = copyout(&osb, ubuf, sizeof(osb));
831 * Implement fstatfs() for (NFS) file handles.
834 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
836 struct freebsd11_statfs osb;
841 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
844 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
845 error = kern_fhstatfs(td, fh, sfp);
847 freebsd11_cvtstatfs(sfp, &osb);
848 error = copyout(&osb, uap->buf, sizeof(osb));
855 * Convert a new format statfs structure to an old format statfs structure.
858 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
861 bzero(osp, sizeof(*osp));
862 osp->f_version = FREEBSD11_STATFS_VERSION;
863 osp->f_type = nsp->f_type;
864 osp->f_flags = nsp->f_flags;
865 osp->f_bsize = nsp->f_bsize;
866 osp->f_iosize = nsp->f_iosize;
867 osp->f_blocks = nsp->f_blocks;
868 osp->f_bfree = nsp->f_bfree;
869 osp->f_bavail = nsp->f_bavail;
870 osp->f_files = nsp->f_files;
871 osp->f_ffree = nsp->f_ffree;
872 osp->f_syncwrites = nsp->f_syncwrites;
873 osp->f_asyncwrites = nsp->f_asyncwrites;
874 osp->f_syncreads = nsp->f_syncreads;
875 osp->f_asyncreads = nsp->f_asyncreads;
876 osp->f_namemax = nsp->f_namemax;
877 osp->f_owner = nsp->f_owner;
878 osp->f_fsid = nsp->f_fsid;
879 strlcpy(osp->f_fstypename, nsp->f_fstypename,
880 MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
881 strlcpy(osp->f_mntonname, nsp->f_mntonname,
882 MIN(MNAMELEN, sizeof(osp->f_mntonname)));
883 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
884 MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
886 #endif /* COMPAT_FREEBSD11 */
889 * Change current working directory to a given file descriptor.
891 #ifndef _SYS_SYSPROTO_H_
897 sys_fchdir(struct thread *td, struct fchdir_args *uap)
899 struct vnode *vp, *tdp;
904 AUDIT_ARG_FD(uap->fd);
905 error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
912 vn_lock(vp, LK_SHARED | LK_RETRY);
913 AUDIT_ARG_VNODE1(vp);
914 error = change_dir(vp, td);
915 while (!error && (mp = vp->v_mountedhere) != NULL) {
918 error = VFS_ROOT(mp, LK_SHARED, &tdp);
935 * Change current working directory (``.'').
937 #ifndef _SYS_SYSPROTO_H_
943 sys_chdir(struct thread *td, struct chdir_args *uap)
946 return (kern_chdir(td, uap->path, UIO_USERSPACE));
950 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
955 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
957 if ((error = namei(&nd)) != 0)
959 if ((error = change_dir(nd.ni_vp, td)) != 0) {
964 VOP_UNLOCK(nd.ni_vp);
966 pwd_chdir(td, nd.ni_vp);
970 static int unprivileged_chroot = 0;
971 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
972 &unprivileged_chroot, 0,
973 "Unprivileged processes can use chroot(2)");
975 * Change notion of root (``/'') directory.
977 #ifndef _SYS_SYSPROTO_H_
983 sys_chroot(struct thread *td, struct chroot_args *uap)
989 error = priv_check(td, PRIV_VFS_CHROOT);
993 if (unprivileged_chroot == 0 ||
994 (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
1000 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
1001 UIO_USERSPACE, uap->path);
1005 error = change_dir(nd.ni_vp, td);
1009 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
1013 VOP_UNLOCK(nd.ni_vp);
1014 error = pwd_chroot(td, nd.ni_vp);
1016 NDFREE_NOTHING(&nd);
1021 NDFREE_NOTHING(&nd);
1026 * Common routine for chroot and chdir. Callers must provide a locked vnode
1030 change_dir(struct vnode *vp, struct thread *td)
1036 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1037 if (vp->v_type != VDIR)
1040 error = mac_vnode_check_chdir(td->td_ucred, vp);
1044 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1047 static __inline void
1048 flags_to_rights(int flags, cap_rights_t *rightsp)
1050 if (flags & O_EXEC) {
1051 cap_rights_set_one(rightsp, CAP_FEXECVE);
1055 switch ((flags & O_ACCMODE)) {
1057 cap_rights_set_one(rightsp, CAP_READ);
1060 cap_rights_set_one(rightsp, CAP_READ);
1063 cap_rights_set_one(rightsp, CAP_WRITE);
1064 if (!(flags & (O_APPEND | O_TRUNC)))
1065 cap_rights_set_one(rightsp, CAP_SEEK);
1070 if (flags & O_CREAT)
1071 cap_rights_set_one(rightsp, CAP_CREATE);
1073 if (flags & O_TRUNC)
1074 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1076 if (flags & (O_SYNC | O_FSYNC))
1077 cap_rights_set_one(rightsp, CAP_FSYNC);
1079 if (flags & (O_EXLOCK | O_SHLOCK))
1080 cap_rights_set_one(rightsp, CAP_FLOCK);
1084 * Check permissions, allocate an open file structure, and call the device
1085 * open routine if any.
1087 #ifndef _SYS_SYSPROTO_H_
1095 sys_open(struct thread *td, struct open_args *uap)
1098 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1099 uap->flags, uap->mode));
1102 #ifndef _SYS_SYSPROTO_H_
1103 struct openat_args {
1111 sys_openat(struct thread *td, struct openat_args *uap)
1114 AUDIT_ARG_FD(uap->fd);
1115 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1120 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1121 int flags, int mode)
1123 struct proc *p = td->td_proc;
1124 struct filedesc *fdp;
1125 struct pwddesc *pdp;
1128 struct nameidata nd;
1129 cap_rights_t rights;
1130 int cmode, error, indx;
1136 AUDIT_ARG_FFLAGS(flags);
1137 AUDIT_ARG_MODE(mode);
1138 cap_rights_init_one(&rights, CAP_LOOKUP);
1139 flags_to_rights(flags, &rights);
1142 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1143 * may be specified. On the other hand, for O_PATH any mode
1144 * except O_EXEC is ignored.
1146 if ((flags & O_PATH) != 0) {
1147 flags &= ~(O_CREAT | O_ACCMODE);
1148 } else if ((flags & O_EXEC) != 0) {
1149 if (flags & O_ACCMODE)
1151 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1154 flags = FFLAGS(flags);
1158 * Allocate a file structure. The descriptor to reference it
1159 * is allocated and used by finstall_refed() below.
1161 error = falloc_noinstall(td, &fp);
1164 /* Set the flags early so the finit in devfs can pick them up. */
1165 fp->f_flag = flags & FMASK;
1166 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1167 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | WANTIOCTLCAPS,
1168 pathseg, path, fd, &rights);
1169 td->td_dupfd = -1; /* XXX check for fdopen */
1170 error = vn_open_cred(&nd, &flags, cmode, VN_OPEN_WANTIOCTLCAPS,
1174 * If the vn_open replaced the method vector, something
1175 * wonderous happened deep below and we just pass it up
1176 * pretending we know what we do.
1178 if (error == ENXIO && fp->f_ops != &badfileops) {
1179 MPASS((flags & O_PATH) == 0);
1184 * Handle special fdopen() case. bleh.
1186 * Don't do this for relative (capability) lookups; we don't
1187 * understand exactly what would happen, and we don't think
1188 * that it ever should.
1190 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1191 (error == ENODEV || error == ENXIO) &&
1192 td->td_dupfd >= 0) {
1193 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1206 * Store the vnode, for any f_type. Typically, the vnode use
1207 * count is decremented by direct call to vn_closefile() for
1208 * files that switched type in the cdevsw fdopen() method.
1213 * If the file wasn't claimed by devfs bind it to the normal
1214 * vnode operations here.
1216 if (fp->f_ops == &badfileops) {
1217 KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1218 ("Unexpected fifo fp %p vp %p", fp, vp));
1219 if ((flags & O_PATH) != 0) {
1220 finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1221 DTYPE_VNODE, NULL, &path_fileops);
1223 finit_vnode(fp, flags, NULL, &vnops);
1228 if (flags & O_TRUNC) {
1229 error = fo_truncate(fp, 0, td->td_ucred, td);
1235 * If we haven't already installed the FD (for dupfdopen), do so now.
