4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * Some corrections by tytso.
11 /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname
14 /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture.
17 #include <linux/init.h>
18 #include <linux/module.h>
19 #include <linux/slab.h>
21 #include <linux/namei.h>
22 #include <linux/quotaops.h>
23 #include <linux/pagemap.h>
24 #include <linux/fsnotify.h>
25 #include <linux/personality.h>
26 #include <linux/security.h>
27 #include <linux/syscalls.h>
28 #include <linux/mount.h>
29 #include <linux/audit.h>
30 #include <linux/capability.h>
31 #include <linux/file.h>
32 #include <linux/fcntl.h>
33 #include <linux/device_cgroup.h>
34 #include <asm/uaccess.h>
36 #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE])
38 /* [Feb-1997 T. Schoebel-Theuer]
39 * Fundamental changes in the pathname lookup mechanisms (namei)
40 * were necessary because of omirr. The reason is that omirr needs
41 * to know the _real_ pathname, not the user-supplied one, in case
42 * of symlinks (and also when transname replacements occur).
44 * The new code replaces the old recursive symlink resolution with
45 * an iterative one (in case of non-nested symlink chains). It does
46 * this with calls to <fs>_follow_link().
47 * As a side effect, dir_namei(), _namei() and follow_link() are now
48 * replaced with a single function lookup_dentry() that can handle all
49 * the special cases of the former code.
51 * With the new dcache, the pathname is stored at each inode, at least as
52 * long as the refcount of the inode is positive. As a side effect, the
53 * size of the dcache depends on the inode cache and thus is dynamic.
55 * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink
56 * resolution to correspond with current state of the code.
58 * Note that the symlink resolution is not *completely* iterative.
59 * There is still a significant amount of tail- and mid- recursion in
60 * the algorithm. Also, note that <fs>_readlink() is not used in
61 * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink()
62 * may return different results than <fs>_follow_link(). Many virtual
63 * filesystems (including /proc) exhibit this behavior.
66 /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation:
67 * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL
68 * and the name already exists in form of a symlink, try to create the new
69 * name indicated by the symlink. The old code always complained that the
70 * name already exists, due to not following the symlink even if its target
71 * is nonexistent. The new semantics affects also mknod() and link() when
72 * the name is a symlink pointing to a non-existant name.
74 * I don't know which semantics is the right one, since I have no access
75 * to standards. But I found by trial that HP-UX 9.0 has the full "new"
76 * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the
77 * "old" one. Personally, I think the new semantics is much more logical.
78 * Note that "ln old new" where "new" is a symlink pointing to a non-existing
79 * file does succeed in both HP-UX and SunOs, but not in Solaris
80 * and in the old Linux semantics.
83 /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink
84 * semantics. See the comments in "open_namei" and "do_link" below.
86 * [10-Sep-98 Alan Modra] Another symlink change.
89 /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks:
90 * inside the path - always follow.
91 * in the last component in creation/removal/renaming - never follow.
92 * if LOOKUP_FOLLOW passed - follow.
93 * if the pathname has trailing slashes - follow.
94 * otherwise - don't follow.
95 * (applied in that order).
97 * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT
98 * restored for 2.4. This is the last surviving part of old 4.2BSD bug.
99 * During the 2.4 we need to fix the userland stuff depending on it -
100 * hopefully we will be able to get rid of that wart in 2.5. So far only
101 * XEmacs seems to be relying on it...
104 * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland)
105 * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives
106 * any extra contention...
109 static int __link_path_walk(const char *name
, struct nameidata
*nd
);
111 /* In order to reduce some races, while at the same time doing additional
112 * checking and hopefully speeding things up, we copy filenames to the
113 * kernel data space before using them..
115 * POSIX.1 2.4: an empty pathname is invalid (ENOENT).
116 * PATH_MAX includes the nul terminator --RR.
