2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
39 * $FreeBSD: src/sys/kern/vfs_syscalls.c,v 1.151.2.18 2003/04/04 20:35:58 tegge Exp $
40 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.133.2.1 2008/09/25 02:20:46 dillon Exp $
43 #include <sys/param.h>
44 #include <sys/systm.h>
47 #include <sys/sysent.h>
48 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/mountctl.h>
51 #include <sys/sysproto.h>
52 #include <sys/filedesc.h>
53 #include <sys/kernel.h>
54 #include <sys/fcntl.h>
56 #include <sys/linker.h>
58 #include <sys/unistd.h>
59 #include <sys/vnode.h>
61 #include <sys/namei.h>
62 #include <sys/nlookup.h>
63 #include <sys/dirent.h>
64 #include <sys/extattr.h>
65 #include <sys/spinlock.h>
66 #include <sys/kern_syscall.h>
67 #include <sys/objcache.h>
68 #include <sys/sysctl.h>
71 #include <sys/file2.h>
72 #include <sys/spinlock2.h>
75 #include <vm/vm_object.h>
76 #include <vm/vm_page.h>
78 #include <machine/limits.h>
79 #include <machine/stdarg.h>
81 #include <vfs/union/union.h>
83 static void mount_warning(struct mount
*mp
, const char *ctl
, ...);
84 static int mount_path(struct proc
*p
, struct mount
*mp
, char **rb
, char **fb
);
85 static int checkvp_chdir (struct vnode
*vn
, struct thread
*td
);
86 static void checkdirs (struct nchandle
*old_nch
, struct nchandle
*new_nch
);
87 static int chroot_refuse_vdir_fds (struct filedesc
*fdp
);
88 static int chroot_visible_mnt(struct mount
*mp
, struct proc
*p
);
89 static int getutimes (const struct timeval
*, struct timespec
*);
90 static int setfown (struct vnode
*, uid_t
, gid_t
);
91 static int setfmode (struct vnode
*, int);
92 static int setfflags (struct vnode
*, int);
93 static int setutimes (struct vnode
*, const struct timespec
*, int);
94 static int usermount
= 0; /* if 1, non-root can mount fs. */
96 int (*union_dircheckp
) (struct thread
*, struct vnode
**, struct file
*);
98 SYSCTL_INT(_vfs
, OID_AUTO
, usermount
, CTLFLAG_RW
, &usermount
, 0, "");
101 * Virtual File System System Calls
105 * Mount a file system.
108 * mount_args(char *type, char *path, int flags, caddr_t data)
112 sys_mount(struct mount_args
*uap
)
114 struct thread
*td
= curthread
;
115 struct proc
*p
= td
->td_proc
;
119 struct vfsconf
*vfsp
;
120 int error
, flag
= 0, flag2
= 0;
123 struct nlookupdata nd
;
124 char fstypename
[MFSNAMELEN
];
125 struct ucred
*cred
= p
->p_ucred
;
128 if (cred
->cr_prison
!= NULL
)
130 if (usermount
== 0 && (error
= suser(td
)))
133 * Do not allow NFS export by non-root users.
135 if (uap
->flags
& MNT_EXPORTED
) {
141 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
144 uap
->flags
|= MNT_NOSUID
| MNT_NODEV
;
147 * Lookup the requested path and extract the nch and vnode.
149 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
151 if ((error
= nlookup(&nd
)) == 0) {
152 if (nd
.nl_nch
.ncp
->nc_vp
== NULL
)
162 * Extract the locked+refd ncp and cleanup the nd structure
165 cache_zero(&nd
.nl_nch
);
168 if ((nch
.ncp
->nc_flag
& NCF_ISMOUNTPT
) && cache_findmount(&nch
))
175 * now we have the locked ref'd nch and unreferenced vnode.
178 if ((error
= vget(vp
, LK_EXCLUSIVE
)) != 0) {
185 * Now we have an unlocked ref'd nch and a locked ref'd vp
187 if (uap
->flags
& MNT_UPDATE
) {
188 if ((vp
->v_flag
& (VROOT
|VPFSROOT
)) == 0) {
195 flag2
= mp
->mnt_kern_flag
;
197 * We only allow the filesystem to be reloaded if it
198 * is currently mounted read-only.
200 if ((uap
->flags
& MNT_RELOAD
) &&
201 ((mp
->mnt_flag
& MNT_RDONLY
) == 0)) {
204 return (EOPNOTSUPP
); /* Needs translation */
207 * Only root, or the user that did the original mount is
208 * permitted to update it.
210 if (mp
->mnt_stat
.f_owner
!= cred
->cr_uid
&&
211 (error
= suser(td
))) {
216 if (vfs_busy(mp
, LK_NOWAIT
)) {
221 if ((vp
->v_flag
& VMOUNT
) != 0 || hasmount
) {
227 vp
->v_flag
|= VMOUNT
;
229 uap
->flags
& (MNT_RELOAD
| MNT_FORCE
| MNT_UPDATE
);
234 * If the user is not root, ensure that they own the directory
235 * onto which we are attempting to mount.
237 if ((error
= VOP_GETATTR(vp
, &va
)) ||
238 (va
.va_uid
!= cred
->cr_uid
&& (error
= suser(td
)))) {
243 if ((error
= vinvalbuf(vp
, V_SAVE
, 0, 0)) != 0) {
248 if (vp
->v_type
!= VDIR
) {
253 if (vp
->v_mount
->mnt_kern_flag
& MNTK_NOSTKMNT
) {
258 if ((error
= copyinstr(uap
->type
, fstypename
, MFSNAMELEN
, NULL
)) != 0) {
263 for (vfsp
= vfsconf
; vfsp
; vfsp
= vfsp
->vfc_next
) {
264 if (!strcmp(vfsp
->vfc_name
, fstypename
))
270 /* Only load modules for root (very important!) */
271 if ((error
= suser(td
)) != 0) {
276 error
= linker_load_file(fstypename
, &lf
);
277 if (error
|| lf
== NULL
) {
285 /* lookup again, see if the VFS was loaded */
286 for (vfsp
= vfsconf
; vfsp
; vfsp
= vfsp
->vfc_next
) {
287 if (!strcmp(vfsp
->vfc_name
, fstypename
))
292 linker_file_unload(lf
);
298 if ((vp
->v_flag
& VMOUNT
) != 0 || hasmount
) {
303 vp
->v_flag
|= VMOUNT
;
306 * Allocate and initialize the filesystem.
308 mp
= kmalloc(sizeof(struct mount
), M_MOUNT
, M_ZERO
|M_WAITOK
);
309 TAILQ_INIT(&mp
->mnt_nvnodelist
);
310 TAILQ_INIT(&mp
->mnt_reservedvnlist
);
311 TAILQ_INIT(&mp
->mnt_jlist
);
312 mp
->mnt_nvnodelistsize
= 0;
313 lockinit(&mp
->mnt_lock
, "vfslock", 0, 0);
314 vfs_busy(mp
, LK_NOWAIT
);
315 mp
->mnt_op
= vfsp
->vfc_vfsops
;
317 vfsp
->vfc_refcount
++;
318 mp
->mnt_stat
.f_type
= vfsp
->vfc_typenum
;
319 mp
->mnt_flag
|= vfsp
->vfc_flags
& MNT_VISFLAGMASK
;
320 strncpy(mp
->mnt_stat
.f_fstypename
, vfsp
->vfc_name
, MFSNAMELEN
);
321 mp
->mnt_stat
.f_owner
= cred
->cr_uid
;
322 mp
->mnt_iosize_max
= DFLTPHYS
;
326 * Set the mount level flags.
328 if (uap
->flags
& MNT_RDONLY
)
329 mp
->mnt_flag
|= MNT_RDONLY
;
330 else if (mp
->mnt_flag
& MNT_RDONLY
)
331 mp
->mnt_kern_flag
|= MNTK_WANTRDWR
;
332 mp
->mnt_flag
&=~ (MNT_NOSUID
| MNT_NOEXEC
| MNT_NODEV
|
333 MNT_SYNCHRONOUS
| MNT_UNION
| MNT_ASYNC
| MNT_NOATIME
|
334 MNT_NOSYMFOLLOW
| MNT_IGNORE
|
335 MNT_NOCLUSTERR
| MNT_NOCLUSTERW
| MNT_SUIDDIR
);
336 mp
->mnt_flag
|= uap
->flags
& (MNT_NOSUID
| MNT_NOEXEC
|
337 MNT_NODEV
| MNT_SYNCHRONOUS
| MNT_UNION
| MNT_ASYNC
| MNT_FORCE
|
338 MNT_NOSYMFOLLOW
| MNT_IGNORE
|
339 MNT_NOATIME
| MNT_NOCLUSTERR
| MNT_NOCLUSTERW
| MNT_SUIDDIR
);
341 * Mount the filesystem.
342 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
345 error
= VFS_MOUNT(mp
, uap
->path
, uap
->data
, cred
);
346 if (mp
->mnt_flag
& MNT_UPDATE
) {
347 if (mp
->mnt_kern_flag
& MNTK_WANTRDWR
)
348 mp
->mnt_flag
&= ~MNT_RDONLY
;
349 mp
->mnt_flag
&=~ (MNT_UPDATE
| MNT_RELOAD
| MNT_FORCE
);
350 mp
->mnt_kern_flag
&=~ MNTK_WANTRDWR
;
353 mp
->mnt_kern_flag
= flag2
;
356 vp
->v_flag
&= ~VMOUNT
;
361 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
363 * Put the new filesystem on the mount list after root. The mount
364 * point gets its own mnt_ncmountpt (unless the VFS already set one
365 * up) which represents the root of the mount. The lookup code
366 * detects the mount point going forward and checks the root of
367 * the mount going backwards.
369 * It is not necessary to invalidate or purge the vnode underneath
370 * because elements under the mount will be given their own glue
374 if (mp
->mnt_ncmountpt
.ncp
== NULL
) {
376 * allocate, then unlock, but leave the ref intact
378 cache_allocroot(&mp
->mnt_ncmountpt
, mp
, NULL
);
379 cache_unlock(&mp
->mnt_ncmountpt
);
381 mp
->mnt_ncmounton
= nch
; /* inherits ref */
382 nch
.ncp
->nc_flag
|= NCF_ISMOUNTPT
;
384 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
385 vp
->v_flag
&= ~VMOUNT
;
386 mountlist_insert(mp
, MNTINS_LAST
);
387 checkdirs(&mp
->mnt_ncmounton
, &mp
->mnt_ncmountpt
);
389 error
= vfs_allocate_syncvnode(mp
);
391 error
= VFS_START(mp
, 0);
394 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_coherency_ops
);
395 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_journal_ops
);
396 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_norm_ops
);
397 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_spec_ops
);
398 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_fifo_ops
);
399 vp
->v_flag
&= ~VMOUNT
;
400 mp
->mnt_vfc
->vfc_refcount
--;
410 * Scan all active processes to see if any of them have a current
411 * or root directory onto which the new filesystem has just been
412 * mounted. If so, replace them with the new mount point.
414 * The passed ncp is ref'd and locked (from the mount code) and
415 * must be associated with the vnode representing the root of the
418 struct checkdirs_info
{
419 struct nchandle old_nch
;
420 struct nchandle new_nch
;
421 struct vnode
*old_vp
;
422 struct vnode
*new_vp
;
425 static int checkdirs_callback(struct proc
*p
, void *data
);
428 checkdirs(struct nchandle
*old_nch
, struct nchandle
*new_nch
)
430 struct checkdirs_info info
;
436 * If the old mount point's vnode has a usecount of 1, it is not
437 * being held as a descriptor anywhere.
439 olddp
= old_nch
->ncp
->nc_vp
;
440 if (olddp
== NULL
|| olddp
->v_sysref
.refcnt
== 1)
444 * Force the root vnode of the new mount point to be resolved
445 * so we can update any matching processes.
448 if (VFS_ROOT(mp
, &newdp
))
449 panic("mount: lost mount");
450 cache_setunresolved(new_nch
);
451 cache_setvp(new_nch
, newdp
);
454 * Special handling of the root node
456 if (rootvnode
== olddp
) {
458 vfs_cache_setroot(newdp
, cache_hold(new_nch
));
462 * Pass newdp separately so the callback does not have to access
463 * it via new_nch->ncp->nc_vp.
