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 1.135 2008/11/11 00:55:49 pavalos 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>
63 #include <sys/namei.h>
64 #include <sys/nlookup.h>
65 #include <sys/dirent.h>
66 #include <sys/extattr.h>
67 #include <sys/spinlock.h>
68 #include <sys/kern_syscall.h>
69 #include <sys/objcache.h>
70 #include <sys/sysctl.h>
73 #include <sys/file2.h>
74 #include <sys/spinlock2.h>
77 #include <vm/vm_object.h>
78 #include <vm/vm_page.h>
80 #include <machine/limits.h>
81 #include <machine/stdarg.h>
83 #include <vfs/union/union.h>
85 static void mount_warning(struct mount
*mp
, const char *ctl
, ...);
86 static int mount_path(struct proc
*p
, struct mount
*mp
, char **rb
, char **fb
);
87 static int checkvp_chdir (struct vnode
*vn
, struct thread
*td
);
88 static void checkdirs (struct nchandle
*old_nch
, struct nchandle
*new_nch
);
89 static int chroot_refuse_vdir_fds (struct filedesc
*fdp
);
90 static int chroot_visible_mnt(struct mount
*mp
, struct proc
*p
);
91 static int getutimes (const struct timeval
*, struct timespec
*);
92 static int setfown (struct vnode
*, uid_t
, gid_t
);
93 static int setfmode (struct vnode
*, int);
94 static int setfflags (struct vnode
*, int);
95 static int setutimes (struct vnode
*, struct vattr
*,
96 const struct timespec
*, int);
97 static int usermount
= 0; /* if 1, non-root can mount fs. */
99 int (*union_dircheckp
) (struct thread
*, struct vnode
**, struct file
*);
101 SYSCTL_INT(_vfs
, OID_AUTO
, usermount
, CTLFLAG_RW
, &usermount
, 0, "");
104 * Virtual File System System Calls
108 * Mount a file system.
111 * mount_args(char *type, char *path, int flags, caddr_t data)
115 sys_mount(struct mount_args
*uap
)
117 struct thread
*td
= curthread
;
118 struct proc
*p
= td
->td_proc
;
122 struct vfsconf
*vfsp
;
123 int error
, flag
= 0, flag2
= 0;
126 struct nlookupdata nd
;
127 char fstypename
[MFSNAMELEN
];
128 struct ucred
*cred
= p
->p_ucred
;
133 if (usermount
== 0 && (error
= priv_check(td
, PRIV_ROOT
)))
136 * Do not allow NFS export by non-root users.
138 if (uap
->flags
& MNT_EXPORTED
) {
139 error
= priv_check(td
, PRIV_ROOT
);
144 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
146 if (priv_check(td
, PRIV_ROOT
))
147 uap
->flags
|= MNT_NOSUID
| MNT_NODEV
;
150 * Lookup the requested path and extract the nch and vnode.
152 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
154 if ((error
= nlookup(&nd
)) == 0) {
155 if (nd
.nl_nch
.ncp
->nc_vp
== NULL
)
165 * Extract the locked+refd ncp and cleanup the nd structure
168 cache_zero(&nd
.nl_nch
);
171 if ((nch
.ncp
->nc_flag
& NCF_ISMOUNTPT
) && cache_findmount(&nch
))
178 * now we have the locked ref'd nch and unreferenced vnode.
181 if ((error
= vget(vp
, LK_EXCLUSIVE
)) != 0) {
188 * Now we have an unlocked ref'd nch and a locked ref'd vp
190 if (uap
->flags
& MNT_UPDATE
) {
191 if ((vp
->v_flag
& (VROOT
|VPFSROOT
)) == 0) {
198 flag2
= mp
->mnt_kern_flag
;
200 * We only allow the filesystem to be reloaded if it
201 * is currently mounted read-only.
203 if ((uap
->flags
& MNT_RELOAD
) &&
204 ((mp
->mnt_flag
& MNT_RDONLY
) == 0)) {
207 return (EOPNOTSUPP
); /* Needs translation */
210 * Only root, or the user that did the original mount is
211 * permitted to update it.
213 if (mp
->mnt_stat
.f_owner
!= cred
->cr_uid
&&
214 (error
= priv_check(td
, PRIV_ROOT
))) {
219 if (vfs_busy(mp
, LK_NOWAIT
)) {
224 if ((vp
->v_flag
& VMOUNT
) != 0 || hasmount
) {
230 vp
->v_flag
|= VMOUNT
;
232 uap
->flags
& (MNT_RELOAD
| MNT_FORCE
| MNT_UPDATE
);
237 * If the user is not root, ensure that they own the directory
238 * onto which we are attempting to mount.
240 if ((error
= VOP_GETATTR(vp
, &va
)) ||
241 (va
.va_uid
!= cred
->cr_uid
&& (error
= priv_check(td
, PRIV_ROOT
)))) {
246 if ((error
= vinvalbuf(vp
, V_SAVE
, 0, 0)) != 0) {
251 if (vp
->v_type
!= VDIR
) {
256 if (vp
->v_mount
->mnt_kern_flag
& MNTK_NOSTKMNT
) {
261 if ((error
= copyinstr(uap
->type
, fstypename
, MFSNAMELEN
, NULL
)) != 0) {
266 vfsp
= vfsconf_find_by_name(fstypename
);
270 /* Only load modules for root (very important!) */
271 if ((error
= priv_check(td
, PRIV_ROOT
)) != 0) {
276 error
= linker_load_file(fstypename
, &lf
, 0);
277 if (error
|| lf
== NULL
) {
285 /* lookup again, see if the VFS was loaded */
286 vfsp
= vfsconf_find_by_name(fstypename
);
289 linker_file_unload(lf
);
295 if ((vp
->v_flag
& VMOUNT
) != 0 || hasmount
) {
300 vp
->v_flag
|= VMOUNT
;
303 * Allocate and initialize the filesystem.
305 mp
= kmalloc(sizeof(struct mount
), M_MOUNT
, M_ZERO
|M_WAITOK
);
306 TAILQ_INIT(&mp
->mnt_nvnodelist
);
307 TAILQ_INIT(&mp
->mnt_reservedvnlist
);
308 TAILQ_INIT(&mp
->mnt_jlist
);
309 mp
->mnt_nvnodelistsize
= 0;
310 lockinit(&mp
->mnt_lock
, "vfslock", 0, 0);
311 vfs_busy(mp
, LK_NOWAIT
);
312 mp
->mnt_op
= vfsp
->vfc_vfsops
;
314 vfsp
->vfc_refcount
++;
315 mp
->mnt_stat
.f_type
= vfsp
->vfc_typenum
;
316 mp
->mnt_flag
|= vfsp
->vfc_flags
& MNT_VISFLAGMASK
;
317 strncpy(mp
->mnt_stat
.f_fstypename
, vfsp
->vfc_name
, MFSNAMELEN
);
318 mp
->mnt_stat
.f_owner
= cred
->cr_uid
;
319 mp
->mnt_iosize_max
= DFLTPHYS
;
323 * Set the mount level flags.
325 if (uap
->flags
& MNT_RDONLY
)
326 mp
->mnt_flag
|= MNT_RDONLY
;
327 else if (mp
->mnt_flag
& MNT_RDONLY
)
328 mp
->mnt_kern_flag
|= MNTK_WANTRDWR
;
329 mp
->mnt_flag
&=~ (MNT_NOSUID
| MNT_NOEXEC
| MNT_NODEV
|
330 MNT_SYNCHRONOUS
| MNT_UNION
| MNT_ASYNC
| MNT_NOATIME
|
331 MNT_NOSYMFOLLOW
| MNT_IGNORE
|
332 MNT_NOCLUSTERR
| MNT_NOCLUSTERW
| MNT_SUIDDIR
);
333 mp
->mnt_flag
|= uap
->flags
& (MNT_NOSUID
| MNT_NOEXEC
|
334 MNT_NODEV
| MNT_SYNCHRONOUS
| MNT_UNION
| MNT_ASYNC
| MNT_FORCE
|
335 MNT_NOSYMFOLLOW
| MNT_IGNORE
|
336 MNT_NOATIME
| MNT_NOCLUSTERR
| MNT_NOCLUSTERW
| MNT_SUIDDIR
);
338 * Mount the filesystem.
339 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
342 error
= VFS_MOUNT(mp
, uap
->path
, uap
->data
, cred
);
343 if (mp
->mnt_flag
& MNT_UPDATE
) {
344 if (mp
->mnt_kern_flag
& MNTK_WANTRDWR
)
345 mp
->mnt_flag
&= ~MNT_RDONLY
;
346 mp
->mnt_flag
&=~ (MNT_UPDATE
| MNT_RELOAD
| MNT_FORCE
);
347 mp
->mnt_kern_flag
&=~ MNTK_WANTRDWR
;
350 mp
->mnt_kern_flag
= flag2
;
353 vp
->v_flag
&= ~VMOUNT
;
358 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
360 * Put the new filesystem on the mount list after root. The mount
361 * point gets its own mnt_ncmountpt (unless the VFS already set one
362 * up) which represents the root of the mount. The lookup code
363 * detects the mount point going forward and checks the root of
364 * the mount going backwards.
366 * It is not necessary to invalidate or purge the vnode underneath
367 * because elements under the mount will be given their own glue
371 if (mp
->mnt_ncmountpt
.ncp
== NULL
) {
373 * allocate, then unlock, but leave the ref intact
375 cache_allocroot(&mp
->mnt_ncmountpt
, mp
, NULL
);
376 cache_unlock(&mp
->mnt_ncmountpt
);
378 mp
->mnt_ncmounton
= nch
; /* inherits ref */
379 nch
.ncp
->nc_flag
|= NCF_ISMOUNTPT
;
381 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
382 vp
->v_flag
&= ~VMOUNT
;
383 mountlist_insert(mp
, MNTINS_LAST
);
385 checkdirs(&mp
->mnt_ncmounton
, &mp
->mnt_ncmountpt
);
386 error
= vfs_allocate_syncvnode(mp
);
388 error
= VFS_START(mp
, 0);
391 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_coherency_ops
);
392 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_journal_ops
);
393 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_norm_ops
);
394 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_spec_ops
);
395 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_fifo_ops
);
396 vp
->v_flag
&= ~VMOUNT
;
397 mp
->mnt_vfc
->vfc_refcount
--;
407 * Scan all active processes to see if any of them have a current
408 * or root directory onto which the new filesystem has just been
409 * mounted. If so, replace them with the new mount point.
411 * The passed ncp is ref'd and locked (from the mount code) and
412 * must be associated with the vnode representing the root of the
415 struct checkdirs_info
{
416 struct nchandle old_nch
;
417 struct nchandle new_nch
;
418 struct vnode
*old_vp
;
419 struct vnode
*new_vp
;
422 static int checkdirs_callback(struct proc
*p
, void *data
);
425 checkdirs(struct nchandle
*old_nch
, struct nchandle
*new_nch
)
427 struct checkdirs_info info
;
433 * If the old mount point's vnode has a usecount of 1, it is not
434 * being held as a descriptor anywhere.
436 olddp
= old_nch
->ncp
->nc_vp
;
437 if (olddp
== NULL
|| olddp
->v_sysref
.refcnt
== 1)
441 * Force the root vnode of the new mount point to be resolved
442 * so we can update any matching processes.
445 if (VFS_ROOT(mp
, &newdp
))
446 panic("mount: lost mount");
447 cache_setunresolved(new_nch
);
448 cache_setvp(new_nch
, newdp
);
451 * Special handling of the root node
453 if (rootvnode
== olddp
) {
455 vfs_cache_setroot(newdp
, cache_hold(new_nch
));
459 * Pass newdp separately so the callback does not have to access
460 * it via new_nch->ncp->nc_vp.
462 info
.old_nch
= *old_nch
;
463 info
.new_nch
= *new_nch
;
465 allproc_scan(checkdirs_callback
, &info
);
470 * NOTE: callback is not MP safe because the scanned process's filedesc
471 * structure can be ripped out from under us, amoung other things.
474 checkdirs_callback(struct proc
*p
, void *data
)
476 struct checkdirs_info
*info
= data
;
477 struct filedesc
*fdp
;
478 struct nchandle ncdrop1
;
479 struct nchandle ncdrop2
;
480 struct vnode
*vprele1
;
481 struct vnode
*vprele2
;
483 if ((fdp
= p
->p_fd
) != NULL
) {
484 cache_zero(&ncdrop1
);
485 cache_zero(&ncdrop2
);
490 * MPUNSAFE - XXX fdp can be pulled out from under a
493 * A shared filedesc is ok, we don't have to copy it
494 * because we are making this change globally.
