2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
49 #include <sys/capsicum.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
72 #include <sys/dirent.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
77 #include <sys/ktrace.h>
80 #include <machine/stdarg.h>
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
90 #include <fs/devfs/devfs.h>
92 #include <ufs/ufs/quota.h>
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97 enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101 struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103 const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107 enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
109 size_t count, struct thread *td);
110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
111 const char *path, enum uio_seg segflag);
114 at2cnpflags(u_int at_flags, u_int mask)
118 MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
119 (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
123 if ((at_flags & AT_BENEATH) != 0)
125 if ((at_flags & AT_RESOLVE_BENEATH) != 0)
127 if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
129 /* NOFOLLOW is pseudo flag */
130 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
131 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
138 kern_sync(struct thread *td)
140 struct mount *mp, *nmp;
143 mtx_lock(&mountlist_mtx);
144 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
145 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
146 nmp = TAILQ_NEXT(mp, mnt_list);
149 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
150 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
151 save = curthread_pflags_set(TDP_SYNCIO);
152 vfs_periodic(mp, MNT_NOWAIT);
153 VFS_SYNC(mp, MNT_NOWAIT);
154 curthread_pflags_restore(save);
155 vn_finished_write(mp);
157 mtx_lock(&mountlist_mtx);
158 nmp = TAILQ_NEXT(mp, mnt_list);
161 mtx_unlock(&mountlist_mtx);
166 * Sync each mounted filesystem.
168 #ifndef _SYS_SYSPROTO_H_
175 sys_sync(struct thread *td, struct sync_args *uap)
178 return (kern_sync(td));
182 * Change filesystem quotas.
184 #ifndef _SYS_SYSPROTO_H_
185 struct quotactl_args {
193 sys_quotactl(struct thread *td, struct quotactl_args *uap)
199 AUDIT_ARG_CMD(uap->cmd);
200 AUDIT_ARG_UID(uap->uid);
201 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
203 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
205 if ((error = namei(&nd)) != 0)
207 NDFREE(&nd, NDF_ONLY_PNBUF);
208 mp = nd.ni_vp->v_mount;
211 error = vfs_busy(mp, 0);
216 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
219 * Since quota on operation typically needs to open quota
220 * file, the Q_QUOTAON handler needs to unbusy the mount point
221 * before calling into namei. Otherwise, unmount might be
222 * started between two vfs_busy() invocations (first is our,
223 * second is from mount point cross-walk code in lookup()),
226 * Require that Q_QUOTAON handles the vfs_busy() reference on
227 * its own, always returning with ubusied mount point.
229 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
230 (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
237 * Used by statfs conversion routines to scale the block size up if
238 * necessary so that all of the block counts are <= 'max_size'. Note
239 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
243 statfs_scale_blocks(struct statfs *sf, long max_size)
248 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
251 * Attempt to scale the block counts to give a more accurate
252 * overview to userland of the ratio of free space to used
253 * space. To do this, find the largest block count and compute
254 * a divisor that lets it fit into a signed integer <= max_size.
256 if (sf->f_bavail < 0)
257 count = -sf->f_bavail;
259 count = sf->f_bavail;
260 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
261 if (count <= max_size)
264 count >>= flsl(max_size);
271 sf->f_bsize <<= shift;
272 sf->f_blocks >>= shift;
273 sf->f_bfree >>= shift;
274 sf->f_bavail >>= shift;
278 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
284 error = vfs_busy(mp, 0);
289 error = mac_mount_check_stat(td->td_ucred, mp);
293 error = VFS_STATFS(mp, buf);
296 if (priv_check_cred_vfs_generation(td->td_ucred)) {
297 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
298 prison_enforce_statfs(td->td_ucred, mp, buf);
306 * Get filesystem statistics.
308 #ifndef _SYS_SYSPROTO_H_
315 sys_statfs(struct thread *td, struct statfs_args *uap)
320 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
321 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
323 error = copyout(sfp, uap->buf, sizeof(struct statfs));
329 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
336 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
341 mp = nd.ni_vp->v_mount;
343 NDFREE(&nd, NDF_ONLY_PNBUF);
345 return (kern_do_statfs(td, mp, buf));
349 * Get filesystem statistics.
351 #ifndef _SYS_SYSPROTO_H_
352 struct fstatfs_args {
358 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
363 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
364 error = kern_fstatfs(td, uap->fd, sfp);
366 error = copyout(sfp, uap->buf, sizeof(struct statfs));
372 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
380 error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
384 vn_lock(vp, LK_SHARED | LK_RETRY);
386 AUDIT_ARG_VNODE1(vp);
393 return (kern_do_statfs(td, mp, buf));
397 * Get statistics on all filesystems.
399 #ifndef _SYS_SYSPROTO_H_
400 struct getfsstat_args {
407 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
412 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
414 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
415 UIO_USERSPACE, uap->mode);
417 td->td_retval[0] = count;
422 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
423 * The caller is responsible for freeing memory which will be allocated
427 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
428 size_t *countp, enum uio_seg bufseg, int mode)
430 struct mount *mp, *nmp;
431 struct statfs *sfsp, *sp, *sptmp, *tofree;
432 size_t count, maxcount;
440 if (bufseg == UIO_SYSSPACE)
445 maxcount = bufsize / sizeof(struct statfs);
449 } else if (bufseg == UIO_USERSPACE) {
452 } else /* if (bufseg == UIO_SYSSPACE) */ {
454 mtx_lock(&mountlist_mtx);
455 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
458 mtx_unlock(&mountlist_mtx);
459 if (maxcount > count)
461 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
468 * If there is no target buffer they only want the count.
470 * This could be TAILQ_FOREACH but it is open-coded to match the original
474 mtx_lock(&mountlist_mtx);
475 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
476 if (prison_canseemount(td->td_ucred, mp) != 0) {
477 nmp = TAILQ_NEXT(mp, mnt_list);
481 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
482 nmp = TAILQ_NEXT(mp, mnt_list);
487 nmp = TAILQ_NEXT(mp, mnt_list);
489 mtx_unlock(&mountlist_mtx);
495 * They want the entire thing.
497 * Short-circuit the corner case of no room for anything, avoids
504 mtx_lock(&mountlist_mtx);
505 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
506 if (prison_canseemount(td->td_ucred, mp) != 0) {
507 nmp = TAILQ_NEXT(mp, mnt_list);
511 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
512 nmp = TAILQ_NEXT(mp, mnt_list);
516 if (mode == MNT_WAIT) {
517 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
519 * If vfs_busy() failed, and MBF_NOWAIT
520 * wasn't passed, then the mp is gone.
521 * Furthermore, because of MBF_MNTLSTLOCK,
522 * the mountlist_mtx was dropped. We have
523 * no other choice than to start over.
525 mtx_unlock(&mountlist_mtx);
526 free(tofree, M_STATFS);
530 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
531 nmp = TAILQ_NEXT(mp, mnt_list);
537 * If MNT_NOWAIT is specified, do not refresh
540 if (mode != MNT_NOWAIT) {
541 error = VFS_STATFS(mp, sp);
543 mtx_lock(&mountlist_mtx);
544 nmp = TAILQ_NEXT(mp, mnt_list);
549 if (priv_check_cred_vfs_generation(td->td_ucred)) {
550 sptmp = malloc(sizeof(struct statfs), M_STATFS,
553 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
554 prison_enforce_statfs(td->td_ucred, mp, sptmp);
558 if (bufseg == UIO_SYSSPACE) {
559 bcopy(sp, sfsp, sizeof(*sp));
560 free(sptmp, M_STATFS);
561 } else /* if (bufseg == UIO_USERSPACE) */ {
562 error = copyout(sp, sfsp, sizeof(*sp));
563 free(sptmp, M_STATFS);
572 if (count == maxcount) {
577 mtx_lock(&mountlist_mtx);
578 nmp = TAILQ_NEXT(mp, mnt_list);
581 mtx_unlock(&mountlist_mtx);
587 #ifdef COMPAT_FREEBSD4
589 * Get old format filesystem statistics.
591 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
593 #ifndef _SYS_SYSPROTO_H_
594 struct freebsd4_statfs_args {
600 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
606 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
607 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
609 freebsd4_cvtstatfs(sfp, &osb);
610 error = copyout(&osb, uap->buf, sizeof(osb));
617 * Get filesystem statistics.
619 #ifndef _SYS_SYSPROTO_H_
620 struct freebsd4_fstatfs_args {
626 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
632 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
633 error = kern_fstatfs(td, uap->fd, sfp);
635 freebsd4_cvtstatfs(sfp, &osb);
636 error = copyout(&osb, uap->buf, sizeof(osb));
643 * Get statistics on all filesystems.
645 #ifndef _SYS_SYSPROTO_H_
646 struct freebsd4_getfsstat_args {
653 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
655 struct statfs *buf, *sp;
660 if (uap->bufsize < 0)
662 count = uap->bufsize / sizeof(struct ostatfs);
663 if (count > SIZE_MAX / sizeof(struct statfs))
665 size = count * sizeof(struct statfs);
666 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
669 td->td_retval[0] = count;
672 while (count != 0 && error == 0) {
673 freebsd4_cvtstatfs(sp, &osb);
674 error = copyout(&osb, uap->buf, sizeof(osb));
685 * Implement fstatfs() for (NFS) file handles.
687 #ifndef _SYS_SYSPROTO_H_
688 struct freebsd4_fhstatfs_args {
689 struct fhandle *u_fhp;
694 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
701 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
704 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
705 error = kern_fhstatfs(td, fh, sfp);
707 freebsd4_cvtstatfs(sfp, &osb);
708 error = copyout(&osb, uap->buf, sizeof(osb));
715 * Convert a new format statfs structure to an old format statfs structure.