118 static int do_getname(const char __user
*filename
, char *page
)
121 unsigned long len
= PATH_MAX
;
123 if (!segment_eq(get_fs(), KERNEL_DS
)) {
124 if ((unsigned long) filename
>= TASK_SIZE
)
126 if (TASK_SIZE
- (unsigned long) filename
< PATH_MAX
)
127 len
= TASK_SIZE
- (unsigned long) filename
;
130 retval
= strncpy_from_user(page
, filename
, len
);
134 return -ENAMETOOLONG
;
140 char * getname(const char __user
* filename
)
144 result
= ERR_PTR(-ENOMEM
);
147 int retval
= do_getname(filename
, tmp
);
152 result
= ERR_PTR(retval
);
155 audit_getname(result
);
159 #ifdef CONFIG_AUDITSYSCALL
160 void putname(const char *name
)
162 if (unlikely(!audit_dummy_context()))
167 EXPORT_SYMBOL(putname
);
172 * generic_permission - check for access rights on a Posix-like filesystem
173 * @inode: inode to check access rights for
174 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
175 * @check_acl: optional callback to check for Posix ACLs
177 * Used to check for read/write/execute permissions on a file.
178 * We use "fsuid" for this, letting us set arbitrary permissions
179 * for filesystem access without changing the "normal" uids which
180 * are used for other things..
182 int generic_permission(struct inode
*inode
, int mask
,
183 int (*check_acl
)(struct inode
*inode
, int mask
))
185 umode_t mode
= inode
->i_mode
;
187 mask
&= MAY_READ
| MAY_WRITE
| MAY_EXEC
;
189 if (current
->fsuid
== inode
->i_uid
)
192 if (IS_POSIXACL(inode
) && (mode
& S_IRWXG
) && check_acl
) {
193 int error
= check_acl(inode
, mask
);
194 if (error
== -EACCES
)
195 goto check_capabilities
;
196 else if (error
!= -EAGAIN
)
200 if (in_group_p(inode
->i_gid
))
205 * If the DACs are ok we don't need any capability check.
207 if ((mask
& ~mode
) == 0)
212 * Read/write DACs are always overridable.
213 * Executable DACs are overridable if at least one exec bit is set.
215 if (!(mask
& MAY_EXEC
) ||
216 (inode
->i_mode
& S_IXUGO
) || S_ISDIR(inode
->i_mode
))
217 if (capable(CAP_DAC_OVERRIDE
))
221 * Searching includes executable on directories, else just read.
223 if (mask
== MAY_READ
|| (S_ISDIR(inode
->i_mode
) && !(mask
& MAY_WRITE
)))
224 if (capable(CAP_DAC_READ_SEARCH
))
230 int permission(struct inode
*inode
, int mask
, struct nameidata
*nd
)
233 struct vfsmount
*mnt
= NULL
;
238 if (mask
& MAY_WRITE
) {
239 umode_t mode
= inode
->i_mode
;
242 * Nobody gets write access to a read-only fs.
244 if (IS_RDONLY(inode
) &&
245 (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
)))
249 * Nobody gets write access to an immutable file.
251 if (IS_IMMUTABLE(inode
))
255 if ((mask
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
257 * MAY_EXEC on regular files is denied if the fs is mounted
258 * with the "noexec" flag.
260 if (mnt
&& (mnt
->mnt_flags
& MNT_NOEXEC
))
264 /* Ordinary permission routines do not understand MAY_APPEND. */
265 if (inode
->i_op
&& inode
->i_op
->permission
) {
268 if (nd
->flags
& LOOKUP_ACCESS
)
270 if (nd
->flags
& LOOKUP_OPEN
)
273 retval
= inode
->i_op
->permission(inode
, mask
| extra
);
276 * Exec permission on a regular file is denied if none
277 * of the execute bits are set.
279 * This check should be done by the ->permission()
282 if ((mask
& MAY_EXEC
) && S_ISREG(inode
->i_mode
) &&
283 !(inode
->i_mode
& S_IXUGO
))
287 retval
= generic_permission(inode
, mask
, NULL
);
292 retval
= devcgroup_inode_permission(inode
, mask
);
296 return security_inode_permission(inode
,
297 mask
& (MAY_READ
|MAY_WRITE
|MAY_EXEC
), nd
);
301 * vfs_permission - check for access rights to a given path
302 * @nd: lookup result that describes the path
303 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
305 * Used to check for read/write/execute permissions on a path.