465 info
.old_nch
= *old_nch
;
466 info
.new_nch
= *new_nch
;
468 allproc_scan(checkdirs_callback
, &info
);
473 * NOTE: callback is not MP safe because the scanned process's filedesc
474 * structure can be ripped out from under us, amoung other things.
477 checkdirs_callback(struct proc
*p
, void *data
)
479 struct checkdirs_info
*info
= data
;
480 struct filedesc
*fdp
;
481 struct nchandle ncdrop1
;
482 struct nchandle ncdrop2
;
483 struct vnode
*vprele1
;
484 struct vnode
*vprele2
;
486 if ((fdp
= p
->p_fd
) != NULL
) {
487 cache_zero(&ncdrop1
);
488 cache_zero(&ncdrop2
);
493 * MPUNSAFE - XXX fdp can be pulled out from under a
496 * A shared filedesc is ok, we don't have to copy it
497 * because we are making this change globally.
499 spin_lock_wr(&fdp
->fd_spin
);
500 if (fdp
->fd_ncdir
.mount
== info
->old_nch
.mount
&&
501 fdp
->fd_ncdir
.ncp
== info
->old_nch
.ncp
) {
502 vprele1
= fdp
->fd_cdir
;
504 fdp
->fd_cdir
= info
->new_vp
;
505 ncdrop1
= fdp
->fd_ncdir
;
506 cache_copy(&info
->new_nch
, &fdp
->fd_ncdir
);
508 if (fdp
->fd_nrdir
.mount
== info
->old_nch
.mount
&&
509 fdp
->fd_nrdir
.ncp
== info
->old_nch
.ncp
) {
510 vprele2
= fdp
->fd_rdir
;
512 fdp
->fd_rdir
= info
->new_vp
;
513 ncdrop2
= fdp
->fd_nrdir
;
514 cache_copy(&info
->new_nch
, &fdp
->fd_nrdir
);
516 spin_unlock_wr(&fdp
->fd_spin
);
518 cache_drop(&ncdrop1
);
520 cache_drop(&ncdrop2
);
530 * Unmount a file system.
532 * Note: unmount takes a path to the vnode mounted on as argument,
533 * not special file (as before).
536 * umount_args(char *path, int flags)
540 sys_unmount(struct unmount_args
*uap
)
542 struct thread
*td
= curthread
;
543 struct proc
*p
= td
->td_proc
;
544 struct mount
*mp
= NULL
;
546 struct nlookupdata nd
;
549 if (p
->p_ucred
->cr_prison
!= NULL
)
551 if (usermount
== 0 && (error
= suser(td
)))
554 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
556 error
= nlookup(&nd
);
560 mp
= nd
.nl_nch
.mount
;
563 * Only root, or the user that did the original mount is
564 * permitted to unmount this filesystem.
566 if ((mp
->mnt_stat
.f_owner
!= p
->p_ucred
->cr_uid
) &&
571 * Don't allow unmounting the root file system.
573 if (mp
->mnt_flag
& MNT_ROOTFS
) {
579 * Must be the root of the filesystem
581 if (nd
.nl_nch
.ncp
!= mp
->mnt_ncmountpt
.ncp
) {
590 return (dounmount(mp
, uap
->flags
));
594 * Do the actual file system unmount.
597 dounmount_interlock(struct mount
*mp
)
599 if (mp
->mnt_kern_flag
& MNTK_UNMOUNT
)
601 mp
->mnt_kern_flag
|= MNTK_UNMOUNT
;
606 dounmount(struct mount
*mp
, int flags
)
608 struct namecache
*ncp
;
617 * Exclusive access for unmounting purposes
619 if ((error
= mountlist_interlock(dounmount_interlock
, mp
)) != 0)
623 * Allow filesystems to detect that a forced unmount is in progress.
625 if (flags
& MNT_FORCE
)
626 mp
->mnt_kern_flag
|= MNTK_UNMOUNTF
;
627 lflags
= LK_EXCLUSIVE
| ((flags
& MNT_FORCE
) ? 0 : LK_NOWAIT
);
628 error
= lockmgr(&mp
->mnt_lock
, lflags
);
630 mp
->mnt_kern_flag
&= ~(MNTK_UNMOUNT
| MNTK_UNMOUNTF
);
631 if (mp
->mnt_kern_flag
& MNTK_MWAIT
)
636 if (mp
->mnt_flag
& MNT_EXPUBLIC
)
637 vfs_setpublicfs(NULL
, NULL
, NULL
);
639 vfs_msync(mp
, MNT_WAIT
);
640 async_flag
= mp
->mnt_flag
& MNT_ASYNC
;
641 mp
->mnt_flag
&=~ MNT_ASYNC
;
644 * If this filesystem isn't aliasing other filesystems,
645 * try to invalidate any remaining namecache entries and
646 * check the count afterwords.
648 if ((mp
->mnt_kern_flag
& MNTK_NCALIASED
) == 0) {
649 cache_lock(&mp
->mnt_ncmountpt
);
650 cache_inval(&mp
->mnt_ncmountpt
, CINV_DESTROY
|CINV_CHILDREN
);
651 cache_unlock(&mp
->mnt_ncmountpt
);
653 if ((ncp
= mp
->mnt_ncmountpt
.ncp
) != NULL
&&
654 (ncp
->nc_refs
!= 1 || TAILQ_FIRST(&ncp
->nc_list
))) {
656 if ((flags
& MNT_FORCE
) == 0) {
658 mount_warning(mp
, "Cannot unmount: "
664 mount_warning(mp
, "Forced unmount: "
675 * nchandle records ref the mount structure. Expect a count of 1
676 * (our mount->mnt_ncmountpt).
678 if (mp
->mnt_refs
!= 1) {
679 if ((flags
& MNT_FORCE
) == 0) {
680 mount_warning(mp
, "Cannot unmount: "
681 "%d process references still "
682 "present", mp
->mnt_refs
);
685 mount_warning(mp
, "Forced unmount: "
686 "%d process references still "
687 "present", mp
->mnt_refs
);
693 * Decomission our special mnt_syncer vnode. This also stops
694 * the vnlru code. If we are unable to unmount we recommission
698 if ((vp
= mp
->mnt_syncer
) != NULL
) {
699 mp
->mnt_syncer
= NULL
;
702 if (((mp
->mnt_flag
& MNT_RDONLY
) ||
703 (error
= VFS_SYNC(mp
, MNT_WAIT
)) == 0) ||
704 (flags
& MNT_FORCE
)) {
705 error
= VFS_UNMOUNT(mp
, flags
);
709 if (mp
->mnt_syncer
== NULL
)
710 vfs_allocate_syncvnode(mp
);
711 mp
->mnt_kern_flag
&= ~(MNTK_UNMOUNT
| MNTK_UNMOUNTF
);
712 mp
->mnt_flag
|= async_flag
;
713 lockmgr(&mp
->mnt_lock
, LK_RELEASE
);
714 if (mp
->mnt_kern_flag
& MNTK_MWAIT
)
719 * Clean up any journals still associated with the mount after
720 * filesystem activity has ceased.
722 journal_remove_all_journals(mp
,
723 ((flags
& MNT_FORCE
) ? MC_JOURNAL_STOP_IMM
: 0));
725 mountlist_remove(mp
);
728 * Remove any installed vnode ops here so the individual VFSs don't
731 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_coherency_ops
);
732 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_journal_ops
);
733 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_norm_ops
);
734 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_spec_ops
);
735 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_fifo_ops
);
737 if (mp
->mnt_ncmountpt
.ncp
!= NULL
) {
738 nch
= mp
->mnt_ncmountpt
;
739 cache_zero(&mp
->mnt_ncmountpt
);
740 cache_clrmountpt(&nch
);
743 if (mp
->mnt_ncmounton
.ncp
!= NULL
) {
744 nch
= mp
->mnt_ncmounton
;
745 cache_zero(&mp
->mnt_ncmounton
);
746 cache_clrmountpt(&nch
);
750 mp
->mnt_vfc
->vfc_refcount
--;
751 if (!TAILQ_EMPTY(&mp
->mnt_nvnodelist
))
752 panic("unmount: dangling vnode");
753 lockmgr(&mp
->mnt_lock
, LK_RELEASE
);
754 if (mp
->mnt_kern_flag
& MNTK_MWAIT
)
763 mount_warning(struct mount
*mp
, const char *ctl
, ...)
770 if (cache_fullpath(NULL
, &mp
->mnt_ncmounton
, &ptr
, &buf
) == 0) {
771 kprintf("unmount(%s): ", ptr
);
776 kprintf("unmount(%p", mp
);
777 if (mp
->mnt_ncmounton
.ncp
&& mp
->mnt_ncmounton
.ncp
->nc_name
)
778 kprintf(",%s", mp
->mnt_ncmounton
.ncp
->nc_name
);
787 * Shim cache_fullpath() to handle the case where a process is chrooted into
788 * a subdirectory of a mount. In this case if the root mount matches the
789 * process root directory's mount we have to specify the process's root
790 * directory instead of the mount point, because the mount point might
791 * be above the root directory.
795 mount_path(struct proc
*p
, struct mount
*mp
, char **rb
, char **fb
)
797 struct nchandle
*nch
;
799 if (p
&& p
->p_fd
->fd_nrdir
.mount
== mp
)
800 nch
= &p
->p_fd
->fd_nrdir
;
802 nch
= &mp
->mnt_ncmountpt
;
803 return(cache_fullpath(p
, nch
, rb
, fb
));
807 * Sync each mounted filesystem.
811 static int syncprt
= 0;
812 SYSCTL_INT(_debug
, OID_AUTO
, syncprt
, CTLFLAG_RW
, &syncprt
, 0, "");
815 static int sync_callback(struct mount
*mp
, void *data
);
819 sys_sync(struct sync_args
*uap
)
821 mountlist_scan(sync_callback
, NULL
, MNTSCAN_FORWARD
);
824 * print out buffer pool stat information on each sync() call.
834 sync_callback(struct mount
*mp
, void *data __unused
)
838 if ((mp
->mnt_flag
& MNT_RDONLY
) == 0) {
839 asyncflag
= mp
->mnt_flag
& MNT_ASYNC
;
840 mp
->mnt_flag
&= ~MNT_ASYNC
;
841 vfs_msync(mp
, MNT_NOWAIT
);
842 VFS_SYNC(mp
, MNT_NOWAIT
);
843 mp
->mnt_flag
|= asyncflag
;
848 /* XXX PRISON: could be per prison flag */
849 static int prison_quotas
;
851 SYSCTL_INT(_kern_prison
, OID_AUTO
, quotas
, CTLFLAG_RW
, &prison_quotas
, 0, "");
855 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
857 * Change filesystem quotas.
861 sys_quotactl(struct quotactl_args
*uap
)
863 struct nlookupdata nd
;
871 if (p
->p_ucred
->cr_prison
&& !prison_quotas
)
874 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
876 error
= nlookup(&nd
);
878 mp
= nd
.nl_nch
.mount
;
879 error
= VFS_QUOTACTL(mp
, uap
->cmd
, uap
->uid
,
880 uap
->arg
, nd
.nl_cred
);
887 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
888 * void *buf, int buflen)
890 * This function operates on a mount point and executes the specified
891 * operation using the specified control data, and possibly returns data.
893 * The actual number of bytes stored in the result buffer is returned, 0
894 * if none, otherwise an error is returned.
898 sys_mountctl(struct mountctl_args
*uap
)
900 struct thread
*td
= curthread
;
901 struct proc
*p
= td
->td_proc
;
909 * Sanity and permissions checks. We must be root.