718 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
721 statfs_scale_blocks(nsp, LONG_MAX);
722 bzero(osp, sizeof(*osp));
723 osp->f_bsize = nsp->f_bsize;
724 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
725 osp->f_blocks = nsp->f_blocks;
726 osp->f_bfree = nsp->f_bfree;
727 osp->f_bavail = nsp->f_bavail;
728 osp->f_files = MIN(nsp->f_files, LONG_MAX);
729 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
730 osp->f_owner = nsp->f_owner;
731 osp->f_type = nsp->f_type;
732 osp->f_flags = nsp->f_flags;
733 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
734 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
735 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
736 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
737 strlcpy(osp->f_fstypename, nsp->f_fstypename,
738 MIN(MFSNAMELEN, OMFSNAMELEN));
739 strlcpy(osp->f_mntonname, nsp->f_mntonname,
740 MIN(MNAMELEN, OMNAMELEN));
741 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
742 MIN(MNAMELEN, OMNAMELEN));
743 osp->f_fsid = nsp->f_fsid;
745 #endif /* COMPAT_FREEBSD4 */
747 #if defined(COMPAT_FREEBSD11)
749 * Get old format filesystem statistics.
751 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
754 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
756 struct freebsd11_statfs osb;
760 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
761 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
763 freebsd11_cvtstatfs(sfp, &osb);
764 error = copyout(&osb, uap->buf, sizeof(osb));
771 * Get filesystem statistics.
774 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
776 struct freebsd11_statfs osb;
780 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
781 error = kern_fstatfs(td, uap->fd, sfp);
783 freebsd11_cvtstatfs(sfp, &osb);
784 error = copyout(&osb, uap->buf, sizeof(osb));
791 * Get statistics on all filesystems.
794 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
796 struct freebsd11_statfs osb;
797 struct statfs *buf, *sp;
801 count = uap->bufsize / sizeof(struct ostatfs);
802 size = count * sizeof(struct statfs);
803 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
806 td->td_retval[0] = count;
809 while (count > 0 && error == 0) {
810 freebsd11_cvtstatfs(sp, &osb);
811 error = copyout(&osb, uap->buf, sizeof(osb));
822 * Implement fstatfs() for (NFS) file handles.
825 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
827 struct freebsd11_statfs osb;
832 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
835 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
836 error = kern_fhstatfs(td, fh, sfp);
838 freebsd11_cvtstatfs(sfp, &osb);
839 error = copyout(&osb, uap->buf, sizeof(osb));
846 * Convert a new format statfs structure to an old format statfs structure.
849 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
852 bzero(osp, sizeof(*osp));
853 osp->f_version = FREEBSD11_STATFS_VERSION;
854 osp->f_type = nsp->f_type;
855 osp->f_flags = nsp->f_flags;
856 osp->f_bsize = nsp->f_bsize;
857 osp->f_iosize = nsp->f_iosize;
858 osp->f_blocks = nsp->f_blocks;
859 osp->f_bfree = nsp->f_bfree;
860 osp->f_bavail = nsp->f_bavail;
861 osp->f_files = nsp->f_files;
862 osp->f_ffree = nsp->f_ffree;
863 osp->f_syncwrites = nsp->f_syncwrites;
864 osp->f_asyncwrites = nsp->f_asyncwrites;
865 osp->f_syncreads = nsp->f_syncreads;
866 osp->f_asyncreads = nsp->f_asyncreads;
867 osp->f_namemax = nsp->f_namemax;
868 osp->f_owner = nsp->f_owner;
869 osp->f_fsid = nsp->f_fsid;
870 strlcpy(osp->f_fstypename, nsp->f_fstypename,
871 MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
872 strlcpy(osp->f_mntonname, nsp->f_mntonname,
873 MIN(MNAMELEN, sizeof(osp->f_mntonname)));
874 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
875 MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
877 #endif /* COMPAT_FREEBSD11 */
880 * Change current working directory to a given file descriptor.
882 #ifndef _SYS_SYSPROTO_H_
888 sys_fchdir(struct thread *td, struct fchdir_args *uap)
890 struct vnode *vp, *tdp;
895 AUDIT_ARG_FD(uap->fd);
896 error = getvnode(td, uap->fd, &cap_fchdir_rights,
903 vn_lock(vp, LK_SHARED | LK_RETRY);
904 AUDIT_ARG_VNODE1(vp);
905 error = change_dir(vp, td);
906 while (!error && (mp = vp->v_mountedhere) != NULL) {
909 error = VFS_ROOT(mp, LK_SHARED, &tdp);
926 * Change current working directory (``.'').
928 #ifndef _SYS_SYSPROTO_H_
934 sys_chdir(struct thread *td, struct chdir_args *uap)
937 return (kern_chdir(td, uap->path, UIO_USERSPACE));
941 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
946 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
948 if ((error = namei(&nd)) != 0)
950 if ((error = change_dir(nd.ni_vp, td)) != 0) {
952 NDFREE(&nd, NDF_ONLY_PNBUF);
955 VOP_UNLOCK(nd.ni_vp);
956 NDFREE(&nd, NDF_ONLY_PNBUF);
957 pwd_chdir(td, nd.ni_vp);
962 * Change notion of root (``/'') directory.
964 #ifndef _SYS_SYSPROTO_H_
970 sys_chroot(struct thread *td, struct chroot_args *uap)
975 error = priv_check(td, PRIV_VFS_CHROOT);
978 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
979 UIO_USERSPACE, uap->path, td);
983 error = change_dir(nd.ni_vp, td);
987 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
991 VOP_UNLOCK(nd.ni_vp);
992 error = pwd_chroot(td, nd.ni_vp);
994 NDFREE(&nd, NDF_ONLY_PNBUF);
999 NDFREE(&nd, NDF_ONLY_PNBUF);
1004 * Common routine for chroot and chdir. Callers must provide a locked vnode
1008 change_dir(struct vnode *vp, struct thread *td)
1014 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1015 if (vp->v_type != VDIR)
1018 error = mac_vnode_check_chdir(td->td_ucred, vp);
1022 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1025 static __inline void
1026 flags_to_rights(int flags, cap_rights_t *rightsp)
1029 if (flags & O_EXEC) {
1030 cap_rights_set_one(rightsp, CAP_FEXECVE);
1032 switch ((flags & O_ACCMODE)) {
1034 cap_rights_set_one(rightsp, CAP_READ);
1037 cap_rights_set_one(rightsp, CAP_READ);
1040 cap_rights_set_one(rightsp, CAP_WRITE);
1041 if (!(flags & (O_APPEND | O_TRUNC)))
1042 cap_rights_set_one(rightsp, CAP_SEEK);
1047 if (flags & O_CREAT)
1048 cap_rights_set_one(rightsp, CAP_CREATE);
1050 if (flags & O_TRUNC)
1051 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1053 if (flags & (O_SYNC | O_FSYNC))
1054 cap_rights_set_one(rightsp, CAP_FSYNC);
1056 if (flags & (O_EXLOCK | O_SHLOCK))
1057 cap_rights_set_one(rightsp, CAP_FLOCK);
1061 * Check permissions, allocate an open file structure, and call the device
1062 * open routine if any.
1064 #ifndef _SYS_SYSPROTO_H_
1072 sys_open(struct thread *td, struct open_args *uap)
1075 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1076 uap->flags, uap->mode));
1079 #ifndef _SYS_SYSPROTO_H_
1080 struct openat_args {
1088 sys_openat(struct thread *td, struct openat_args *uap)
1091 AUDIT_ARG_FD(uap->fd);
1092 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1097 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1098 int flags, int mode)
1100 struct proc *p = td->td_proc;
1101 struct filedesc *fdp = p->p_fd;
1104 struct nameidata nd;
1105 cap_rights_t rights;
1106 int cmode, error, indx;
1110 AUDIT_ARG_FFLAGS(flags);
1111 AUDIT_ARG_MODE(mode);
1112 cap_rights_init_one(&rights, CAP_LOOKUP);
1113 flags_to_rights(flags, &rights);
1115 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1118 if (flags & O_EXEC) {
1119 if (flags & O_ACCMODE)
1121 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1124 flags = FFLAGS(flags);
1128 * Allocate a file structure. The descriptor to reference it
1129 * is allocated and set by finstall() below.
1131 error = falloc_noinstall(td, &fp);
1135 * An extra reference on `fp' has been held for us by
1136 * falloc_noinstall().
1138 /* Set the flags early so the finit in devfs can pick them up. */
1139 fp->f_flag = flags & FMASK;
1140 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1141 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1143 td->td_dupfd = -1; /* XXX check for fdopen */
1144 error = vn_open(&nd, &flags, cmode, fp);
1147 * If the vn_open replaced the method vector, something
1148 * wonderous happened deep below and we just pass it up
1149 * pretending we know what we do.
1151 if (error == ENXIO && fp->f_ops != &badfileops)
1155 * Handle special fdopen() case. bleh.
1157 * Don't do this for relative (capability) lookups; we don't
1158 * understand exactly what would happen, and we don't think
1159 * that it ever should.
1161 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1162 (error == ENODEV || error == ENXIO) &&
1163 td->td_dupfd >= 0) {
1164 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1173 NDFREE(&nd, NDF_ONLY_PNBUF);
1177 * Store the vnode, for any f_type. Typically, the vnode use
1178 * count is decremented by direct call to vn_closefile() for
1179 * files that switched type in the cdevsw fdopen() method.
1183 * If the file wasn't claimed by devfs bind it to the normal
1184 * vnode operations here.
1186 if (fp->f_ops == &badfileops) {
1187 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1188 finit_vnode(fp, flags, NULL, &vnops);
1192 if (flags & O_TRUNC) {
1193 error = fo_truncate(fp, 0, td->td_ucred, td);
1199 * If we haven't already installed the FD (for dupfdopen), do so now.