1238 struct filecaps *fcaps;
1241 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1242 fcaps = &nd.ni_filecaps;
1246 error = finstall_refed(td, fp, &indx, flags, fcaps);
1247 /* On success finstall_refed() consumes fcaps. */
1252 NDFREE_IOCTLCAPS(&nd);
1253 falloc_abort(td, fp);
1256 td->td_retval[0] = indx;
1259 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1260 NDFREE_IOCTLCAPS(&nd);
1261 falloc_abort(td, fp);
1269 #ifndef _SYS_SYSPROTO_H_
1270 struct ocreat_args {
1276 ocreat(struct thread *td, struct ocreat_args *uap)
1279 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1280 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1282 #endif /* COMPAT_43 */
1285 * Create a special file.
1287 #ifndef _SYS_SYSPROTO_H_
1288 struct mknodat_args {
1296 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1299 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1303 #if defined(COMPAT_FREEBSD11)
1305 freebsd11_mknod(struct thread *td,
1306 struct freebsd11_mknod_args *uap)
1309 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1310 uap->mode, uap->dev));
1314 freebsd11_mknodat(struct thread *td,
1315 struct freebsd11_mknodat_args *uap)
1318 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1321 #endif /* COMPAT_FREEBSD11 */
1324 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1325 int mode, dev_t dev)
1330 struct nameidata nd;
1331 int error, whiteout = 0;
1333 AUDIT_ARG_MODE(mode);
1335 switch (mode & S_IFMT) {
1338 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1339 if (error == 0 && dev == VNOVAL)
1343 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1347 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1358 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1359 NOCACHE, pathseg, path, fd, &cap_mknodat_rights);
1360 if ((error = namei(&nd)) != 0)
1365 if (vp == nd.ni_dvp)
1373 vattr.va_mode = (mode & ALLPERMS) &
1374 ~td->td_proc->p_pd->pd_cmask;
1375 vattr.va_rdev = dev;
1378 switch (mode & S_IFMT) {
1380 vattr.va_type = VCHR;
1383 vattr.va_type = VBLK;
1389 panic("kern_mknod: invalid mode");
1392 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1395 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1400 if (error == 0 && !whiteout)
1401 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1402 &nd.ni_cnd, &vattr);
1406 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1408 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1409 &nd.ni_cnd, &vattr);
1412 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1414 vn_finished_write(mp);
1416 if (error == ERELOOKUP)
1422 * Create a named pipe.
1424 #ifndef _SYS_SYSPROTO_H_
1425 struct mkfifo_args {
1431 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1434 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1438 #ifndef _SYS_SYSPROTO_H_
1439 struct mkfifoat_args {
1446 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1449 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1454 kern_mkfifoat(struct thread *td, int fd, const char *path,
1455 enum uio_seg pathseg, int mode)
1459 struct nameidata nd;
1462 AUDIT_ARG_MODE(mode);
1466 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1467 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights);
1468 if ((error = namei(&nd)) != 0)
1470 if (nd.ni_vp != NULL) {
1472 if (nd.ni_vp == nd.ni_dvp)
1479 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1482 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1487 vattr.va_type = VFIFO;
1488 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1490 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1495 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1499 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1500 vn_finished_write(mp);
1502 if (error == ERELOOKUP)
1508 * Make a hard file link.
1510 #ifndef _SYS_SYSPROTO_H_
1517 sys_link(struct thread *td, struct link_args *uap)
1520 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1521 UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1524 #ifndef _SYS_SYSPROTO_H_
1525 struct linkat_args {
1534 sys_linkat(struct thread *td, struct linkat_args *uap)
1537 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1538 UIO_USERSPACE, uap->flag));
1541 int hardlink_check_uid = 0;
1542 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1543 &hardlink_check_uid, 0,
1544 "Unprivileged processes cannot create hard links to files owned by other "
1546 static int hardlink_check_gid = 0;
1547 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1548 &hardlink_check_gid, 0,
1549 "Unprivileged processes cannot create hard links to files owned by other "
1553 can_hardlink(struct vnode *vp, struct ucred *cred)
1558 if (!hardlink_check_uid && !hardlink_check_gid)
1561 error = VOP_GETATTR(vp, &va, cred);
1565 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1566 error = priv_check_cred(cred, PRIV_VFS_LINK);
1571 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1572 error = priv_check_cred(cred, PRIV_VFS_LINK);
1581 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1582 const char *path2, enum uio_seg segflag, int flag)
1584 struct nameidata nd;
1587 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1588 AT_EMPTY_PATH)) != 0)
1594 NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1595 AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1596 segflag, path1, fd1, &cap_linkat_source_rights);
1597 if ((error = namei(&nd)) != 0)
1600 if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1601 error = priv_check(td, PRIV_VFS_FHOPEN);
1607 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1608 } while (error == EAGAIN || error == ERELOOKUP);
1613 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1614 enum uio_seg segflag)
1616 struct nameidata nd;
1620 if (vp->v_type == VDIR) {
1622 return (EPERM); /* POSIX */
1624 NDINIT_ATRIGHTS(&nd, CREATE,
1625 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1626 &cap_linkat_target_rights);
1627 if ((error = namei(&nd)) == 0) {
1628 if (nd.ni_vp != NULL) {
1630 if (nd.ni_dvp == nd.ni_vp)
1637 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1639 * Cross-device link. No need to recheck
1640 * vp->v_type, since it cannot change, except
1647 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1648 error = can_hardlink(vp, td->td_ucred);
1651 error = mac_vnode_check_link(td->td_ucred,
1652 nd.ni_dvp, vp, &nd.ni_cnd);
1660 error = vn_start_write(vp, &mp, V_NOWAIT);
1665 error = vn_start_write(NULL, &mp,
1671 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1672 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1673 vn_finished_write(mp);
1689 * Make a symbolic link.
1691 #ifndef _SYS_SYSPROTO_H_
1692 struct symlink_args {
1698 sys_symlink(struct thread *td, struct symlink_args *uap)
1701 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1705 #ifndef _SYS_SYSPROTO_H_
1706 struct symlinkat_args {
1713 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1716 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1721 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1722 enum uio_seg segflg)
1726 const char *syspath;
1728 struct nameidata nd;
1731 if (segflg == UIO_SYSSPACE) {
1734 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1735 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1739 AUDIT_ARG_TEXT(syspath);
1743 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1744 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights);
1745 if ((error = namei(&nd)) != 0)
1749 if (nd.ni_vp == nd.ni_dvp)
1758 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1761 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1766 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1768 vattr.va_type = VLNK;
1769 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1774 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1778 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1779 vn_finished_write(mp);
1781 if (error == ERELOOKUP)
1784 if (segflg != UIO_SYSSPACE)
1785 uma_zfree(namei_zone, tmppath);
1790 * Delete a whiteout from the filesystem.
1792 #ifndef _SYS_SYSPROTO_H_
1793 struct undelete_args {
1798 sys_undelete(struct thread *td, struct undelete_args *uap)
1801 struct nameidata nd;
1807 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1808 UIO_USERSPACE, uap->path);
1813 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1815 if (nd.ni_vp == nd.ni_dvp)
1823 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1826 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1830 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1833 vn_finished_write(mp);
1834 if (error == ERELOOKUP)
1840 * Delete a name from the filesystem.