912 if (p
->p_ucred
->cr_prison
!= NULL
)
914 if ((error
= suser(td
)) != 0)
918 * Argument length checks
920 if (uap
->ctllen
< 0 || uap
->ctllen
> 1024)
922 if (uap
->buflen
< 0 || uap
->buflen
> 16 * 1024)
924 if (uap
->path
== NULL
)
928 * Allocate the necessary buffers and copyin data
930 path
= objcache_get(namei_oc
, M_WAITOK
);
931 error
= copyinstr(uap
->path
, path
, MAXPATHLEN
, NULL
);
936 ctl
= kmalloc(uap
->ctllen
+ 1, M_TEMP
, M_WAITOK
|M_ZERO
);
937 error
= copyin(uap
->ctl
, ctl
, uap
->ctllen
);
942 buf
= kmalloc(uap
->buflen
+ 1, M_TEMP
, M_WAITOK
|M_ZERO
);
945 * Validate the descriptor
948 fp
= holdfp(p
->p_fd
, uap
->fd
, -1);
958 * Execute the internal kernel function and clean up.
960 error
= kern_mountctl(path
, uap
->op
, fp
, ctl
, uap
->ctllen
, buf
, uap
->buflen
, &uap
->sysmsg_result
);
963 if (error
== 0 && uap
->sysmsg_result
> 0)
964 error
= copyout(buf
, uap
->buf
, uap
->sysmsg_result
);
967 objcache_put(namei_oc
, path
);
976 * Execute a mount control operation by resolving the path to a mount point
977 * and calling vop_mountctl().
979 * Use the mount point from the nch instead of the vnode so nullfs mounts
980 * can properly spike the VOP.
983 kern_mountctl(const char *path
, int op
, struct file
*fp
,
984 const void *ctl
, int ctllen
,
985 void *buf
, int buflen
, int *res
)
989 struct nlookupdata nd
;
994 error
= nlookup_init(&nd
, path
, UIO_SYSSPACE
, NLC_FOLLOW
);
996 error
= nlookup(&nd
);
998 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
999 mp
= nd
.nl_nch
.mount
;
1005 * Must be the root of the filesystem
1007 if ((vp
->v_flag
& (VROOT
|VPFSROOT
)) == 0) {
1011 error
= vop_mountctl(mp
->mnt_vn_use_ops
, op
, fp
, ctl
, ctllen
,
1018 kern_statfs(struct nlookupdata
*nd
, struct statfs
*buf
)
1020 struct thread
*td
= curthread
;
1021 struct proc
*p
= td
->td_proc
;
1024 char *fullpath
, *freepath
;
1027 if ((error
= nlookup(nd
)) != 0)
1029 mp
= nd
->nl_nch
.mount
;
1031 if ((error
= VFS_STATFS(mp
, sp
, nd
->nl_cred
)) != 0)
1034 error
= mount_path(p
, mp
, &fullpath
, &freepath
);
1037 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
1038 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
1039 kfree(freepath
, M_TEMP
);
1041 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
1042 bcopy(sp
, buf
, sizeof(*buf
));
1043 /* Only root should have access to the fsid's. */
1045 buf
->f_fsid
.val
[0] = buf
->f_fsid
.val
[1] = 0;
1050 * statfs_args(char *path, struct statfs *buf)
1052 * Get filesystem statistics.
1055 sys_statfs(struct statfs_args
*uap
)
1057 struct nlookupdata nd
;
1061 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1063 error
= kern_statfs(&nd
, &buf
);
1066 error
= copyout(&buf
, uap
->buf
, sizeof(*uap
->buf
));
1071 kern_fstatfs(int fd
, struct statfs
*buf
)
1073 struct thread
*td
= curthread
;
1074 struct proc
*p
= td
->td_proc
;
1078 char *fullpath
, *freepath
;
1082 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
1084 mp
= ((struct vnode
*)fp
->f_data
)->v_mount
;
1089 if (fp
->f_cred
== NULL
) {
1094 if ((error
= VFS_STATFS(mp
, sp
, fp
->f_cred
)) != 0)
1097 if ((error
= mount_path(p
, mp
, &fullpath
, &freepath
)) != 0)
1099 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
1100 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
1101 kfree(freepath
, M_TEMP
);
1103 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
1104 bcopy(sp
, buf
, sizeof(*buf
));
1106 /* Only root should have access to the fsid's. */
1108 buf
->f_fsid
.val
[0] = buf
->f_fsid
.val
[1] = 0;
1116 * fstatfs_args(int fd, struct statfs *buf)
1118 * Get filesystem statistics.
1121 sys_fstatfs(struct fstatfs_args
*uap
)
1126 error
= kern_fstatfs(uap
->fd
, &buf
);
1129 error
= copyout(&buf
, uap
->buf
, sizeof(*uap
->buf
));
1134 kern_statvfs(struct nlookupdata
*nd
, struct statvfs
*buf
)
1140 if ((error
= nlookup(nd
)) != 0)
1142 mp
= nd
->nl_nch
.mount
;
1143 sp
= &mp
->mnt_vstat
;
1144 if ((error
= VFS_STATVFS(mp
, sp
, nd
->nl_cred
)) != 0)
1148 if (mp
->mnt_flag
& MNT_RDONLY
)
1149 sp
->f_flag
|= ST_RDONLY
;
1150 if (mp
->mnt_flag
& MNT_NOSUID
)
1151 sp
->f_flag
|= ST_NOSUID
;
1152 bcopy(sp
, buf
, sizeof(*buf
));
1157 * statfs_args(char *path, struct statfs *buf)
1159 * Get filesystem statistics.
1162 sys_statvfs(struct statvfs_args
*uap
)
1164 struct nlookupdata nd
;
1168 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1170 error
= kern_statvfs(&nd
, &buf
);
1173 error
= copyout(&buf
, uap
->buf
, sizeof(*uap
->buf
));
1178 kern_fstatvfs(int fd
, struct statvfs
*buf
)
1180 struct thread
*td
= curthread
;
1181 struct proc
*p
= td
->td_proc
;
1188 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
1190 mp
= ((struct vnode
*)fp
->f_data
)->v_mount
;
1195 if (fp
->f_cred
== NULL
) {
1199 sp
= &mp
->mnt_vstat
;
1200 if ((error
= VFS_STATVFS(mp
, sp
, fp
->f_cred
)) != 0)
1204 if (mp
->mnt_flag
& MNT_RDONLY
)
1205 sp
->f_flag
|= ST_RDONLY
;
1206 if (mp
->mnt_flag
& MNT_NOSUID
)
1207 sp
->f_flag
|= ST_NOSUID
;
1209 bcopy(sp
, buf
, sizeof(*buf
));
1217 * fstatfs_args(int fd, struct statfs *buf)
1219 * Get filesystem statistics.
1222 sys_fstatvfs(struct fstatvfs_args
*uap
)
1227 error
= kern_fstatvfs(uap
->fd
, &buf
);
1230 error
= copyout(&buf
, uap
->buf
, sizeof(*uap
->buf
));
1235 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
1237 * Get statistics on all filesystems.
1240 struct getfsstat_info
{
1241 struct statfs
*sfsp
;
1249 static int getfsstat_callback(struct mount
*, void *);
1253 sys_getfsstat(struct getfsstat_args
*uap
)
1255 struct thread
*td
= curthread
;
1256 struct proc
*p
= td
->td_proc
;
1257 struct getfsstat_info info
;
1259 bzero(&info
, sizeof(info
));
1261 info
.maxcount
= uap
->bufsize
/ sizeof(struct statfs
);
1262 info
.sfsp
= uap
->buf
;
1264 info
.flags
= uap
->flags
;
1267 mountlist_scan(getfsstat_callback
, &info
, MNTSCAN_FORWARD
);
1268 if (info
.sfsp
&& info
.count
> info
.maxcount
)
1269 uap
->sysmsg_result
= info
.maxcount
;
1271 uap
->sysmsg_result
= info
.count
;
1272 return (info
.error
);
1276 getfsstat_callback(struct mount
*mp
, void *data
)
1278 struct getfsstat_info
*info
= data
;
1284 if (info
->sfsp
&& info
->count
< info
->maxcount
) {
1285 if (info
->p
&& !chroot_visible_mnt(mp
, info
->p
))
1290 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1291 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1292 * overrides MNT_WAIT.
1294 if (((info
->flags
& (MNT_LAZY
|MNT_NOWAIT
)) == 0 ||
1295 (info
->flags
& MNT_WAIT
)) &&
1296 (error
= VFS_STATFS(mp
, sp
, info
->p
->p_ucred
))) {
1299 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
1301 error
= mount_path(info
->p
, mp
, &fullpath
, &freepath
);
1303 info
->error
= error
;
1306 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
1307 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
1308 kfree(freepath
, M_TEMP
);
1310 error
= copyout(sp
, info
->sfsp
, sizeof(*sp
));
1312 info
->error
= error
;
1322 * getvfsstat_args(struct statfs *buf, struct statvfs *vbuf,
1323 long bufsize, int flags)
1325 * Get statistics on all filesystems.
1328 struct getvfsstat_info
{
1329 struct statfs
*sfsp
;
1330 struct statvfs
*vsfsp
;
1338 static int getvfsstat_callback(struct mount
*, void *);
1342 sys_getvfsstat(struct getvfsstat_args
*uap
)
1344 struct thread
*td
= curthread
;
1345 struct proc
*p
= td
->td_proc
;
1346 struct getvfsstat_info info
;
1348 bzero(&info
, sizeof(info
));
1350 info
.maxcount
= uap
->vbufsize
/ sizeof(struct statvfs
);
1351 info
.sfsp
= uap
->buf
;
1352 info
.vsfsp
= uap
->vbuf
;
1354 info
.flags
= uap
->flags
;
1357 mountlist_scan(getvfsstat_callback
, &info
, MNTSCAN_FORWARD
);
1358 if (info
.vsfsp
&& info
.count
> info
.maxcount
)
1359 uap
->sysmsg_result
= info
.maxcount
;
1361 uap
->sysmsg_result
= info
.count
;
1362 return (info
.error
);
1366 getvfsstat_callback(struct mount
*mp
, void *data
)
1368 struct getvfsstat_info
*info
= data
;
1370 struct statvfs
*vsp
;
1375 if (info
->vsfsp
&& info
->count
< info
->maxcount
) {
1376 if (info
->p
&& !chroot_visible_mnt(mp
, info
->p
))
1379 vsp
= &mp
->mnt_vstat
;
1382 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1383 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1384 * overrides MNT_WAIT.
1386 if (((info
->flags
& (MNT_LAZY
|MNT_NOWAIT
)) == 0 ||
1387 (info
->flags
& MNT_WAIT
)) &&
1388 (error
= VFS_STATFS(mp
, sp
, info
->p
->p_ucred
))) {
1391 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
1393 if (((info
->flags
& (MNT_LAZY
|MNT_NOWAIT
)) == 0 ||
1394 (info
->flags
& MNT_WAIT
)) &&
1395 (error
= VFS_STATVFS(mp
, vsp
, info
->p
->p_ucred
))) {
1399 if (mp
->mnt_flag
& MNT_RDONLY
)
1400 vsp
->f_flag
|= ST_RDONLY
;
1401 if (mp
->mnt_flag
& MNT_NOSUID
)
1402 vsp
->f_flag
|= ST_NOSUID
;
1404 error
= mount_path(info
->p
, mp
, &fullpath
, &freepath
);
1406 info
->error
= error
;
1409 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
1410 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
1411 kfree(freepath
, M_TEMP
);
1413 error
= copyout(sp
, info
->sfsp
, sizeof(*sp
));
1415 error
= copyout(vsp
, info
->vsfsp
, sizeof(*vsp
));
1417 info
->error
= error
;
1429 * fchdir_args(int fd)
1431 * Change current working directory to a given file descriptor.