1202 struct filecaps *fcaps;
1205 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1206 fcaps = &nd.ni_filecaps;
1210 error = finstall(td, fp, &indx, flags, fcaps);
1211 /* On success finstall() consumes fcaps. */
1213 filecaps_free(&nd.ni_filecaps);
1217 filecaps_free(&nd.ni_filecaps);
1221 * Release our private reference, leaving the one associated with
1222 * the descriptor table intact.
1225 td->td_retval[0] = indx;
1228 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1237 #ifndef _SYS_SYSPROTO_H_
1238 struct ocreat_args {
1244 ocreat(struct thread *td, struct ocreat_args *uap)
1247 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1248 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1250 #endif /* COMPAT_43 */
1253 * Create a special file.
1255 #ifndef _SYS_SYSPROTO_H_
1256 struct mknodat_args {
1264 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1267 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1271 #if defined(COMPAT_FREEBSD11)
1273 freebsd11_mknod(struct thread *td,
1274 struct freebsd11_mknod_args *uap)
1277 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1278 uap->mode, uap->dev));
1282 freebsd11_mknodat(struct thread *td,
1283 struct freebsd11_mknodat_args *uap)
1286 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1289 #endif /* COMPAT_FREEBSD11 */
1292 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1293 int mode, dev_t dev)
1298 struct nameidata nd;
1299 int error, whiteout = 0;
1301 AUDIT_ARG_MODE(mode);
1303 switch (mode & S_IFMT) {
1306 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1307 if (error == 0 && dev == VNOVAL)
1311 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1315 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1325 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1326 NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1328 if ((error = namei(&nd)) != 0)
1332 NDFREE(&nd, NDF_ONLY_PNBUF);
1333 if (vp == nd.ni_dvp)
1341 vattr.va_mode = (mode & ALLPERMS) &
1342 ~td->td_proc->p_fd->fd_cmask;
1343 vattr.va_rdev = dev;
1346 switch (mode & S_IFMT) {
1348 vattr.va_type = VCHR;
1351 vattr.va_type = VBLK;
1357 panic("kern_mknod: invalid mode");
1360 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1361 NDFREE(&nd, NDF_ONLY_PNBUF);
1363 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1368 if (error == 0 && !whiteout)
1369 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1370 &nd.ni_cnd, &vattr);
1374 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1376 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1377 &nd.ni_cnd, &vattr);
1382 NDFREE(&nd, NDF_ONLY_PNBUF);
1384 vn_finished_write(mp);
1385 if (error == ERELOOKUP)
1391 * Create a named pipe.
1393 #ifndef _SYS_SYSPROTO_H_
1394 struct mkfifo_args {
1400 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1403 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1407 #ifndef _SYS_SYSPROTO_H_
1408 struct mkfifoat_args {
1415 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1418 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1423 kern_mkfifoat(struct thread *td, int fd, const char *path,
1424 enum uio_seg pathseg, int mode)
1428 struct nameidata nd;
1431 AUDIT_ARG_MODE(mode);
1434 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1435 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1437 if ((error = namei(&nd)) != 0)
1439 if (nd.ni_vp != NULL) {
1440 NDFREE(&nd, NDF_ONLY_PNBUF);
1441 if (nd.ni_vp == nd.ni_dvp)
1448 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1449 NDFREE(&nd, NDF_ONLY_PNBUF);
1451 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1456 vattr.va_type = VFIFO;
1457 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1459 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1464 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1471 vn_finished_write(mp);
1472 NDFREE(&nd, NDF_ONLY_PNBUF);
1473 if (error == ERELOOKUP)
1479 * Make a hard file link.
1481 #ifndef _SYS_SYSPROTO_H_
1488 sys_link(struct thread *td, struct link_args *uap)
1491 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1492 UIO_USERSPACE, FOLLOW));
1495 #ifndef _SYS_SYSPROTO_H_
1496 struct linkat_args {
1505 sys_linkat(struct thread *td, struct linkat_args *uap)
1510 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH |
1511 AT_RESOLVE_BENEATH)) != 0)
1514 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1515 UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW | AT_BENEATH |
1516 AT_RESOLVE_BENEATH)));
1519 int hardlink_check_uid = 0;
1520 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1521 &hardlink_check_uid, 0,
1522 "Unprivileged processes cannot create hard links to files owned by other "
1524 static int hardlink_check_gid = 0;
1525 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1526 &hardlink_check_gid, 0,
1527 "Unprivileged processes cannot create hard links to files owned by other "
1531 can_hardlink(struct vnode *vp, struct ucred *cred)
1536 if (!hardlink_check_uid && !hardlink_check_gid)
1539 error = VOP_GETATTR(vp, &va, cred);
1543 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1544 error = priv_check_cred(cred, PRIV_VFS_LINK);
1549 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1550 error = priv_check_cred(cred, PRIV_VFS_LINK);
1559 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1560 const char *path2, enum uio_seg segflag, int follow)
1562 struct nameidata nd;
1567 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1568 path1, fd1, &cap_linkat_source_rights, td);
1569 if ((error = namei(&nd)) != 0)
1571 NDFREE(&nd, NDF_ONLY_PNBUF);
1572 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1573 } while (error == EAGAIN || error == ERELOOKUP);
1578 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1579 enum uio_seg segflag)
1581 struct nameidata nd;
1585 if (vp->v_type == VDIR) {
1587 return (EPERM); /* POSIX */
1589 NDINIT_ATRIGHTS(&nd, CREATE,
1590 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1591 &cap_linkat_target_rights, td);
1592 if ((error = namei(&nd)) == 0) {
1593 if (nd.ni_vp != NULL) {
1594 NDFREE(&nd, NDF_ONLY_PNBUF);
1595 if (nd.ni_dvp == nd.ni_vp)
1602 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1604 * Cross-device link. No need to recheck
1605 * vp->v_type, since it cannot change, except
1608 NDFREE(&nd, NDF_ONLY_PNBUF);
1612 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1613 error = can_hardlink(vp, td->td_ucred);
1616 error = mac_vnode_check_link(td->td_ucred,
1617 nd.ni_dvp, vp, &nd.ni_cnd);
1622 NDFREE(&nd, NDF_ONLY_PNBUF);
1625 error = vn_start_write(vp, &mp, V_NOWAIT);
1629 NDFREE(&nd, NDF_ONLY_PNBUF);
1630 error = vn_start_write(NULL, &mp,
1636 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1639 vn_finished_write(mp);
1640 NDFREE(&nd, NDF_ONLY_PNBUF);
1643 NDFREE(&nd, NDF_ONLY_PNBUF);
1653 * Make a symbolic link.
1655 #ifndef _SYS_SYSPROTO_H_
1656 struct symlink_args {
1662 sys_symlink(struct thread *td, struct symlink_args *uap)
1665 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1669 #ifndef _SYS_SYSPROTO_H_
1670 struct symlinkat_args {
1677 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1680 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1685 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1686 enum uio_seg segflg)
1690 const char *syspath;
1692 struct nameidata nd;
1695 if (segflg == UIO_SYSSPACE) {
1698 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1699 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1703 AUDIT_ARG_TEXT(syspath);
1706 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1707 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1709 if ((error = namei(&nd)) != 0)
1712 NDFREE(&nd, NDF_ONLY_PNBUF);
1713 if (nd.ni_vp == nd.ni_dvp)
1721 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1722 NDFREE(&nd, NDF_ONLY_PNBUF);
1724 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1729 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1731 vattr.va_type = VLNK;
1732 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1737 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1743 NDFREE(&nd, NDF_ONLY_PNBUF);
1745 vn_finished_write(mp);
1746 if (error == ERELOOKUP)
1749 if (segflg != UIO_SYSSPACE)
1750 uma_zfree(namei_zone, tmppath);
1755 * Delete a whiteout from the filesystem.
1757 #ifndef _SYS_SYSPROTO_H_
1758 struct undelete_args {
1763 sys_undelete(struct thread *td, struct undelete_args *uap)
1766 struct nameidata nd;
1771 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1772 UIO_USERSPACE, uap->path, td);
1777 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1778 NDFREE(&nd, NDF_ONLY_PNBUF);
1779 if (nd.ni_vp == nd.ni_dvp)
1787 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1788 NDFREE(&nd, NDF_ONLY_PNBUF);
1790 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1794 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1795 NDFREE(&nd, NDF_ONLY_PNBUF);
1797 vn_finished_write(mp);
1798 if (error == ERELOOKUP)
1804 * Delete a name from the filesystem.
1806 #ifndef _SYS_SYSPROTO_H_
1807 struct unlink_args {
1812 sys_unlink(struct thread *td, struct unlink_args *uap)
1815 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1820 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1821 int flag, enum uio_seg pathseg, ino_t oldinum)
1824 if ((flag & ~AT_REMOVEDIR) != 0)
1827 if ((flag & AT_REMOVEDIR) != 0)
1828 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1830 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1833 #ifndef _SYS_SYSPROTO_H_
1834 struct unlinkat_args {
1841 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1844 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1848 #ifndef _SYS_SYSPROTO_H_
1849 struct funlinkat_args {
1857 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1860 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1865 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1866 enum uio_seg pathseg, int flag, ino_t oldinum)
1871 struct nameidata nd;
1876 if (fd != FD_NONE) {
1877 error = getvnode(td, fd, &cap_no_rights, &fp);
1884 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1885 at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
1886 pathseg, path, dfd, &cap_unlinkat_rights, td);
1887 if ((error = namei(&nd)) != 0) {
1888 if (error == EINVAL)
1893 if (vp->v_type == VDIR && oldinum == 0) {
1894 error = EPERM; /* POSIX */
1895 } else if (oldinum != 0 &&
1896 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1897 sb.st_ino != oldinum) {
1898 error = EIDRM; /* Identifier removed */
1899 } else if (fp != NULL && fp->f_vnode != vp) {
1900 if (VN_IS_DOOMED(fp->f_vnode))
1906 * The root of a mounted filesystem cannot be deleted.