1842 #ifndef _SYS_SYSPROTO_H_
1843 struct unlink_args {
1848 sys_unlink(struct thread *td, struct unlink_args *uap)
1851 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1856 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1857 int flag, enum uio_seg pathseg, ino_t oldinum)
1860 if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1863 if ((flag & AT_REMOVEDIR) != 0)
1864 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1866 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1869 #ifndef _SYS_SYSPROTO_H_
1870 struct unlinkat_args {
1877 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1880 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1884 #ifndef _SYS_SYSPROTO_H_
1885 struct funlinkat_args {
1893 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1896 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1901 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1902 enum uio_seg pathseg, int flag, ino_t oldinum)
1907 struct nameidata nd;
1912 if (fd != FD_NONE) {
1913 error = getvnode_path(td, fd, &cap_no_rights, &fp);
1921 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1922 at2cnpflags(flag, AT_RESOLVE_BENEATH),
1923 pathseg, path, dfd, &cap_unlinkat_rights);
1924 if ((error = namei(&nd)) != 0) {
1925 if (error == EINVAL)
1930 if (vp->v_type == VDIR && oldinum == 0) {
1931 error = EPERM; /* POSIX */
1932 } else if (oldinum != 0 &&
1933 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
1934 sb.st_ino != oldinum) {
1935 error = EIDRM; /* Identifier removed */
1936 } else if (fp != NULL && fp->f_vnode != vp) {
1937 if (VN_IS_DOOMED(fp->f_vnode))
1943 * The root of a mounted filesystem cannot be deleted.
1945 * XXX: can this only be a VDIR case?
1947 if (vp->v_vflag & VV_ROOT)
1951 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1954 if (vp == nd.ni_dvp)
1958 if ((error = vn_start_write(NULL, &mp,
1959 V_XSLEEP | PCATCH)) != 0) {
1965 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1970 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1971 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1975 vn_finished_write(mp);
1979 if (vp == nd.ni_dvp)
1983 if (error == ERELOOKUP)
1992 * Reposition read/write file offset.
1994 #ifndef _SYS_SYSPROTO_H_
2003 sys_lseek(struct thread *td, struct lseek_args *uap)
2006 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2010 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2016 error = fget(td, fd, &cap_seek_rights, &fp);
2019 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2020 fo_seek(fp, offset, whence, td) : ESPIPE;
2025 #if defined(COMPAT_43)
2027 * Reposition read/write file offset.
2029 #ifndef _SYS_SYSPROTO_H_
2030 struct olseek_args {
2037 olseek(struct thread *td, struct olseek_args *uap)
2040 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2042 #endif /* COMPAT_43 */
2044 #if defined(COMPAT_FREEBSD6)
2045 /* Version with the 'pad' argument */
2047 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2050 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2055 * Check access permissions using passed credentials.
2058 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2064 /* Flags == 0 means only check for existence. */
2065 if (user_flags == 0)
2069 if (user_flags & R_OK)
2071 if (user_flags & W_OK)
2073 if (user_flags & X_OK)
2076 error = mac_vnode_check_access(cred, vp, accmode);
2080 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2081 error = VOP_ACCESS(vp, accmode, cred, td);
2086 * Check access permissions using "real" credentials.
2088 #ifndef _SYS_SYSPROTO_H_
2089 struct access_args {
2095 sys_access(struct thread *td, struct access_args *uap)
2098 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2102 #ifndef _SYS_SYSPROTO_H_
2103 struct faccessat_args {
2111 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2114 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2119 kern_accessat(struct thread *td, int fd, const char *path,
2120 enum uio_seg pathseg, int flag, int amode)
2122 struct ucred *cred, *usecred;
2124 struct nameidata nd;
2127 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2129 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2133 * Create and modify a temporary credential instead of one that
2134 * is potentially shared (if we need one).
2136 cred = td->td_ucred;
2137 if ((flag & AT_EACCESS) == 0 &&
2138 ((cred->cr_uid != cred->cr_ruid ||
2139 cred->cr_rgid != cred->cr_groups[0]))) {
2140 usecred = crdup(cred);
2141 usecred->cr_uid = cred->cr_ruid;
2142 usecred->cr_groups[0] = cred->cr_rgid;
2143 td->td_ucred = usecred;
2146 AUDIT_ARG_VALUE(amode);
2147 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2148 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2149 AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights);
2150 if ((error = namei(&nd)) != 0)
2154 error = vn_access(vp, amode, usecred, td);
2155 NDFREE_NOTHING(&nd);
2158 if (usecred != cred) {
2159 td->td_ucred = cred;
2166 * Check access permissions using "effective" credentials.
2168 #ifndef _SYS_SYSPROTO_H_
2169 struct eaccess_args {
2175 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2178 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2179 AT_EACCESS, uap->amode));
2182 #if defined(COMPAT_43)
2184 * Get file status; this version follows links.
2186 #ifndef _SYS_SYSPROTO_H_
2193 ostat(struct thread *td, struct ostat_args *uap)
2199 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2204 return (copyout(&osb, uap->ub, sizeof (osb)));
2208 * Get file status; this version does not follow links.
2210 #ifndef _SYS_SYSPROTO_H_
2211 struct olstat_args {
2217 olstat(struct thread *td, struct olstat_args *uap)
2223 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2224 UIO_USERSPACE, &sb, NULL);
2228 return (copyout(&osb, uap->ub, sizeof (osb)));
2232 * Convert from an old to a new stat structure.
2233 * XXX: many values are blindly truncated.
2236 cvtstat(struct stat *st, struct ostat *ost)
2239 bzero(ost, sizeof(*ost));
2240 ost->st_dev = st->st_dev;
2241 ost->st_ino = st->st_ino;
2242 ost->st_mode = st->st_mode;
2243 ost->st_nlink = st->st_nlink;
2244 ost->st_uid = st->st_uid;
2245 ost->st_gid = st->st_gid;
2246 ost->st_rdev = st->st_rdev;
2247 ost->st_size = MIN(st->st_size, INT32_MAX);
2248 ost->st_atim = st->st_atim;
2249 ost->st_mtim = st->st_mtim;
2250 ost->st_ctim = st->st_ctim;
2251 ost->st_blksize = st->st_blksize;
2252 ost->st_blocks = st->st_blocks;
2253 ost->st_flags = st->st_flags;
2254 ost->st_gen = st->st_gen;
2256 #endif /* COMPAT_43 */
2258 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2259 int ino64_trunc_error;
2260 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2261 &ino64_trunc_error, 0,
2262 "Error on truncation of device, file or inode number, or link count");
2265 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2268 ost->st_dev = st->st_dev;
2269 if (ost->st_dev != st->st_dev) {
2270 switch (ino64_trunc_error) {
2273 * Since dev_t is almost raw, don't clamp to the
2274 * maximum for case 2, but ignore the error.
2281 ost->st_ino = st->st_ino;
2282 if (ost->st_ino != st->st_ino) {
2283 switch (ino64_trunc_error) {
2290 ost->st_ino = UINT32_MAX;
2294 ost->st_mode = st->st_mode;
2295 ost->st_nlink = st->st_nlink;
2296 if (ost->st_nlink != st->st_nlink) {
2297 switch (ino64_trunc_error) {
2304 ost->st_nlink = UINT16_MAX;
2308 ost->st_uid = st->st_uid;
2309 ost->st_gid = st->st_gid;
2310 ost->st_rdev = st->st_rdev;
2311 if (ost->st_rdev != st->st_rdev) {
2312 switch (ino64_trunc_error) {
2319 ost->st_atim = st->st_atim;
2320 ost->st_mtim = st->st_mtim;
2321 ost->st_ctim = st->st_ctim;
2322 ost->st_size = st->st_size;
2323 ost->st_blocks = st->st_blocks;
2324 ost->st_blksize = st->st_blksize;
2325 ost->st_flags = st->st_flags;
2326 ost->st_gen = st->st_gen;
2328 ost->st_birthtim = st->st_birthtim;
2329 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2330 sizeof(*ost) - offsetof(struct freebsd11_stat,
2331 st_birthtim) - sizeof(ost->st_birthtim));
2336 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2339 struct freebsd11_stat osb;
2342 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2346 error = freebsd11_cvtstat(&sb, &osb);
2348 error = copyout(&osb, uap->ub, sizeof(osb));
2353 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2356 struct freebsd11_stat osb;
2359 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2360 UIO_USERSPACE, &sb, NULL);
2363 error = freebsd11_cvtstat(&sb, &osb);
2365 error = copyout(&osb, uap->ub, sizeof(osb));
2370 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2374 struct freebsd11_stat osb;
2377 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2380 error = kern_fhstat(td, fh, &sb);
2383 error = freebsd11_cvtstat(&sb, &osb);
2385 error = copyout(&osb, uap->sb, sizeof(osb));
2390 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2393 struct freebsd11_stat osb;
2396 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2397 UIO_USERSPACE, &sb, NULL);
2400 error = freebsd11_cvtstat(&sb, &osb);
2402 error = copyout(&osb, uap->buf, sizeof(osb));
2405 #endif /* COMPAT_FREEBSD11 */
2410 #ifndef _SYS_SYSPROTO_H_
2411 struct fstatat_args {
2419 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2424 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2425 UIO_USERSPACE, &sb, NULL);
2427 error = copyout(&sb, uap->buf, sizeof (sb));
2432 kern_statat(struct thread *td, int flag, int fd, const char *path,
2433 enum uio_seg pathseg, struct stat *sbp,
2434 void (*hook)(struct vnode *vp, struct stat *sbp))
2436 struct nameidata nd;
2439 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2440 AT_EMPTY_PATH)) != 0)
2443 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2444 AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2445 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights);
2447 if ((error = namei(&nd)) != 0) {
2448 if (error == ENOTDIR &&
2449 (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2450 error = kern_fstat(td, fd, sbp);
2453 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2455 if (__predict_false(hook != NULL))
2456 hook(nd.ni_vp, sbp);
2458 NDFREE_NOTHING(&nd);
2460 #ifdef __STAT_TIME_T_EXT
2461 sbp->st_atim_ext = 0;
2462 sbp->st_mtim_ext = 0;
2463 sbp->st_ctim_ext = 0;
2464 sbp->st_btim_ext = 0;
2467 if (KTRPOINT(td, KTR_STRUCT))
2468 ktrstat_error(sbp, error);
2473 #if defined(COMPAT_FREEBSD11)
2475 * Implementation of the NetBSD [l]stat() functions.