1435 sys_fchdir(struct fchdir_args
*uap
)
1437 struct thread
*td
= curthread
;
1438 struct proc
*p
= td
->td_proc
;
1439 struct filedesc
*fdp
= p
->p_fd
;
1440 struct vnode
*vp
, *ovp
;
1443 struct nchandle nch
, onch
, tnch
;
1446 if ((error
= holdvnode(fdp
, uap
->fd
, &fp
)) != 0)
1448 vp
= (struct vnode
*)fp
->f_data
;
1450 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
1451 if (vp
->v_type
!= VDIR
|| fp
->f_nchandle
.ncp
== NULL
)
1454 error
= VOP_ACCESS(vp
, VEXEC
, p
->p_ucred
);
1460 cache_copy(&fp
->f_nchandle
, &nch
);
1463 * If the ncp has become a mount point, traverse through
1467 while (!error
&& (nch
.ncp
->nc_flag
& NCF_ISMOUNTPT
) &&
1468 (mp
= cache_findmount(&nch
)) != NULL
1470 error
= nlookup_mp(mp
, &tnch
);
1472 cache_unlock(&tnch
); /* leave ref intact */
1474 vp
= tnch
.ncp
->nc_vp
;
1475 error
= vget(vp
, LK_SHARED
);
1476 KKASSERT(error
== 0);
1483 onch
= fdp
->fd_ncdir
;
1484 vn_unlock(vp
); /* leave ref intact */
1486 fdp
->fd_ncdir
= nch
;
1498 kern_chdir(struct nlookupdata
*nd
)
1500 struct thread
*td
= curthread
;
1501 struct proc
*p
= td
->td_proc
;
1502 struct filedesc
*fdp
= p
->p_fd
;
1503 struct vnode
*vp
, *ovp
;
1504 struct nchandle onch
;
1507 if ((error
= nlookup(nd
)) != 0)
1509 if ((vp
= nd
->nl_nch
.ncp
->nc_vp
) == NULL
)
1511 if ((error
= vget(vp
, LK_SHARED
)) != 0)
1514 error
= checkvp_chdir(vp
, td
);
1518 onch
= fdp
->fd_ncdir
;
1519 cache_unlock(&nd
->nl_nch
); /* leave reference intact */
1520 fdp
->fd_ncdir
= nd
->nl_nch
;
1524 cache_zero(&nd
->nl_nch
);
1532 * chdir_args(char *path)
1534 * Change current working directory (``.'').
1537 sys_chdir(struct chdir_args
*uap
)
1539 struct nlookupdata nd
;
1542 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1544 error
= kern_chdir(&nd
);
1550 * Helper function for raised chroot(2) security function: Refuse if
1551 * any filedescriptors are open directories.
1554 chroot_refuse_vdir_fds(struct filedesc
*fdp
)
1561 for (fd
= 0; fd
< fdp
->fd_nfiles
; fd
++) {
1562 if ((error
= holdvnode(fdp
, fd
, &fp
)) != 0)
1564 vp
= (struct vnode
*)fp
->f_data
;
1565 if (vp
->v_type
!= VDIR
) {
1576 * This sysctl determines if we will allow a process to chroot(2) if it
1577 * has a directory open:
1578 * 0: disallowed for all processes.
1579 * 1: allowed for processes that were not already chroot(2)'ed.
1580 * 2: allowed for all processes.
1583 static int chroot_allow_open_directories
= 1;
1585 SYSCTL_INT(_kern
, OID_AUTO
, chroot_allow_open_directories
, CTLFLAG_RW
,
1586 &chroot_allow_open_directories
, 0, "");
1589 * chroot to the specified namecache entry. We obtain the vp from the
1590 * namecache data. The passed ncp must be locked and referenced and will
1591 * remain locked and referenced on return.
1594 kern_chroot(struct nchandle
*nch
)
1596 struct thread
*td
= curthread
;
1597 struct proc
*p
= td
->td_proc
;
1598 struct filedesc
*fdp
= p
->p_fd
;
1603 * Only root can chroot
1605 if ((error
= suser_cred(p
->p_ucred
, PRISON_ROOT
)) != 0)
1609 * Disallow open directory descriptors (fchdir() breakouts).
1611 if (chroot_allow_open_directories
== 0 ||
1612 (chroot_allow_open_directories
== 1 && fdp
->fd_rdir
!= rootvnode
)) {
1613 if ((error
= chroot_refuse_vdir_fds(fdp
)) != 0)
1616 if ((vp
= nch
->ncp
->nc_vp
) == NULL
)
1619 if ((error
= vget(vp
, LK_SHARED
)) != 0)
1623 * Check the validity of vp as a directory to change to and
1624 * associate it with rdir/jdir.
1626 error
= checkvp_chdir(vp
, td
);
1627 vn_unlock(vp
); /* leave reference intact */
1629 vrele(fdp
->fd_rdir
);
1630 fdp
->fd_rdir
= vp
; /* reference inherited by fd_rdir */
1631 cache_drop(&fdp
->fd_nrdir
);
1632 cache_copy(nch
, &fdp
->fd_nrdir
);
1633 if (fdp
->fd_jdir
== NULL
) {
1636 cache_copy(nch
, &fdp
->fd_njdir
);
1645 * chroot_args(char *path)
1647 * Change notion of root (``/'') directory.
1651 sys_chroot(struct chroot_args
*uap
)
1653 struct thread
*td
= curthread
;
1654 struct nlookupdata nd
;
1657 KKASSERT(td
->td_proc
);
1658 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1663 error
= nlookup(&nd
);
1665 error
= kern_chroot(&nd
.nl_nch
);
1671 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1672 * determine whether it is legal to chdir to the vnode. The vnode's state
1673 * is not changed by this call.
1676 checkvp_chdir(struct vnode
*vp
, struct thread
*td
)
1680 if (vp
->v_type
!= VDIR
)
1683 error
= VOP_ACCESS(vp
, VEXEC
, td
->td_proc
->p_ucred
);
1688 kern_open(struct nlookupdata
*nd
, int oflags
, int mode
, int *res
)
1690 struct thread
*td
= curthread
;
1691 struct proc
*p
= td
->td_proc
;
1692 struct lwp
*lp
= td
->td_lwp
;
1693 struct filedesc
*fdp
= p
->p_fd
;
1698 int type
, indx
, error
;
1701 if ((oflags
& O_ACCMODE
) == O_ACCMODE
)
1703 flags
= FFLAGS(oflags
);
1704 error
= falloc(p
, &nfp
, NULL
);
1708 cmode
= ((mode
&~ fdp
->fd_cmask
) & ALLPERMS
) &~ S_ISTXT
;
1711 * XXX p_dupfd is a real mess. It allows a device to return a
1712 * file descriptor to be duplicated rather then doing the open
1718 * Call vn_open() to do the lookup and assign the vnode to the
1719 * file pointer. vn_open() does not change the ref count on fp
1720 * and the vnode, on success, will be inherited by the file pointer
1723 nd
->nl_flags
|= NLC_LOCKVP
;
1724 error
= vn_open(nd
, fp
, flags
, cmode
);
1728 * handle special fdopen() case. bleh. dupfdopen() is
1729 * responsible for dropping the old contents of ofiles[indx]
1732 * Note that fsetfd() will add a ref to fp which represents
1733 * the fd_files[] assignment. We must still drop our
1736 if ((error
== ENODEV
|| error
== ENXIO
) && lp
->lwp_dupfd
>= 0) {
1737 if (fdalloc(p
, 0, &indx
) == 0) {
1738 error
= dupfdopen(p
, indx
, lp
->lwp_dupfd
, flags
, error
);
1741 fdrop(fp
); /* our ref */
1744 fsetfd(p
, NULL
, indx
);
1747 fdrop(fp
); /* our ref */
1748 if (error
== ERESTART
)
1754 * ref the vnode for ourselves so it can't be ripped out from under
1755 * is. XXX need an ND flag to request that the vnode be returned
1758 * Reserve a file descriptor but do not assign it until the open
1761 vp
= (struct vnode
*)fp
->f_data
;
1763 if ((error
= fdalloc(p
, 0, &indx
)) != 0) {
1770 * If no error occurs the vp will have been assigned to the file
1775 if (flags
& (O_EXLOCK
| O_SHLOCK
)) {
1776 lf
.l_whence
= SEEK_SET
;
1779 if (flags
& O_EXLOCK
)
1780 lf
.l_type
= F_WRLCK
;
1782 lf
.l_type
= F_RDLCK
;
1783 if (flags
& FNONBLOCK
)
1788 if ((error
= VOP_ADVLOCK(vp
, (caddr_t
)fp
, F_SETLK
, &lf
, type
)) != 0) {
1790 * lock request failed. Clean up the reserved
1794 fsetfd(p
, NULL
, indx
);
1798 fp
->f_flag
|= FHASLOCK
;
1802 * Assert that all regular file vnodes were created with a object.
1804 KASSERT(vp
->v_type
!= VREG
|| vp
->v_object
!= NULL
,
1805 ("open: regular file has no backing object after vn_open"));
1811 * release our private reference, leaving the one associated with the
1812 * descriptor table intact.
1814 fsetfd(p
, fp
, indx
);
1821 * open_args(char *path, int flags, int mode)
1823 * Check permissions, allocate an open file structure,
1824 * and call the device open routine if any.
1827 sys_open(struct open_args
*uap
)
1829 struct nlookupdata nd
;
1832 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1834 error
= kern_open(&nd
, uap
->flags
,
1835 uap
->mode
, &uap
->sysmsg_result
);
1842 kern_mknod(struct nlookupdata
*nd
, int mode
, int rmajor
, int rminor
)
1844 struct thread
*td
= curthread
;
1845 struct proc
*p
= td
->td_proc
;
1853 switch (mode
& S_IFMT
) {
1859 error
= suser_cred(p
->p_ucred
, PRISON_ROOT
);
1866 nd
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
1867 if ((error
= nlookup(nd
)) != 0)
1869 if (nd
->nl_nch
.ncp
->nc_vp
)
1871 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
1875 vattr
.va_mode
= (mode
& ALLPERMS
) &~ p
->p_fd
->fd_cmask
;
1876 vattr
.va_rmajor
= rmajor
;
1877 vattr
.va_rminor
= rminor
;
1880 switch (mode
& S_IFMT
) {
1881 case S_IFMT
: /* used by badsect to flag bad sectors */
1882 vattr
.va_type
= VBAD
;
1885 vattr
.va_type
= VCHR
;
1888 vattr
.va_type
= VBLK
;
1894 /* special directories support for HAMMER */
1895 vattr
.va_type
= VDIR
;
1903 error
= VOP_NWHITEOUT(&nd
->nl_nch
, nd
->nl_dvp
,
1904 nd
->nl_cred
, NAMEI_CREATE
);
1907 error
= VOP_NMKNOD(&nd
->nl_nch
, nd
->nl_dvp
,
1908 &vp
, nd
->nl_cred
, &vattr
);
1917 * mknod_args(char *path, int mode, int dev)
1919 * Create a special file.
1922 sys_mknod(struct mknod_args
*uap
)
1924 struct nlookupdata nd
;
1927 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
1929 error
= kern_mknod(&nd
, uap
->mode
,
1930 umajor(uap
->dev
), uminor(uap
->dev
));
1937 kern_mkfifo(struct nlookupdata
*nd
, int mode
)
1939 struct thread
*td
= curthread
;
1940 struct proc
*p
= td
->td_proc
;
1947 nd
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
1948 if ((error
= nlookup(nd
)) != 0)
1950 if (nd
->nl_nch
.ncp
->nc_vp
)
1952 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
1956 vattr
.va_type
= VFIFO
;
1957 vattr
.va_mode
= (mode
& ALLPERMS
) &~ p
->p_fd
->fd_cmask
;
1959 error
= VOP_NMKNOD(&nd
->nl_nch
, nd
->nl_dvp
, &vp
, nd
->nl_cred
, &vattr
);
1966 * mkfifo_args(char *path, int mode)
1968 * Create a named pipe.
1971 sys_mkfifo(struct mkfifo_args
*uap
)
1973 struct nlookupdata nd
;
1976 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
1978 error
= kern_mkfifo(&nd
, uap
->mode
);
1983 static int hardlink_check_uid
= 0;
1984 SYSCTL_INT(_security
, OID_AUTO
, hardlink_check_uid
, CTLFLAG_RW
,
1985 &hardlink_check_uid
, 0,
1986 "Unprivileged processes cannot create hard links to files owned by other "
1988 static int hardlink_check_gid
= 0;
1989 SYSCTL_INT(_security
, OID_AUTO
, hardlink_check_gid
, CTLFLAG_RW
,
1990 &hardlink_check_gid
, 0,
1991 "Unprivileged processes cannot create hard links to files owned by other "
1995 can_hardlink(struct vnode
*vp
, struct thread
*td
, struct ucred
*cred
)
2001 * Shortcut if disabled
2003 if (hardlink_check_uid
== 0 && hardlink_check_gid
== 0)
2007 * root cred can always hardlink
2009 if (suser_cred(cred
, PRISON_ROOT
) == 0)
2013 * Otherwise only if the originating file is owned by the
2014 * same user or group. Note that any group is allowed if
2015 * the file is owned by the caller.