1908 * XXX: can this only be a VDIR case?
1910 if (vp->v_vflag & VV_ROOT)
1914 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1915 NDFREE(&nd, NDF_ONLY_PNBUF);
1917 if (vp == nd.ni_dvp)
1921 if ((error = vn_start_write(NULL, &mp,
1922 V_XSLEEP | PCATCH)) != 0) {
1928 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1933 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1934 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1938 vn_finished_write(mp);
1940 NDFREE(&nd, NDF_ONLY_PNBUF);
1942 if (vp == nd.ni_dvp)
1946 if (error == ERELOOKUP)
1955 * Reposition read/write file offset.
1957 #ifndef _SYS_SYSPROTO_H_
1966 sys_lseek(struct thread *td, struct lseek_args *uap)
1969 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1973 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1979 error = fget(td, fd, &cap_seek_rights, &fp);
1982 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1983 fo_seek(fp, offset, whence, td) : ESPIPE;
1988 #if defined(COMPAT_43)
1990 * Reposition read/write file offset.
1992 #ifndef _SYS_SYSPROTO_H_
1993 struct olseek_args {
2000 olseek(struct thread *td, struct olseek_args *uap)
2003 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2005 #endif /* COMPAT_43 */
2007 #if defined(COMPAT_FREEBSD6)
2008 /* Version with the 'pad' argument */
2010 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2013 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2018 * Check access permissions using passed credentials.
2021 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2027 /* Flags == 0 means only check for existence. */
2028 if (user_flags == 0)
2032 if (user_flags & R_OK)
2034 if (user_flags & W_OK)
2036 if (user_flags & X_OK)
2039 error = mac_vnode_check_access(cred, vp, accmode);
2043 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2044 error = VOP_ACCESS(vp, accmode, cred, td);
2049 * Check access permissions using "real" credentials.
2051 #ifndef _SYS_SYSPROTO_H_
2052 struct access_args {
2058 sys_access(struct thread *td, struct access_args *uap)
2061 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2065 #ifndef _SYS_SYSPROTO_H_
2066 struct faccessat_args {
2074 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2077 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2082 kern_accessat(struct thread *td, int fd, const char *path,
2083 enum uio_seg pathseg, int flag, int amode)
2085 struct ucred *cred, *usecred;
2087 struct nameidata nd;
2090 if ((flag & ~(AT_EACCESS | AT_BENEATH | AT_RESOLVE_BENEATH)) != 0)
2092 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2096 * Create and modify a temporary credential instead of one that
2097 * is potentially shared (if we need one).
2099 cred = td->td_ucred;
2100 if ((flag & AT_EACCESS) == 0 &&
2101 ((cred->cr_uid != cred->cr_ruid ||
2102 cred->cr_rgid != cred->cr_groups[0]))) {
2103 usecred = crdup(cred);
2104 usecred->cr_uid = cred->cr_ruid;
2105 usecred->cr_groups[0] = cred->cr_rgid;
2106 td->td_ucred = usecred;
2109 AUDIT_ARG_VALUE(amode);
2110 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2111 AUDITVNODE1 | at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
2112 pathseg, path, fd, &cap_fstat_rights, td);
2113 if ((error = namei(&nd)) != 0)
2117 error = vn_access(vp, amode, usecred, td);
2118 NDFREE(&nd, NDF_ONLY_PNBUF);
2121 if (usecred != cred) {
2122 td->td_ucred = cred;
2129 * Check access permissions using "effective" credentials.
2131 #ifndef _SYS_SYSPROTO_H_
2132 struct eaccess_args {
2138 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2141 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2142 AT_EACCESS, uap->amode));
2145 #if defined(COMPAT_43)
2147 * Get file status; this version follows links.
2149 #ifndef _SYS_SYSPROTO_H_
2156 ostat(struct thread *td, struct ostat_args *uap)
2162 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2167 return (copyout(&osb, uap->ub, sizeof (osb)));
2171 * Get file status; this version does not follow links.
2173 #ifndef _SYS_SYSPROTO_H_
2174 struct olstat_args {
2180 olstat(struct thread *td, struct olstat_args *uap)
2186 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2187 UIO_USERSPACE, &sb, NULL);
2191 return (copyout(&osb, uap->ub, sizeof (osb)));
2195 * Convert from an old to a new stat structure.
2196 * XXX: many values are blindly truncated.
2199 cvtstat(struct stat *st, struct ostat *ost)
2202 bzero(ost, sizeof(*ost));
2203 ost->st_dev = st->st_dev;
2204 ost->st_ino = st->st_ino;
2205 ost->st_mode = st->st_mode;
2206 ost->st_nlink = st->st_nlink;
2207 ost->st_uid = st->st_uid;
2208 ost->st_gid = st->st_gid;
2209 ost->st_rdev = st->st_rdev;
2210 ost->st_size = MIN(st->st_size, INT32_MAX);
2211 ost->st_atim = st->st_atim;
2212 ost->st_mtim = st->st_mtim;
2213 ost->st_ctim = st->st_ctim;
2214 ost->st_blksize = st->st_blksize;
2215 ost->st_blocks = st->st_blocks;
2216 ost->st_flags = st->st_flags;
2217 ost->st_gen = st->st_gen;
2219 #endif /* COMPAT_43 */
2221 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2222 int ino64_trunc_error;
2223 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2224 &ino64_trunc_error, 0,
2225 "Error on truncation of device, file or inode number, or link count");
2228 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2231 ost->st_dev = st->st_dev;
2232 if (ost->st_dev != st->st_dev) {
2233 switch (ino64_trunc_error) {
2236 * Since dev_t is almost raw, don't clamp to the
2237 * maximum for case 2, but ignore the error.
2244 ost->st_ino = st->st_ino;
2245 if (ost->st_ino != st->st_ino) {
2246 switch (ino64_trunc_error) {
2253 ost->st_ino = UINT32_MAX;
2257 ost->st_mode = st->st_mode;
2258 ost->st_nlink = st->st_nlink;
2259 if (ost->st_nlink != st->st_nlink) {
2260 switch (ino64_trunc_error) {
2267 ost->st_nlink = UINT16_MAX;
2271 ost->st_uid = st->st_uid;
2272 ost->st_gid = st->st_gid;
2273 ost->st_rdev = st->st_rdev;
2274 if (ost->st_rdev != st->st_rdev) {
2275 switch (ino64_trunc_error) {
2282 ost->st_atim = st->st_atim;
2283 ost->st_mtim = st->st_mtim;
2284 ost->st_ctim = st->st_ctim;
2285 ost->st_size = st->st_size;
2286 ost->st_blocks = st->st_blocks;
2287 ost->st_blksize = st->st_blksize;
2288 ost->st_flags = st->st_flags;
2289 ost->st_gen = st->st_gen;
2291 ost->st_birthtim = st->st_birthtim;
2292 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2293 sizeof(*ost) - offsetof(struct freebsd11_stat,
2294 st_birthtim) - sizeof(ost->st_birthtim));
2299 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2302 struct freebsd11_stat osb;
2305 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2309 error = freebsd11_cvtstat(&sb, &osb);
2311 error = copyout(&osb, uap->ub, sizeof(osb));
2316 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2319 struct freebsd11_stat osb;
2322 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2323 UIO_USERSPACE, &sb, NULL);
2326 error = freebsd11_cvtstat(&sb, &osb);
2328 error = copyout(&osb, uap->ub, sizeof(osb));
2333 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2337 struct freebsd11_stat osb;
2340 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2343 error = kern_fhstat(td, fh, &sb);
2346 error = freebsd11_cvtstat(&sb, &osb);
2348 error = copyout(&osb, uap->sb, sizeof(osb));
2353 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2356 struct freebsd11_stat osb;
2359 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2360 UIO_USERSPACE, &sb, NULL);
2363 error = freebsd11_cvtstat(&sb, &osb);
2365 error = copyout(&osb, uap->buf, sizeof(osb));
2368 #endif /* COMPAT_FREEBSD11 */
2373 #ifndef _SYS_SYSPROTO_H_
2374 struct fstatat_args {
2382 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2387 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2388 UIO_USERSPACE, &sb, NULL);
2390 error = copyout(&sb, uap->buf, sizeof (sb));
2395 kern_statat(struct thread *td, int flag, int fd, const char *path,
2396 enum uio_seg pathseg, struct stat *sbp,
2397 void (*hook)(struct vnode *vp, struct stat *sbp))
2399 struct nameidata nd;
2402 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2403 AT_RESOLVE_BENEATH)) != 0)
2406 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_BENEATH |
2407 AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF |
2408 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2410 if ((error = namei(&nd)) != 0)
2412 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2414 if (__predict_false(hook != NULL))
2415 hook(nd.ni_vp, sbp);
2417 NDFREE(&nd, NDF_ONLY_PNBUF);
2419 #ifdef __STAT_TIME_T_EXT
2420 sbp->st_atim_ext = 0;
2421 sbp->st_mtim_ext = 0;
2422 sbp->st_ctim_ext = 0;
2423 sbp->st_btim_ext = 0;
2426 if (KTRPOINT(td, KTR_STRUCT))
2427 ktrstat_error(sbp, error);
2432 #if defined(COMPAT_FREEBSD11)
2434 * Implementation of the NetBSD [l]stat() functions.