2478 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2480 struct freebsd11_stat sb11;
2483 error = freebsd11_cvtstat(sb, &sb11);
2487 bzero(nsb, sizeof(*nsb));
2488 CP(sb11, *nsb, st_dev);
2489 CP(sb11, *nsb, st_ino);
2490 CP(sb11, *nsb, st_mode);
2491 CP(sb11, *nsb, st_nlink);
2492 CP(sb11, *nsb, st_uid);
2493 CP(sb11, *nsb, st_gid);
2494 CP(sb11, *nsb, st_rdev);
2495 CP(sb11, *nsb, st_atim);
2496 CP(sb11, *nsb, st_mtim);
2497 CP(sb11, *nsb, st_ctim);
2498 CP(sb11, *nsb, st_size);
2499 CP(sb11, *nsb, st_blocks);
2500 CP(sb11, *nsb, st_blksize);
2501 CP(sb11, *nsb, st_flags);
2502 CP(sb11, *nsb, st_gen);
2503 CP(sb11, *nsb, st_birthtim);
2507 #ifndef _SYS_SYSPROTO_H_
2508 struct freebsd11_nstat_args {
2514 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2520 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2524 error = freebsd11_cvtnstat(&sb, &nsb);
2526 error = copyout(&nsb, uap->ub, sizeof (nsb));
2531 * NetBSD lstat. Get file status; this version does not follow links.
2533 #ifndef _SYS_SYSPROTO_H_
2534 struct freebsd11_nlstat_args {
2540 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2546 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2547 UIO_USERSPACE, &sb, NULL);
2550 error = freebsd11_cvtnstat(&sb, &nsb);
2552 error = copyout(&nsb, uap->ub, sizeof (nsb));
2555 #endif /* COMPAT_FREEBSD11 */
2558 * Get configurable pathname variables.
2560 #ifndef _SYS_SYSPROTO_H_
2561 struct pathconf_args {
2567 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2572 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2575 td->td_retval[0] = value;
2579 #ifndef _SYS_SYSPROTO_H_
2580 struct lpathconf_args {
2586 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2591 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2594 td->td_retval[0] = value;
2599 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2600 int name, u_long flags, long *valuep)
2602 struct nameidata nd;
2605 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2607 if ((error = namei(&nd)) != 0)
2609 NDFREE_NOTHING(&nd);
2611 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2617 * Return target name of a symbolic link.
2619 #ifndef _SYS_SYSPROTO_H_
2620 struct readlink_args {
2627 sys_readlink(struct thread *td, struct readlink_args *uap)
2630 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2631 uap->buf, UIO_USERSPACE, uap->count));
2633 #ifndef _SYS_SYSPROTO_H_
2634 struct readlinkat_args {
2642 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2645 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2646 uap->buf, UIO_USERSPACE, uap->bufsize));
2650 kern_readlinkat(struct thread *td, int fd, const char *path,
2651 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2654 struct nameidata nd;
2657 if (count > IOSIZE_MAX)
2660 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2661 EMPTYPATH, pathseg, path, fd);
2663 if ((error = namei(&nd)) != 0)
2665 NDFREE_NOTHING(&nd);
2668 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2675 * Helper function to readlink from a vnode
2678 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2685 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2687 error = mac_vnode_check_readlink(td->td_ucred, vp);
2691 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2694 aiov.iov_base = buf;
2695 aiov.iov_len = count;
2696 auio.uio_iov = &aiov;
2697 auio.uio_iovcnt = 1;
2698 auio.uio_offset = 0;
2699 auio.uio_rw = UIO_READ;
2700 auio.uio_segflg = bufseg;
2702 auio.uio_resid = count;
2703 error = VOP_READLINK(vp, &auio, td->td_ucred);
2704 td->td_retval[0] = count - auio.uio_resid;
2709 * Common implementation code for chflags() and fchflags().
2712 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2718 /* We can't support the value matching VNOVAL. */
2719 if (flags == VNOVAL)
2720 return (EOPNOTSUPP);
2723 * Prevent non-root users from setting flags on devices. When
2724 * a device is reused, users can retain ownership of the device
2725 * if they are allowed to set flags and programs assume that
2726 * chown can't fail when done as root.
2728 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2729 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2734 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2737 vattr.va_flags = flags;
2738 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2740 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2743 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2745 vn_finished_write(mp);
2750 * Change flags of a file given a path name.
2752 #ifndef _SYS_SYSPROTO_H_
2753 struct chflags_args {
2759 sys_chflags(struct thread *td, struct chflags_args *uap)
2762 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2766 #ifndef _SYS_SYSPROTO_H_
2767 struct chflagsat_args {
2775 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2778 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2779 uap->flags, uap->atflag));
2783 * Same as chflags() but doesn't follow symlinks.
2785 #ifndef _SYS_SYSPROTO_H_
2786 struct lchflags_args {
2792 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2795 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2796 uap->flags, AT_SYMLINK_NOFOLLOW));
2800 kern_chflagsat(struct thread *td, int fd, const char *path,
2801 enum uio_seg pathseg, u_long flags, int atflag)
2803 struct nameidata nd;
2806 if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2807 AT_EMPTY_PATH)) != 0)
2810 AUDIT_ARG_FFLAGS(flags);
2811 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2812 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2813 fd, &cap_fchflags_rights);
2814 if ((error = namei(&nd)) != 0)
2816 NDFREE_NOTHING(&nd);
2817 error = setfflags(td, nd.ni_vp, flags);
2823 * Change flags of a file given a file descriptor.
2825 #ifndef _SYS_SYSPROTO_H_
2826 struct fchflags_args {
2832 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2837 AUDIT_ARG_FD(uap->fd);
2838 AUDIT_ARG_FFLAGS(uap->flags);
2839 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2844 if (AUDITING_TD(td)) {
2845 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2846 AUDIT_ARG_VNODE1(fp->f_vnode);
2847 VOP_UNLOCK(fp->f_vnode);
2850 error = setfflags(td, fp->f_vnode, uap->flags);
2856 * Common implementation code for chmod(), lchmod() and fchmod().
2859 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2865 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2867 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2869 vattr.va_mode = mode & ALLPERMS;
2871 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2874 error = VOP_SETATTR(vp, &vattr, cred);
2876 vn_finished_write(mp);
2881 * Change mode of a file given path name.