2017 error
= VOP_GETATTR(vp
, &va
);
2021 if (hardlink_check_uid
) {
2022 if (cred
->cr_uid
!= va
.va_uid
)
2026 if (hardlink_check_gid
) {
2027 if (cred
->cr_uid
!= va
.va_uid
&& !groupmember(va
.va_gid
, cred
))
2035 kern_link(struct nlookupdata
*nd
, struct nlookupdata
*linknd
)
2037 struct thread
*td
= curthread
;
2042 * Lookup the source and obtained a locked vnode.
2044 * XXX relookup on vget failure / race ?
2047 if ((error
= nlookup(nd
)) != 0)
2049 vp
= nd
->nl_nch
.ncp
->nc_vp
;
2050 KKASSERT(vp
!= NULL
);
2051 if (vp
->v_type
== VDIR
)
2052 return (EPERM
); /* POSIX */
2053 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2055 if ((error
= vget(vp
, LK_EXCLUSIVE
)) != 0)
2059 * Unlock the source so we can lookup the target without deadlocking
2060 * (XXX vp is locked already, possible other deadlock?). The target
2063 KKASSERT(nd
->nl_flags
& NLC_NCPISLOCKED
);
2064 nd
->nl_flags
&= ~NLC_NCPISLOCKED
;
2065 cache_unlock(&nd
->nl_nch
);
2067 linknd
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
2068 if ((error
= nlookup(linknd
)) != 0) {
2072 if (linknd
->nl_nch
.ncp
->nc_vp
) {
2078 * Finally run the new API VOP.
2080 error
= can_hardlink(vp
, td
, td
->td_proc
->p_ucred
);
2082 error
= VOP_NLINK(&linknd
->nl_nch
, linknd
->nl_dvp
,
2083 vp
, linknd
->nl_cred
);
2090 * link_args(char *path, char *link)
2092 * Make a hard file link.
2095 sys_link(struct link_args
*uap
)
2097 struct nlookupdata nd
, linknd
;
2100 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2102 error
= nlookup_init(&linknd
, uap
->link
, UIO_USERSPACE
, 0);
2104 error
= kern_link(&nd
, &linknd
);
2105 nlookup_done(&linknd
);
2112 kern_symlink(struct nlookupdata
*nd
, char *path
, int mode
)
2120 nd
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
2121 if ((error
= nlookup(nd
)) != 0)
2123 if (nd
->nl_nch
.ncp
->nc_vp
)
2125 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2129 vattr
.va_mode
= mode
;
2130 error
= VOP_NSYMLINK(&nd
->nl_nch
, dvp
, &vp
, nd
->nl_cred
, &vattr
, path
);
2137 * symlink(char *path, char *link)
2139 * Make a symbolic link.
2142 sys_symlink(struct symlink_args
*uap
)
2144 struct thread
*td
= curthread
;
2145 struct nlookupdata nd
;
2150 path
= objcache_get(namei_oc
, M_WAITOK
);
2151 error
= copyinstr(uap
->path
, path
, MAXPATHLEN
, NULL
);
2153 error
= nlookup_init(&nd
, uap
->link
, UIO_USERSPACE
, 0);
2155 mode
= ACCESSPERMS
& ~td
->td_proc
->p_fd
->fd_cmask
;
2156 error
= kern_symlink(&nd
, path
, mode
);
2160 objcache_put(namei_oc
, path
);
2165 * undelete_args(char *path)
2167 * Delete a whiteout from the filesystem.
2171 sys_undelete(struct undelete_args
*uap
)
2173 struct nlookupdata nd
;
2176 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2178 nd
.nl_flags
|= NLC_DELETE
| NLC_REFDVP
;
2180 error
= nlookup(&nd
);
2182 error
= ncp_writechk(&nd
.nl_nch
);
2184 error
= VOP_NWHITEOUT(&nd
.nl_nch
, nd
.nl_dvp
, nd
.nl_cred
,
2192 kern_unlink(struct nlookupdata
*nd
)
2197 nd
->nl_flags
|= NLC_DELETE
| NLC_REFDVP
;
2198 if ((error
= nlookup(nd
)) != 0)
2200 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2202 error
= VOP_NREMOVE(&nd
->nl_nch
, nd
->nl_dvp
, nd
->nl_cred
);
2207 * unlink_args(char *path)
2209 * Delete a name from the filesystem.
2212 sys_unlink(struct unlink_args
*uap
)
2214 struct nlookupdata nd
;
2217 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2219 error
= kern_unlink(&nd
);
2225 kern_lseek(int fd
, off_t offset
, int whence
, off_t
*res
)
2227 struct thread
*td
= curthread
;
2228 struct proc
*p
= td
->td_proc
;
2235 fp
= holdfp(p
->p_fd
, fd
, -1);
2238 if (fp
->f_type
!= DTYPE_VNODE
) {
2242 vp
= (struct vnode
*)fp
->f_data
;
2246 new_offset
= fp
->f_offset
+ offset
;
2250 error
= VOP_GETATTR(vp
, &vattr
);
2251 new_offset
= offset
+ vattr
.va_size
;
2254 new_offset
= offset
;
2264 * Validate the seek position. Negative offsets are not allowed
2265 * for regular files, block specials, or directories.
2268 if (new_offset
< 0 &&
2269 (vp
->v_type
== VREG
|| vp
->v_type
== VDIR
||
2270 vp
->v_type
== VCHR
|| vp
->v_type
== VBLK
)) {
2273 fp
->f_offset
= new_offset
;
2276 *res
= fp
->f_offset
;
2283 * lseek_args(int fd, int pad, off_t offset, int whence)
2285 * Reposition read/write file offset.
2288 sys_lseek(struct lseek_args
*uap
)
2292 error
= kern_lseek(uap
->fd
, uap
->offset
, uap
->whence
,
2293 &uap
->sysmsg_offset
);
2299 kern_access(struct nlookupdata
*nd
, int aflags
)
2304 if ((error
= nlookup(nd
)) != 0)
2307 error
= cache_vget(&nd
->nl_nch
, nd
->nl_cred
, LK_EXCLUSIVE
, &vp
);
2311 /* Flags == 0 means only check for existence. */
2320 if ((flags
& VWRITE
) == 0 ||
2321 (error
= vn_writechk(vp
, &nd
->nl_nch
)) == 0)
2322 error
= VOP_ACCESS(vp
, flags
, nd
->nl_cred
);
2325 * If the file handle is stale we have to re-resolve the
2326 * entry. This is a hack at the moment.
2328 if (error
== ESTALE
) {
2330 cache_setunresolved(&nd
->nl_nch
);
2331 error
= cache_resolve(&nd
->nl_nch
, nd
->nl_cred
);
2344 * access_args(char *path, int flags)
2346 * Check access permissions.
2349 sys_access(struct access_args
*uap
)
2351 struct nlookupdata nd
;
2354 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2356 error
= kern_access(&nd
, uap
->flags
);
2362 kern_stat(struct nlookupdata
*nd
, struct stat
*st
)
2368 if ((error
= nlookup(nd
)) != 0)
2371 if ((vp
= nd
->nl_nch
.ncp
->nc_vp
) == NULL
)
2375 if ((error
= vget(vp
, LK_SHARED
)) != 0)
2377 error
= vn_stat(vp
, st
, nd
->nl_cred
);
2380 * If the file handle is stale we have to re-resolve the entry. This
2381 * is a hack at the moment.
2383 if (error
== ESTALE
) {
2385 cache_setunresolved(&nd
->nl_nch
);
2386 error
= cache_resolve(&nd
->nl_nch
, nd
->nl_cred
);
2396 * stat_args(char *path, struct stat *ub)
2398 * Get file status; this version follows links.
2401 sys_stat(struct stat_args
*uap
)
2403 struct nlookupdata nd
;
2407 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2409 error
= kern_stat(&nd
, &st
);
2411 error
= copyout(&st
, uap
->ub
, sizeof(*uap
->ub
));
2418 * lstat_args(char *path, struct stat *ub)
2420 * Get file status; this version does not follow links.
2423 sys_lstat(struct lstat_args
*uap
)
2425 struct nlookupdata nd
;
2429 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2431 error
= kern_stat(&nd
, &st
);
2433 error
= copyout(&st
, uap
->ub
, sizeof(*uap
->ub
));
2440 * pathconf_Args(char *path, int name)
2442 * Get configurable pathname variables.
2446 sys_pathconf(struct pathconf_args
*uap
)
2448 struct nlookupdata nd
;
2453 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2455 error
= nlookup(&nd
);
2457 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
2460 error
= VOP_PATHCONF(vp
, uap
->name
, uap
->sysmsg_fds
);
2468 * kern_readlink isn't properly split yet. There is a copyin burried
2469 * in VOP_READLINK().
2472 kern_readlink(struct nlookupdata
*nd
, char *buf
, int count
, int *res
)
2474 struct thread
*td
= curthread
;
2475 struct proc
*p
= td
->td_proc
;
2481 if ((error
= nlookup(nd
)) != 0)
2483 error
= cache_vget(&nd
->nl_nch
, nd
->nl_cred
, LK_EXCLUSIVE
, &vp
);
2486 if (vp
->v_type
!= VLNK
) {
2489 aiov
.iov_base
= buf
;
2490 aiov
.iov_len
= count
;
2491 auio
.uio_iov
= &aiov
;
2492 auio
.uio_iovcnt
= 1;
2493 auio
.uio_offset
= 0;
2494 auio
.uio_rw
= UIO_READ
;
2495 auio
.uio_segflg
= UIO_USERSPACE
;
2497 auio
.uio_resid
= count
;
2498 error
= VOP_READLINK(vp
, &auio
, p
->p_ucred
);
2501 *res
= count
- auio
.uio_resid
;
2506 * readlink_args(char *path, char *buf, int count)
2508 * Return target name of a symbolic link.
2511 sys_readlink(struct readlink_args
*uap
)
2513 struct nlookupdata nd
;
2516 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2518 error
= kern_readlink(&nd
, uap
->buf
, uap
->count
,
2519 &uap
->sysmsg_result
);
2526 setfflags(struct vnode
*vp
, int flags
)
2528 struct thread
*td
= curthread
;
2529 struct proc
*p
= td
->td_proc
;
2534 * Prevent non-root users from setting flags on devices. When
2535 * a device is reused, users can retain ownership of the device
2536 * if they are allowed to set flags and programs assume that
2537 * chown can't fail when done as root.
2539 if ((vp
->v_type
== VCHR
|| vp
->v_type
== VBLK
) &&
2540 ((error
= suser_cred(p
->p_ucred
, PRISON_ROOT
)) != 0))
2544 * note: vget is required for any operation that might mod the vnode
2545 * so VINACTIVE is properly cleared.
2547 if ((error
= vget(vp
, LK_EXCLUSIVE
)) == 0) {
2549 vattr
.va_flags
= flags
;
2550 error
= VOP_SETATTR(vp
, &vattr
, p
->p_ucred
);
2557 * chflags(char *path, int flags)
2559 * Change flags of a file given a path name.
2563 sys_chflags(struct chflags_args
*uap
)
2565 struct nlookupdata nd
;
2570 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2571 /* XXX Add NLC flag indicating modifying operation? */
2573 error
= nlookup(&nd
);
2575 error
= ncp_writechk(&nd
.nl_nch
);
2577 error
= cache_vref(&nd
.nl_nch
, nd
.nl_cred
, &vp
);
2580 error
= setfflags(vp
, uap
->flags
);
2587 * fchflags_args(int fd, int flags)
2589 * Change flags of a file given a file descriptor.