2437 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2440 bzero(nsb, sizeof(*nsb));
2441 nsb->st_dev = sb->st_dev;
2442 nsb->st_ino = sb->st_ino;
2443 nsb->st_mode = sb->st_mode;
2444 nsb->st_nlink = sb->st_nlink;
2445 nsb->st_uid = sb->st_uid;
2446 nsb->st_gid = sb->st_gid;
2447 nsb->st_rdev = sb->st_rdev;
2448 nsb->st_atim = sb->st_atim;
2449 nsb->st_mtim = sb->st_mtim;
2450 nsb->st_ctim = sb->st_ctim;
2451 nsb->st_size = sb->st_size;
2452 nsb->st_blocks = sb->st_blocks;
2453 nsb->st_blksize = sb->st_blksize;
2454 nsb->st_flags = sb->st_flags;
2455 nsb->st_gen = sb->st_gen;
2456 nsb->st_birthtim = sb->st_birthtim;
2459 #ifndef _SYS_SYSPROTO_H_
2460 struct freebsd11_nstat_args {
2466 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2472 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2476 freebsd11_cvtnstat(&sb, &nsb);
2477 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2481 * NetBSD lstat. Get file status; this version does not follow links.
2483 #ifndef _SYS_SYSPROTO_H_
2484 struct freebsd11_nlstat_args {
2490 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2496 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2497 UIO_USERSPACE, &sb, NULL);
2500 freebsd11_cvtnstat(&sb, &nsb);
2501 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2503 #endif /* COMPAT_FREEBSD11 */
2506 * Get configurable pathname variables.
2508 #ifndef _SYS_SYSPROTO_H_
2509 struct pathconf_args {
2515 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2520 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2523 td->td_retval[0] = value;
2527 #ifndef _SYS_SYSPROTO_H_
2528 struct lpathconf_args {
2534 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2539 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2542 td->td_retval[0] = value;
2547 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2548 int name, u_long flags, long *valuep)
2550 struct nameidata nd;
2553 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2555 if ((error = namei(&nd)) != 0)
2557 NDFREE(&nd, NDF_ONLY_PNBUF);
2559 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2565 * Return target name of a symbolic link.
2567 #ifndef _SYS_SYSPROTO_H_
2568 struct readlink_args {
2575 sys_readlink(struct thread *td, struct readlink_args *uap)
2578 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2579 uap->buf, UIO_USERSPACE, uap->count));
2581 #ifndef _SYS_SYSPROTO_H_
2582 struct readlinkat_args {
2590 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2593 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2594 uap->buf, UIO_USERSPACE, uap->bufsize));
2598 kern_readlinkat(struct thread *td, int fd, const char *path,
2599 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2602 struct nameidata nd;
2605 if (count > IOSIZE_MAX)
2608 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2609 pathseg, path, fd, td);
2611 if ((error = namei(&nd)) != 0)
2613 NDFREE(&nd, NDF_ONLY_PNBUF);
2616 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2623 * Helper function to readlink from a vnode
2626 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2633 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2635 error = mac_vnode_check_readlink(td->td_ucred, vp);
2639 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2642 aiov.iov_base = buf;
2643 aiov.iov_len = count;
2644 auio.uio_iov = &aiov;
2645 auio.uio_iovcnt = 1;
2646 auio.uio_offset = 0;
2647 auio.uio_rw = UIO_READ;
2648 auio.uio_segflg = bufseg;
2650 auio.uio_resid = count;
2651 error = VOP_READLINK(vp, &auio, td->td_ucred);
2652 td->td_retval[0] = count - auio.uio_resid;
2657 * Common implementation code for chflags() and fchflags().
2660 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2666 /* We can't support the value matching VNOVAL. */
2667 if (flags == VNOVAL)
2668 return (EOPNOTSUPP);
2671 * Prevent non-root users from setting flags on devices. When
2672 * a device is reused, users can retain ownership of the device
2673 * if they are allowed to set flags and programs assume that
2674 * chown can't fail when done as root.
2676 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2677 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2682 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2685 vattr.va_flags = flags;
2686 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2688 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2691 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2693 vn_finished_write(mp);
2698 * Change flags of a file given a path name.
2700 #ifndef _SYS_SYSPROTO_H_
2701 struct chflags_args {
2707 sys_chflags(struct thread *td, struct chflags_args *uap)
2710 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2714 #ifndef _SYS_SYSPROTO_H_
2715 struct chflagsat_args {
2723 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2726 if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2727 AT_RESOLVE_BENEATH)) != 0)
2730 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2731 uap->flags, uap->atflag));
2735 * Same as chflags() but doesn't follow symlinks.
2737 #ifndef _SYS_SYSPROTO_H_
2738 struct lchflags_args {
2744 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2747 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2748 uap->flags, AT_SYMLINK_NOFOLLOW));
2752 kern_chflagsat(struct thread *td, int fd, const char *path,
2753 enum uio_seg pathseg, u_long flags, int atflag)
2755 struct nameidata nd;
2758 AUDIT_ARG_FFLAGS(flags);
2759 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2760 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2761 &cap_fchflags_rights, td);
2762 if ((error = namei(&nd)) != 0)
2764 NDFREE(&nd, NDF_ONLY_PNBUF);
2765 error = setfflags(td, nd.ni_vp, flags);
2771 * Change flags of a file given a file descriptor.
2773 #ifndef _SYS_SYSPROTO_H_
2774 struct fchflags_args {
2780 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2785 AUDIT_ARG_FD(uap->fd);
2786 AUDIT_ARG_FFLAGS(uap->flags);
2787 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2792 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2793 AUDIT_ARG_VNODE1(fp->f_vnode);
2794 VOP_UNLOCK(fp->f_vnode);
2796 error = setfflags(td, fp->f_vnode, uap->flags);
2802 * Common implementation code for chmod(), lchmod() and fchmod().
2805 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2811 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2813 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2815 vattr.va_mode = mode & ALLPERMS;
2817 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2820 error = VOP_SETATTR(vp, &vattr, cred);
2822 vn_finished_write(mp);
2827 * Change mode of a file given path name.
2829 #ifndef _SYS_SYSPROTO_H_
2836 sys_chmod(struct thread *td, struct chmod_args *uap)
2839 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2843 #ifndef _SYS_SYSPROTO_H_
2844 struct fchmodat_args {
2852 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2855 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2856 AT_RESOLVE_BENEATH)) != 0)
2859 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2860 uap->mode, uap->flag));
2864 * Change mode of a file given path name (don't follow links.)
2866 #ifndef _SYS_SYSPROTO_H_
2867 struct lchmod_args {
2873 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2876 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2877 uap->mode, AT_SYMLINK_NOFOLLOW));
2881 kern_fchmodat(struct thread *td, int fd, const char *path,
2882 enum uio_seg pathseg, mode_t mode, int flag)
2884 struct nameidata nd;
2887 AUDIT_ARG_MODE(mode);
2888 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2889 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2890 &cap_fchmod_rights, td);
2891 if ((error = namei(&nd)) != 0)
2893 NDFREE(&nd, NDF_ONLY_PNBUF);
2894 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2900 * Change mode of a file given a file descriptor.
2902 #ifndef _SYS_SYSPROTO_H_
2903 struct fchmod_args {
2909 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2914 AUDIT_ARG_FD(uap->fd);
2915 AUDIT_ARG_MODE(uap->mode);
2917 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2920 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2926 * Common implementation for chown(), lchown(), and fchown()
2929 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2936 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2938 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2943 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2947 error = VOP_SETATTR(vp, &vattr, cred);
2949 vn_finished_write(mp);
2954 * Set ownership given a path name.
2956 #ifndef _SYS_SYSPROTO_H_
2964 sys_chown(struct thread *td, struct chown_args *uap)
2967 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2971 #ifndef _SYS_SYSPROTO_H_
2972 struct fchownat_args {
2981 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2984 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2985 AT_RESOLVE_BENEATH)) != 0)
2988 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2989 uap->gid, uap->flag));
2993 kern_fchownat(struct thread *td, int fd, const char *path,
2994 enum uio_seg pathseg, int uid, int gid, int flag)
2996 struct nameidata nd;
2999 AUDIT_ARG_OWNER(uid, gid);
3000 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3001 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
3002 &cap_fchown_rights, td);
3004 if ((error = namei(&nd)) != 0)
3006 NDFREE(&nd, NDF_ONLY_PNBUF);
3007 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3013 * Set ownership given a path name, do not cross symlinks.
3015 #ifndef _SYS_SYSPROTO_H_
3016 struct lchown_args {
3023 sys_lchown(struct thread *td, struct lchown_args *uap)
3026 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3027 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3031 * Set ownership given a file descriptor.
3033 #ifndef _SYS_SYSPROTO_H_
3034 struct fchown_args {
3041 sys_fchown(struct thread *td, struct fchown_args *uap)
3046 AUDIT_ARG_FD(uap->fd);
3047 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3048 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3051 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3057 * Common implementation code for utimes(), lutimes(), and futimes().
3060 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3061 struct timespec *tsp)
3063 struct timeval tv[2];
3064 const struct timeval *tvp;
3067 if (usrtvp == NULL) {
3068 vfs_timestamp(&tsp[0]);
3071 if (tvpseg == UIO_SYSSPACE) {
3074 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3079 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3080 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3082 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3083 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3089 * Common implementation code for futimens(), utimensat().