306 * We use "fsuid" for this, letting us set arbitrary permissions
307 * for filesystem access without changing the "normal" uids which
308 * are used for other things.
310 int vfs_permission(struct nameidata
*nd
, int mask
)
312 return permission(nd
->path
.dentry
->d_inode
, mask
, nd
);
316 * file_permission - check for additional access rights to a given file
317 * @file: file to check access rights for
318 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
320 * Used to check for read/write/execute permissions on an already opened
324 * Do not use this function in new code. All access checks should
325 * be done using vfs_permission().
327 int file_permission(struct file
*file
, int mask
)
329 return permission(file
->f_path
.dentry
->d_inode
, mask
, NULL
);
333 * get_write_access() gets write permission for a file.
334 * put_write_access() releases this write permission.
335 * This is used for regular files.
336 * We cannot support write (and maybe mmap read-write shared) accesses and
337 * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode
338 * can have the following values:
339 * 0: no writers, no VM_DENYWRITE mappings
340 * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist
341 * > 0: (i_writecount) users are writing to the file.
343 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
344 * except for the cases where we don't hold i_writecount yet. Then we need to
345 * use {get,deny}_write_access() - these functions check the sign and refuse
346 * to do the change if sign is wrong. Exclusion between them is provided by
347 * the inode->i_lock spinlock.
350 int get_write_access(struct inode
* inode
)
352 spin_lock(&inode
->i_lock
);
353 if (atomic_read(&inode
->i_writecount
) < 0) {
354 spin_unlock(&inode
->i_lock
);
357 atomic_inc(&inode
->i_writecount
);
358 spin_unlock(&inode
->i_lock
);
363 int deny_write_access(struct file
* file
)
365 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
367 spin_lock(&inode
->i_lock
);
368 if (atomic_read(&inode
->i_writecount
) > 0) {
369 spin_unlock(&inode
->i_lock
);
372 atomic_dec(&inode
->i_writecount
);
373 spin_unlock(&inode
->i_lock
);
379 * path_get - get a reference to a path
380 * @path: path to get the reference to
382 * Given a path increment the reference count to the dentry and the vfsmount.
384 void path_get(struct path
*path
)
389 EXPORT_SYMBOL(path_get
);
392 * path_put - put a reference to a path
393 * @path: path to put the reference to
395 * Given a path decrement the reference count to the dentry and the vfsmount.
397 void path_put(struct path
*path
)
402 EXPORT_SYMBOL(path_put
);
405 * release_open_intent - free up open intent resources
406 * @nd: pointer to nameidata
408 void release_open_intent(struct nameidata
*nd
)
410 if (nd
->intent
.open
.file
->f_path
.dentry
== NULL
)
411 put_filp(nd
->intent
.open
.file
);
413 fput(nd
->intent
.open
.file
);
416 static inline struct dentry
*
417 do_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
419 int status
= dentry
->d_op
->d_revalidate(dentry
, nd
);
420 if (unlikely(status
<= 0)) {
422 * The dentry failed validation.
423 * If d_revalidate returned 0 attempt to invalidate
424 * the dentry otherwise d_revalidate is asking us
425 * to return a fail status.
428 if (!d_invalidate(dentry
)) {
434 dentry
= ERR_PTR(status
);
441 * Internal lookup() using the new generic dcache.
444 static struct dentry
* cached_lookup(struct dentry
* parent
, struct qstr
* name
, struct nameidata
*nd
)
446 struct dentry
* dentry
= __d_lookup(parent
, name
);
448 /* lockess __d_lookup may fail due to concurrent d_move()
449 * in some unrelated directory, so try with d_lookup
452 dentry
= d_lookup(parent
, name
);
454 if (dentry
&& dentry
->d_op
&& dentry
->d_op
->d_revalidate
)
455 dentry
= do_revalidate(dentry
, nd
);
461 * Short-cut version of permission(), for calling by
462 * path_walk(), when dcache lock is held. Combines parts
463 * of permission() and generic_permission(), and tests ONLY for
464 * MAY_EXEC permission.