2593 sys_fchflags(struct fchflags_args
*uap
)
2595 struct thread
*td
= curthread
;
2596 struct proc
*p
= td
->td_proc
;
2600 if ((error
= holdvnode(p
->p_fd
, uap
->fd
, &fp
)) != 0)
2602 if (fp
->f_nchandle
.ncp
)
2603 error
= ncp_writechk(&fp
->f_nchandle
);
2605 error
= setfflags((struct vnode
*) fp
->f_data
, uap
->flags
);
2611 setfmode(struct vnode
*vp
, int mode
)
2613 struct thread
*td
= curthread
;
2614 struct proc
*p
= td
->td_proc
;
2619 * note: vget is required for any operation that might mod the vnode
2620 * so VINACTIVE is properly cleared.
2622 if ((error
= vget(vp
, LK_EXCLUSIVE
)) == 0) {
2624 vattr
.va_mode
= mode
& ALLPERMS
;
2625 error
= VOP_SETATTR(vp
, &vattr
, p
->p_ucred
);
2632 kern_chmod(struct nlookupdata
*nd
, int mode
)
2637 /* XXX Add NLC flag indicating modifying operation? */
2638 if ((error
= nlookup(nd
)) != 0)
2640 if ((error
= cache_vref(&nd
->nl_nch
, nd
->nl_cred
, &vp
)) != 0)
2642 if ((error
= ncp_writechk(&nd
->nl_nch
)) == 0)
2643 error
= setfmode(vp
, mode
);
2649 * chmod_args(char *path, int mode)
2651 * Change mode of a file given path name.
2655 sys_chmod(struct chmod_args
*uap
)
2657 struct nlookupdata nd
;
2660 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2662 error
= kern_chmod(&nd
, uap
->mode
);
2668 * lchmod_args(char *path, int mode)
2670 * Change mode of a file given path name (don't follow links.)
2674 sys_lchmod(struct lchmod_args
*uap
)
2676 struct nlookupdata nd
;
2679 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2681 error
= kern_chmod(&nd
, uap
->mode
);
2687 * fchmod_args(int fd, int mode)
2689 * Change mode of a file given a file descriptor.
2693 sys_fchmod(struct fchmod_args
*uap
)
2695 struct thread
*td
= curthread
;
2696 struct proc
*p
= td
->td_proc
;
2700 if ((error
= holdvnode(p
->p_fd
, uap
->fd
, &fp
)) != 0)
2702 if (fp
->f_nchandle
.ncp
)
2703 error
= ncp_writechk(&fp
->f_nchandle
);
2705 error
= setfmode((struct vnode
*)fp
->f_data
, uap
->mode
);
2711 setfown(struct vnode
*vp
, uid_t uid
, gid_t gid
)
2713 struct thread
*td
= curthread
;
2714 struct proc
*p
= td
->td_proc
;
2719 * note: vget is required for any operation that might mod the vnode
2720 * so VINACTIVE is properly cleared.
2722 if ((error
= vget(vp
, LK_EXCLUSIVE
)) == 0) {
2726 error
= VOP_SETATTR(vp
, &vattr
, p
->p_ucred
);
2733 kern_chown(struct nlookupdata
*nd
, int uid
, int gid
)
2738 /* XXX Add NLC flag indicating modifying operation? */
2739 if ((error
= nlookup(nd
)) != 0)
2741 if ((error
= cache_vref(&nd
->nl_nch
, nd
->nl_cred
, &vp
)) != 0)
2743 if ((error
= ncp_writechk(&nd
->nl_nch
)) == 0)
2744 error
= setfown(vp
, uid
, gid
);
2750 * chown(char *path, int uid, int gid)
2752 * Set ownership given a path name.
2755 sys_chown(struct chown_args
*uap
)
2757 struct nlookupdata nd
;
2760 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2762 error
= kern_chown(&nd
, uap
->uid
, uap
->gid
);
2768 * lchown_args(char *path, int uid, int gid)
2770 * Set ownership given a path name, do not cross symlinks.
2773 sys_lchown(struct lchown_args
*uap
)
2775 struct nlookupdata nd
;
2778 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2780 error
= kern_chown(&nd
, uap
->uid
, uap
->gid
);
2786 * fchown_args(int fd, int uid, int gid)
2788 * Set ownership given a file descriptor.
2792 sys_fchown(struct fchown_args
*uap
)
2794 struct thread
*td
= curthread
;
2795 struct proc
*p
= td
->td_proc
;
2799 if ((error
= holdvnode(p
->p_fd
, uap
->fd
, &fp
)) != 0)
2801 if (fp
->f_nchandle
.ncp
)
2802 error
= ncp_writechk(&fp
->f_nchandle
);
2804 error
= setfown((struct vnode
*)fp
->f_data
, uap
->uid
, uap
->gid
);
2810 getutimes(const struct timeval
*tvp
, struct timespec
*tsp
)
2812 struct timeval tv
[2];
2816 TIMEVAL_TO_TIMESPEC(&tv
[0], &tsp
[0]);
2819 TIMEVAL_TO_TIMESPEC(&tvp
[0], &tsp
[0]);
2820 TIMEVAL_TO_TIMESPEC(&tvp
[1], &tsp
[1]);
2826 setutimes(struct vnode
*vp
, const struct timespec
*ts
, int nullflag
)
2828 struct thread
*td
= curthread
;
2829 struct proc
*p
= td
->td_proc
;
2834 * note: vget is required for any operation that might mod the vnode
2835 * so VINACTIVE is properly cleared.
2837 if ((error
= vget(vp
, LK_EXCLUSIVE
)) == 0) {
2839 vattr
.va_atime
= ts
[0];
2840 vattr
.va_mtime
= ts
[1];
2842 vattr
.va_vaflags
|= VA_UTIMES_NULL
;
2843 error
= VOP_SETATTR(vp
, &vattr
, p
->p_ucred
);
2850 kern_utimes(struct nlookupdata
*nd
, struct timeval
*tptr
)
2852 struct timespec ts
[2];
2856 if ((error
= getutimes(tptr
, ts
)) != 0)
2858 /* XXX Add NLC flag indicating modifying operation? */
2859 if ((error
= nlookup(nd
)) != 0)
2861 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2863 if ((error
= cache_vref(&nd
->nl_nch
, nd
->nl_cred
, &vp
)) != 0)
2865 error
= setutimes(vp
, ts
, tptr
== NULL
);
2871 * utimes_args(char *path, struct timeval *tptr)
2873 * Set the access and modification times of a file.
2876 sys_utimes(struct utimes_args
*uap
)
2878 struct timeval tv
[2];
2879 struct nlookupdata nd
;
2883 error
= copyin(uap
->tptr
, tv
, sizeof(tv
));
2887 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2889 error
= kern_utimes(&nd
, uap
->tptr
? tv
: NULL
);
2895 * lutimes_args(char *path, struct timeval *tptr)
2897 * Set the access and modification times of a file.
2900 sys_lutimes(struct lutimes_args
*uap
)
2902 struct timeval tv
[2];
2903 struct nlookupdata nd
;
2907 error
= copyin(uap
->tptr
, tv
, sizeof(tv
));
2911 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2913 error
= kern_utimes(&nd
, uap
->tptr
? tv
: NULL
);
2919 kern_futimes(int fd
, struct timeval
*tptr
)
2921 struct thread
*td
= curthread
;
2922 struct proc
*p
= td
->td_proc
;
2923 struct timespec ts
[2];
2927 error
= getutimes(tptr
, ts
);
2930 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
2932 if (fp
->f_nchandle
.ncp
)
2933 error
= ncp_writechk(&fp
->f_nchandle
);
2935 error
= setutimes((struct vnode
*)fp
->f_data
, ts
, tptr
== NULL
);
2941 * futimes_args(int fd, struct timeval *tptr)
2943 * Set the access and modification times of a file.
2946 sys_futimes(struct futimes_args
*uap
)
2948 struct timeval tv
[2];
2952 error
= copyin(uap
->tptr
, tv
, sizeof(tv
));
2957 error
= kern_futimes(uap
->fd
, uap
->tptr
? tv
: NULL
);
2963 kern_truncate(struct nlookupdata
*nd
, off_t length
)
2971 /* XXX Add NLC flag indicating modifying operation? */
2972 if ((error
= nlookup(nd
)) != 0)
2974 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2976 if ((error
= cache_vref(&nd
->nl_nch
, nd
->nl_cred
, &vp
)) != 0)
2978 if ((error
= vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
)) != 0) {
2982 if (vp
->v_type
== VDIR
) {
2984 } else if ((error
= vn_writechk(vp
, &nd
->nl_nch
)) == 0 &&
2985 (error
= VOP_ACCESS(vp
, VWRITE
, nd
->nl_cred
)) == 0) {
2987 vattr
.va_size
= length
;
2988 error
= VOP_SETATTR(vp
, &vattr
, nd
->nl_cred
);
2995 * truncate(char *path, int pad, off_t length)
2997 * Truncate a file given its path name.
3000 sys_truncate(struct truncate_args
*uap
)
3002 struct nlookupdata nd
;
3005 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
3007 error
= kern_truncate(&nd
, uap
->length
);
3013 kern_ftruncate(int fd
, off_t length
)
3015 struct thread
*td
= curthread
;
3016 struct proc
*p
= td
->td_proc
;
3024 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
3026 if (fp
->f_nchandle
.ncp
) {
3027 error
= ncp_writechk(&fp
->f_nchandle
);
3031 if ((fp
->f_flag
& FWRITE
) == 0) {
3035 vp
= (struct vnode
*)fp
->f_data
;
3036 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
3037 if (vp
->v_type
== VDIR
) {
3039 } else if ((error
= vn_writechk(vp
, NULL
)) == 0) {
3041 vattr
.va_size
= length
;
3042 error
= VOP_SETATTR(vp
, &vattr
, fp
->f_cred
);
3051 * ftruncate_args(int fd, int pad, off_t length)
3053 * Truncate a file given a file descriptor.
3056 sys_ftruncate(struct ftruncate_args
*uap
)
3060 error
= kern_ftruncate(uap
->fd
, uap
->length
);
3068 * Sync an open file.
3072 sys_fsync(struct fsync_args
*uap
)
3074 struct thread
*td
= curthread
;
3075 struct proc
*p
= td
->td_proc
;
3081 if ((error
= holdvnode(p
->p_fd
, uap
->fd
, &fp
)) != 0)
3083 vp
= (struct vnode
*)fp
->f_data
;
3084 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
3085 if ((obj
= vp
->v_object
) != NULL
)
3086 vm_object_page_clean(obj
, 0, 0, 0);
3087 if ((error
= VOP_FSYNC(vp
, MNT_WAIT
)) == 0 && vp
->v_mount
)
3088 error
= buf_fsync(vp
);
3095 kern_rename(struct nlookupdata
*fromnd
, struct nlookupdata
*tond
)
3097 struct nchandle fnchd
;
3098 struct nchandle tnchd
;
3099 struct namecache
*ncp
;
3106 fromnd
->nl_flags
|= NLC_REFDVP
;
3107 if ((error
= nlookup(fromnd
)) != 0)
3109 if ((fnchd
.ncp
= fromnd
->nl_nch
.ncp
->nc_parent
) == NULL
)
3111 fnchd
.mount
= fromnd
->nl_nch
.mount
;
3115 * unlock the source nch so we can lookup the target nch without
3116 * deadlocking. The target may or may not exist so we do not check
3117 * for a target vp like kern_mkdir() and other creation functions do.