3091 #define UTIMENS_NULL 0x1
3092 #define UTIMENS_EXIT 0x2
3094 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3095 struct timespec *tsp, int *retflags)
3097 struct timespec tsnow;
3100 vfs_timestamp(&tsnow);
3102 if (usrtsp == NULL) {
3105 *retflags |= UTIMENS_NULL;
3108 if (tspseg == UIO_SYSSPACE) {
3111 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3113 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3114 *retflags |= UTIMENS_EXIT;
3115 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3116 *retflags |= UTIMENS_NULL;
3117 if (tsp[0].tv_nsec == UTIME_OMIT)
3118 tsp[0].tv_sec = VNOVAL;
3119 else if (tsp[0].tv_nsec == UTIME_NOW)
3121 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3123 if (tsp[1].tv_nsec == UTIME_OMIT)
3124 tsp[1].tv_sec = VNOVAL;
3125 else if (tsp[1].tv_nsec == UTIME_NOW)
3127 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3134 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3138 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3139 int numtimes, int nullflag)
3143 int error, setbirthtime;
3145 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3147 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3149 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3150 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3153 vattr.va_atime = ts[0];
3154 vattr.va_mtime = ts[1];
3156 vattr.va_birthtime = ts[1];
3158 vattr.va_birthtime = ts[2];
3160 vattr.va_vaflags |= VA_UTIMES_NULL;
3162 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3166 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3168 vn_finished_write(mp);
3173 * Set the access and modification times of a file.
3175 #ifndef _SYS_SYSPROTO_H_
3176 struct utimes_args {
3178 struct timeval *tptr;
3182 sys_utimes(struct thread *td, struct utimes_args *uap)
3185 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3186 uap->tptr, UIO_USERSPACE));
3189 #ifndef _SYS_SYSPROTO_H_
3190 struct futimesat_args {
3193 const struct timeval * times;
3197 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3200 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3201 uap->times, UIO_USERSPACE));
3205 kern_utimesat(struct thread *td, int fd, const char *path,
3206 enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3208 struct nameidata nd;
3209 struct timespec ts[2];
3212 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3214 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3215 &cap_futimes_rights, td);
3217 if ((error = namei(&nd)) != 0)
3219 NDFREE(&nd, NDF_ONLY_PNBUF);
3220 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3226 * Set the access and modification times of a file.
3228 #ifndef _SYS_SYSPROTO_H_
3229 struct lutimes_args {
3231 struct timeval *tptr;
3235 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3238 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3243 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3244 struct timeval *tptr, enum uio_seg tptrseg)
3246 struct timespec ts[2];
3247 struct nameidata nd;
3250 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3252 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3253 if ((error = namei(&nd)) != 0)
3255 NDFREE(&nd, NDF_ONLY_PNBUF);
3256 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3262 * Set the access and modification times of a file.
3264 #ifndef _SYS_SYSPROTO_H_
3265 struct futimes_args {
3267 struct timeval *tptr;
3271 sys_futimes(struct thread *td, struct futimes_args *uap)
3274 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3278 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3279 enum uio_seg tptrseg)
3281 struct timespec ts[2];
3286 error = getutimes(tptr, tptrseg, ts);
3289 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3293 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3294 AUDIT_ARG_VNODE1(fp->f_vnode);
3295 VOP_UNLOCK(fp->f_vnode);
3297 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3303 sys_futimens(struct thread *td, struct futimens_args *uap)
3306 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3310 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3311 enum uio_seg tptrseg)
3313 struct timespec ts[2];
3318 error = getutimens(tptr, tptrseg, ts, &flags);
3321 if (flags & UTIMENS_EXIT)
3323 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3327 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3328 AUDIT_ARG_VNODE1(fp->f_vnode);
3329 VOP_UNLOCK(fp->f_vnode);
3331 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3337 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3340 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3341 uap->times, UIO_USERSPACE, uap->flag));
3345 kern_utimensat(struct thread *td, int fd, const char *path,
3346 enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3349 struct nameidata nd;
3350 struct timespec ts[2];
3353 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
3354 AT_RESOLVE_BENEATH)) != 0)
3357 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3359 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3360 AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1,
3361 pathseg, path, fd, &cap_futimes_rights, td);
3362 if ((error = namei(&nd)) != 0)
3365 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3367 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3368 * "Search permission is denied by a component of the path prefix."
3370 NDFREE(&nd, NDF_ONLY_PNBUF);
3371 if ((flags & UTIMENS_EXIT) == 0)
3372 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3378 * Truncate a file given its path name.
3380 #ifndef _SYS_SYSPROTO_H_
3381 struct truncate_args {
3388 sys_truncate(struct thread *td, struct truncate_args *uap)
3391 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3395 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3402 struct nameidata nd;
3408 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3409 if ((error = namei(&nd)) != 0)
3412 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3413 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3414 vn_rangelock_unlock(vp, rl_cookie);
3418 NDFREE(&nd, NDF_ONLY_PNBUF);
3419 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3420 if (vp->v_type == VDIR)
3423 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3426 else if ((error = vn_writechk(vp)) == 0 &&
3427 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3429 vattr.va_size = length;
3430 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3433 vn_finished_write(mp);
3434 vn_rangelock_unlock(vp, rl_cookie);
3436 if (error == ERELOOKUP)
3441 #if defined(COMPAT_43)
3443 * Truncate a file given its path name.
3445 #ifndef _SYS_SYSPROTO_H_
3446 struct otruncate_args {
3452 otruncate(struct thread *td, struct otruncate_args *uap)
3455 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3457 #endif /* COMPAT_43 */
3459 #if defined(COMPAT_FREEBSD6)
3460 /* Versions with the pad argument */
3462 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3465 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3469 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3472 return (kern_ftruncate(td, uap->fd, uap->length));
3477 kern_fsync(struct thread *td, int fd, bool fullsync)
3482 int error, lock_flags;
3485 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3491 /* XXXKIB: compete outstanding aio writes */;
3494 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3497 if (MNT_SHARED_WRITES(mp) ||
3498 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3499 lock_flags = LK_SHARED;
3501 lock_flags = LK_EXCLUSIVE;
3503 vn_lock(vp, lock_flags | LK_RETRY);
3504 AUDIT_ARG_VNODE1(vp);
3505 if (vp->v_object != NULL) {
3506 VM_OBJECT_WLOCK(vp->v_object);
3507 vm_object_page_clean(vp->v_object, 0, 0, 0);
3508 VM_OBJECT_WUNLOCK(vp->v_object);
3510 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3512 vn_finished_write(mp);
3513 if (error == ERELOOKUP)
3521 * Sync an open file.
3523 #ifndef _SYS_SYSPROTO_H_
3529 sys_fsync(struct thread *td, struct fsync_args *uap)
3532 return (kern_fsync(td, uap->fd, true));
3536 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3539 return (kern_fsync(td, uap->fd, false));
3543 * Rename files. Source and destination must either both be directories, or
3544 * both not be directories. If target is a directory, it must be empty.
3546 #ifndef _SYS_SYSPROTO_H_
3547 struct rename_args {
3553 sys_rename(struct thread *td, struct rename_args *uap)
3556 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3557 uap->to, UIO_USERSPACE));
3560 #ifndef _SYS_SYSPROTO_H_
3561 struct renameat_args {
3569 sys_renameat(struct thread *td, struct renameat_args *uap)
3572 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3578 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3579 const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3583 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3584 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3585 if ((error = namei(fromnd)) != 0)
3587 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3588 fromnd->ni_vp, &fromnd->ni_cnd);
3589 VOP_UNLOCK(fromnd->ni_dvp);
3590 if (fromnd->ni_dvp != fromnd->ni_vp)
3591 VOP_UNLOCK(fromnd->ni_vp);
3593 NDFREE(fromnd, NDF_ONLY_PNBUF);
3594 vrele(fromnd->ni_dvp);
3595 vrele(fromnd->ni_vp);
3596 if (fromnd->ni_startdir)
3597 vrele(fromnd->ni_startdir);
3604 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3605 const char *new, enum uio_seg pathseg)
3607 struct mount *mp = NULL;
3608 struct vnode *tvp, *fvp, *tdvp;
3609 struct nameidata fromnd, tond;
3615 if (mac_vnode_check_rename_from_enabled()) {
3616 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3622 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3623 pathseg, old, oldfd, &cap_renameat_source_rights, td);
3624 if ((error = namei(&fromnd)) != 0)
3630 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3631 SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3632 &cap_renameat_target_rights, td);
3633 if (fromnd.ni_vp->v_type == VDIR)
3634 tond.ni_cnd.cn_flags |= WILLBEDIR;
3635 if ((error = namei(&tond)) != 0) {
3636 /* Translate error code for rename("dir1", "dir2/."). */
3637 if (error == EISDIR && fvp->v_type == VDIR)
3639 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3640 vrele(fromnd.ni_dvp);
3646 error = vn_start_write(fvp, &mp, V_NOWAIT);
3648 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3649 NDFREE(&tond, NDF_ONLY_PNBUF);
3656 vrele(fromnd.ni_dvp);
3658 vrele(tond.ni_startdir);
3659 if (fromnd.ni_startdir != NULL)
3660 vrele(fromnd.ni_startdir);
3661 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3667 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3670 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3675 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3677 * If the target already exists we require CAP_UNLINKAT
3678 * from 'newfd', when newfd was used for the lookup.
3680 error = cap_check(&tond.ni_filecaps.fc_rights,
3681 &cap_unlinkat_rights);
3692 * If the source is the same as the destination (that is, if they
3693 * are links to the same vnode), then there is nothing to do.
3699 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3700 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3704 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3705 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3706 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3707 NDFREE(&tond, NDF_ONLY_PNBUF);
3709 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3710 NDFREE(&tond, NDF_ONLY_PNBUF);
3717 vrele(fromnd.ni_dvp);
3720 vrele(tond.ni_startdir);
3721 vn_finished_write(mp);
3723 if (fromnd.ni_startdir)
3724 vrele(fromnd.ni_startdir);
3725 if (error == ERESTART)
3727 if (error == ERELOOKUP)
3733 * Make a directory file.