496 spin_lock_wr(&fdp
->fd_spin
);
497 if (fdp
->fd_ncdir
.mount
== info
->old_nch
.mount
&&
498 fdp
->fd_ncdir
.ncp
== info
->old_nch
.ncp
) {
499 vprele1
= fdp
->fd_cdir
;
501 fdp
->fd_cdir
= info
->new_vp
;
502 ncdrop1
= fdp
->fd_ncdir
;
503 cache_copy(&info
->new_nch
, &fdp
->fd_ncdir
);
505 if (fdp
->fd_nrdir
.mount
== info
->old_nch
.mount
&&
506 fdp
->fd_nrdir
.ncp
== info
->old_nch
.ncp
) {
507 vprele2
= fdp
->fd_rdir
;
509 fdp
->fd_rdir
= info
->new_vp
;
510 ncdrop2
= fdp
->fd_nrdir
;
511 cache_copy(&info
->new_nch
, &fdp
->fd_nrdir
);
513 spin_unlock_wr(&fdp
->fd_spin
);
515 cache_drop(&ncdrop1
);
517 cache_drop(&ncdrop2
);
527 * Unmount a file system.
529 * Note: unmount takes a path to the vnode mounted on as argument,
530 * not special file (as before).
533 * umount_args(char *path, int flags)
537 sys_unmount(struct unmount_args
*uap
)
539 struct thread
*td
= curthread
;
540 struct proc
*p
= td
->td_proc
;
541 struct mount
*mp
= NULL
;
543 struct nlookupdata nd
;
546 if (p
->p_ucred
->cr_prison
!= NULL
)
548 if (usermount
== 0 && (error
= priv_check(td
, PRIV_ROOT
)))
551 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
553 error
= nlookup(&nd
);
557 mp
= nd
.nl_nch
.mount
;
560 * Only root, or the user that did the original mount is
561 * permitted to unmount this filesystem.
563 if ((mp
->mnt_stat
.f_owner
!= p
->p_ucred
->cr_uid
) &&
564 (error
= priv_check(td
, PRIV_ROOT
)))
568 * Don't allow unmounting the root file system.
570 if (mp
->mnt_flag
& MNT_ROOTFS
) {
576 * Must be the root of the filesystem
578 if (nd
.nl_nch
.ncp
!= mp
->mnt_ncmountpt
.ncp
) {
587 return (dounmount(mp
, uap
->flags
));
591 * Do the actual file system unmount.
594 dounmount_interlock(struct mount
*mp
)
596 if (mp
->mnt_kern_flag
& MNTK_UNMOUNT
)
598 mp
->mnt_kern_flag
|= MNTK_UNMOUNT
;
603 dounmount(struct mount
*mp
, int flags
)
605 struct namecache
*ncp
;
614 * Exclusive access for unmounting purposes
616 if ((error
= mountlist_interlock(dounmount_interlock
, mp
)) != 0)
620 * Allow filesystems to detect that a forced unmount is in progress.
622 if (flags
& MNT_FORCE
)
623 mp
->mnt_kern_flag
|= MNTK_UNMOUNTF
;
624 lflags
= LK_EXCLUSIVE
| ((flags
& MNT_FORCE
) ? 0 : LK_NOWAIT
);
625 error
= lockmgr(&mp
->mnt_lock
, lflags
);
627 mp
->mnt_kern_flag
&= ~(MNTK_UNMOUNT
| MNTK_UNMOUNTF
);
628 if (mp
->mnt_kern_flag
& MNTK_MWAIT
)
633 if (mp
->mnt_flag
& MNT_EXPUBLIC
)
634 vfs_setpublicfs(NULL
, NULL
, NULL
);
636 vfs_msync(mp
, MNT_WAIT
);
637 async_flag
= mp
->mnt_flag
& MNT_ASYNC
;
638 mp
->mnt_flag
&=~ MNT_ASYNC
;
641 * If this filesystem isn't aliasing other filesystems,
642 * try to invalidate any remaining namecache entries and
643 * check the count afterwords.
645 if ((mp
->mnt_kern_flag
& MNTK_NCALIASED
) == 0) {
646 cache_lock(&mp
->mnt_ncmountpt
);
647 cache_inval(&mp
->mnt_ncmountpt
, CINV_DESTROY
|CINV_CHILDREN
);
648 cache_unlock(&mp
->mnt_ncmountpt
);
650 if ((ncp
= mp
->mnt_ncmountpt
.ncp
) != NULL
&&
651 (ncp
->nc_refs
!= 1 || TAILQ_FIRST(&ncp
->nc_list
))) {
653 if ((flags
& MNT_FORCE
) == 0) {
655 mount_warning(mp
, "Cannot unmount: "
661 mount_warning(mp
, "Forced unmount: "
672 * nchandle records ref the mount structure. Expect a count of 1
673 * (our mount->mnt_ncmountpt).
675 if (mp
->mnt_refs
!= 1) {
676 if ((flags
& MNT_FORCE
) == 0) {
677 mount_warning(mp
, "Cannot unmount: "
678 "%d process references still "
679 "present", mp
->mnt_refs
);
682 mount_warning(mp
, "Forced unmount: "
683 "%d process references still "
684 "present", mp
->mnt_refs
);
690 * Decomission our special mnt_syncer vnode. This also stops
691 * the vnlru code. If we are unable to unmount we recommission
695 if ((vp
= mp
->mnt_syncer
) != NULL
) {
696 mp
->mnt_syncer
= NULL
;
699 if (((mp
->mnt_flag
& MNT_RDONLY
) ||
700 (error
= VFS_SYNC(mp
, MNT_WAIT
)) == 0) ||
701 (flags
& MNT_FORCE
)) {
702 error
= VFS_UNMOUNT(mp
, flags
);
706 if (mp
->mnt_syncer
== NULL
)
707 vfs_allocate_syncvnode(mp
);
708 mp
->mnt_kern_flag
&= ~(MNTK_UNMOUNT
| MNTK_UNMOUNTF
);
709 mp
->mnt_flag
|= async_flag
;
710 lockmgr(&mp
->mnt_lock
, LK_RELEASE
);
711 if (mp
->mnt_kern_flag
& MNTK_MWAIT
)
716 * Clean up any journals still associated with the mount after
717 * filesystem activity has ceased.
719 journal_remove_all_journals(mp
,
720 ((flags
& MNT_FORCE
) ? MC_JOURNAL_STOP_IMM
: 0));
722 mountlist_remove(mp
);
725 * Remove any installed vnode ops here so the individual VFSs don't
728 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_coherency_ops
);
729 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_journal_ops
);
730 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_norm_ops
);
731 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_spec_ops
);
732 vfs_rm_vnodeops(mp
, NULL
, &mp
->mnt_vn_fifo_ops
);
734 if (mp
->mnt_ncmountpt
.ncp
!= NULL
) {
735 nch
= mp
->mnt_ncmountpt
;
736 cache_zero(&mp
->mnt_ncmountpt
);
737 cache_clrmountpt(&nch
);
740 if (mp
->mnt_ncmounton
.ncp
!= NULL
) {
741 nch
= mp
->mnt_ncmounton
;
742 cache_zero(&mp
->mnt_ncmounton
);
743 cache_clrmountpt(&nch
);
747 mp
->mnt_vfc
->vfc_refcount
--;
748 if (!TAILQ_EMPTY(&mp
->mnt_nvnodelist
))
749 panic("unmount: dangling vnode");
750 lockmgr(&mp
->mnt_lock
, LK_RELEASE
);
751 if (mp
->mnt_kern_flag
& MNTK_MWAIT
)
760 mount_warning(struct mount
*mp
, const char *ctl
, ...)
767 if (cache_fullpath(NULL
, &mp
->mnt_ncmounton
, &ptr
, &buf
) == 0) {
768 kprintf("unmount(%s): ", ptr
);
773 kprintf("unmount(%p", mp
);
774 if (mp
->mnt_ncmounton
.ncp
&& mp
->mnt_ncmounton
.ncp
->nc_name
)
775 kprintf(",%s", mp
->mnt_ncmounton
.ncp
->nc_name
);
784 * Shim cache_fullpath() to handle the case where a process is chrooted into
785 * a subdirectory of a mount. In this case if the root mount matches the
786 * process root directory's mount we have to specify the process's root
787 * directory instead of the mount point, because the mount point might
788 * be above the root directory.
792 mount_path(struct proc
*p
, struct mount
*mp
, char **rb
, char **fb
)
794 struct nchandle
*nch
;
796 if (p
&& p
->p_fd
->fd_nrdir
.mount
== mp
)
797 nch
= &p
->p_fd
->fd_nrdir
;
799 nch
= &mp
->mnt_ncmountpt
;
800 return(cache_fullpath(p
, nch
, rb
, fb
));
804 * Sync each mounted filesystem.
808 static int syncprt
= 0;
809 SYSCTL_INT(_debug
, OID_AUTO
, syncprt
, CTLFLAG_RW
, &syncprt
, 0, "");
812 static int sync_callback(struct mount
*mp
, void *data
);
816 sys_sync(struct sync_args
*uap
)
818 mountlist_scan(sync_callback
, NULL
, MNTSCAN_FORWARD
);
821 * print out buffer pool stat information on each sync() call.
831 sync_callback(struct mount
*mp
, void *data __unused
)
835 if ((mp
->mnt_flag
& MNT_RDONLY
) == 0) {
836 asyncflag
= mp
->mnt_flag
& MNT_ASYNC
;
837 mp
->mnt_flag
&= ~MNT_ASYNC
;
838 vfs_msync(mp
, MNT_NOWAIT
);
839 VFS_SYNC(mp
, MNT_NOWAIT
);
840 mp
->mnt_flag
|= asyncflag
;
845 /* XXX PRISON: could be per prison flag */
846 static int prison_quotas
;
848 SYSCTL_INT(_kern_prison
, OID_AUTO
, quotas
, CTLFLAG_RW
, &prison_quotas
, 0, "");
852 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
854 * Change filesystem quotas.
858 sys_quotactl(struct quotactl_args
*uap
)
860 struct nlookupdata nd
;
868 if (p
->p_ucred
->cr_prison
&& !prison_quotas
)
871 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
873 error
= nlookup(&nd
);
875 mp
= nd
.nl_nch
.mount
;
876 error
= VFS_QUOTACTL(mp
, uap
->cmd
, uap
->uid
,
877 uap
->arg
, nd
.nl_cred
);
884 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
885 * void *buf, int buflen)
887 * This function operates on a mount point and executes the specified
888 * operation using the specified control data, and possibly returns data.
890 * The actual number of bytes stored in the result buffer is returned, 0
891 * if none, otherwise an error is returned.
895 sys_mountctl(struct mountctl_args
*uap
)
897 struct thread
*td
= curthread
;
898 struct proc
*p
= td
->td_proc
;
906 * Sanity and permissions checks. We must be root.
909 if (p
->p_ucred
->cr_prison
!= NULL
)
911 if ((uap
->op
!= MOUNTCTL_MOUNTFLAGS
) &&
912 (error
= priv_check(td
, PRIV_ROOT
)) != 0)
916 * Argument length checks
918 if (uap
->ctllen
< 0 || uap
->ctllen
> 1024)
920 if (uap
->buflen
< 0 || uap
->buflen
> 16 * 1024)
922 if (uap
->path
== NULL
)
926 * Allocate the necessary buffers and copyin data
928 path
= objcache_get(namei_oc
, M_WAITOK
);
929 error
= copyinstr(uap
->path
, path
, MAXPATHLEN
, NULL
);
934 ctl
= kmalloc(uap
->ctllen
+ 1, M_TEMP
, M_WAITOK
|M_ZERO
);
935 error
= copyin(uap
->ctl
, ctl
, uap
->ctllen
);
940 buf
= kmalloc(uap
->buflen
+ 1, M_TEMP
, M_WAITOK
|M_ZERO
);
943 * Validate the descriptor
946 fp
= holdfp(p
->p_fd
, uap
->fd
, -1);
956 * Execute the internal kernel function and clean up.