2883 #ifndef _SYS_SYSPROTO_H_
2890 sys_chmod(struct thread *td, struct chmod_args *uap)
2893 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2897 #ifndef _SYS_SYSPROTO_H_
2898 struct fchmodat_args {
2906 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2909 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2910 uap->mode, uap->flag));
2914 * Change mode of a file given path name (don't follow links.)
2916 #ifndef _SYS_SYSPROTO_H_
2917 struct lchmod_args {
2923 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2926 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2927 uap->mode, AT_SYMLINK_NOFOLLOW));
2931 kern_fchmodat(struct thread *td, int fd, const char *path,
2932 enum uio_seg pathseg, mode_t mode, int flag)
2934 struct nameidata nd;
2937 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2938 AT_EMPTY_PATH)) != 0)
2941 AUDIT_ARG_MODE(mode);
2942 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2943 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2944 fd, &cap_fchmod_rights);
2945 if ((error = namei(&nd)) != 0)
2947 NDFREE_NOTHING(&nd);
2948 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2954 * Change mode of a file given a file descriptor.
2956 #ifndef _SYS_SYSPROTO_H_
2957 struct fchmod_args {
2963 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2968 AUDIT_ARG_FD(uap->fd);
2969 AUDIT_ARG_MODE(uap->mode);
2971 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2974 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2980 * Common implementation for chown(), lchown(), and fchown()
2983 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2990 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2992 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2997 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3001 error = VOP_SETATTR(vp, &vattr, cred);
3003 vn_finished_write(mp);
3008 * Set ownership given a path name.
3010 #ifndef _SYS_SYSPROTO_H_
3018 sys_chown(struct thread *td, struct chown_args *uap)
3021 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3025 #ifndef _SYS_SYSPROTO_H_
3026 struct fchownat_args {
3035 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3038 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3039 uap->gid, uap->flag));
3043 kern_fchownat(struct thread *td, int fd, const char *path,
3044 enum uio_seg pathseg, int uid, int gid, int flag)
3046 struct nameidata nd;
3049 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3050 AT_EMPTY_PATH)) != 0)
3053 AUDIT_ARG_OWNER(uid, gid);
3054 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3055 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3056 fd, &cap_fchown_rights);
3058 if ((error = namei(&nd)) != 0)
3060 NDFREE_NOTHING(&nd);
3061 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3067 * Set ownership given a path name, do not cross symlinks.
3069 #ifndef _SYS_SYSPROTO_H_
3070 struct lchown_args {
3077 sys_lchown(struct thread *td, struct lchown_args *uap)
3080 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3081 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3085 * Set ownership given a file descriptor.
3087 #ifndef _SYS_SYSPROTO_H_
3088 struct fchown_args {
3095 sys_fchown(struct thread *td, struct fchown_args *uap)
3100 AUDIT_ARG_FD(uap->fd);
3101 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3102 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3105 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3111 * Common implementation code for utimes(), lutimes(), and futimes().
3114 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3115 struct timespec *tsp)
3117 struct timeval tv[2];
3118 const struct timeval *tvp;
3121 if (usrtvp == NULL) {
3122 vfs_timestamp(&tsp[0]);
3125 if (tvpseg == UIO_SYSSPACE) {
3128 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3133 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3134 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3136 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3137 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3143 * Common implementation code for futimens(), utimensat().
3145 #define UTIMENS_NULL 0x1
3146 #define UTIMENS_EXIT 0x2
3148 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3149 struct timespec *tsp, int *retflags)
3151 struct timespec tsnow;
3154 vfs_timestamp(&tsnow);
3156 if (usrtsp == NULL) {
3159 *retflags |= UTIMENS_NULL;
3162 if (tspseg == UIO_SYSSPACE) {
3165 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3167 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3168 *retflags |= UTIMENS_EXIT;
3169 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3170 *retflags |= UTIMENS_NULL;
3171 if (tsp[0].tv_nsec == UTIME_OMIT)
3172 tsp[0].tv_sec = VNOVAL;
3173 else if (tsp[0].tv_nsec == UTIME_NOW)
3175 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3177 if (tsp[1].tv_nsec == UTIME_OMIT)
3178 tsp[1].tv_sec = VNOVAL;
3179 else if (tsp[1].tv_nsec == UTIME_NOW)
3181 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3188 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3192 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3193 int numtimes, int nullflag)
3200 setbirthtime = false;
3201 vattr.va_birthtime.tv_sec = VNOVAL;
3202 vattr.va_birthtime.tv_nsec = 0;
3204 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3206 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3207 if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3208 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3209 setbirthtime = true;
3211 vattr.va_atime = ts[0];
3212 vattr.va_mtime = ts[1];
3214 vattr.va_birthtime = ts[1];
3216 vattr.va_birthtime = ts[2];
3218 vattr.va_vaflags |= VA_UTIMES_NULL;
3220 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3224 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3226 vn_finished_write(mp);
3231 * Set the access and modification times of a file.
3233 #ifndef _SYS_SYSPROTO_H_
3234 struct utimes_args {
3236 struct timeval *tptr;
3240 sys_utimes(struct thread *td, struct utimes_args *uap)
3243 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3244 uap->tptr, UIO_USERSPACE));
3247 #ifndef _SYS_SYSPROTO_H_
3248 struct futimesat_args {
3251 const struct timeval * times;
3255 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3258 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3259 uap->times, UIO_USERSPACE));
3263 kern_utimesat(struct thread *td, int fd, const char *path,
3264 enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg)
3266 struct nameidata nd;
3267 struct timespec ts[2];
3270 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3272 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3273 &cap_futimes_rights);
3275 if ((error = namei(&nd)) != 0)
3277 NDFREE_NOTHING(&nd);
3278 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3284 * Set the access and modification times of a file.
3286 #ifndef _SYS_SYSPROTO_H_
3287 struct lutimes_args {
3289 struct timeval *tptr;
3293 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3296 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3301 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3302 const struct timeval *tptr, enum uio_seg tptrseg)
3304 struct timespec ts[2];
3305 struct nameidata nd;
3308 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3310 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path);
3311 if ((error = namei(&nd)) != 0)
3313 NDFREE_NOTHING(&nd);
3314 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3320 * Set the access and modification times of a file.
3322 #ifndef _SYS_SYSPROTO_H_
3323 struct futimes_args {
3325 struct timeval *tptr;
3329 sys_futimes(struct thread *td, struct futimes_args *uap)
3332 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3336 kern_futimes(struct thread *td, int fd, const struct timeval *tptr,
3337 enum uio_seg tptrseg)
3339 struct timespec ts[2];
3344 error = getutimes(tptr, tptrseg, ts);
3347 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3351 if (AUDITING_TD(td)) {
3352 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3353 AUDIT_ARG_VNODE1(fp->f_vnode);
3354 VOP_UNLOCK(fp->f_vnode);
3357 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3363 sys_futimens(struct thread *td, struct futimens_args *uap)
3366 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3370 kern_futimens(struct thread *td, int fd, const struct timespec *tptr,
3371 enum uio_seg tptrseg)
3373 struct timespec ts[2];
3378 error = getutimens(tptr, tptrseg, ts, &flags);
3381 if (flags & UTIMENS_EXIT)
3383 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3387 if (AUDITING_TD(td)) {
3388 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3389 AUDIT_ARG_VNODE1(fp->f_vnode);
3390 VOP_UNLOCK(fp->f_vnode);
3393 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3399 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3402 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3403 uap->times, UIO_USERSPACE, uap->flag));
3407 kern_utimensat(struct thread *td, int fd, const char *path,
3408 enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg,
3411 struct nameidata nd;
3412 struct timespec ts[2];
3415 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3416 AT_EMPTY_PATH)) != 0)
3419 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3421 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3422 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3423 pathseg, path, fd, &cap_futimes_rights);
3424 if ((error = namei(&nd)) != 0)
3427 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3429 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3430 * "Search permission is denied by a component of the path prefix."
3432 NDFREE_NOTHING(&nd);
3433 if ((flags & UTIMENS_EXIT) == 0)
3434 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3440 * Truncate a file given its path name.