3735 #ifndef _SYS_SYSPROTO_H_
3742 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3745 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3749 #ifndef _SYS_SYSPROTO_H_
3750 struct mkdirat_args {
3757 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3760 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3764 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3770 struct nameidata nd;
3773 AUDIT_ARG_MODE(mode);
3776 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3777 NC_NOMAKEENTRY | NC_KEEPPOSENTRY, segflg, path, fd,
3778 &cap_mkdirat_rights, td);
3779 nd.ni_cnd.cn_flags |= WILLBEDIR;
3780 if ((error = namei(&nd)) != 0)
3784 NDFREE(&nd, NDF_ONLY_PNBUF);
3786 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3787 * the strange behaviour of leaving the vnode unlocked
3788 * if the target is the same vnode as the parent.
3790 if (vp == nd.ni_dvp)
3797 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3798 NDFREE(&nd, NDF_ONLY_PNBUF);
3800 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3805 vattr.va_type = VDIR;
3806 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3808 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3813 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3817 NDFREE(&nd, NDF_ONLY_PNBUF);
3821 vn_finished_write(mp);
3822 if (error == ERELOOKUP)
3828 * Remove a directory file.
3830 #ifndef _SYS_SYSPROTO_H_
3836 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3839 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3844 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3845 enum uio_seg pathseg, int flag)
3850 struct nameidata nd;
3851 cap_rights_t rights;
3855 if (fd != FD_NONE) {
3856 error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3864 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3865 at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
3866 pathseg, path, dfd, &cap_unlinkat_rights, td);
3867 if ((error = namei(&nd)) != 0)
3870 if (vp->v_type != VDIR) {
3875 * No rmdir "." please.
3877 if (nd.ni_dvp == vp) {
3882 * The root of a mounted filesystem cannot be deleted.
3884 if (vp->v_vflag & VV_ROOT) {
3889 if (fp != NULL && fp->f_vnode != vp) {
3890 if (VN_IS_DOOMED(fp->f_vnode))
3898 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3903 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3904 NDFREE(&nd, NDF_ONLY_PNBUF);
3906 if (nd.ni_dvp == vp)
3910 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3914 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3915 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3916 vn_finished_write(mp);
3918 NDFREE(&nd, NDF_ONLY_PNBUF);
3920 if (nd.ni_dvp == vp)
3924 if (error == ERELOOKUP)
3932 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3934 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3935 long *basep, void (*func)(struct freebsd11_dirent *))
3937 struct freebsd11_dirent dstdp;
3938 struct dirent *dp, *edp;
3941 ssize_t resid, ucount;
3944 /* XXX arbitrary sanity limit on `count'. */
3945 count = min(count, 64 * 1024);
3947 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3949 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3957 for (dp = (struct dirent *)dirbuf,
3958 edp = (struct dirent *)&dirbuf[count - resid];
3959 ucount < count && dp < edp; ) {
3960 if (dp->d_reclen == 0)
3962 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3963 if (dp->d_namlen >= sizeof(dstdp.d_name))
3965 dstdp.d_type = dp->d_type;
3966 dstdp.d_namlen = dp->d_namlen;
3967 dstdp.d_fileno = dp->d_fileno; /* truncate */
3968 if (dstdp.d_fileno != dp->d_fileno) {
3969 switch (ino64_trunc_error) {
3977 dstdp.d_fileno = UINT32_MAX;
3981 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3982 ((dp->d_namlen + 1 + 3) &~ 3);
3983 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3984 bzero(dstdp.d_name + dstdp.d_namlen,
3985 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3987 MPASS(dstdp.d_reclen <= dp->d_reclen);
3988 MPASS(ucount + dstdp.d_reclen <= count);
3991 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3994 dp = (struct dirent *)((char *)dp + dp->d_reclen);
3995 ucount += dstdp.d_reclen;
3999 free(dirbuf, M_TEMP);
4001 td->td_retval[0] = ucount;
4008 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4010 #if (BYTE_ORDER == LITTLE_ENDIAN)
4012 * The expected low byte of dp->d_namlen is our dp->d_type.
4013 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4015 dp->d_type = dp->d_namlen;
4019 * The dp->d_type is the high byte of the expected dp->d_namlen,
4020 * so must be zero'ed.
4027 * Read a block of directory entries in a filesystem independent format.
4029 #ifndef _SYS_SYSPROTO_H_
4030 struct ogetdirentries_args {
4038 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4043 error = kern_ogetdirentries(td, uap, &loff);
4045 error = copyout(&loff, uap->basep, sizeof(long));
4050 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4056 /* XXX arbitrary sanity limit on `count'. */
4057 if (uap->count > 64 * 1024)
4060 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4061 &base, ogetdirentries_cvt);
4063 if (error == 0 && uap->basep != NULL)
4064 error = copyout(&base, uap->basep, sizeof(long));
4068 #endif /* COMPAT_43 */
4070 #if defined(COMPAT_FREEBSD11)
4071 #ifndef _SYS_SYSPROTO_H_
4072 struct freebsd11_getdirentries_args {
4080 freebsd11_getdirentries(struct thread *td,
4081 struct freebsd11_getdirentries_args *uap)
4086 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4089 if (error == 0 && uap->basep != NULL)
4090 error = copyout(&base, uap->basep, sizeof(long));
4095 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4097 struct freebsd11_getdirentries_args ap;
4101 ap.count = uap->count;
4103 return (freebsd11_getdirentries(td, &ap));
4105 #endif /* COMPAT_FREEBSD11 */
4108 * Read a block of directory entries in a filesystem independent format.
4111 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4116 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4117 NULL, UIO_USERSPACE);
4120 if (uap->basep != NULL)
4121 error = copyout(&base, uap->basep, sizeof(off_t));
4126 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4127 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4138 if (count > IOSIZE_MAX)
4140 auio.uio_resid = count;
4141 error = getvnode(td, fd, &cap_read_rights, &fp);
4144 if ((fp->f_flag & FREAD) == 0) {
4149 foffset = foffset_lock(fp, 0);
4151 if (vp->v_type != VDIR) {
4155 aiov.iov_base = buf;
4156 aiov.iov_len = count;
4157 auio.uio_iov = &aiov;
4158 auio.uio_iovcnt = 1;
4159 auio.uio_rw = UIO_READ;
4160 auio.uio_segflg = bufseg;
4162 vn_lock(vp, LK_SHARED | LK_RETRY);
4163 AUDIT_ARG_VNODE1(vp);
4164 loff = auio.uio_offset = foffset;
4166 error = mac_vnode_check_readdir(td->td_ucred, vp);
4169 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4171 foffset = auio.uio_offset;
4176 if (count == auio.uio_resid &&
4177 (vp->v_vflag & VV_ROOT) &&
4178 (vp->v_mount->mnt_flag & MNT_UNION)) {
4179 struct vnode *tvp = vp;
4181 vp = vp->v_mount->mnt_vnodecovered;
4191 *residp = auio.uio_resid;
4192 td->td_retval[0] = count - auio.uio_resid;
4194 foffset_unlock(fp, foffset, 0);
4200 * Set the mode mask for creation of filesystem nodes.
4202 #ifndef _SYS_SYSPROTO_H_
4208 sys_umask(struct thread *td, struct umask_args *uap)
4210 struct filedesc *fdp;
4212 fdp = td->td_proc->p_fd;
4213 FILEDESC_XLOCK(fdp);
4214 td->td_retval[0] = fdp->fd_cmask;
4215 fdp->fd_cmask = uap->newmask & ALLPERMS;
4216 FILEDESC_XUNLOCK(fdp);
4221 * Void all references to file by ripping underlying filesystem away from
4224 #ifndef _SYS_SYSPROTO_H_
4225 struct revoke_args {
4230 sys_revoke(struct thread *td, struct revoke_args *uap)
4234 struct nameidata nd;
4237 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4239 if ((error = namei(&nd)) != 0)
4242 NDFREE(&nd, NDF_ONLY_PNBUF);
4243 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4248 error = mac_vnode_check_revoke(td->td_ucred, vp);
4252 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4255 if (td->td_ucred->cr_uid != vattr.va_uid) {
4256 error = priv_check(td, PRIV_VFS_ADMIN);
4260 if (devfs_usecount(vp) > 0)
4261 VOP_REVOKE(vp, REVOKEALL);
4268 * Convert a user file descriptor to a kernel file entry and check that, if it
4269 * is a capability, the correct rights are present. A reference on the file
4270 * entry is held upon returning.
4273 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4278 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4283 * The file could be not of the vnode type, or it may be not
4284 * yet fully initialized, in which case the f_vnode pointer
4285 * may be set, but f_ops is still badfileops. E.g.,
4286 * devfs_open() transiently create such situation to
4287 * facilitate csw d_fdopen().
4289 * Dupfdopen() handling in kern_openat() installs the
4290 * half-baked file into the process descriptor table, allowing
4291 * other thread to dereference it. Guard against the race by
4294 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4303 * Get an (NFS) file handle.
4305 #ifndef _SYS_SYSPROTO_H_
4306 struct lgetfh_args {
4312 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4315 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4316 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4319 #ifndef _SYS_SYSPROTO_H_
4326 sys_getfh(struct thread *td, struct getfh_args *uap)
4329 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4330 uap->fhp, UIO_USERSPACE));
4334 * syscall for the rpc.lockd to use to translate an open descriptor into
4335 * a NFS file handle.