466 * If appropriate, check DAC only. If not appropriate, or
467 * short-cut DAC fails, then call permission() to do more
468 * complete permission check.
470 static int exec_permission_lite(struct inode
*inode
,
471 struct nameidata
*nd
)
473 umode_t mode
= inode
->i_mode
;
475 if (inode
->i_op
&& inode
->i_op
->permission
)
478 if (current
->fsuid
== inode
->i_uid
)
480 else if (in_group_p(inode
->i_gid
))
486 if ((inode
->i_mode
& S_IXUGO
) && capable(CAP_DAC_OVERRIDE
))
489 if (S_ISDIR(inode
->i_mode
) && capable(CAP_DAC_OVERRIDE
))
492 if (S_ISDIR(inode
->i_mode
) && capable(CAP_DAC_READ_SEARCH
))
497 return security_inode_permission(inode
, MAY_EXEC
, nd
);
501 * This is called when everything else fails, and we actually have
502 * to go to the low-level filesystem to find out what we should do..
504 * We get the directory semaphore, and after getting that we also
505 * make sure that nobody added the entry to the dcache in the meantime..
508 static struct dentry
* real_lookup(struct dentry
* parent
, struct qstr
* name
, struct nameidata
*nd
)
510 struct dentry
* result
;
511 struct inode
*dir
= parent
->d_inode
;
513 mutex_lock(&dir
->i_mutex
);
515 * First re-do the cached lookup just in case it was created
516 * while we waited for the directory semaphore..
518 * FIXME! This could use version numbering or similar to
519 * avoid unnecessary cache lookups.
521 * The "dcache_lock" is purely to protect the RCU list walker
522 * from concurrent renames at this point (we mustn't get false
523 * negatives from the RCU list walk here, unlike the optimistic
526 * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup
528 result
= d_lookup(parent
, name
);
530 struct dentry
*dentry
;
532 /* Don't create child dentry for a dead directory. */
533 result
= ERR_PTR(-ENOENT
);
537 dentry
= d_alloc(parent
, name
);
538 result
= ERR_PTR(-ENOMEM
);
540 result
= dir
->i_op
->lookup(dir
, dentry
, nd
);
547 mutex_unlock(&dir
->i_mutex
);
552 * Uhhuh! Nasty case: the cache was re-populated while
553 * we waited on the semaphore. Need to revalidate.
555 mutex_unlock(&dir
->i_mutex
);
556 if (result
->d_op
&& result
->d_op
->d_revalidate
) {
557 result
= do_revalidate(result
, nd
);
559 result
= ERR_PTR(-ENOENT
);
565 static __always_inline
void
566 walk_init_root(const char *name
, struct nameidata
*nd
)
568 struct fs_struct
*fs
= current
->fs
;
570 read_lock(&fs
->lock
);
573 read_unlock(&fs
->lock
);
577 * Wrapper to retry pathname resolution whenever the underlying
578 * file system returns an ESTALE.
580 * Retry the whole path once, forcing real lookup requests
581 * instead of relying on the dcache.
583 static __always_inline
int link_path_walk(const char *name
, struct nameidata
*nd
)
585 struct path save
= nd
->path
;
588 /* make sure the stuff we saved doesn't go away */
591 result
= __link_path_walk(name
, nd
);
592 if (result
== -ESTALE
) {
593 /* nd->path had been dropped */
596 nd
->flags
|= LOOKUP_REVAL
;
597 result
= __link_path_walk(name
, nd
);
605 static __always_inline
int __vfs_follow_link(struct nameidata
*nd
, const char *link
)
614 walk_init_root(link
, nd
);
616 res
= link_path_walk(link
, nd
);
617 if (nd
->depth
|| res
|| nd
->last_type
!=LAST_NORM
)
620 * If it is an iterative symlinks resolution in open_namei() we
621 * have to copy the last component. And all that crap because of
622 * bloody create() on broken symlinks. Furrfu...