958 error
= kern_mountctl(path
, uap
->op
, fp
, ctl
, uap
->ctllen
, buf
, uap
->buflen
, &uap
->sysmsg_result
);
961 if (error
== 0 && uap
->sysmsg_result
> 0)
962 error
= copyout(buf
, uap
->buf
, uap
->sysmsg_result
);
965 objcache_put(namei_oc
, path
);
974 * Execute a mount control operation by resolving the path to a mount point
975 * and calling vop_mountctl().
977 * Use the mount point from the nch instead of the vnode so nullfs mounts
978 * can properly spike the VOP.
981 kern_mountctl(const char *path
, int op
, struct file
*fp
,
982 const void *ctl
, int ctllen
,
983 void *buf
, int buflen
, int *res
)
987 struct nlookupdata nd
;
992 error
= nlookup_init(&nd
, path
, UIO_SYSSPACE
, NLC_FOLLOW
);
994 error
= nlookup(&nd
);
996 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
997 mp
= nd
.nl_nch
.mount
;
1003 * Must be the root of the filesystem
1005 if ((vp
->v_flag
& (VROOT
|VPFSROOT
)) == 0) {
1009 error
= vop_mountctl(mp
->mnt_vn_use_ops
, op
, fp
, ctl
, ctllen
,
1016 kern_statfs(struct nlookupdata
*nd
, struct statfs
*buf
)
1018 struct thread
*td
= curthread
;
1019 struct proc
*p
= td
->td_proc
;
1022 char *fullpath
, *freepath
;
1025 if ((error
= nlookup(nd
)) != 0)
1027 mp
= nd
->nl_nch
.mount
;
1029 if ((error
= VFS_STATFS(mp
, sp
, nd
->nl_cred
)) != 0)
1032 error
= mount_path(p
, mp
, &fullpath
, &freepath
);
1035 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
1036 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
1037 kfree(freepath
, M_TEMP
);
1039 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
1040 bcopy(sp
, buf
, sizeof(*buf
));
1041 /* Only root should have access to the fsid's. */
1042 if (priv_check(td
, PRIV_ROOT
))
1043 buf
->f_fsid
.val
[0] = buf
->f_fsid
.val
[1] = 0;
1048 * statfs_args(char *path, struct statfs *buf)
1050 * Get filesystem statistics.
1053 sys_statfs(struct statfs_args
*uap
)
1055 struct nlookupdata nd
;
1059 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1061 error
= kern_statfs(&nd
, &buf
);
1064 error
= copyout(&buf
, uap
->buf
, sizeof(*uap
->buf
));
1069 kern_fstatfs(int fd
, struct statfs
*buf
)
1071 struct thread
*td
= curthread
;
1072 struct proc
*p
= td
->td_proc
;
1076 char *fullpath
, *freepath
;
1080 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
1082 mp
= ((struct vnode
*)fp
->f_data
)->v_mount
;
1087 if (fp
->f_cred
== NULL
) {
1092 if ((error
= VFS_STATFS(mp
, sp
, fp
->f_cred
)) != 0)
1095 if ((error
= mount_path(p
, mp
, &fullpath
, &freepath
)) != 0)
1097 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
1098 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
1099 kfree(freepath
, M_TEMP
);
1101 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
1102 bcopy(sp
, buf
, sizeof(*buf
));
1104 /* Only root should have access to the fsid's. */
1105 if (priv_check(td
, PRIV_ROOT
))
1106 buf
->f_fsid
.val
[0] = buf
->f_fsid
.val
[1] = 0;
1114 * fstatfs_args(int fd, struct statfs *buf)
1116 * Get filesystem statistics.
1119 sys_fstatfs(struct fstatfs_args
*uap
)
1124 error
= kern_fstatfs(uap
->fd
, &buf
);
1127 error
= copyout(&buf
, uap
->buf
, sizeof(*uap
->buf
));
1132 kern_statvfs(struct nlookupdata
*nd
, struct statvfs
*buf
)
1138 if ((error
= nlookup(nd
)) != 0)
1140 mp
= nd
->nl_nch
.mount
;
1141 sp
= &mp
->mnt_vstat
;
1142 if ((error
= VFS_STATVFS(mp
, sp
, nd
->nl_cred
)) != 0)
1146 if (mp
->mnt_flag
& MNT_RDONLY
)
1147 sp
->f_flag
|= ST_RDONLY
;
1148 if (mp
->mnt_flag
& MNT_NOSUID
)
1149 sp
->f_flag
|= ST_NOSUID
;
1150 bcopy(sp
, buf
, sizeof(*buf
));
1155 * statfs_args(char *path, struct statfs *buf)
1157 * Get filesystem statistics.
1160 sys_statvfs(struct statvfs_args
*uap
)
1162 struct nlookupdata nd
;
1166 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1168 error
= kern_statvfs(&nd
, &buf
);
1171 error
= copyout(&buf
, uap
->buf
, sizeof(*uap
->buf
));
1176 kern_fstatvfs(int fd
, struct statvfs
*buf
)
1178 struct thread
*td
= curthread
;
1179 struct proc
*p
= td
->td_proc
;
1186 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
1188 mp
= ((struct vnode
*)fp
->f_data
)->v_mount
;
1193 if (fp
->f_cred
== NULL
) {
1197 sp
= &mp
->mnt_vstat
;
1198 if ((error
= VFS_STATVFS(mp
, sp
, fp
->f_cred
)) != 0)
1202 if (mp
->mnt_flag
& MNT_RDONLY
)
1203 sp
->f_flag
|= ST_RDONLY
;
1204 if (mp
->mnt_flag
& MNT_NOSUID
)
1205 sp
->f_flag
|= ST_NOSUID
;
1207 bcopy(sp
, buf
, sizeof(*buf
));
1215 * fstatfs_args(int fd, struct statfs *buf)
1217 * Get filesystem statistics.
1220 sys_fstatvfs(struct fstatvfs_args
*uap
)
1225 error
= kern_fstatvfs(uap
->fd
, &buf
);
1228 error
= copyout(&buf
, uap
->buf
, sizeof(*uap
->buf
));
1233 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
1235 * Get statistics on all filesystems.
1238 struct getfsstat_info
{
1239 struct statfs
*sfsp
;
1247 static int getfsstat_callback(struct mount
*, void *);
1251 sys_getfsstat(struct getfsstat_args
*uap
)
1253 struct thread
*td
= curthread
;
1254 struct proc
*p
= td
->td_proc
;
1255 struct getfsstat_info info
;
1257 bzero(&info
, sizeof(info
));
1259 info
.maxcount
= uap
->bufsize
/ sizeof(struct statfs
);
1260 info
.sfsp
= uap
->buf
;
1262 info
.flags
= uap
->flags
;
1265 mountlist_scan(getfsstat_callback
, &info
, MNTSCAN_FORWARD
);
1266 if (info
.sfsp
&& info
.count
> info
.maxcount
)
1267 uap
->sysmsg_result
= info
.maxcount
;
1269 uap
->sysmsg_result
= info
.count
;
1270 return (info
.error
);
1274 getfsstat_callback(struct mount
*mp
, void *data
)
1276 struct getfsstat_info
*info
= data
;
1282 if (info
->sfsp
&& info
->count
< info
->maxcount
) {
1283 if (info
->p
&& !chroot_visible_mnt(mp
, info
->p
))
1288 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1289 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1290 * overrides MNT_WAIT.
1292 if (((info
->flags
& (MNT_LAZY
|MNT_NOWAIT
)) == 0 ||
1293 (info
->flags
& MNT_WAIT
)) &&
1294 (error
= VFS_STATFS(mp
, sp
, info
->p
->p_ucred
))) {
1297 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
1299 error
= mount_path(info
->p
, mp
, &fullpath
, &freepath
);
1301 info
->error
= error
;
1304 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
1305 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
1306 kfree(freepath
, M_TEMP
);
1308 error
= copyout(sp
, info
->sfsp
, sizeof(*sp
));
1310 info
->error
= error
;
1320 * getvfsstat_args(struct statfs *buf, struct statvfs *vbuf,
1321 long bufsize, int flags)
1323 * Get statistics on all filesystems.
1326 struct getvfsstat_info
{
1327 struct statfs
*sfsp
;
1328 struct statvfs
*vsfsp
;
1336 static int getvfsstat_callback(struct mount
*, void *);
1340 sys_getvfsstat(struct getvfsstat_args
*uap
)
1342 struct thread
*td
= curthread
;
1343 struct proc
*p
= td
->td_proc
;
1344 struct getvfsstat_info info
;
1346 bzero(&info
, sizeof(info
));
1348 info
.maxcount
= uap
->vbufsize
/ sizeof(struct statvfs
);
1349 info
.sfsp
= uap
->buf
;
1350 info
.vsfsp
= uap
->vbuf
;
1352 info
.flags
= uap
->flags
;
1355 mountlist_scan(getvfsstat_callback
, &info
, MNTSCAN_FORWARD
);
1356 if (info
.vsfsp
&& info
.count
> info
.maxcount
)
1357 uap
->sysmsg_result
= info
.maxcount
;
1359 uap
->sysmsg_result
= info
.count
;
1360 return (info
.error
);
1364 getvfsstat_callback(struct mount
*mp
, void *data
)
1366 struct getvfsstat_info
*info
= data
;
1368 struct statvfs
*vsp
;
1373 if (info
->vsfsp
&& info
->count
< info
->maxcount
) {
1374 if (info
->p
&& !chroot_visible_mnt(mp
, info
->p
))
1377 vsp
= &mp
->mnt_vstat
;
1380 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1381 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1382 * overrides MNT_WAIT.
1384 if (((info
->flags
& (MNT_LAZY
|MNT_NOWAIT
)) == 0 ||
1385 (info
->flags
& MNT_WAIT
)) &&
1386 (error
= VFS_STATFS(mp
, sp
, info
->p
->p_ucred
))) {
1389 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
1391 if (((info
->flags
& (MNT_LAZY
|MNT_NOWAIT
)) == 0 ||
1392 (info
->flags
& MNT_WAIT
)) &&
1393 (error
= VFS_STATVFS(mp
, vsp
, info
->p
->p_ucred
))) {
1397 if (mp
->mnt_flag
& MNT_RDONLY
)
1398 vsp
->f_flag
|= ST_RDONLY
;
1399 if (mp
->mnt_flag
& MNT_NOSUID
)
1400 vsp
->f_flag
|= ST_NOSUID
;
1402 error
= mount_path(info
->p
, mp
, &fullpath
, &freepath
);
1404 info
->error
= error
;
1407 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
1408 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
1409 kfree(freepath
, M_TEMP
);
1411 error
= copyout(sp
, info
->sfsp
, sizeof(*sp
));
1413 error
= copyout(vsp
, info
->vsfsp
, sizeof(*vsp
));
1415 info
->error
= error
;
1427 * fchdir_args(int fd)
1429 * Change current working directory to a given file descriptor.
1433 sys_fchdir(struct fchdir_args
*uap
)
1435 struct thread
*td
= curthread
;
1436 struct proc
*p
= td
->td_proc
;
1437 struct filedesc
*fdp
= p
->p_fd
;
1438 struct vnode
*vp
, *ovp
;
1441 struct nchandle nch
, onch
, tnch
;
1444 if ((error
= holdvnode(fdp
, uap
->fd
, &fp
)) != 0)
1446 vp
= (struct vnode
*)fp
->f_data
;
1448 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
1449 if (vp
->v_type
!= VDIR
|| fp
->f_nchandle
.ncp
== NULL
)
1452 error
= VOP_ACCESS(vp
, VEXEC
, p
->p_ucred
);
1458 cache_copy(&fp
->f_nchandle
, &nch
);
1461 * If the ncp has become a mount point, traverse through
1465 while (!error
&& (nch
.ncp
->nc_flag
& NCF_ISMOUNTPT
) &&
1466 (mp
= cache_findmount(&nch
)) != NULL
1468 error
= nlookup_mp(mp
, &tnch
);
1470 cache_unlock(&tnch
); /* leave ref intact */
1472 vp
= tnch
.ncp
->nc_vp
;
1473 error
= vget(vp
, LK_SHARED
);
1474 KKASSERT(error
== 0);
1481 onch
= fdp
->fd_ncdir
;
1482 vn_unlock(vp
); /* leave ref intact */
1484 fdp
->fd_ncdir
= nch
;
1496 kern_chdir(struct nlookupdata
*nd
)
1498 struct thread
*td
= curthread
;
1499 struct proc
*p
= td
->td_proc
;
1500 struct filedesc
*fdp
= p
->p_fd
;
1501 struct vnode
*vp
, *ovp
;
1502 struct nchandle onch
;
1505 if ((error
= nlookup(nd
)) != 0)
1507 if ((vp
= nd
->nl_nch
.ncp
->nc_vp
) == NULL
)
1509 if ((error
= vget(vp
, LK_SHARED
)) != 0)
1512 error
= checkvp_chdir(vp
, td
);
1516 onch
= fdp
->fd_ncdir
;
1517 cache_unlock(&nd
->nl_nch
); /* leave reference intact */
1518 fdp
->fd_ncdir
= nd
->nl_nch
;
1522 cache_zero(&nd
->nl_nch
);
1530 * chdir_args(char *path)
1532 * Change current working directory (``.'').