4337 * warning: do not remove the priv_check() call or this becomes one giant
4340 #ifndef _SYS_SYSPROTO_H_
4341 struct getfhat_args {
4349 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4352 if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
4353 AT_RESOLVE_BENEATH)) != 0)
4355 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4356 uap->fhp, UIO_USERSPACE));
4360 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4361 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4363 struct nameidata nd;
4368 error = priv_check(td, PRIV_VFS_GETFH);
4371 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4372 AT_BENEATH | AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1,
4373 pathseg, path, fd, td);
4377 NDFREE(&nd, NDF_ONLY_PNBUF);
4379 bzero(&fh, sizeof(fh));
4380 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4381 error = VOP_VPTOFH(vp, &fh.fh_fid);
4384 if (fhseg == UIO_USERSPACE)
4385 error = copyout(&fh, fhp, sizeof (fh));
4387 memcpy(fhp, &fh, sizeof(fh));
4392 #ifndef _SYS_SYSPROTO_H_
4393 struct fhlink_args {
4399 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4402 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4405 #ifndef _SYS_SYSPROTO_H_
4406 struct fhlinkat_args {
4413 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4416 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4420 kern_fhlinkat(struct thread *td, int fd, const char *path,
4421 enum uio_seg pathseg, fhandle_t *fhp)
4428 error = priv_check(td, PRIV_VFS_GETFH);
4431 error = copyin(fhp, &fh, sizeof(fh));
4436 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4438 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4443 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4444 } while (error == EAGAIN || error == ERELOOKUP);
4448 #ifndef _SYS_SYSPROTO_H_
4449 struct fhreadlink_args {
4456 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4463 error = priv_check(td, PRIV_VFS_GETFH);
4466 if (uap->bufsize > IOSIZE_MAX)
4468 error = copyin(uap->fhp, &fh, sizeof(fh));
4471 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4473 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4477 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4483 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4486 * warning: do not remove the priv_check() call or this becomes one giant
4489 #ifndef _SYS_SYSPROTO_H_
4490 struct fhopen_args {
4491 const struct fhandle *u_fhp;
4496 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4498 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4502 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4511 error = priv_check(td, PRIV_VFS_FHOPEN);
4515 fmode = FFLAGS(flags);
4516 /* why not allow a non-read/write open for our lockd? */
4517 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4519 error = copyin(u_fhp, &fhp, sizeof(fhp));
4522 /* find the mount point */
4523 mp = vfs_busyfs(&fhp.fh_fsid);
4526 /* now give me my vnode, it gets returned to me locked */
4527 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4532 error = falloc_noinstall(td, &fp);
4538 * An extra reference on `fp' has been held for us by
4539 * falloc_noinstall().
4545 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4547 KASSERT(fp->f_ops == &badfileops,
4548 ("VOP_OPEN in fhopen() set f_ops"));
4549 KASSERT(td->td_dupfd < 0,
4550 ("fhopen() encountered fdopen()"));
4559 finit_vnode(fp, fmode, NULL, &vnops);
4561 if ((fmode & O_TRUNC) != 0) {
4562 error = fo_truncate(fp, 0, td->td_ucred, td);
4567 error = finstall(td, fp, &indx, fmode, NULL);
4570 td->td_retval[0] = indx;
4575 * Stat an (NFS) file handle.
4577 #ifndef _SYS_SYSPROTO_H_
4578 struct fhstat_args {
4579 struct fhandle *u_fhp;
4584 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4590 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4593 error = kern_fhstat(td, fh, &sb);
4595 error = copyout(&sb, uap->sb, sizeof(sb));
4600 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4606 error = priv_check(td, PRIV_VFS_FHSTAT);
4609 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4611 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4615 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4621 * Implement fstatfs() for (NFS) file handles.
4623 #ifndef _SYS_SYSPROTO_H_
4624 struct fhstatfs_args {
4625 struct fhandle *u_fhp;
4630 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4636 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4639 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4640 error = kern_fhstatfs(td, fh, sfp);
4642 error = copyout(sfp, uap->buf, sizeof(*sfp));
4643 free(sfp, M_STATFS);
4648 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4654 error = priv_check(td, PRIV_VFS_FHSTATFS);
4657 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4659 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4665 error = prison_canseemount(td->td_ucred, mp);
4669 error = mac_mount_check_stat(td->td_ucred, mp);
4673 error = VFS_STATFS(mp, buf);
4680 * Unlike madvise(2), we do not make a best effort to remember every
4681 * possible caching hint. Instead, we remember the last setting with
4682 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4683 * region of any current setting.
4686 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4689 struct fadvise_info *fa, *new;
4695 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4697 AUDIT_ARG_VALUE(advice);
4699 case POSIX_FADV_SEQUENTIAL:
4700 case POSIX_FADV_RANDOM:
4701 case POSIX_FADV_NOREUSE:
4702 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4704 case POSIX_FADV_NORMAL:
4705 case POSIX_FADV_WILLNEED:
4706 case POSIX_FADV_DONTNEED:
4712 /* XXX: CAP_POSIX_FADVISE? */
4714 error = fget(td, fd, &cap_no_rights, &fp);
4717 AUDIT_ARG_FILE(td->td_proc, fp);
4718 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4722 if (fp->f_type != DTYPE_VNODE) {
4727 if (vp->v_type != VREG) {
4734 end = offset + len - 1;
4736 case POSIX_FADV_SEQUENTIAL:
4737 case POSIX_FADV_RANDOM:
4738 case POSIX_FADV_NOREUSE:
4740 * Try to merge any existing non-standard region with
4741 * this new region if possible, otherwise create a new
4742 * non-standard region for this request.
4744 mtx_pool_lock(mtxpool_sleep, fp);
4746 if (fa != NULL && fa->fa_advice == advice &&
4747 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4748 (end != OFF_MAX && fa->fa_start == end + 1) ||
4749 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4750 if (offset < fa->fa_start)
4751 fa->fa_start = offset;
4752 if (end > fa->fa_end)
4755 new->fa_advice = advice;
4756 new->fa_start = offset;
4761 mtx_pool_unlock(mtxpool_sleep, fp);
4763 case POSIX_FADV_NORMAL:
4765 * If a the "normal" region overlaps with an existing
4766 * non-standard region, trim or remove the
4767 * non-standard region.
4769 mtx_pool_lock(mtxpool_sleep, fp);
4772 if (offset <= fa->fa_start && end >= fa->fa_end) {
4774 fp->f_advice = NULL;
4775 } else if (offset <= fa->fa_start &&
4776 end >= fa->fa_start)
4777 fa->fa_start = end + 1;
4778 else if (offset <= fa->fa_end && end >= fa->fa_end)
4779 fa->fa_end = offset - 1;
4780 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4782 * If the "normal" region is a middle
4783 * portion of the existing
4784 * non-standard region, just remove
4785 * the whole thing rather than picking
4786 * one side or the other to
4790 fp->f_advice = NULL;
4793 mtx_pool_unlock(mtxpool_sleep, fp);
4795 case POSIX_FADV_WILLNEED:
4796 case POSIX_FADV_DONTNEED:
4797 error = VOP_ADVISE(vp, offset, end, advice);
4803 free(new, M_FADVISE);
4808 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4812 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4814 return (kern_posix_error(td, error));
4818 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4819 off_t *outoffp, size_t len, unsigned int flags)
4821 struct file *infp, *outfp;
4822 struct vnode *invp, *outvp;
4825 void *rl_rcookie, *rl_wcookie;
4826 off_t savinoff, savoutoff;
4828 infp = outfp = NULL;
4829 rl_rcookie = rl_wcookie = NULL;
4838 if (len > SSIZE_MAX)
4840 * Although the len argument is size_t, the return argument
4841 * is ssize_t (which is signed). Therefore a size that won't
4842 * fit in ssize_t can't be returned.
4846 /* Get the file structures for the file descriptors. */
4847 error = fget_read(td, infd, &cap_read_rights, &infp);
4850 if (infp->f_ops == &badfileops) {
4854 if (infp->f_vnode == NULL) {
4858 error = fget_write(td, outfd, &cap_write_rights, &outfp);
4861 if (outfp->f_ops == &badfileops) {
4865 if (outfp->f_vnode == NULL) {
4870 /* Set the offset pointers to the correct place. */
4872 inoffp = &infp->f_offset;
4873 if (outoffp == NULL)
4874 outoffp = &outfp->f_offset;
4876 savoutoff = *outoffp;
4878 invp = infp->f_vnode;
4879 outvp = outfp->f_vnode;
4880 /* Sanity check the f_flag bits. */
4881 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4882 (infp->f_flag & FREAD) == 0) {
4887 /* If len == 0, just return 0. */
4892 * If infp and outfp refer to the same file, the byte ranges cannot
4895 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4896 savoutoff) || (savinoff > savoutoff && savoutoff + len >
4902 /* Range lock the byte ranges for both invp and outvp. */
4904 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4906 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4908 if (rl_rcookie != NULL)
4910 vn_rangelock_unlock(outvp, rl_wcookie);
4911 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4912 vn_rangelock_unlock(invp, rl_rcookie);
4916 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4917 flags, infp->f_cred, outfp->f_cred, td);
4919 if (rl_rcookie != NULL)
4920 vn_rangelock_unlock(invp, rl_rcookie);
4921 if (rl_wcookie != NULL)
4922 vn_rangelock_unlock(outvp, rl_wcookie);
4923 if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4925 *outoffp = savoutoff;
4931 td->td_retval[0] = retlen;
4936 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4938 off_t inoff, outoff, *inoffp, *outoffp;
4941 inoffp = outoffp = NULL;
4942 if (uap->inoffp != NULL) {
4943 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4948 if (uap->outoffp != NULL) {
4949 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4954 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4955 outoffp, uap->len, uap->flags);
4956 if (error == 0 && uap->inoffp != NULL)
4957 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4958 if (error == 0 && uap->outoffp != NULL)
4959 error = copyout(outoffp, uap->outoffp, sizeof(off_t));