3119 * The source and target directories are ref'd and rechecked after
3120 * everything is relocked to determine if the source or target file
3123 KKASSERT(fromnd
->nl_flags
& NLC_NCPISLOCKED
);
3124 fromnd
->nl_flags
&= ~NLC_NCPISLOCKED
;
3125 cache_unlock(&fromnd
->nl_nch
);
3127 tond
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
3128 if ((error
= nlookup(tond
)) != 0) {
3132 if ((tnchd
.ncp
= tond
->nl_nch
.ncp
->nc_parent
) == NULL
) {
3136 tnchd
.mount
= tond
->nl_nch
.mount
;
3140 * If the source and target are the same there is nothing to do
3142 if (fromnd
->nl_nch
.ncp
== tond
->nl_nch
.ncp
) {
3149 * Mount points cannot be renamed or overwritten
3151 if ((fromnd
->nl_nch
.ncp
->nc_flag
| tond
->nl_nch
.ncp
->nc_flag
) &
3160 * relock the source ncp. NOTE AFTER RELOCKING: the source ncp
3161 * may have become invalid while it was unlocked, nc_vp and nc_mount
3164 if (cache_lock_nonblock(&fromnd
->nl_nch
) == 0) {
3165 cache_resolve(&fromnd
->nl_nch
, fromnd
->nl_cred
);
3166 } else if (fromnd
->nl_nch
.ncp
> tond
->nl_nch
.ncp
) {
3167 cache_lock(&fromnd
->nl_nch
);
3168 cache_resolve(&fromnd
->nl_nch
, fromnd
->nl_cred
);
3170 cache_unlock(&tond
->nl_nch
);
3171 cache_lock(&fromnd
->nl_nch
);
3172 cache_resolve(&fromnd
->nl_nch
, fromnd
->nl_cred
);
3173 cache_lock(&tond
->nl_nch
);
3174 cache_resolve(&tond
->nl_nch
, tond
->nl_cred
);
3176 fromnd
->nl_flags
|= NLC_NCPISLOCKED
;
3179 * make sure the parent directories linkages are the same
3181 if (fnchd
.ncp
!= fromnd
->nl_nch
.ncp
->nc_parent
||
3182 tnchd
.ncp
!= tond
->nl_nch
.ncp
->nc_parent
) {
3189 * Both the source and target must be within the same filesystem and
3190 * in the same filesystem as their parent directories within the
3191 * namecache topology.
3193 * NOTE: fromnd's nc_mount or nc_vp could be NULL.
3196 if (mp
!= tnchd
.mount
|| mp
!= fromnd
->nl_nch
.mount
||
3197 mp
!= tond
->nl_nch
.mount
) {
3204 * Make sure the mount point is writable
3206 if ((error
= ncp_writechk(&tond
->nl_nch
)) != 0) {
3213 * If the target exists and either the source or target is a directory,
3214 * then both must be directories.
3216 * Due to relocking of the source, fromnd->nl_nch.ncp->nc_vp might h
3219 if (tond
->nl_nch
.ncp
->nc_vp
) {
3220 if (fromnd
->nl_nch
.ncp
->nc_vp
== NULL
) {
3222 } else if (fromnd
->nl_nch
.ncp
->nc_vp
->v_type
== VDIR
) {
3223 if (tond
->nl_nch
.ncp
->nc_vp
->v_type
!= VDIR
)
3225 } else if (tond
->nl_nch
.ncp
->nc_vp
->v_type
== VDIR
) {
3231 * You cannot rename a source into itself or a subdirectory of itself.
3232 * We check this by travsersing the target directory upwards looking
3233 * for a match against the source.
3236 for (ncp
= tnchd
.ncp
; ncp
; ncp
= ncp
->nc_parent
) {
3237 if (fromnd
->nl_nch
.ncp
== ncp
) {
3248 * Even though the namespaces are different, they may still represent
3249 * hardlinks to the same file. The filesystem might have a hard time
3250 * with this so we issue a NREMOVE of the source instead of a NRENAME
3251 * when we detect the situation.
3254 fdvp
= fromnd
->nl_dvp
;
3255 tdvp
= tond
->nl_dvp
;
3256 if (fdvp
== NULL
|| tdvp
== NULL
) {
3258 } else if (fromnd
->nl_nch
.ncp
->nc_vp
== tond
->nl_nch
.ncp
->nc_vp
) {
3259 error
= VOP_NREMOVE(&fromnd
->nl_nch
, fdvp
,
3262 error
= VOP_NRENAME(&fromnd
->nl_nch
, &tond
->nl_nch
,
3263 fdvp
, tdvp
, tond
->nl_cred
);
3270 * rename_args(char *from, char *to)
3272 * Rename files. Source and destination must either both be directories,
3273 * or both not be directories. If target is a directory, it must be empty.
3276 sys_rename(struct rename_args
*uap
)
3278 struct nlookupdata fromnd
, tond
;
3281 error
= nlookup_init(&fromnd
, uap
->from
, UIO_USERSPACE
, 0);
3283 error
= nlookup_init(&tond
, uap
->to
, UIO_USERSPACE
, 0);
3285 error
= kern_rename(&fromnd
, &tond
);
3286 nlookup_done(&tond
);
3288 nlookup_done(&fromnd
);
3293 kern_mkdir(struct nlookupdata
*nd
, int mode
)
3295 struct thread
*td
= curthread
;
3296 struct proc
*p
= td
->td_proc
;
3302 nd
->nl_flags
|= NLC_WILLBEDIR
| NLC_CREATE
| NLC_REFDVP
;
3303 if ((error
= nlookup(nd
)) != 0)
3306 if (nd
->nl_nch
.ncp
->nc_vp
)
3308 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
3311 vattr
.va_type
= VDIR
;
3312 vattr
.va_mode
= (mode
& ACCESSPERMS
) &~ p
->p_fd
->fd_cmask
;
3315 error
= VOP_NMKDIR(&nd
->nl_nch
, nd
->nl_dvp
, &vp
, p
->p_ucred
, &vattr
);
3322 * mkdir_args(char *path, int mode)
3324 * Make a directory file.
3328 sys_mkdir(struct mkdir_args
*uap
)
3330 struct nlookupdata nd
;
3333 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
3335 error
= kern_mkdir(&nd
, uap
->mode
);
3341 kern_rmdir(struct nlookupdata
*nd
)
3346 nd
->nl_flags
|= NLC_DELETE
| NLC_REFDVP
;
3347 if ((error
= nlookup(nd
)) != 0)
3351 * Do not allow directories representing mount points to be
3352 * deleted, even if empty. Check write perms on mount point
3353 * in case the vnode is aliased (aka nullfs).
3355 if (nd
->nl_nch
.ncp
->nc_flag
& (NCF_ISMOUNTPT
))
3357 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
3359 error
= VOP_NRMDIR(&nd
->nl_nch
, nd
->nl_dvp
, nd
->nl_cred
);
3364 * rmdir_args(char *path)
3366 * Remove a directory file.
3370 sys_rmdir(struct rmdir_args
*uap
)
3372 struct nlookupdata nd
;
3375 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
3377 error
= kern_rmdir(&nd
);
3383 kern_getdirentries(int fd
, char *buf
, u_int count
, long *basep
, int *res
,
3384 enum uio_seg direction
)
3386 struct thread
*td
= curthread
;
3387 struct proc
*p
= td
->td_proc
;
3395 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
3397 if ((fp
->f_flag
& FREAD
) == 0) {
3401 vp
= (struct vnode
*)fp
->f_data
;
3403 if (vp
->v_type
!= VDIR
) {
3407 aiov
.iov_base
= buf
;
3408 aiov
.iov_len
= count
;
3409 auio
.uio_iov
= &aiov
;
3410 auio
.uio_iovcnt
= 1;
3411 auio
.uio_rw
= UIO_READ
;
3412 auio
.uio_segflg
= direction
;
3414 auio
.uio_resid
= count
;
3415 loff
= auio
.uio_offset
= fp
->f_offset
;
3416 error
= VOP_READDIR(vp
, &auio
, fp
->f_cred
, &eofflag
, NULL
, NULL
);
3417 fp
->f_offset
= auio
.uio_offset
;
3420 if (count
== auio
.uio_resid
) {
3421 if (union_dircheckp
) {
3422 error
= union_dircheckp(td
, &vp
, fp
);
3429 if ((vp
->v_flag
& VROOT
) &&
3430 (vp
->v_mount
->mnt_flag
& MNT_UNION
)) {
3431 struct vnode
*tvp
= vp
;
3432 vp
= vp
->v_mount
->mnt_vnodecovered
;
3443 * WARNING! *basep may not be wide enough to accomodate the
3444 * seek offset. XXX should we hack this to return the upper 32 bits
3445 * for offsets greater then 4G?
3448 *basep
= (long)loff
;
3450 *res
= count
- auio
.uio_resid
;
3457 * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
3459 * Read a block of directory entries in a file system independent format.
3462 sys_getdirentries(struct getdirentries_args
*uap
)
3467 error
= kern_getdirentries(uap
->fd
, uap
->buf
, uap
->count
, &base
,
3468 &uap
->sysmsg_result
, UIO_USERSPACE
);
3470 if (error
== 0 && uap
->basep
)
3471 error
= copyout(&base
, uap
->basep
, sizeof(*uap
->basep
));
3476 * getdents_args(int fd, char *buf, size_t count)
3479 sys_getdents(struct getdents_args
*uap
)
3483 error
= kern_getdirentries(uap
->fd
, uap
->buf
, uap
->count
, NULL
,
3484 &uap
->sysmsg_result
, UIO_USERSPACE
);
3490 * umask(int newmask)
3492 * Set the mode mask for creation of filesystem nodes.
3497 sys_umask(struct umask_args
*uap
)
3499 struct thread
*td
= curthread
;
3500 struct proc
*p
= td
->td_proc
;
3501 struct filedesc
*fdp
;
3504 uap
->sysmsg_result
= fdp
->fd_cmask
;
3505 fdp
->fd_cmask
= uap
->newmask
& ALLPERMS
;
3510 * revoke(char *path)
3512 * Void all references to file by ripping underlying filesystem
3517 sys_revoke(struct revoke_args
*uap
)
3519 struct nlookupdata nd
;
3526 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
3528 error
= nlookup(&nd
);
3530 error
= cache_vref(&nd
.nl_nch
, nd
.nl_cred
, &vp
);
3531 cred
= crhold(nd
.nl_cred
);
3534 if (vp
->v_type
!= VCHR
&& vp
->v_type
!= VBLK
)
3537 error
= VOP_GETATTR(vp
, &vattr
);
3538 if (error
== 0 && cred
->cr_uid
!= vattr
.va_uid
)
3539 error
= suser_cred(cred
, PRISON_ROOT
);
3540 if (error
== 0 && count_udev(vp
->v_umajor
, vp
->v_uminor
) > 0) {
3543 VOP_REVOKE(vp
, REVOKEALL
);
3554 * getfh_args(char *fname, fhandle_t *fhp)
3556 * Get (NFS) file handle
3558 * NOTE: We use the fsid of the covering mount, even if it is a nullfs
3559 * mount. This allows nullfs mounts to be explicitly exported.
3561 * WARNING: nullfs mounts of HAMMER PFS ROOTs are safe.
3563 * nullfs mounts of subdirectories are not safe. That is, it will
3564 * work, but you do not really have protection against access to
3565 * the related parent directories.
3568 sys_getfh(struct getfh_args
*uap
)
3570 struct thread
*td
= curthread
;
3571 struct nlookupdata nd
;
3578 * Must be super user
3580 if ((error
= suser(td
)) != 0)
3584 error
= nlookup_init(&nd
, uap
->fname
, UIO_USERSPACE
, NLC_FOLLOW
);
3586 error
= nlookup(&nd
);
3588 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
3589 mp
= nd
.nl_nch
.mount
;
3592 bzero(&fh
, sizeof(fh
));
3593 fh
.fh_fsid
= mp
->mnt_stat
.f_fsid
;
3594 error
= VFS_VPTOFH(vp
, &fh
.fh_fid
);
3597 error
= copyout(&fh
, uap
->fhp
, sizeof(fh
));
3603 * fhopen_args(const struct fhandle *u_fhp, int flags)
3605 * syscall for the rpc.lockd to use to translate a NFS file handle into
3606 * an open descriptor.
3608 * warning: do not remove the suser() call or this becomes one giant
3612 sys_fhopen(struct fhopen_args
*uap
)
3614 struct thread
*td
= curthread
;
3615 struct proc
*p
= td
->td_proc
;
3620 struct vattr
*vap
= &vat
;
3622 int fmode
, mode
, error
, type
;
3628 * Must be super user
3634 fmode
= FFLAGS(uap
->flags
);
3635 /* why not allow a non-read/write open for our lockd? */
3636 if (((fmode
& (FREAD
| FWRITE
)) == 0) || (fmode
& O_CREAT
))
3638 error
= copyin(uap
->u_fhp
, &fhp
, sizeof(fhp
));
3641 /* find the mount point */
3642 mp
= vfs_getvfs(&fhp
.fh_fsid
);
3645 /* now give me my vnode, it gets returned to me locked */
3646 error
= VFS_FHTOVP(mp
, NULL
, &fhp
.fh_fid
, &vp
);
3650 * from now on we have to make sure not
3651 * to forget about the vnode
3652 * any error that causes an abort must vput(vp)
3653 * just set error = err and 'goto bad;'.