1535 sys_chdir(struct chdir_args
*uap
)
1537 struct nlookupdata nd
;
1540 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1542 error
= kern_chdir(&nd
);
1548 * Helper function for raised chroot(2) security function: Refuse if
1549 * any filedescriptors are open directories.
1552 chroot_refuse_vdir_fds(struct filedesc
*fdp
)
1559 for (fd
= 0; fd
< fdp
->fd_nfiles
; fd
++) {
1560 if ((error
= holdvnode(fdp
, fd
, &fp
)) != 0)
1562 vp
= (struct vnode
*)fp
->f_data
;
1563 if (vp
->v_type
!= VDIR
) {
1574 * This sysctl determines if we will allow a process to chroot(2) if it
1575 * has a directory open:
1576 * 0: disallowed for all processes.
1577 * 1: allowed for processes that were not already chroot(2)'ed.
1578 * 2: allowed for all processes.
1581 static int chroot_allow_open_directories
= 1;
1583 SYSCTL_INT(_kern
, OID_AUTO
, chroot_allow_open_directories
, CTLFLAG_RW
,
1584 &chroot_allow_open_directories
, 0, "");
1587 * chroot to the specified namecache entry. We obtain the vp from the
1588 * namecache data. The passed ncp must be locked and referenced and will
1589 * remain locked and referenced on return.
1592 kern_chroot(struct nchandle
*nch
)
1594 struct thread
*td
= curthread
;
1595 struct proc
*p
= td
->td_proc
;
1596 struct filedesc
*fdp
= p
->p_fd
;
1601 * Only privileged user can chroot
1603 error
= priv_check_cred(p
->p_ucred
, PRIV_VFS_CHROOT
, 0);
1608 * Disallow open directory descriptors (fchdir() breakouts).
1610 if (chroot_allow_open_directories
== 0 ||
1611 (chroot_allow_open_directories
== 1 && fdp
->fd_rdir
!= rootvnode
)) {
1612 if ((error
= chroot_refuse_vdir_fds(fdp
)) != 0)
1615 if ((vp
= nch
->ncp
->nc_vp
) == NULL
)
1618 if ((error
= vget(vp
, LK_SHARED
)) != 0)
1622 * Check the validity of vp as a directory to change to and
1623 * associate it with rdir/jdir.
1625 error
= checkvp_chdir(vp
, td
);
1626 vn_unlock(vp
); /* leave reference intact */
1628 vrele(fdp
->fd_rdir
);
1629 fdp
->fd_rdir
= vp
; /* reference inherited by fd_rdir */
1630 cache_drop(&fdp
->fd_nrdir
);
1631 cache_copy(nch
, &fdp
->fd_nrdir
);
1632 if (fdp
->fd_jdir
== NULL
) {
1635 cache_copy(nch
, &fdp
->fd_njdir
);
1644 * chroot_args(char *path)
1646 * Change notion of root (``/'') directory.
1650 sys_chroot(struct chroot_args
*uap
)
1652 struct thread
*td
= curthread
;
1653 struct nlookupdata nd
;
1656 KKASSERT(td
->td_proc
);
1657 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
1662 nd
.nl_flags
|= NLC_EXEC
;
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
, 0);
1834 error
= kern_open(&nd
, uap
->flags
,
1835 uap
->mode
, &uap
->sysmsg_result
);
1842 * openat_args(int fd, char *path, int flags, int mode)
1845 sys_openat(struct openat_args
*uap
)
1847 struct nlookupdata nd
;
1851 error
= nlookup_init_at(&nd
, &fp
, uap
->fd
, uap
->path
, UIO_USERSPACE
, 0);
1853 error
= kern_open(&nd
, uap
->flags
, uap
->mode
,
1854 &uap
->sysmsg_result
);
1856 nlookup_done_at(&nd
, fp
);
1861 kern_mknod(struct nlookupdata
*nd
, int mode
, int rmajor
, int rminor
)
1863 struct thread
*td
= curthread
;
1864 struct proc
*p
= td
->td_proc
;
1873 vattr
.va_mode
= (mode
& ALLPERMS
) &~ p
->p_fd
->fd_cmask
;
1874 vattr
.va_rmajor
= rmajor
;
1875 vattr
.va_rminor
= rminor
;
1877 switch (mode
& S_IFMT
) {
1878 case S_IFMT
: /* used by badsect to flag bad sectors */
1879 error
= priv_check_cred(p
->p_ucred
, PRIV_VFS_MKNOD_BAD
, 0);
1880 vattr
.va_type
= VBAD
;
1883 error
= priv_check(td
, PRIV_VFS_MKNOD_DEV
);
1884 vattr
.va_type
= VCHR
;
1887 error
= priv_check(td
, PRIV_VFS_MKNOD_DEV
);
1888 vattr
.va_type
= VBLK
;
1891 error
= priv_check_cred(p
->p_ucred
, PRIV_VFS_MKNOD_WHT
, 0);
1894 case S_IFDIR
: /* special directories support for HAMMER */
1895 error
= priv_check_cred(p
->p_ucred
, PRIV_VFS_MKNOD_DIR
, 0);
1896 vattr
.va_type
= VDIR
;
1907 nd
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
1908 if ((error
= nlookup(nd
)) != 0)
1910 if (nd
->nl_nch
.ncp
->nc_vp
)
1912 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
1916 error
= VOP_NWHITEOUT(&nd
->nl_nch
, nd
->nl_dvp
,
1917 nd
->nl_cred
, NAMEI_CREATE
);
1920 error
= VOP_NMKNOD(&nd
->nl_nch
, nd
->nl_dvp
,
1921 &vp
, nd
->nl_cred
, &vattr
);
1929 * mknod_args(char *path, int mode, int dev)
1931 * Create a special file.
1934 sys_mknod(struct mknod_args
*uap
)
1936 struct nlookupdata nd
;
1939 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
1941 error
= kern_mknod(&nd
, uap
->mode
,
1942 umajor(uap
->dev
), uminor(uap
->dev
));
1949 kern_mkfifo(struct nlookupdata
*nd
, int mode
)
1951 struct thread
*td
= curthread
;
1952 struct proc
*p
= td
->td_proc
;
1959 nd
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
1960 if ((error
= nlookup(nd
)) != 0)
1962 if (nd
->nl_nch
.ncp
->nc_vp
)
1964 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
1968 vattr
.va_type
= VFIFO
;
1969 vattr
.va_mode
= (mode
& ALLPERMS
) &~ p
->p_fd
->fd_cmask
;
1971 error
= VOP_NMKNOD(&nd
->nl_nch
, nd
->nl_dvp
, &vp
, nd
->nl_cred
, &vattr
);
1978 * mkfifo_args(char *path, int mode)
1980 * Create a named pipe.
1983 sys_mkfifo(struct mkfifo_args
*uap
)
1985 struct nlookupdata nd
;
1988 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
1990 error
= kern_mkfifo(&nd
, uap
->mode
);
1995 static int hardlink_check_uid
= 0;
1996 SYSCTL_INT(_security
, OID_AUTO
, hardlink_check_uid
, CTLFLAG_RW
,
1997 &hardlink_check_uid
, 0,
1998 "Unprivileged processes cannot create hard links to files owned by other "
2000 static int hardlink_check_gid
= 0;
2001 SYSCTL_INT(_security
, OID_AUTO
, hardlink_check_gid
, CTLFLAG_RW
,
2002 &hardlink_check_gid
, 0,
2003 "Unprivileged processes cannot create hard links to files owned by other "
2007 can_hardlink(struct vnode
*vp
, struct thread
*td
, struct ucred
*cred
)
2013 * Shortcut if disabled
2015 if (hardlink_check_uid
== 0 && hardlink_check_gid
== 0)
2019 * Privileged user can always hardlink
2021 if (priv_check_cred(cred
, PRIV_VFS_LINK
, 0) == 0)
2025 * Otherwise only if the originating file is owned by the
2026 * same user or group. Note that any group is allowed if
2027 * the file is owned by the caller.
2029 error
= VOP_GETATTR(vp
, &va
);
2033 if (hardlink_check_uid
) {
2034 if (cred
->cr_uid
!= va
.va_uid
)
2038 if (hardlink_check_gid
) {
2039 if (cred
->cr_uid
!= va
.va_uid
&& !groupmember(va
.va_gid
, cred
))
2047 kern_link(struct nlookupdata
*nd
, struct nlookupdata
*linknd
)
2049 struct thread
*td
= curthread
;
2054 * Lookup the source and obtained a locked vnode.
2056 * You may only hardlink a file which you have write permission
2057 * on or which you own.
2059 * XXX relookup on vget failure / race ?
2062 nd
->nl_flags
|= NLC_WRITE
| NLC_OWN
| NLC_HLINK
;
2063 if ((error
= nlookup(nd
)) != 0)
2065 vp
= nd
->nl_nch
.ncp
->nc_vp
;
2066 KKASSERT(vp
!= NULL
);
2067 if (vp
->v_type
== VDIR
)
2068 return (EPERM
); /* POSIX */
2069 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2071 if ((error
= vget(vp
, LK_EXCLUSIVE
)) != 0)
2075 * Unlock the source so we can lookup the target without deadlocking
2076 * (XXX vp is locked already, possible other deadlock?). The target
2079 KKASSERT(nd
->nl_flags
& NLC_NCPISLOCKED
);
2080 nd
->nl_flags
&= ~NLC_NCPISLOCKED
;
2081 cache_unlock(&nd
->nl_nch
);
2083 linknd
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
2084 if ((error
= nlookup(linknd
)) != 0) {
2088 if (linknd
->nl_nch
.ncp
->nc_vp
) {
2094 * Finally run the new API VOP.
2096 error
= can_hardlink(vp
, td
, td
->td_proc
->p_ucred
);
2098 error
= VOP_NLINK(&linknd
->nl_nch
, linknd
->nl_dvp
,
2099 vp
, linknd
->nl_cred
);
2106 * link_args(char *path, char *link)
2108 * Make a hard file link.
2111 sys_link(struct link_args
*uap
)
2113 struct nlookupdata nd
, linknd
;
2116 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2118 error
= nlookup_init(&linknd
, uap
->link
, UIO_USERSPACE
, 0);
2120 error
= kern_link(&nd
, &linknd
);
2121 nlookup_done(&linknd
);
2128 kern_symlink(struct nlookupdata
*nd
, char *path
, int mode
)
2136 nd
->nl_flags
|= NLC_CREATE
| NLC_REFDVP
;
2137 if ((error
= nlookup(nd
)) != 0)
2139 if (nd
->nl_nch
.ncp
->nc_vp
)
2141 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2145 vattr
.va_mode
= mode
;
2146 error
= VOP_NSYMLINK(&nd
->nl_nch
, dvp
, &vp
, nd
->nl_cred
, &vattr
, path
);
2153 * symlink(char *path, char *link)
2155 * Make a symbolic link.
2158 sys_symlink(struct symlink_args
*uap
)
2160 struct thread
*td
= curthread
;
2161 struct nlookupdata nd
;
2166 path
= objcache_get(namei_oc
, M_WAITOK
);
2167 error
= copyinstr(uap
->path
, path
, MAXPATHLEN
, NULL
);
2169 error
= nlookup_init(&nd
, uap
->link
, UIO_USERSPACE
, 0);
2171 mode
= ACCESSPERMS
& ~td
->td_proc
->p_fd
->fd_cmask
;
2172 error
= kern_symlink(&nd
, path
, mode
);
2176 objcache_put(namei_oc
, path
);
2181 * undelete_args(char *path)
2183 * Delete a whiteout from the filesystem.