625 if (unlikely(!name
)) {
629 strcpy(name
, nd
->last
.name
);
630 nd
->last
.name
= name
;
634 return PTR_ERR(link
);
637 static void path_put_conditional(struct path
*path
, struct nameidata
*nd
)
640 if (path
->mnt
!= nd
->path
.mnt
)
644 static inline void path_to_nameidata(struct path
*path
, struct nameidata
*nd
)
646 dput(nd
->path
.dentry
);
647 if (nd
->path
.mnt
!= path
->mnt
)
648 mntput(nd
->path
.mnt
);
649 nd
->path
.mnt
= path
->mnt
;
650 nd
->path
.dentry
= path
->dentry
;
653 static __always_inline
int __do_follow_link(struct path
*path
, struct nameidata
*nd
)
657 struct dentry
*dentry
= path
->dentry
;
659 touch_atime(path
->mnt
, dentry
);
660 nd_set_link(nd
, NULL
);
662 if (path
->mnt
!= nd
->path
.mnt
) {
663 path_to_nameidata(path
, nd
);
667 cookie
= dentry
->d_inode
->i_op
->follow_link(dentry
, nd
);
668 error
= PTR_ERR(cookie
);
669 if (!IS_ERR(cookie
)) {
670 char *s
= nd_get_link(nd
);
673 error
= __vfs_follow_link(nd
, s
);
674 if (dentry
->d_inode
->i_op
->put_link
)
675 dentry
->d_inode
->i_op
->put_link(dentry
, nd
, cookie
);
683 * This limits recursive symlink follows to 8, while
684 * limiting consecutive symlinks to 40.
686 * Without that kind of total limit, nasty chains of consecutive
687 * symlinks can cause almost arbitrarily long lookups.
689 static inline int do_follow_link(struct path
*path
, struct nameidata
*nd
)
692 if (current
->link_count
>= MAX_NESTED_LINKS
)
694 if (current
->total_link_count
>= 40)
696 BUG_ON(nd
->depth
>= MAX_NESTED_LINKS
);
698 err
= security_inode_follow_link(path
->dentry
, nd
);
701 current
->link_count
++;
702 current
->total_link_count
++;
704 err
= __do_follow_link(path
, nd
);
705 current
->link_count
--;
709 path_put_conditional(path
, nd
);
714 int follow_up(struct vfsmount
**mnt
, struct dentry
**dentry
)
716 struct vfsmount
*parent
;
717 struct dentry
*mountpoint
;
718 spin_lock(&vfsmount_lock
);
719 parent
=(*mnt
)->mnt_parent
;
720 if (parent
== *mnt
) {
721 spin_unlock(&vfsmount_lock
);
725 mountpoint
=dget((*mnt
)->mnt_mountpoint
);
726 spin_unlock(&vfsmount_lock
);
728 *dentry
= mountpoint
;
734 /* no need for dcache_lock, as serialization is taken care in
737 static int __follow_mount(struct path
*path
)
740 while (d_mountpoint(path
->dentry
)) {
741 struct vfsmount
*mounted
= lookup_mnt(path
->mnt
, path
->dentry
);
748 path
->dentry
= dget(mounted
->mnt_root
);
754 static void follow_mount(struct vfsmount
**mnt
, struct dentry
**dentry
)
756 while (d_mountpoint(*dentry
)) {
757 struct vfsmount
*mounted
= lookup_mnt(*mnt
, *dentry
);
763 *dentry
= dget(mounted
->mnt_root
);
767 /* no need for dcache_lock, as serialization is taken care in
770 int follow_down(struct vfsmount
**mnt
, struct dentry
**dentry
)
772 struct vfsmount
*mounted
;
774 mounted
= lookup_mnt(*mnt
, *dentry
);
779 *dentry
= dget(mounted
->mnt_root
);
785 static __always_inline
void follow_dotdot(struct nameidata