3442 #ifndef _SYS_SYSPROTO_H_
3443 struct truncate_args {
3450 sys_truncate(struct thread *td, struct truncate_args *uap)
3453 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3457 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3464 struct nameidata nd;
3471 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
3472 if ((error = namei(&nd)) != 0)
3475 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3476 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3477 vn_rangelock_unlock(vp, rl_cookie);
3482 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3483 if (vp->v_type == VDIR)
3486 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3489 else if ((error = vn_writechk(vp)) == 0 &&
3490 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3492 vattr.va_size = length;
3493 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3496 vn_finished_write(mp);
3497 vn_rangelock_unlock(vp, rl_cookie);
3499 if (error == ERELOOKUP)
3504 #if defined(COMPAT_43)
3506 * Truncate a file given its path name.
3508 #ifndef _SYS_SYSPROTO_H_
3509 struct otruncate_args {
3515 otruncate(struct thread *td, struct otruncate_args *uap)
3518 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3520 #endif /* COMPAT_43 */
3522 #if defined(COMPAT_FREEBSD6)
3523 /* Versions with the pad argument */
3525 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3528 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3532 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3535 return (kern_ftruncate(td, uap->fd, uap->length));
3540 kern_fsync(struct thread *td, int fd, bool fullsync)
3548 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3554 /* XXXKIB: compete outstanding aio writes */;
3557 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3560 vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3561 AUDIT_ARG_VNODE1(vp);
3562 if (vp->v_object != NULL) {
3563 VM_OBJECT_WLOCK(vp->v_object);
3564 vm_object_page_clean(vp->v_object, 0, 0, 0);
3565 VM_OBJECT_WUNLOCK(vp->v_object);
3567 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3569 vn_finished_write(mp);
3570 if (error == ERELOOKUP)
3578 * Sync an open file.
3580 #ifndef _SYS_SYSPROTO_H_
3586 sys_fsync(struct thread *td, struct fsync_args *uap)
3589 return (kern_fsync(td, uap->fd, true));
3593 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3596 return (kern_fsync(td, uap->fd, false));
3600 * Rename files. Source and destination must either both be directories, or
3601 * both not be directories. If target is a directory, it must be empty.
3603 #ifndef _SYS_SYSPROTO_H_
3604 struct rename_args {
3610 sys_rename(struct thread *td, struct rename_args *uap)
3613 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3614 uap->to, UIO_USERSPACE));
3617 #ifndef _SYS_SYSPROTO_H_
3618 struct renameat_args {
3626 sys_renameat(struct thread *td, struct renameat_args *uap)
3629 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3635 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3636 const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3640 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3641 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights);
3642 if ((error = namei(fromnd)) != 0)
3644 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3645 fromnd->ni_vp, &fromnd->ni_cnd);
3646 VOP_UNLOCK(fromnd->ni_dvp);
3647 if (fromnd->ni_dvp != fromnd->ni_vp)
3648 VOP_UNLOCK(fromnd->ni_vp);
3650 NDFREE_PNBUF(fromnd);
3651 vrele(fromnd->ni_dvp);
3652 vrele(fromnd->ni_vp);
3653 if (fromnd->ni_startdir)
3654 vrele(fromnd->ni_startdir);
3661 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3662 const char *new, enum uio_seg pathseg)
3664 struct mount *mp = NULL;
3665 struct vnode *tvp, *fvp, *tdvp;
3666 struct nameidata fromnd, tond;
3673 if (mac_vnode_check_rename_from_enabled()) {
3674 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3680 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3681 pathseg, old, oldfd, &cap_renameat_source_rights);
3682 if ((error = namei(&fromnd)) != 0)
3688 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3689 if (fromnd.ni_vp->v_type == VDIR)
3690 tondflags |= WILLBEDIR;
3691 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3692 &cap_renameat_target_rights);
3693 if ((error = namei(&tond)) != 0) {
3694 /* Translate error code for rename("dir1", "dir2/."). */
3695 if (error == EISDIR && fvp->v_type == VDIR)
3697 NDFREE_PNBUF(&fromnd);
3698 vrele(fromnd.ni_dvp);
3704 error = vn_start_write(fvp, &mp, V_NOWAIT);
3706 NDFREE_PNBUF(&fromnd);
3707 NDFREE_PNBUF(&tond);
3714 vrele(fromnd.ni_dvp);
3716 vrele(tond.ni_startdir);
3717 if (fromnd.ni_startdir != NULL)
3718 vrele(fromnd.ni_startdir);
3719 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3725 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3728 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3733 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3735 * If the target already exists we require CAP_UNLINKAT
3736 * from 'newfd', when newfd was used for the lookup.
3738 error = cap_check(&tond.ni_filecaps.fc_rights,
3739 &cap_unlinkat_rights);
3750 * If the source is the same as the destination (that is, if they
3751 * are links to the same vnode), then there is nothing to do.
3757 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3758 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3762 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3763 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3764 NDFREE_PNBUF(&fromnd);
3765 NDFREE_PNBUF(&tond);
3767 NDFREE_PNBUF(&fromnd);
3768 NDFREE_PNBUF(&tond);
3775 vrele(fromnd.ni_dvp);
3778 vrele(tond.ni_startdir);
3779 vn_finished_write(mp);
3781 if (fromnd.ni_startdir)
3782 vrele(fromnd.ni_startdir);
3783 if (error == ERESTART)
3785 if (error == ERELOOKUP)
3791 * Make a directory file.
3793 #ifndef _SYS_SYSPROTO_H_
3800 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3803 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3807 #ifndef _SYS_SYSPROTO_H_
3808 struct mkdirat_args {
3815 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3818 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3822 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3827 struct nameidata nd;
3830 AUDIT_ARG_MODE(mode);
3834 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3835 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3836 segflg, path, fd, &cap_mkdirat_rights);
3837 if ((error = namei(&nd)) != 0)
3839 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3842 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3847 vattr.va_type = VDIR;
3848 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3850 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3855 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3860 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3861 vn_finished_write(mp);
3862 if (error == ERELOOKUP)
3868 * Remove a directory file.
3870 #ifndef _SYS_SYSPROTO_H_
3876 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3879 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3884 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3885 enum uio_seg pathseg, int flag)
3890 struct nameidata nd;
3891 cap_rights_t rights;
3895 if (fd != FD_NONE) {
3896 error = getvnode(td, fd, cap_rights_init_one(&rights,
3905 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3906 at2cnpflags(flag, AT_RESOLVE_BENEATH),
3907 pathseg, path, dfd, &cap_unlinkat_rights);
3908 if ((error = namei(&nd)) != 0)
3911 if (vp->v_type != VDIR) {
3916 * No rmdir "." please.
3918 if (nd.ni_dvp == vp) {
3923 * The root of a mounted filesystem cannot be deleted.
3925 if (vp->v_vflag & VV_ROOT) {
3930 if (fp != NULL && fp->f_vnode != vp) {
3931 if (VN_IS_DOOMED(fp->f_vnode))
3939 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3944 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3947 if (nd.ni_dvp == vp)
3951 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3955 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3956 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3957 vn_finished_write(mp);
3961 if (nd.ni_dvp == vp)
3965 if (error == ERELOOKUP)
3973 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3975 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3976 long *basep, void (*func)(struct freebsd11_dirent *))
3978 struct freebsd11_dirent dstdp;
3979 struct dirent *dp, *edp;
3982 ssize_t resid, ucount;
3985 /* XXX arbitrary sanity limit on `count'. */
3986 count = min(count, 64 * 1024);
3988 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3990 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3998 for (dp = (struct dirent *)dirbuf,
3999 edp = (struct dirent *)&dirbuf[count - resid];
4000 ucount < count && dp < edp; ) {
4001 if (dp->d_reclen == 0)
4003 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
4004 if (dp->d_namlen >= sizeof(dstdp.d_name))
4006 dstdp.d_type = dp->d_type;
4007 dstdp.d_namlen = dp->d_namlen;
4008 dstdp.d_fileno = dp->d_fileno; /* truncate */
4009 if (dstdp.d_fileno != dp->d_fileno) {
4010 switch (ino64_trunc_error) {
4018 dstdp.d_fileno = UINT32_MAX;
4022 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4023 ((dp->d_namlen + 1 + 3) &~ 3);
4024 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4025 bzero(dstdp.d_name + dstdp.d_namlen,
4026 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4028 MPASS(dstdp.d_reclen <= dp->d_reclen);
4029 MPASS(ucount + dstdp.d_reclen <= count);
4032 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4035 dp = (struct dirent *)((char *)dp + dp->d_reclen);
4036 ucount += dstdp.d_reclen;
4040 free(dirbuf, M_TEMP);
4042 td->td_retval[0] = ucount;
4049 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4051 #if (BYTE_ORDER == LITTLE_ENDIAN)
4053 * The expected low byte of dp->d_namlen is our dp->d_type.