2187 sys_undelete(struct undelete_args
*uap
)
2189 struct nlookupdata nd
;
2192 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2194 nd
.nl_flags
|= NLC_DELETE
| NLC_REFDVP
;
2196 error
= nlookup(&nd
);
2198 error
= ncp_writechk(&nd
.nl_nch
);
2200 error
= VOP_NWHITEOUT(&nd
.nl_nch
, nd
.nl_dvp
, nd
.nl_cred
,
2208 kern_unlink(struct nlookupdata
*nd
)
2213 nd
->nl_flags
|= NLC_DELETE
| NLC_REFDVP
;
2214 if ((error
= nlookup(nd
)) != 0)
2216 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2218 error
= VOP_NREMOVE(&nd
->nl_nch
, nd
->nl_dvp
, nd
->nl_cred
);
2223 * unlink_args(char *path)
2225 * Delete a name from the filesystem.
2228 sys_unlink(struct unlink_args
*uap
)
2230 struct nlookupdata nd
;
2233 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2235 error
= kern_unlink(&nd
);
2241 kern_lseek(int fd
, off_t offset
, int whence
, off_t
*res
)
2243 struct thread
*td
= curthread
;
2244 struct proc
*p
= td
->td_proc
;
2251 fp
= holdfp(p
->p_fd
, fd
, -1);
2254 if (fp
->f_type
!= DTYPE_VNODE
) {
2258 vp
= (struct vnode
*)fp
->f_data
;
2262 new_offset
= fp
->f_offset
+ offset
;
2266 error
= VOP_GETATTR(vp
, &vattr
);
2267 new_offset
= offset
+ vattr
.va_size
;
2270 new_offset
= offset
;
2280 * Validate the seek position. Negative offsets are not allowed
2281 * for regular files or directories.
2283 * Normally we would also not want to allow negative offsets for
2284 * character and block-special devices. However kvm addresses
2285 * on 64 bit architectures might appear to be negative and must
2289 if (new_offset
< 0 &&
2290 (vp
->v_type
== VREG
|| vp
->v_type
== VDIR
)) {
2293 fp
->f_offset
= new_offset
;
2296 *res
= fp
->f_offset
;
2303 * lseek_args(int fd, int pad, off_t offset, int whence)
2305 * Reposition read/write file offset.
2308 sys_lseek(struct lseek_args
*uap
)
2312 error
= kern_lseek(uap
->fd
, uap
->offset
, uap
->whence
,
2313 &uap
->sysmsg_offset
);
2319 kern_access(struct nlookupdata
*nd
, int aflags
)
2324 if ((error
= nlookup(nd
)) != 0)
2327 error
= cache_vget(&nd
->nl_nch
, nd
->nl_cred
, LK_EXCLUSIVE
, &vp
);
2331 /* Flags == 0 means only check for existence. */
2340 if ((flags
& VWRITE
) == 0 ||
2341 (error
= vn_writechk(vp
, &nd
->nl_nch
)) == 0)
2342 error
= VOP_ACCESS(vp
, flags
, nd
->nl_cred
);
2345 * If the file handle is stale we have to re-resolve the
2346 * entry. This is a hack at the moment.
2348 if (error
== ESTALE
) {
2350 cache_setunresolved(&nd
->nl_nch
);
2351 error
= cache_resolve(&nd
->nl_nch
, nd
->nl_cred
);
2364 * access_args(char *path, int flags)
2366 * Check access permissions.
2369 sys_access(struct access_args
*uap
)
2371 struct nlookupdata nd
;
2374 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2376 error
= kern_access(&nd
, uap
->flags
);
2382 kern_stat(struct nlookupdata
*nd
, struct stat
*st
)
2388 if ((error
= nlookup(nd
)) != 0)
2391 if ((vp
= nd
->nl_nch
.ncp
->nc_vp
) == NULL
)
2395 if ((error
= vget(vp
, LK_SHARED
)) != 0)
2397 error
= vn_stat(vp
, st
, nd
->nl_cred
);
2400 * If the file handle is stale we have to re-resolve the entry. This
2401 * is a hack at the moment.
2403 if (error
== ESTALE
) {
2405 cache_setunresolved(&nd
->nl_nch
);
2406 error
= cache_resolve(&nd
->nl_nch
, nd
->nl_cred
);
2416 * stat_args(char *path, struct stat *ub)
2418 * Get file status; this version follows links.
2421 sys_stat(struct stat_args
*uap
)
2423 struct nlookupdata nd
;
2427 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2429 error
= kern_stat(&nd
, &st
);
2431 error
= copyout(&st
, uap
->ub
, sizeof(*uap
->ub
));
2438 * lstat_args(char *path, struct stat *ub)
2440 * Get file status; this version does not follow links.
2443 sys_lstat(struct lstat_args
*uap
)
2445 struct nlookupdata nd
;
2449 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2451 error
= kern_stat(&nd
, &st
);
2453 error
= copyout(&st
, uap
->ub
, sizeof(*uap
->ub
));
2460 * fstatat_args(int fd, char *path, struct stat *sb, int flags)
2462 * Get status of file pointed to by fd/path.
2465 sys_fstatat(struct fstatat_args
*uap
)
2467 struct nlookupdata nd
;
2473 if (uap
->flags
& ~_AT_SYMLINK_MASK
)
2476 flags
= (uap
->flags
& AT_SYMLINK_NOFOLLOW
) ? 0 : NLC_FOLLOW
;
2478 error
= nlookup_init_at(&nd
, &fp
, uap
->fd
, uap
->path
,
2479 UIO_USERSPACE
, flags
);
2481 error
= kern_stat(&nd
, &st
);
2483 error
= copyout(&st
, uap
->sb
, sizeof(*uap
->sb
));
2485 nlookup_done_at(&nd
, fp
);
2490 * pathconf_Args(char *path, int name)
2492 * Get configurable pathname variables.
2496 sys_pathconf(struct pathconf_args
*uap
)
2498 struct nlookupdata nd
;
2503 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2505 error
= nlookup(&nd
);
2507 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
2510 error
= VOP_PATHCONF(vp
, uap
->name
, &uap
->sysmsg_reg
);
2518 * kern_readlink isn't properly split yet. There is a copyin burried
2519 * in VOP_READLINK().
2522 kern_readlink(struct nlookupdata
*nd
, char *buf
, int count
, int *res
)
2524 struct thread
*td
= curthread
;
2525 struct proc
*p
= td
->td_proc
;
2531 if ((error
= nlookup(nd
)) != 0)
2533 error
= cache_vget(&nd
->nl_nch
, nd
->nl_cred
, LK_EXCLUSIVE
, &vp
);
2536 if (vp
->v_type
!= VLNK
) {
2539 aiov
.iov_base
= buf
;
2540 aiov
.iov_len
= count
;
2541 auio
.uio_iov
= &aiov
;
2542 auio
.uio_iovcnt
= 1;
2543 auio
.uio_offset
= 0;
2544 auio
.uio_rw
= UIO_READ
;
2545 auio
.uio_segflg
= UIO_USERSPACE
;
2547 auio
.uio_resid
= count
;
2548 error
= VOP_READLINK(vp
, &auio
, p
->p_ucred
);
2551 *res
= count
- auio
.uio_resid
;
2556 * readlink_args(char *path, char *buf, int count)
2558 * Return target name of a symbolic link.
2561 sys_readlink(struct readlink_args
*uap
)
2563 struct nlookupdata nd
;
2566 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2568 error
= kern_readlink(&nd
, uap
->buf
, uap
->count
,
2569 &uap
->sysmsg_result
);
2576 setfflags(struct vnode
*vp
, int flags
)
2578 struct thread
*td
= curthread
;
2579 struct proc
*p
= td
->td_proc
;
2584 * Prevent non-root users from setting flags on devices. When
2585 * a device is reused, users can retain ownership of the device
2586 * if they are allowed to set flags and programs assume that
2587 * chown can't fail when done as root.
2589 if ((vp
->v_type
== VCHR
|| vp
->v_type
== VBLK
) &&
2590 ((error
= priv_check_cred(p
->p_ucred
, PRIV_VFS_CHFLAGS_DEV
, 0)) != 0))
2594 * note: vget is required for any operation that might mod the vnode
2595 * so VINACTIVE is properly cleared.
2597 if ((error
= vget(vp
, LK_EXCLUSIVE
)) == 0) {
2599 vattr
.va_flags
= flags
;
2600 error
= VOP_SETATTR(vp
, &vattr
, p
->p_ucred
);
2607 * chflags(char *path, int flags)
2609 * Change flags of a file given a path name.
2613 sys_chflags(struct chflags_args
*uap
)
2615 struct nlookupdata nd
;
2620 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2622 error
= nlookup(&nd
);
2624 error
= ncp_writechk(&nd
.nl_nch
);
2626 error
= cache_vref(&nd
.nl_nch
, nd
.nl_cred
, &vp
);
2629 error
= setfflags(vp
, uap
->flags
);
2636 * lchflags(char *path, int flags)
2638 * Change flags of a file given a path name, but don't follow symlinks.
2642 sys_lchflags(struct lchflags_args
*uap
)
2644 struct nlookupdata nd
;
2649 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2651 error
= nlookup(&nd
);
2653 error
= ncp_writechk(&nd
.nl_nch
);
2655 error
= cache_vref(&nd
.nl_nch
, nd
.nl_cred
, &vp
);
2658 error
= setfflags(vp
, uap
->flags
);
2665 * fchflags_args(int fd, int flags)
2667 * Change flags of a file given a file descriptor.
2671 sys_fchflags(struct fchflags_args
*uap
)
2673 struct thread
*td
= curthread
;
2674 struct proc
*p
= td
->td_proc
;
2678 if ((error
= holdvnode(p
->p_fd
, uap
->fd
, &fp
)) != 0)
2680 if (fp
->f_nchandle
.ncp
)
2681 error
= ncp_writechk(&fp
->f_nchandle
);
2683 error
= setfflags((struct vnode
*) fp
->f_data
, uap
->flags
);
2689 setfmode(struct vnode
*vp
, int mode
)
2691 struct thread
*td
= curthread
;
2692 struct proc
*p
= td
->td_proc
;
2697 * note: vget is required for any operation that might mod the vnode
2698 * so VINACTIVE is properly cleared.
2700 if ((error
= vget(vp
, LK_EXCLUSIVE
)) == 0) {
2702 vattr
.va_mode
= mode
& ALLPERMS
;
2703 error
= VOP_SETATTR(vp
, &vattr
, p
->p_ucred
);
2710 kern_chmod(struct nlookupdata
*nd
, int mode
)
2715 if ((error
= nlookup(nd
)) != 0)
2717 if ((error
= cache_vref(&nd
->nl_nch
, nd
->nl_cred
, &vp
)) != 0)
2719 if ((error
= ncp_writechk(&nd
->nl_nch
)) == 0)
2720 error
= setfmode(vp
, mode
);
2726 * chmod_args(char *path, int mode)
2728 * Change mode of a file given path name.
2732 sys_chmod(struct chmod_args
*uap
)
2734 struct nlookupdata nd
;
2737 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2739 error
= kern_chmod(&nd
, uap
->mode
);
2745 * lchmod_args(char *path, int mode)
2747 * Change mode of a file given path name (don't follow links.)
2751 sys_lchmod(struct lchmod_args
*uap
)
2753 struct nlookupdata nd
;
2756 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2758 error
= kern_chmod(&nd
, uap
->mode
);
2764 * fchmod_args(int fd, int mode)
2766 * Change mode of a file given a file descriptor.
2770 sys_fchmod(struct fchmod_args
*uap
)
2772 struct thread
*td
= curthread
;
2773 struct proc
*p
= td
->td_proc
;
2777 if ((error
= holdvnode(p
->p_fd
, uap
->fd
, &fp
)) != 0)
2779 if (fp
->f_nchandle
.ncp
)
2780 error
= ncp_writechk(&fp
->f_nchandle
);
2782 error
= setfmode((struct vnode
*)fp
->f_data
, uap
->mode
);
2788 * fchmodat_args(char *path, int mode)
2790 * Change mode of a file pointed to by fd/path.