*nd
)
787 struct fs_struct
*fs
= current
->fs
;
790 struct vfsmount
*parent
;
791 struct dentry
*old
= nd
->path
.dentry
;
793 read_lock(&fs
->lock
);
794 if (nd
->path
.dentry
== fs
->root
.dentry
&&
795 nd
->path
.mnt
== fs
->root
.mnt
) {
796 read_unlock(&fs
->lock
);
799 read_unlock(&fs
->lock
);
800 spin_lock(&dcache_lock
);
801 if (nd
->path
.dentry
!= nd
->path
.mnt
->mnt_root
) {
802 nd
->path
.dentry
= dget(nd
->path
.dentry
->d_parent
);
803 spin_unlock(&dcache_lock
);
807 spin_unlock(&dcache_lock
);
808 spin_lock(&vfsmount_lock
);
809 parent
= nd
->path
.mnt
->mnt_parent
;
810 if (parent
== nd
->path
.mnt
) {
811 spin_unlock(&vfsmount_lock
);
815 nd
->path
.dentry
= dget(nd
->path
.mnt
->mnt_mountpoint
);
816 spin_unlock(&vfsmount_lock
);
818 mntput(nd
->path
.mnt
);
819 nd
->path
.mnt
= parent
;
821 follow_mount(&nd
->path
.mnt
, &nd
->path
.dentry
);
825 * It's more convoluted than I'd like it to be, but... it's still fairly
826 * small and for now I'd prefer to have fast path as straight as possible.
827 * It _is_ time-critical.
829 static int do_lookup(struct nameidata
*nd
, struct qstr
*name
,
832 struct vfsmount
*mnt
= nd
->path
.mnt
;
833 struct dentry
*dentry
= __d_lookup(nd
->path
.dentry
, name
);
837 if (dentry
->d_op
&& dentry
->d_op
->d_revalidate
)
838 goto need_revalidate
;
841 path
->dentry
= dentry
;
842 __follow_mount(path
);
846 dentry
= real_lookup(nd
->path
.dentry
, name
, nd
);
852 dentry
= do_revalidate(dentry
, nd
);
860 return PTR_ERR(dentry
);
865 * This is the basic name resolution function, turning a pathname into
866 * the final dentry. We expect 'base' to be positive and a directory.
868 * Returns 0 and nd will have valid dentry and mnt on success.
869 * Returns error and drops reference to input namei data on failure.
871 static int __link_path_walk(const char *name
, struct nameidata
*nd
)
876 unsigned int lookup_flags
= nd
->flags
;
883 inode
= nd
->path
.dentry
->d_inode
;
885 lookup_flags
= LOOKUP_FOLLOW
| (nd
->flags
& LOOKUP_CONTINUE
);
887 /* At this point we know we have a real path component. */
893 nd
->flags
|= LOOKUP_CONTINUE
;
894 err
= exec_permission_lite(inode
, nd
);
896 err
= vfs_permission(nd
, MAY_EXEC
);
901 c
= *(const unsigned char *)name
;
903 hash
= init_name_hash();
906 hash
= partial_name_hash(c
, hash
);
907 c
= *(const unsigned char *)name
;
908 } while (c
&& (c
!= '/'));
909 this.len
= name
- (const char *) this.name
;
910 this.hash
= end_name_hash(hash
);
912 /* remove trailing slashes? */
915 while (*++name
== '/');
917 goto last_with_slashes
;
920 * "." and ".." are special - ".." especially so because it has
921 * to be able to know about the current root directory and
922 * parent relationships.
924 if (this.name
[0] == '.') switch (this.len
) {
928 if (this.name
[1] != '.')
931 inode
= nd
->path
.dentry
->d_inode
;
937 * See if the low-level filesystem might want
938 * to use its own hash..