4054 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4056 dp->d_type = dp->d_namlen;
4060 * The dp->d_type is the high byte of the expected dp->d_namlen,
4061 * so must be zero'ed.
4068 * Read a block of directory entries in a filesystem independent format.
4070 #ifndef _SYS_SYSPROTO_H_
4071 struct ogetdirentries_args {
4079 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4084 error = kern_ogetdirentries(td, uap, &loff);
4086 error = copyout(&loff, uap->basep, sizeof(long));
4091 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4097 /* XXX arbitrary sanity limit on `count'. */
4098 if (uap->count > 64 * 1024)
4101 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4102 &base, ogetdirentries_cvt);
4104 if (error == 0 && uap->basep != NULL)
4105 error = copyout(&base, uap->basep, sizeof(long));
4109 #endif /* COMPAT_43 */
4111 #if defined(COMPAT_FREEBSD11)
4112 #ifndef _SYS_SYSPROTO_H_
4113 struct freebsd11_getdirentries_args {
4121 freebsd11_getdirentries(struct thread *td,
4122 struct freebsd11_getdirentries_args *uap)
4127 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4130 if (error == 0 && uap->basep != NULL)
4131 error = copyout(&base, uap->basep, sizeof(long));
4136 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4138 struct freebsd11_getdirentries_args ap;
4142 ap.count = uap->count;
4144 return (freebsd11_getdirentries(td, &ap));
4146 #endif /* COMPAT_FREEBSD11 */
4149 * Read a block of directory entries in a filesystem independent format.
4152 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4157 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4158 NULL, UIO_USERSPACE);
4161 if (uap->basep != NULL)
4162 error = copyout(&base, uap->basep, sizeof(off_t));
4167 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4168 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4179 if (count > IOSIZE_MAX)
4181 auio.uio_resid = count;
4182 error = getvnode(td, fd, &cap_read_rights, &fp);
4185 if ((fp->f_flag & FREAD) == 0) {
4190 foffset = foffset_lock(fp, 0);
4192 if (vp->v_type != VDIR) {
4196 if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) {
4200 aiov.iov_base = buf;
4201 aiov.iov_len = count;
4202 auio.uio_iov = &aiov;
4203 auio.uio_iovcnt = 1;
4204 auio.uio_rw = UIO_READ;
4205 auio.uio_segflg = bufseg;
4207 vn_lock(vp, LK_SHARED | LK_RETRY);
4208 AUDIT_ARG_VNODE1(vp);
4209 loff = auio.uio_offset = foffset;
4211 error = mac_vnode_check_readdir(td->td_ucred, vp);
4214 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4216 foffset = auio.uio_offset;
4221 if (count == auio.uio_resid &&
4222 (vp->v_vflag & VV_ROOT) &&
4223 (vp->v_mount->mnt_flag & MNT_UNION)) {
4224 struct vnode *tvp = vp;
4226 vp = vp->v_mount->mnt_vnodecovered;
4236 *residp = auio.uio_resid;
4237 td->td_retval[0] = count - auio.uio_resid;
4239 foffset_unlock(fp, foffset, 0);
4245 * Set the mode mask for creation of filesystem nodes.
4247 #ifndef _SYS_SYSPROTO_H_
4253 sys_umask(struct thread *td, struct umask_args *uap)
4255 struct pwddesc *pdp;
4257 pdp = td->td_proc->p_pd;
4259 td->td_retval[0] = pdp->pd_cmask;
4260 pdp->pd_cmask = uap->newmask & ALLPERMS;
4261 PWDDESC_XUNLOCK(pdp);
4266 * Void all references to file by ripping underlying filesystem away from
4269 #ifndef _SYS_SYSPROTO_H_
4270 struct revoke_args {
4275 sys_revoke(struct thread *td, struct revoke_args *uap)
4279 struct nameidata nd;
4282 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4284 if ((error = namei(&nd)) != 0)
4287 NDFREE_NOTHING(&nd);
4288 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4293 error = mac_vnode_check_revoke(td->td_ucred, vp);
4297 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4300 if (td->td_ucred->cr_uid != vattr.va_uid) {
4301 error = priv_check(td, PRIV_VFS_ADMIN);
4305 if (devfs_usecount(vp) > 0)
4306 VOP_REVOKE(vp, REVOKEALL);
4313 * This variant of getvnode() allows O_PATH files. Caller should
4314 * ensure that returned file and vnode are only used for compatible
4318 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4324 error = fget_unlocked(td, fd, rightsp, &fp);
4329 * The file could be not of the vnode type, or it may be not
4330 * yet fully initialized, in which case the f_vnode pointer
4331 * may be set, but f_ops is still badfileops. E.g.,
4332 * devfs_open() transiently create such situation to
4333 * facilitate csw d_fdopen().
4335 * Dupfdopen() handling in kern_openat() installs the
4336 * half-baked file into the process descriptor table, allowing
4337 * other thread to dereference it. Guard against the race by
4340 if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4351 * Convert a user file descriptor to a kernel file entry and check
4352 * that, if it is a capability, the correct rights are present.
4353 * A reference on the file entry is held upon returning.
4356 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4360 error = getvnode_path(td, fd, rightsp, fpp);
4361 if (__predict_false(error != 0))
4365 * Filter out O_PATH file descriptors, most getvnode() callers
4366 * do not call fo_ methods.
4368 if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4378 * Get an (NFS) file handle.
4380 #ifndef _SYS_SYSPROTO_H_
4381 struct lgetfh_args {
4387 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4390 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4391 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4394 #ifndef _SYS_SYSPROTO_H_
4401 sys_getfh(struct thread *td, struct getfh_args *uap)
4404 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4405 uap->fhp, UIO_USERSPACE));
4409 * syscall for the rpc.lockd to use to translate an open descriptor into
4410 * a NFS file handle.
4412 * warning: do not remove the priv_check() call or this becomes one giant
4415 #ifndef _SYS_SYSPROTO_H_
4416 struct getfhat_args {
4424 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4427 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4428 uap->fhp, UIO_USERSPACE));
4432 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4433 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4435 struct nameidata nd;
4440 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4442 error = priv_check(td, PRIV_VFS_GETFH);
4445 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4446 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4451 NDFREE_NOTHING(&nd);
4453 bzero(&fh, sizeof(fh));
4454 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4455 error = VOP_VPTOFH(vp, &fh.fh_fid);
4458 if (fhseg == UIO_USERSPACE)
4459 error = copyout(&fh, fhp, sizeof (fh));
4461 memcpy(fhp, &fh, sizeof(fh));
4466 #ifndef _SYS_SYSPROTO_H_
4467 struct fhlink_args {
4473 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4476 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4479 #ifndef _SYS_SYSPROTO_H_
4480 struct fhlinkat_args {
4487 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4490 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4494 kern_fhlinkat(struct thread *td, int fd, const char *path,
4495 enum uio_seg pathseg, fhandle_t *fhp)
4502 error = priv_check(td, PRIV_VFS_GETFH);
4505 error = copyin(fhp, &fh, sizeof(fh));
4510 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4512 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4517 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4518 } while (error == EAGAIN || error == ERELOOKUP);
4522 #ifndef _SYS_SYSPROTO_H_
4523 struct fhreadlink_args {
4530 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4537 error = priv_check(td, PRIV_VFS_GETFH);
4540 if (uap->bufsize > IOSIZE_MAX)
4542 error = copyin(uap->fhp, &fh, sizeof(fh));
4545 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4547 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4551 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4557 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4560 * warning: do not remove the priv_check() call or this becomes one giant
4563 #ifndef _SYS_SYSPROTO_H_
4564 struct fhopen_args {
4565 const struct fhandle *u_fhp;
4570 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4572 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4576 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4585 error = priv_check(td, PRIV_VFS_FHOPEN);
4589 fmode = FFLAGS(flags);
4590 /* why not allow a non-read/write open for our lockd? */
4591 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4593 error = copyin(u_fhp, &fhp, sizeof(fhp));
4596 /* find the mount point */
4597 mp = vfs_busyfs(&fhp.fh_fsid);
4600 /* now give me my vnode, it gets returned to me locked */
4601 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4606 error = falloc_noinstall(td, &fp);
4612 * An extra reference on `fp' has been held for us by
4613 * falloc_noinstall().