2793 sys_fchmodat(struct fchmodat_args
*uap
)
2795 struct nlookupdata nd
;
2800 if (uap
->flags
& ~_AT_SYMLINK_MASK
)
2802 flags
= (uap
->flags
& AT_SYMLINK_NOFOLLOW
) ? 0 : NLC_FOLLOW
;
2804 error
= nlookup_init_at(&nd
, &fp
, uap
->fd
, uap
->path
,
2805 UIO_USERSPACE
, flags
);
2807 error
= kern_chmod(&nd
, uap
->mode
);
2808 nlookup_done_at(&nd
, fp
);
2813 setfown(struct vnode
*vp
, uid_t uid
, gid_t gid
)
2815 struct thread
*td
= curthread
;
2816 struct proc
*p
= td
->td_proc
;
2821 * note: vget is required for any operation that might mod the vnode
2822 * so VINACTIVE is properly cleared.
2824 if ((error
= vget(vp
, LK_EXCLUSIVE
)) == 0) {
2828 error
= VOP_SETATTR(vp
, &vattr
, p
->p_ucred
);
2835 kern_chown(struct nlookupdata
*nd
, int uid
, int gid
)
2840 if ((error
= nlookup(nd
)) != 0)
2842 if ((error
= cache_vref(&nd
->nl_nch
, nd
->nl_cred
, &vp
)) != 0)
2844 if ((error
= ncp_writechk(&nd
->nl_nch
)) == 0)
2845 error
= setfown(vp
, uid
, gid
);
2851 * chown(char *path, int uid, int gid)
2853 * Set ownership given a path name.
2856 sys_chown(struct chown_args
*uap
)
2858 struct nlookupdata nd
;
2861 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
2863 error
= kern_chown(&nd
, uap
->uid
, uap
->gid
);
2869 * lchown_args(char *path, int uid, int gid)
2871 * Set ownership given a path name, do not cross symlinks.
2874 sys_lchown(struct lchown_args
*uap
)
2876 struct nlookupdata nd
;
2879 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
2881 error
= kern_chown(&nd
, uap
->uid
, uap
->gid
);
2887 * fchown_args(int fd, int uid, int gid)
2889 * Set ownership given a file descriptor.
2893 sys_fchown(struct fchown_args
*uap
)
2895 struct thread
*td
= curthread
;
2896 struct proc
*p
= td
->td_proc
;
2900 if ((error
= holdvnode(p
->p_fd
, uap
->fd
, &fp
)) != 0)
2902 if (fp
->f_nchandle
.ncp
)
2903 error
= ncp_writechk(&fp
->f_nchandle
);
2905 error
= setfown((struct vnode
*)fp
->f_data
, uap
->uid
, uap
->gid
);
2911 * fchownat(int fd, char *path, int uid, int gid, int flags)
2913 * Set ownership of file pointed to by fd/path.
2916 sys_fchownat(struct fchownat_args
*uap
)
2918 struct nlookupdata nd
;
2923 if (uap
->flags
& ~_AT_SYMLINK_MASK
)
2925 flags
= (uap
->flags
& AT_SYMLINK_NOFOLLOW
) ? 0 : NLC_FOLLOW
;
2927 error
= nlookup_init_at(&nd
, &fp
, uap
->fd
, uap
->path
,
2928 UIO_USERSPACE
, flags
);
2930 error
= kern_chown(&nd
, uap
->uid
, uap
->gid
);
2931 nlookup_done_at(&nd
, fp
);
2937 getutimes(const struct timeval
*tvp
, struct timespec
*tsp
)
2939 struct timeval tv
[2];
2943 TIMEVAL_TO_TIMESPEC(&tv
[0], &tsp
[0]);
2946 TIMEVAL_TO_TIMESPEC(&tvp
[0], &tsp
[0]);
2947 TIMEVAL_TO_TIMESPEC(&tvp
[1], &tsp
[1]);
2953 setutimes(struct vnode
*vp
, struct vattr
*vattr
,
2954 const struct timespec
*ts
, int nullflag
)
2956 struct thread
*td
= curthread
;
2957 struct proc
*p
= td
->td_proc
;
2961 vattr
->va_atime
= ts
[0];
2962 vattr
->va_mtime
= ts
[1];
2964 vattr
->va_vaflags
|= VA_UTIMES_NULL
;
2965 error
= VOP_SETATTR(vp
, vattr
, p
->p_ucred
);
2971 kern_utimes(struct nlookupdata
*nd
, struct timeval
*tptr
)
2973 struct timespec ts
[2];
2978 if ((error
= getutimes(tptr
, ts
)) != 0)
2982 * NOTE: utimes() succeeds for the owner even if the file
2983 * is not user-writable.
2985 nd
->nl_flags
|= NLC_OWN
| NLC_WRITE
;
2987 if ((error
= nlookup(nd
)) != 0)
2989 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
2991 if ((error
= cache_vref(&nd
->nl_nch
, nd
->nl_cred
, &vp
)) != 0)
2995 * note: vget is required for any operation that might mod the vnode
2996 * so VINACTIVE is properly cleared.
2998 if ((error
= vn_writechk(vp
, &nd
->nl_nch
)) == 0) {
2999 error
= vget(vp
, LK_EXCLUSIVE
);
3001 error
= setutimes(vp
, &vattr
, ts
, (tptr
== NULL
));
3010 * utimes_args(char *path, struct timeval *tptr)
3012 * Set the access and modification times of a file.
3015 sys_utimes(struct utimes_args
*uap
)
3017 struct timeval tv
[2];
3018 struct nlookupdata nd
;
3022 error
= copyin(uap
->tptr
, tv
, sizeof(tv
));
3026 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
3028 error
= kern_utimes(&nd
, uap
->tptr
? tv
: NULL
);
3034 * lutimes_args(char *path, struct timeval *tptr)
3036 * Set the access and modification times of a file.
3039 sys_lutimes(struct lutimes_args
*uap
)
3041 struct timeval tv
[2];
3042 struct nlookupdata nd
;
3046 error
= copyin(uap
->tptr
, tv
, sizeof(tv
));
3050 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
3052 error
= kern_utimes(&nd
, uap
->tptr
? tv
: NULL
);
3058 * Set utimes on a file descriptor. The creds used to open the
3059 * file are used to determine whether the operation is allowed
3063 kern_futimes(int fd
, struct timeval
*tptr
)
3065 struct thread
*td
= curthread
;
3066 struct proc
*p
= td
->td_proc
;
3067 struct timespec ts
[2];
3073 error
= getutimes(tptr
, ts
);
3076 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
3078 if (fp
->f_nchandle
.ncp
)
3079 error
= ncp_writechk(&fp
->f_nchandle
);
3082 error
= vget(vp
, LK_EXCLUSIVE
);
3084 error
= VOP_GETATTR(vp
, &vattr
);
3086 error
= naccess_va(&vattr
, NLC_OWN
| NLC_WRITE
,
3090 error
= setutimes(vp
, &vattr
, ts
,
3101 * futimes_args(int fd, struct timeval *tptr)
3103 * Set the access and modification times of a file.
3106 sys_futimes(struct futimes_args
*uap
)
3108 struct timeval tv
[2];
3112 error
= copyin(uap
->tptr
, tv
, sizeof(tv
));
3117 error
= kern_futimes(uap
->fd
, uap
->tptr
? tv
: NULL
);
3123 kern_truncate(struct nlookupdata
*nd
, off_t length
)
3131 nd
->nl_flags
|= NLC_WRITE
| NLC_TRUNCATE
;
3132 if ((error
= nlookup(nd
)) != 0)
3134 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
3136 if ((error
= cache_vref(&nd
->nl_nch
, nd
->nl_cred
, &vp
)) != 0)
3138 if ((error
= vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
)) != 0) {
3142 if (vp
->v_type
== VDIR
) {
3144 } else if ((error
= vn_writechk(vp
, &nd
->nl_nch
)) == 0) {
3146 vattr
.va_size
= length
;
3147 error
= VOP_SETATTR(vp
, &vattr
, nd
->nl_cred
);
3154 * truncate(char *path, int pad, off_t length)
3156 * Truncate a file given its path name.
3159 sys_truncate(struct truncate_args
*uap
)
3161 struct nlookupdata nd
;
3164 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
3166 error
= kern_truncate(&nd
, uap
->length
);
3172 kern_ftruncate(int fd
, off_t length
)
3174 struct thread
*td
= curthread
;
3175 struct proc
*p
= td
->td_proc
;
3183 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
3185 if (fp
->f_nchandle
.ncp
) {
3186 error
= ncp_writechk(&fp
->f_nchandle
);
3190 if ((fp
->f_flag
& FWRITE
) == 0) {
3194 if (fp
->f_flag
& FAPPENDONLY
) { /* inode was set s/uapnd */
3198 vp
= (struct vnode
*)fp
->f_data
;
3199 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
3200 if (vp
->v_type
== VDIR
) {
3202 } else if ((error
= vn_writechk(vp
, NULL
)) == 0) {
3204 vattr
.va_size
= length
;
3205 error
= VOP_SETATTR(vp
, &vattr
, fp
->f_cred
);
3214 * ftruncate_args(int fd, int pad, off_t length)
3216 * Truncate a file given a file descriptor.
3219 sys_ftruncate(struct ftruncate_args
*uap
)
3223 error
= kern_ftruncate(uap
->fd
, uap
->length
);
3231 * Sync an open file.
3235 sys_fsync(struct fsync_args
*uap
)
3237 struct thread
*td
= curthread
;
3238 struct proc
*p
= td
->td_proc
;
3244 if ((error
= holdvnode(p
->p_fd
, uap
->fd
, &fp
)) != 0)
3246 vp
= (struct vnode
*)fp
->f_data
;
3247 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
3248 if ((obj
= vp
->v_object
) != NULL
)
3249 vm_object_page_clean(obj
, 0, 0, 0);
3250 if ((error
= VOP_FSYNC(vp
, MNT_WAIT
)) == 0 && vp
->v_mount
)
3251 error
= buf_fsync(vp
);
3258 kern_rename(struct nlookupdata
*fromnd
, struct nlookupdata
*tond
)
3260 struct nchandle fnchd
;
3261 struct nchandle tnchd
;
3262 struct namecache
*ncp
;
3269 fromnd
->nl_flags
|= NLC_REFDVP
| NLC_RENAME_SRC
;
3270 if ((error
= nlookup(fromnd
)) != 0)
3272 if ((fnchd
.ncp
= fromnd
->nl_nch
.ncp
->nc_parent
) == NULL
)
3274 fnchd
.mount
= fromnd
->nl_nch
.mount
;
3278 * unlock the source nch so we can lookup the target nch without
3279 * deadlocking. The target may or may not exist so we do not check
3280 * for a target vp like kern_mkdir() and other creation functions do.
3282 * The source and target directories are ref'd and rechecked after
3283 * everything is relocked to determine if the source or target file
3286 KKASSERT(fromnd
->nl_flags
& NLC_NCPISLOCKED
);
3287 fromnd
->nl_flags
&= ~NLC_NCPISLOCKED
;
3288 cache_unlock(&fromnd
->nl_nch
);
3290 tond
->nl_flags
|= NLC_RENAME_DST
| NLC_REFDVP
;
3291 if ((error
= nlookup(tond
)) != 0) {
3295 if ((tnchd
.ncp
= tond
->nl_nch
.ncp
->nc_parent
) == NULL
) {
3299 tnchd
.mount
= tond
->nl_nch
.mount
;
3303 * If the source and target are the same there is nothing to do
3305 if (fromnd
->nl_nch
.ncp
== tond
->nl_nch
.ncp
) {
3312 * Mount points cannot be renamed or overwritten
3314 if ((fromnd
->nl_nch
.ncp
->nc_flag
| tond
->nl_nch
.ncp
->nc_flag
) &
3323 * relock the source ncp. NOTE AFTER RELOCKING: the source ncp
3324 * may have become invalid while it was unlocked, nc_vp and nc_mount
3327 if (cache_lock_nonblock(&fromnd
->nl_nch
) == 0) {
3328 cache_resolve(&fromnd
->nl_nch
, fromnd
->nl_cred
);
3329 } else if (fromnd
->nl_nch
.ncp
> tond
->nl_nch
.ncp
) {
3330 cache_lock(&fromnd
->nl_nch
);
3331 cache_resolve(&fromnd
->nl_nch
, fromnd
->nl_cred
);
3333 cache_unlock(&tond
->nl_nch
);
3334 cache_lock(&fromnd
->nl_nch
);
3335 cache_resolve(&fromnd
->nl_nch
, fromnd
->nl_cred
);
3336 cache_lock(&tond
->nl_nch
);
3337 cache_resolve(&tond
->nl_nch
, tond
->nl_cred
);
3339 fromnd
->nl_flags
|= NLC_NCPISLOCKED
;
3342 * make sure the parent directories linkages are the same
3344 if (fnchd
.ncp
!= fromnd
->nl_nch
.ncp
->nc_parent
||
3345 tnchd
.ncp
!= tond
->nl_nch
.ncp
->nc_parent
) {
3352 * Both the source and target must be within the same filesystem and
3353 * in the same filesystem as their parent directories within the
3354 * namecache topology.