940 if (nd
->path
.dentry
->d_op
&& nd
->path
.dentry
->d_op
->d_hash
) {
941 err
= nd
->path
.dentry
->d_op
->d_hash(nd
->path
.dentry
,
946 /* This does the actual lookups.. */
947 err
= do_lookup(nd
, &this, &next
);
952 inode
= next
.dentry
->d_inode
;
959 if (inode
->i_op
->follow_link
) {
960 err
= do_follow_link(&next
, nd
);
964 inode
= nd
->path
.dentry
->d_inode
;
971 path_to_nameidata(&next
, nd
);
973 if (!inode
->i_op
->lookup
)
976 /* here ends the main loop */
979 lookup_flags
|= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
981 /* Clear LOOKUP_CONTINUE iff it was previously unset */
982 nd
->flags
&= lookup_flags
| ~LOOKUP_CONTINUE
;
983 if (lookup_flags
& LOOKUP_PARENT
)
985 if (this.name
[0] == '.') switch (this.len
) {
989 if (this.name
[1] != '.')
992 inode
= nd
->path
.dentry
->d_inode
;
997 if (nd
->path
.dentry
->d_op
&& nd
->path
.dentry
->d_op
->d_hash
) {
998 err
= nd
->path
.dentry
->d_op
->d_hash(nd
->path
.dentry
,
1003 err
= do_lookup(nd
, &this, &next
);
1006 inode
= next
.dentry
->d_inode
;
1007 if ((lookup_flags
& LOOKUP_FOLLOW
)
1008 && inode
&& inode
->i_op
&& inode
->i_op
->follow_link
) {
1009 err
= do_follow_link(&next
, nd
);
1012 inode
= nd
->path
.dentry
->d_inode
;
1014 path_to_nameidata(&next
, nd
);
1018 if (lookup_flags
& LOOKUP_DIRECTORY
) {
1020 if (!inode
->i_op
|| !inode
->i_op
->lookup
)
1026 nd
->last_type
= LAST_NORM
;
1027 if (this.name
[0] != '.')
1030 nd
->last_type
= LAST_DOT
;
1031 else if (this.len
== 2 && this.name
[1] == '.')
1032 nd
->last_type
= LAST_DOTDOT
;
1037 * We bypassed the ordinary revalidation routines.
1038 * We may need to check the cached dentry for staleness.
1040 if (nd
->path
.dentry
&& nd
->path
.dentry
->d_sb
&&
1041 (nd
->path
.dentry
->d_sb
->s_type
->fs_flags
& FS_REVAL_DOT
)) {
1043 /* Note: we do not d_invalidate() */
1044 if (!nd
->path
.dentry
->d_op
->d_revalidate(
1045 nd
->path
.dentry
, nd
))
1051 path_put_conditional(&next
, nd
);
1054 path_put(&nd
->path
);
1059 static int path_walk(const char *name
, struct nameidata
*nd
)
1061 current
->total_link_count
= 0;
1062 return link_path_walk(name
, nd
);
1065 /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */
1066 static int do_path_lookup(int dfd
, const char *name
,
1067 unsigned int flags
, struct nameidata
*nd
)
1072 struct fs_struct
*fs
= current
->fs
;
1074 nd
->last_type
= LAST_ROOT
; /* if there are only slashes... */
1079 read_lock(&fs
->lock
);
1080 nd
->path
= fs
->root
;
1081 path_get(&fs
->root
);
1082 read_unlock(&fs
->lock
);
1083 } else if (dfd
== AT_FDCWD
) {
1084 read_lock(&fs
->lock
);
1087 read_unlock(&fs
->lock
);
1089 struct dentry
*dentry
;
1091 file
= fget_light(dfd
, &fput_needed
);
1096 dentry
= file
->f_path
.dentry
;
1099 if (!S_ISDIR(dentry
->d_inode
->i_mode
))
1102 retval
= file_permission(file
, MAY_EXEC
);
1106 nd
->path
= file
->f_path
;
1107 path_get(&file
->f_path
);
1109 fput_light(file
, fput_needed
);
1112 retval
= path_walk(name
, nd
);
1113 if (unlikely(!retval
&& !audit_dummy_context() && nd
->path
.dentry
&&
1114 nd
->path
.dentry
->d_inode
))
1115 audit_inode(name
, nd
->path
.dentry
);
1120 fput_light(file
, fput_needed
);
1124 int path_lookup(const char *name
, unsigned int flags
,