4619 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4621 KASSERT(fp->f_ops == &badfileops,
4622 ("VOP_OPEN in fhopen() set f_ops"));
4623 KASSERT(td->td_dupfd < 0,
4624 ("fhopen() encountered fdopen()"));
4633 finit_vnode(fp, fmode, NULL, &vnops);
4635 if ((fmode & O_TRUNC) != 0) {
4636 error = fo_truncate(fp, 0, td->td_ucred, td);
4641 error = finstall(td, fp, &indx, fmode, NULL);
4644 td->td_retval[0] = indx;
4649 * Stat an (NFS) file handle.
4651 #ifndef _SYS_SYSPROTO_H_
4652 struct fhstat_args {
4653 struct fhandle *u_fhp;
4658 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4664 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4667 error = kern_fhstat(td, fh, &sb);
4669 error = copyout(&sb, uap->sb, sizeof(sb));
4674 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4680 error = priv_check(td, PRIV_VFS_FHSTAT);
4683 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4685 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4689 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4695 * Implement fstatfs() for (NFS) file handles.
4697 #ifndef _SYS_SYSPROTO_H_
4698 struct fhstatfs_args {
4699 struct fhandle *u_fhp;
4704 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4710 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4713 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4714 error = kern_fhstatfs(td, fh, sfp);
4716 error = copyout(sfp, uap->buf, sizeof(*sfp));
4717 free(sfp, M_STATFS);
4722 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4728 error = priv_check(td, PRIV_VFS_FHSTATFS);
4731 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4733 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4739 error = prison_canseemount(td->td_ucred, mp);
4743 error = mac_mount_check_stat(td->td_ucred, mp);
4747 error = VFS_STATFS(mp, buf);
4754 * Unlike madvise(2), we do not make a best effort to remember every
4755 * possible caching hint. Instead, we remember the last setting with
4756 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4757 * region of any current setting.
4760 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4763 struct fadvise_info *fa, *new;
4769 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4771 AUDIT_ARG_VALUE(advice);
4773 case POSIX_FADV_SEQUENTIAL:
4774 case POSIX_FADV_RANDOM:
4775 case POSIX_FADV_NOREUSE:
4776 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4778 case POSIX_FADV_NORMAL:
4779 case POSIX_FADV_WILLNEED:
4780 case POSIX_FADV_DONTNEED:
4786 /* XXX: CAP_POSIX_FADVISE? */
4788 error = fget(td, fd, &cap_no_rights, &fp);
4791 AUDIT_ARG_FILE(td->td_proc, fp);
4792 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4796 if (fp->f_type != DTYPE_VNODE) {
4801 if (vp->v_type != VREG) {
4808 end = offset + len - 1;
4810 case POSIX_FADV_SEQUENTIAL:
4811 case POSIX_FADV_RANDOM:
4812 case POSIX_FADV_NOREUSE:
4814 * Try to merge any existing non-standard region with
4815 * this new region if possible, otherwise create a new
4816 * non-standard region for this request.
4818 mtx_pool_lock(mtxpool_sleep, fp);
4820 if (fa != NULL && fa->fa_advice == advice &&
4821 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4822 (end != OFF_MAX && fa->fa_start == end + 1) ||
4823 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4824 if (offset < fa->fa_start)
4825 fa->fa_start = offset;
4826 if (end > fa->fa_end)
4829 new->fa_advice = advice;
4830 new->fa_start = offset;
4835 mtx_pool_unlock(mtxpool_sleep, fp);
4837 case POSIX_FADV_NORMAL:
4839 * If a the "normal" region overlaps with an existing
4840 * non-standard region, trim or remove the
4841 * non-standard region.
4843 mtx_pool_lock(mtxpool_sleep, fp);
4846 if (offset <= fa->fa_start && end >= fa->fa_end) {
4848 fp->f_advice = NULL;
4849 } else if (offset <= fa->fa_start &&
4850 end >= fa->fa_start)
4851 fa->fa_start = end + 1;
4852 else if (offset <= fa->fa_end && end >= fa->fa_end)
4853 fa->fa_end = offset - 1;
4854 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4856 * If the "normal" region is a middle
4857 * portion of the existing
4858 * non-standard region, just remove
4859 * the whole thing rather than picking
4860 * one side or the other to
4864 fp->f_advice = NULL;
4867 mtx_pool_unlock(mtxpool_sleep, fp);
4869 case POSIX_FADV_WILLNEED:
4870 case POSIX_FADV_DONTNEED:
4871 error = VOP_ADVISE(vp, offset, end, advice);
4877 free(new, M_FADVISE);
4882 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4886 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4888 return (kern_posix_error(td, error));
4892 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4893 off_t *outoffp, size_t len, unsigned int flags)
4895 struct file *infp, *outfp;
4896 struct vnode *invp, *outvp;
4899 void *rl_rcookie, *rl_wcookie;
4900 off_t savinoff, savoutoff;
4902 infp = outfp = NULL;
4903 rl_rcookie = rl_wcookie = NULL;
4912 if (len > SSIZE_MAX)
4914 * Although the len argument is size_t, the return argument
4915 * is ssize_t (which is signed). Therefore a size that won't
4916 * fit in ssize_t can't be returned.
4920 /* Get the file structures for the file descriptors. */
4921 error = fget_read(td, infd, &cap_read_rights, &infp);
4924 if (infp->f_ops == &badfileops) {
4928 if (infp->f_vnode == NULL) {
4932 error = fget_write(td, outfd, &cap_write_rights, &outfp);
4935 if (outfp->f_ops == &badfileops) {
4939 if (outfp->f_vnode == NULL) {
4944 /* Set the offset pointers to the correct place. */
4946 inoffp = &infp->f_offset;
4947 if (outoffp == NULL)
4948 outoffp = &outfp->f_offset;
4950 savoutoff = *outoffp;
4952 invp = infp->f_vnode;
4953 outvp = outfp->f_vnode;
4954 /* Sanity check the f_flag bits. */
4955 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4956 (infp->f_flag & FREAD) == 0) {
4961 /* If len == 0, just return 0. */
4966 * If infp and outfp refer to the same file, the byte ranges cannot
4969 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4970 savoutoff) || (savinoff > savoutoff && savoutoff + len >
4976 /* Range lock the byte ranges for both invp and outvp. */
4978 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4980 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4982 if (rl_rcookie != NULL)
4984 vn_rangelock_unlock(outvp, rl_wcookie);
4985 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4986 vn_rangelock_unlock(invp, rl_rcookie);
4990 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4991 flags, infp->f_cred, outfp->f_cred, td);
4993 if (rl_rcookie != NULL)
4994 vn_rangelock_unlock(invp, rl_rcookie);
4995 if (rl_wcookie != NULL)
4996 vn_rangelock_unlock(outvp, rl_wcookie);
4997 if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4999 *outoffp = savoutoff;
5005 td->td_retval[0] = retlen;
5010 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
5012 off_t inoff, outoff, *inoffp, *outoffp;
5015 inoffp = outoffp = NULL;
5016 if (uap->inoffp != NULL) {
5017 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
5022 if (uap->outoffp != NULL) {
5023 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
5028 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5029 outoffp, uap->len, uap->flags);
5030 if (error == 0 && uap->inoffp != NULL)
5031 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5032 if (error == 0 && uap->outoffp != NULL)
5033 error = copyout(outoffp, uap->outoffp, sizeof(off_t));