3356 * NOTE: fromnd's nc_mount or nc_vp could be NULL.
3359 if (mp
!= tnchd
.mount
|| mp
!= fromnd
->nl_nch
.mount
||
3360 mp
!= tond
->nl_nch
.mount
) {
3367 * Make sure the mount point is writable
3369 if ((error
= ncp_writechk(&tond
->nl_nch
)) != 0) {
3376 * If the target exists and either the source or target is a directory,
3377 * then both must be directories.
3379 * Due to relocking of the source, fromnd->nl_nch.ncp->nc_vp might h
3382 if (tond
->nl_nch
.ncp
->nc_vp
) {
3383 if (fromnd
->nl_nch
.ncp
->nc_vp
== NULL
) {
3385 } else if (fromnd
->nl_nch
.ncp
->nc_vp
->v_type
== VDIR
) {
3386 if (tond
->nl_nch
.ncp
->nc_vp
->v_type
!= VDIR
)
3388 } else if (tond
->nl_nch
.ncp
->nc_vp
->v_type
== VDIR
) {
3394 * You cannot rename a source into itself or a subdirectory of itself.
3395 * We check this by travsersing the target directory upwards looking
3396 * for a match against the source.
3399 for (ncp
= tnchd
.ncp
; ncp
; ncp
= ncp
->nc_parent
) {
3400 if (fromnd
->nl_nch
.ncp
== ncp
) {
3411 * Even though the namespaces are different, they may still represent
3412 * hardlinks to the same file. The filesystem might have a hard time
3413 * with this so we issue a NREMOVE of the source instead of a NRENAME
3414 * when we detect the situation.
3417 fdvp
= fromnd
->nl_dvp
;
3418 tdvp
= tond
->nl_dvp
;
3419 if (fdvp
== NULL
|| tdvp
== NULL
) {
3421 } else if (fromnd
->nl_nch
.ncp
->nc_vp
== tond
->nl_nch
.ncp
->nc_vp
) {
3422 error
= VOP_NREMOVE(&fromnd
->nl_nch
, fdvp
,
3425 error
= VOP_NRENAME(&fromnd
->nl_nch
, &tond
->nl_nch
,
3426 fdvp
, tdvp
, tond
->nl_cred
);
3433 * rename_args(char *from, char *to)
3435 * Rename files. Source and destination must either both be directories,
3436 * or both not be directories. If target is a directory, it must be empty.
3439 sys_rename(struct rename_args
*uap
)
3441 struct nlookupdata fromnd
, tond
;
3444 error
= nlookup_init(&fromnd
, uap
->from
, UIO_USERSPACE
, 0);
3446 error
= nlookup_init(&tond
, uap
->to
, UIO_USERSPACE
, 0);
3448 error
= kern_rename(&fromnd
, &tond
);
3449 nlookup_done(&tond
);
3451 nlookup_done(&fromnd
);
3456 kern_mkdir(struct nlookupdata
*nd
, int mode
)
3458 struct thread
*td
= curthread
;
3459 struct proc
*p
= td
->td_proc
;
3465 nd
->nl_flags
|= NLC_WILLBEDIR
| NLC_CREATE
| NLC_REFDVP
;
3466 if ((error
= nlookup(nd
)) != 0)
3469 if (nd
->nl_nch
.ncp
->nc_vp
)
3471 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
3474 vattr
.va_type
= VDIR
;
3475 vattr
.va_mode
= (mode
& ACCESSPERMS
) &~ p
->p_fd
->fd_cmask
;
3478 error
= VOP_NMKDIR(&nd
->nl_nch
, nd
->nl_dvp
, &vp
, p
->p_ucred
, &vattr
);
3485 * mkdir_args(char *path, int mode)
3487 * Make a directory file.
3491 sys_mkdir(struct mkdir_args
*uap
)
3493 struct nlookupdata nd
;
3496 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
3498 error
= kern_mkdir(&nd
, uap
->mode
);
3504 kern_rmdir(struct nlookupdata
*nd
)
3509 nd
->nl_flags
|= NLC_DELETE
| NLC_REFDVP
;
3510 if ((error
= nlookup(nd
)) != 0)
3514 * Do not allow directories representing mount points to be
3515 * deleted, even if empty. Check write perms on mount point
3516 * in case the vnode is aliased (aka nullfs).
3518 if (nd
->nl_nch
.ncp
->nc_flag
& (NCF_ISMOUNTPT
))
3520 if ((error
= ncp_writechk(&nd
->nl_nch
)) != 0)
3522 error
= VOP_NRMDIR(&nd
->nl_nch
, nd
->nl_dvp
, nd
->nl_cred
);
3527 * rmdir_args(char *path)
3529 * Remove a directory file.
3533 sys_rmdir(struct rmdir_args
*uap
)
3535 struct nlookupdata nd
;
3538 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, 0);
3540 error
= kern_rmdir(&nd
);
3546 kern_getdirentries(int fd
, char *buf
, u_int count
, long *basep
, int *res
,
3547 enum uio_seg direction
)
3549 struct thread
*td
= curthread
;
3550 struct proc
*p
= td
->td_proc
;
3558 if ((error
= holdvnode(p
->p_fd
, fd
, &fp
)) != 0)
3560 if ((fp
->f_flag
& FREAD
) == 0) {
3564 vp
= (struct vnode
*)fp
->f_data
;
3566 if (vp
->v_type
!= VDIR
) {
3570 aiov
.iov_base
= buf
;
3571 aiov
.iov_len
= count
;
3572 auio
.uio_iov
= &aiov
;
3573 auio
.uio_iovcnt
= 1;
3574 auio
.uio_rw
= UIO_READ
;
3575 auio
.uio_segflg
= direction
;
3577 auio
.uio_resid
= count
;
3578 loff
= auio
.uio_offset
= fp
->f_offset
;
3579 error
= VOP_READDIR(vp
, &auio
, fp
->f_cred
, &eofflag
, NULL
, NULL
);
3580 fp
->f_offset
= auio
.uio_offset
;
3583 if (count
== auio
.uio_resid
) {
3584 if (union_dircheckp
) {
3585 error
= union_dircheckp(td
, &vp
, fp
);
3592 if ((vp
->v_flag
& VROOT
) &&
3593 (vp
->v_mount
->mnt_flag
& MNT_UNION
)) {
3594 struct vnode
*tvp
= vp
;
3595 vp
= vp
->v_mount
->mnt_vnodecovered
;
3606 * WARNING! *basep may not be wide enough to accomodate the
3607 * seek offset. XXX should we hack this to return the upper 32 bits
3608 * for offsets greater then 4G?
3611 *basep
= (long)loff
;
3613 *res
= count
- auio
.uio_resid
;
3620 * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
3622 * Read a block of directory entries in a file system independent format.
3625 sys_getdirentries(struct getdirentries_args
*uap
)
3630 error
= kern_getdirentries(uap
->fd
, uap
->buf
, uap
->count
, &base
,
3631 &uap
->sysmsg_result
, UIO_USERSPACE
);
3633 if (error
== 0 && uap
->basep
)
3634 error
= copyout(&base
, uap
->basep
, sizeof(*uap
->basep
));
3639 * getdents_args(int fd, char *buf, size_t count)
3642 sys_getdents(struct getdents_args
*uap
)
3646 error
= kern_getdirentries(uap
->fd
, uap
->buf
, uap
->count
, NULL
,
3647 &uap
->sysmsg_result
, UIO_USERSPACE
);
3653 * umask(int newmask)
3655 * Set the mode mask for creation of filesystem nodes.
3660 sys_umask(struct umask_args
*uap
)
3662 struct thread
*td
= curthread
;
3663 struct proc
*p
= td
->td_proc
;
3664 struct filedesc
*fdp
;
3667 uap
->sysmsg_result
= fdp
->fd_cmask
;
3668 fdp
->fd_cmask
= uap
->newmask
& ALLPERMS
;
3673 * revoke(char *path)
3675 * Void all references to file by ripping underlying filesystem
3680 sys_revoke(struct revoke_args
*uap
)
3682 struct nlookupdata nd
;
3689 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
3691 error
= nlookup(&nd
);
3693 error
= cache_vref(&nd
.nl_nch
, nd
.nl_cred
, &vp
);
3694 cred
= crhold(nd
.nl_cred
);
3698 error
= VOP_GETATTR(vp
, &vattr
);
3699 if (error
== 0 && cred
->cr_uid
!= vattr
.va_uid
)
3700 error
= priv_check_cred(cred
, PRIV_VFS_REVOKE
, 0);
3701 if (error
== 0 && (vp
->v_type
== VCHR
|| vp
->v_type
== VBLK
)) {
3703 error
= vrevoke(vp
, cred
);
3704 } else if (error
== 0) {
3705 error
= vrevoke(vp
, cred
);
3715 * getfh_args(char *fname, fhandle_t *fhp)
3717 * Get (NFS) file handle
3719 * NOTE: We use the fsid of the covering mount, even if it is a nullfs
3720 * mount. This allows nullfs mounts to be explicitly exported.
3722 * WARNING: nullfs mounts of HAMMER PFS ROOTs are safe.
3724 * nullfs mounts of subdirectories are not safe. That is, it will
3725 * work, but you do not really have protection against access to
3726 * the related parent directories.
3729 sys_getfh(struct getfh_args
*uap
)
3731 struct thread
*td
= curthread
;
3732 struct nlookupdata nd
;
3739 * Must be super user
3741 if ((error
= priv_check(td
, PRIV_ROOT
)) != 0)
3745 error
= nlookup_init(&nd
, uap
->fname
, UIO_USERSPACE
, NLC_FOLLOW
);
3747 error
= nlookup(&nd
);
3749 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
3750 mp
= nd
.nl_nch
.mount
;
3753 bzero(&fh
, sizeof(fh
));
3754 fh
.fh_fsid
= mp
->mnt_stat
.f_fsid
;
3755 error
= VFS_VPTOFH(vp
, &fh
.fh_fid
);
3758 error
= copyout(&fh
, uap
->fhp
, sizeof(fh
));
3764 * fhopen_args(const struct fhandle *u_fhp, int flags)
3766 * syscall for the rpc.lockd to use to translate a NFS file handle into
3767 * an open descriptor.
3769 * warning: do not remove the priv_check() call or this becomes one giant
3773 sys_fhopen(struct fhopen_args
*uap
)
3775 struct thread
*td
= curthread
;
3776 struct proc
*p
= td
->td_proc
;
3781 struct vattr
*vap
= &vat
;
3783 int fmode
, mode
, error
, type
;
3789 * Must be super user
3791 error
= priv_check(td
, PRIV_ROOT
);
3795 fmode
= FFLAGS(uap
->flags
);
3796 /* why not allow a non-read/write open for our lockd? */
3797 if (((fmode
& (FREAD
| FWRITE
)) == 0) || (fmode
& O_CREAT
))
3799 error
= copyin(uap
->u_fhp
, &fhp
, sizeof(fhp
));
3802 /* find the mount point */
3803 mp
= vfs_getvfs(&fhp
.fh_fsid
);
3806 /* now give me my vnode, it gets returned to me locked */
3807 error
= VFS_FHTOVP(mp
, NULL
, &fhp
.fh_fid
, &vp
);
3811 * from now on we have to make sure not
3812 * to forget about the vnode
3813 * any error that causes an abort must vput(vp)
3814 * just set error = err and 'goto bad;'.