3659 if (vp
->v_type
== VLNK
) {
3663 if (vp
->v_type
== VSOCK
) {
3668 if (fmode
& (FWRITE
| O_TRUNC
)) {
3669 if (vp
->v_type
== VDIR
) {
3673 error
= vn_writechk(vp
, NULL
);
3681 error
= VOP_ACCESS(vp
, mode
, p
->p_ucred
);
3685 if (fmode
& O_TRUNC
) {
3686 vn_unlock(vp
); /* XXX */
3687 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
); /* XXX */
3690 error
= VOP_SETATTR(vp
, vap
, p
->p_ucred
);
3696 * VOP_OPEN needs the file pointer so it can potentially override
3699 * WARNING! no f_nchandle will be associated when fhopen()ing a
3702 if ((error
= falloc(p
, &nfp
, &indx
)) != 0)
3706 error
= VOP_OPEN(vp
, fmode
, p
->p_ucred
, fp
);
3709 * setting f_ops this way prevents VOP_CLOSE from being
3710 * called or fdrop() releasing the vp from v_data. Since
3711 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3713 fp
->f_ops
= &badfileops
;
3719 * The fp is given its own reference, we still have our ref and lock.
3721 * Assert that all regular files must be created with a VM object.
3723 if (vp
->v_type
== VREG
&& vp
->v_object
== NULL
) {
3724 kprintf("fhopen: regular file did not have VM object: %p\n", vp
);
3729 * The open was successful. Handle any locking requirements.
3731 if (fmode
& (O_EXLOCK
| O_SHLOCK
)) {
3732 lf
.l_whence
= SEEK_SET
;
3735 if (fmode
& O_EXLOCK
)
3736 lf
.l_type
= F_WRLCK
;
3738 lf
.l_type
= F_RDLCK
;
3739 if (fmode
& FNONBLOCK
)
3744 if ((error
= VOP_ADVLOCK(vp
, (caddr_t
)fp
, F_SETLK
, &lf
, type
)) != 0) {
3746 * release our private reference.
3748 fsetfd(p
, NULL
, indx
);
3753 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
3754 fp
->f_flag
|= FHASLOCK
;
3758 * Clean up. Associate the file pointer with the previously
3759 * reserved descriptor and return it.
3762 fsetfd(p
, fp
, indx
);
3764 uap
->sysmsg_result
= indx
;
3768 fsetfd(p
, NULL
, indx
);
3776 * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3779 sys_fhstat(struct fhstat_args
*uap
)
3781 struct thread
*td
= curthread
;
3789 * Must be super user
3795 error
= copyin(uap
->u_fhp
, &fh
, sizeof(fhandle_t
));
3799 if ((mp
= vfs_getvfs(&fh
.fh_fsid
)) == NULL
)
3801 if ((error
= VFS_FHTOVP(mp
, NULL
, &fh
.fh_fid
, &vp
)))
3803 error
= vn_stat(vp
, &sb
, td
->td_proc
->p_ucred
);
3807 error
= copyout(&sb
, uap
->sb
, sizeof(sb
));
3812 * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3815 sys_fhstatfs(struct fhstatfs_args
*uap
)
3817 struct thread
*td
= curthread
;
3818 struct proc
*p
= td
->td_proc
;
3823 char *fullpath
, *freepath
;
3828 * Must be super user
3830 if ((error
= suser(td
)))
3833 if ((error
= copyin(uap
->u_fhp
, &fh
, sizeof(fhandle_t
))) != 0)
3836 if ((mp
= vfs_getvfs(&fh
.fh_fsid
)) == NULL
)
3839 if (p
!= NULL
&& !chroot_visible_mnt(mp
, p
))
3842 if ((error
= VFS_FHTOVP(mp
, NULL
, &fh
.fh_fid
, &vp
)))
3847 if ((error
= VFS_STATFS(mp
, sp
, p
->p_ucred
)) != 0)
3850 error
= mount_path(p
, mp
, &fullpath
, &freepath
);
3853 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
3854 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
3855 kfree(freepath
, M_TEMP
);
3857 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
3859 bcopy(sp
, &sb
, sizeof(sb
));
3860 sb
.f_fsid
.val
[0] = sb
.f_fsid
.val
[1] = 0;
3863 return (copyout(sp
, uap
->buf
, sizeof(*sp
)));
3867 * fhstatvfs_args(struct fhandle *u_fhp, struct statvfs *buf)
3870 sys_fhstatvfs(struct fhstatvfs_args
*uap
)
3872 struct thread
*td
= curthread
;
3873 struct proc
*p
= td
->td_proc
;
3881 * Must be super user
3883 if ((error
= suser(td
)))
3886 if ((error
= copyin(uap
->u_fhp
, &fh
, sizeof(fhandle_t
))) != 0)
3889 if ((mp
= vfs_getvfs(&fh
.fh_fsid
)) == NULL
)
3892 if (p
!= NULL
&& !chroot_visible_mnt(mp
, p
))
3895 if ((error
= VFS_FHTOVP(mp
, NULL
, &fh
.fh_fid
, &vp
)))
3898 sp
= &mp
->mnt_vstat
;
3900 if ((error
= VFS_STATVFS(mp
, sp
, p
->p_ucred
)) != 0)
3904 if (mp
->mnt_flag
& MNT_RDONLY
)
3905 sp
->f_flag
|= ST_RDONLY
;
3906 if (mp
->mnt_flag
& MNT_NOSUID
)
3907 sp
->f_flag
|= ST_NOSUID
;
3909 return (copyout(sp
, uap
->buf
, sizeof(*sp
)));
3914 * Syscall to push extended attribute configuration information into the
3915 * VFS. Accepts a path, which it converts to a mountpoint, as well as
3916 * a command (int cmd), and attribute name and misc data. For now, the
3917 * attribute name is left in userspace for consumption by the VFS_op.
3918 * It will probably be changed to be copied into sysspace by the
3919 * syscall in the future, once issues with various consumers of the
3920 * attribute code have raised their hands.
3922 * Currently this is used only by UFS Extended Attributes.
3925 sys_extattrctl(struct extattrctl_args
*uap
)
3927 struct nlookupdata nd
;
3933 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
3935 error
= nlookup(&nd
);
3937 mp
= nd
.nl_nch
.mount
;
3938 error
= VFS_EXTATTRCTL(mp
, uap
->cmd
,
3939 uap
->attrname
, uap
->arg
,
3947 * Syscall to set a named extended attribute on a file or directory.
3948 * Accepts attribute name, and a uio structure pointing to the data to set.
3949 * The uio is consumed in the style of writev(). The real work happens
3950 * in VOP_SETEXTATTR().
3953 sys_extattr_set_file(struct extattr_set_file_args
*uap
)
3955 char attrname
[EXTATTR_MAXNAMELEN
];
3956 struct iovec aiov
[UIO_SMALLIOV
];
3957 struct iovec
*needfree
;
3958 struct nlookupdata nd
;
3967 error
= copyin(uap
->attrname
, attrname
, EXTATTR_MAXNAMELEN
);
3972 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
3974 error
= nlookup(&nd
);
3976 error
= ncp_writechk(&nd
.nl_nch
);
3978 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
3985 iovlen
= uap
->iovcnt
* sizeof(struct iovec
);
3986 if (uap
->iovcnt
> UIO_SMALLIOV
) {
3987 if (uap
->iovcnt
> UIO_MAXIOV
) {
3991 MALLOC(iov
, struct iovec
*, iovlen
, M_IOV
, M_WAITOK
);
3997 auio
.uio_iovcnt
= uap
->iovcnt
;
3998 auio
.uio_rw
= UIO_WRITE
;
3999 auio
.uio_segflg
= UIO_USERSPACE
;
4000 auio
.uio_td
= nd
.nl_td
;
4001 auio
.uio_offset
= 0;
4002 if ((error
= copyin(uap
->iovp
, iov
, iovlen
)))
4005 for (i
= 0; i
< uap
->iovcnt
; i
++) {
4006 if (iov
->iov_len
> INT_MAX
- auio
.uio_resid
) {
4010 auio
.uio_resid
+= iov
->iov_len
;
4013 cnt
= auio
.uio_resid
;
4014 error
= VOP_SETEXTATTR(vp
, attrname
, &auio
, nd
.nl_cred
);
4015 cnt
-= auio
.uio_resid
;
4016 uap
->sysmsg_result
= cnt
;
4021 FREE(needfree
, M_IOV
);
4026 * Syscall to get a named extended attribute on a file or directory.
4027 * Accepts attribute name, and a uio structure pointing to a buffer for the
4028 * data. The uio is consumed in the style of readv(). The real work
4029 * happens in VOP_GETEXTATTR();
4032 sys_extattr_get_file(struct extattr_get_file_args
*uap
)
4034 char attrname
[EXTATTR_MAXNAMELEN
];
4035 struct iovec aiov
[UIO_SMALLIOV
];
4036 struct iovec
*needfree
;
4037 struct nlookupdata nd
;
4046 error
= copyin(uap
->attrname
, attrname
, EXTATTR_MAXNAMELEN
);
4051 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
4053 error
= nlookup(&nd
);
4055 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
4061 iovlen
= uap
->iovcnt
* sizeof (struct iovec
);
4063 if (uap
->iovcnt
> UIO_SMALLIOV
) {
4064 if (uap
->iovcnt
> UIO_MAXIOV
) {
4068 MALLOC(iov
, struct iovec
*, iovlen
, M_IOV
, M_WAITOK
);
4074 auio
.uio_iovcnt
= uap
->iovcnt
;
4075 auio
.uio_rw
= UIO_READ
;
4076 auio
.uio_segflg
= UIO_USERSPACE
;
4077 auio
.uio_td
= nd
.nl_td
;
4078 auio
.uio_offset
= 0;
4079 if ((error
= copyin(uap
->iovp
, iov
, iovlen
)))
4082 for (i
= 0; i
< uap
->iovcnt
; i
++) {
4083 if (iov
->iov_len
> INT_MAX
- auio
.uio_resid
) {
4087 auio
.uio_resid
+= iov
->iov_len
;
4090 cnt
= auio
.uio_resid
;
4091 error
= VOP_GETEXTATTR(vp
, attrname
, &auio
, nd
.nl_cred
);
4092 cnt
-= auio
.uio_resid
;
4093 uap
->sysmsg_result
= cnt
;
4098 FREE(needfree
, M_IOV
);
4103 * Syscall to delete a named extended attribute from a file or directory.
4104 * Accepts attribute name. The real work happens in VOP_SETEXTATTR().
4107 sys_extattr_delete_file(struct extattr_delete_file_args
*uap
)
4109 char attrname
[EXTATTR_MAXNAMELEN
];
4110 struct nlookupdata nd
;
4114 error
= copyin(uap
->attrname
, attrname
, EXTATTR_MAXNAMELEN
);
4119 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
4121 error
= nlookup(&nd
);
4123 error
= ncp_writechk(&nd
.nl_nch
);
4125 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
4131 error
= VOP_SETEXTATTR(vp
, attrname
, NULL
, nd
.nl_cred
);
4138 * Determine if the mount is visible to the process.
4141 chroot_visible_mnt(struct mount
*mp
, struct proc
*p
)
4143 struct nchandle nch
;
4146 * Traverse from the mount point upwards. If we hit the process
4147 * root then the mount point is visible to the process.
4149 nch
= mp
->mnt_ncmountpt
;
4151 if (nch
.mount
== p
->p_fd
->fd_nrdir
.mount
&&
4152 nch
.ncp
== p
->p_fd
->fd_nrdir
.ncp
) {
4155 if (nch
.ncp
== nch
.mount
->mnt_ncmountpt
.ncp
) {
4156 nch
= nch
.mount
->mnt_ncmounton
;
4158 nch
.ncp
= nch
.ncp
->nc_parent
;
4163 * If the mount point is not visible to the process, but the
4164 * process root is in a subdirectory of the mount, return
4167 if (p
->p_fd
->fd_nrdir
.mount
== mp
)