3820 if (vp
->v_type
== VLNK
) {
3824 if (vp
->v_type
== VSOCK
) {
3829 if (fmode
& (FWRITE
| O_TRUNC
)) {
3830 if (vp
->v_type
== VDIR
) {
3834 error
= vn_writechk(vp
, NULL
);
3842 error
= VOP_ACCESS(vp
, mode
, p
->p_ucred
);
3846 if (fmode
& O_TRUNC
) {
3847 vn_unlock(vp
); /* XXX */
3848 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
); /* XXX */
3851 error
= VOP_SETATTR(vp
, vap
, p
->p_ucred
);
3857 * VOP_OPEN needs the file pointer so it can potentially override
3860 * WARNING! no f_nchandle will be associated when fhopen()ing a
3863 if ((error
= falloc(p
, &nfp
, &indx
)) != 0)
3867 error
= VOP_OPEN(vp
, fmode
, p
->p_ucred
, fp
);
3870 * setting f_ops this way prevents VOP_CLOSE from being
3871 * called or fdrop() releasing the vp from v_data. Since
3872 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3874 fp
->f_ops
= &badfileops
;
3880 * The fp is given its own reference, we still have our ref and lock.
3882 * Assert that all regular files must be created with a VM object.
3884 if (vp
->v_type
== VREG
&& vp
->v_object
== NULL
) {
3885 kprintf("fhopen: regular file did not have VM object: %p\n", vp
);
3890 * The open was successful. Handle any locking requirements.
3892 if (fmode
& (O_EXLOCK
| O_SHLOCK
)) {
3893 lf
.l_whence
= SEEK_SET
;
3896 if (fmode
& O_EXLOCK
)
3897 lf
.l_type
= F_WRLCK
;
3899 lf
.l_type
= F_RDLCK
;
3900 if (fmode
& FNONBLOCK
)
3905 if ((error
= VOP_ADVLOCK(vp
, (caddr_t
)fp
, F_SETLK
, &lf
, type
)) != 0) {
3907 * release our private reference.
3909 fsetfd(p
, NULL
, indx
);
3914 vn_lock(vp
, LK_EXCLUSIVE
| LK_RETRY
);
3915 fp
->f_flag
|= FHASLOCK
;
3919 * Clean up. Associate the file pointer with the previously
3920 * reserved descriptor and return it.
3923 fsetfd(p
, fp
, indx
);
3925 uap
->sysmsg_result
= indx
;
3929 fsetfd(p
, NULL
, indx
);
3937 * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3940 sys_fhstat(struct fhstat_args
*uap
)
3942 struct thread
*td
= curthread
;
3950 * Must be super user
3952 error
= priv_check(td
, PRIV_ROOT
);
3956 error
= copyin(uap
->u_fhp
, &fh
, sizeof(fhandle_t
));
3960 if ((mp
= vfs_getvfs(&fh
.fh_fsid
)) == NULL
)
3962 if ((error
= VFS_FHTOVP(mp
, NULL
, &fh
.fh_fid
, &vp
)))
3964 error
= vn_stat(vp
, &sb
, td
->td_proc
->p_ucred
);
3968 error
= copyout(&sb
, uap
->sb
, sizeof(sb
));
3973 * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3976 sys_fhstatfs(struct fhstatfs_args
*uap
)
3978 struct thread
*td
= curthread
;
3979 struct proc
*p
= td
->td_proc
;
3984 char *fullpath
, *freepath
;
3989 * Must be super user
3991 if ((error
= priv_check(td
, PRIV_ROOT
)))
3994 if ((error
= copyin(uap
->u_fhp
, &fh
, sizeof(fhandle_t
))) != 0)
3997 if ((mp
= vfs_getvfs(&fh
.fh_fsid
)) == NULL
)
4000 if (p
!= NULL
&& !chroot_visible_mnt(mp
, p
))
4003 if ((error
= VFS_FHTOVP(mp
, NULL
, &fh
.fh_fid
, &vp
)))
4008 if ((error
= VFS_STATFS(mp
, sp
, p
->p_ucred
)) != 0)
4011 error
= mount_path(p
, mp
, &fullpath
, &freepath
);
4014 bzero(sp
->f_mntonname
, sizeof(sp
->f_mntonname
));
4015 strlcpy(sp
->f_mntonname
, fullpath
, sizeof(sp
->f_mntonname
));
4016 kfree(freepath
, M_TEMP
);
4018 sp
->f_flags
= mp
->mnt_flag
& MNT_VISFLAGMASK
;
4019 if (priv_check(td
, PRIV_ROOT
)) {
4020 bcopy(sp
, &sb
, sizeof(sb
));
4021 sb
.f_fsid
.val
[0] = sb
.f_fsid
.val
[1] = 0;
4024 return (copyout(sp
, uap
->buf
, sizeof(*sp
)));
4028 * fhstatvfs_args(struct fhandle *u_fhp, struct statvfs *buf)
4031 sys_fhstatvfs(struct fhstatvfs_args
*uap
)
4033 struct thread
*td
= curthread
;
4034 struct proc
*p
= td
->td_proc
;
4042 * Must be super user
4044 if ((error
= priv_check(td
, PRIV_ROOT
)))
4047 if ((error
= copyin(uap
->u_fhp
, &fh
, sizeof(fhandle_t
))) != 0)
4050 if ((mp
= vfs_getvfs(&fh
.fh_fsid
)) == NULL
)
4053 if (p
!= NULL
&& !chroot_visible_mnt(mp
, p
))
4056 if ((error
= VFS_FHTOVP(mp
, NULL
, &fh
.fh_fid
, &vp
)))
4059 sp
= &mp
->mnt_vstat
;
4061 if ((error
= VFS_STATVFS(mp
, sp
, p
->p_ucred
)) != 0)
4065 if (mp
->mnt_flag
& MNT_RDONLY
)
4066 sp
->f_flag
|= ST_RDONLY
;
4067 if (mp
->mnt_flag
& MNT_NOSUID
)
4068 sp
->f_flag
|= ST_NOSUID
;
4070 return (copyout(sp
, uap
->buf
, sizeof(*sp
)));
4075 * Syscall to push extended attribute configuration information into the
4076 * VFS. Accepts a path, which it converts to a mountpoint, as well as
4077 * a command (int cmd), and attribute name and misc data. For now, the
4078 * attribute name is left in userspace for consumption by the VFS_op.
4079 * It will probably be changed to be copied into sysspace by the
4080 * syscall in the future, once issues with various consumers of the
4081 * attribute code have raised their hands.
4083 * Currently this is used only by UFS Extended Attributes.
4086 sys_extattrctl(struct extattrctl_args
*uap
)
4088 struct nlookupdata nd
;
4094 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
4096 error
= nlookup(&nd
);
4098 mp
= nd
.nl_nch
.mount
;
4099 error
= VFS_EXTATTRCTL(mp
, uap
->cmd
,
4100 uap
->attrname
, uap
->arg
,
4108 * Syscall to set a named extended attribute on a file or directory.
4109 * Accepts attribute name, and a uio structure pointing to the data to set.
4110 * The uio is consumed in the style of writev(). The real work happens
4111 * in VOP_SETEXTATTR().
4114 sys_extattr_set_file(struct extattr_set_file_args
*uap
)
4116 char attrname
[EXTATTR_MAXNAMELEN
];
4117 struct iovec aiov
[UIO_SMALLIOV
];
4118 struct iovec
*needfree
;
4119 struct nlookupdata nd
;
4128 error
= copyin(uap
->attrname
, attrname
, EXTATTR_MAXNAMELEN
);
4133 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
4135 error
= nlookup(&nd
);
4137 error
= ncp_writechk(&nd
.nl_nch
);
4139 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
4146 iovlen
= uap
->iovcnt
* sizeof(struct iovec
);
4147 if (uap
->iovcnt
> UIO_SMALLIOV
) {
4148 if (uap
->iovcnt
> UIO_MAXIOV
) {
4152 MALLOC(iov
, struct iovec
*, iovlen
, M_IOV
, M_WAITOK
);
4158 auio
.uio_iovcnt
= uap
->iovcnt
;
4159 auio
.uio_rw
= UIO_WRITE
;
4160 auio
.uio_segflg
= UIO_USERSPACE
;
4161 auio
.uio_td
= nd
.nl_td
;
4162 auio
.uio_offset
= 0;
4163 if ((error
= copyin(uap
->iovp
, iov
, iovlen
)))
4166 for (i
= 0; i
< uap
->iovcnt
; i
++) {
4167 if (iov
->iov_len
> INT_MAX
- auio
.uio_resid
) {
4171 auio
.uio_resid
+= iov
->iov_len
;
4174 cnt
= auio
.uio_resid
;
4175 error
= VOP_SETEXTATTR(vp
, attrname
, &auio
, nd
.nl_cred
);
4176 cnt
-= auio
.uio_resid
;
4177 uap
->sysmsg_result
= cnt
;
4182 FREE(needfree
, M_IOV
);
4187 * Syscall to get a named extended attribute on a file or directory.
4188 * Accepts attribute name, and a uio structure pointing to a buffer for the
4189 * data. The uio is consumed in the style of readv(). The real work
4190 * happens in VOP_GETEXTATTR();
4193 sys_extattr_get_file(struct extattr_get_file_args
*uap
)
4195 char attrname
[EXTATTR_MAXNAMELEN
];
4196 struct iovec aiov
[UIO_SMALLIOV
];
4197 struct iovec
*needfree
;
4198 struct nlookupdata nd
;
4207 error
= copyin(uap
->attrname
, attrname
, EXTATTR_MAXNAMELEN
);
4212 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
4214 error
= nlookup(&nd
);
4216 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
4222 iovlen
= uap
->iovcnt
* sizeof (struct iovec
);
4224 if (uap
->iovcnt
> UIO_SMALLIOV
) {
4225 if (uap
->iovcnt
> UIO_MAXIOV
) {
4229 MALLOC(iov
, struct iovec
*, iovlen
, M_IOV
, M_WAITOK
);
4235 auio
.uio_iovcnt
= uap
->iovcnt
;
4236 auio
.uio_rw
= UIO_READ
;
4237 auio
.uio_segflg
= UIO_USERSPACE
;
4238 auio
.uio_td
= nd
.nl_td
;
4239 auio
.uio_offset
= 0;
4240 if ((error
= copyin(uap
->iovp
, iov
, iovlen
)))
4243 for (i
= 0; i
< uap
->iovcnt
; i
++) {
4244 if (iov
->iov_len
> INT_MAX
- auio
.uio_resid
) {
4248 auio
.uio_resid
+= iov
->iov_len
;
4251 cnt
= auio
.uio_resid
;
4252 error
= VOP_GETEXTATTR(vp
, attrname
, &auio
, nd
.nl_cred
);
4253 cnt
-= auio
.uio_resid
;
4254 uap
->sysmsg_result
= cnt
;
4259 FREE(needfree
, M_IOV
);
4264 * Syscall to delete a named extended attribute from a file or directory.
4265 * Accepts attribute name. The real work happens in VOP_SETEXTATTR().
4268 sys_extattr_delete_file(struct extattr_delete_file_args
*uap
)
4270 char attrname
[EXTATTR_MAXNAMELEN
];
4271 struct nlookupdata nd
;
4275 error
= copyin(uap
->attrname
, attrname
, EXTATTR_MAXNAMELEN
);
4280 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
, NLC_FOLLOW
);
4282 error
= nlookup(&nd
);
4284 error
= ncp_writechk(&nd
.nl_nch
);
4286 error
= cache_vget(&nd
.nl_nch
, nd
.nl_cred
, LK_EXCLUSIVE
, &vp
);
4292 error
= VOP_SETEXTATTR(vp
, attrname
, NULL
, nd
.nl_cred
);
4299 * Determine if the mount is visible to the process.
4302 chroot_visible_mnt(struct mount
*mp
, struct proc
*p
)
4304 struct nchandle nch
;
4307 * Traverse from the mount point upwards. If we hit the process
4308 * root then the mount point is visible to the process.
4310 nch
= mp
->mnt_ncmountpt
;
4312 if (nch
.mount
== p
->p_fd
->fd_nrdir
.mount
&&
4313 nch
.ncp
== p
->p_fd
->fd_nrdir
.ncp
) {
4316 if (nch
.ncp
== nch
.mount
->mnt_ncmountpt
.ncp
) {
4317 nch
= nch
.mount
->mnt_ncmounton
;
4319 nch
.ncp
= nch
.ncp
->nc_parent
;
4324 * If the mount point is not visible to the process, but the
4325 * process root is in a subdirectory of the mount, return
4328 if (p
->p_fd
->fd_nrdir
.mount
== mp
)