1125 struct nameidata
*nd
)
1127 return do_path_lookup(AT_FDCWD
, name
, flags
, nd
);
1131 * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair
1132 * @dentry: pointer to dentry of the base directory
1133 * @mnt: pointer to vfs mount of the base directory
1134 * @name: pointer to file name
1135 * @flags: lookup flags
1136 * @nd: pointer to nameidata
1138 int vfs_path_lookup(struct dentry
*dentry
, struct vfsmount
*mnt
,
1139 const char *name
, unsigned int flags
,
1140 struct nameidata
*nd
)
1144 /* same as do_path_lookup */
1145 nd
->last_type
= LAST_ROOT
;
1149 nd
->path
.dentry
= dentry
;
1151 path_get(&nd
->path
);
1153 retval
= path_walk(name
, nd
);
1154 if (unlikely(!retval
&& !audit_dummy_context() && nd
->path
.dentry
&&
1155 nd
->path
.dentry
->d_inode
))
1156 audit_inode(name
, nd
->path
.dentry
);
1162 static int __path_lookup_intent_open(int dfd
, const char *name
,
1163 unsigned int lookup_flags
, struct nameidata
*nd
,
1164 int open_flags
, int create_mode
)
1166 struct file
*filp
= get_empty_filp();
1171 nd
->intent
.open
.file
= filp
;
1172 nd
->intent
.open
.flags
= open_flags
;
1173 nd
->intent
.open
.create_mode
= create_mode
;
1174 err
= do_path_lookup(dfd
, name
, lookup_flags
|LOOKUP_OPEN
, nd
);
1175 if (IS_ERR(nd
->intent
.open
.file
)) {
1177 err
= PTR_ERR(nd
->intent
.open
.file
);
1178 path_put(&nd
->path
);
1180 } else if (err
!= 0)
1181 release_open_intent(nd
);
1186 * path_lookup_open - lookup a file path with open intent
1187 * @dfd: the directory to use as base, or AT_FDCWD
1188 * @name: pointer to file name
1189 * @lookup_flags: lookup intent flags
1190 * @nd: pointer to nameidata
1191 * @open_flags: open intent flags
1193 int path_lookup_open(int dfd
, const char *name
, unsigned int lookup_flags
,
1194 struct nameidata
*nd
, int open_flags
)
1196 return __path_lookup_intent_open(dfd
, name
, lookup_flags
, nd
,
1201 * path_lookup_create - lookup a file path with open + create intent
1202 * @dfd: the directory to use as base, or AT_FDCWD
1203 * @name: pointer to file name
1204 * @lookup_flags: lookup intent flags
1205 * @nd: pointer to nameidata
1206 * @open_flags: open intent flags
1207 * @create_mode: create intent flags
1209 static int path_lookup_create(int dfd
, const char *name
,
1210 unsigned int lookup_flags
, struct nameidata
*nd
,
1211 int open_flags
, int create_mode
)
1213 return __path_lookup_intent_open(dfd
, name
, lookup_flags
|LOOKUP_CREATE
,
1214 nd
, open_flags
, create_mode
);
1217 int __user_path_lookup_open(const char __user
*name
, unsigned int lookup_flags
,
1218 struct nameidata
*nd
, int open_flags
)
1220 char *tmp
= getname(name
);
1221 int err
= PTR_ERR(tmp
);
1224 err
= __path_lookup_intent_open(AT_FDCWD
, tmp
, lookup_flags
, nd
, open_flags
, 0);
1230 static struct dentry
*__lookup_hash(struct qstr
*name
,
1231 struct dentry
*base
, struct nameidata
*nd
)
1233 struct dentry
*dentry
;
1234 struct inode
*inode
;
1237 inode
= base
->d_inode
;
1240 * See if the low-level filesystem might want
1241 * to use its own hash..
1243 if (base
->d_op
&& base
->d_op
->d_hash
) {
1244 err
= base
->d_op
->d_hash(base
, name
);
1245 dentry
= ERR_PTR(err
);
1250 dentry
= cached_lookup(base
, name
, nd
);
1254 /* Don't create child dentry for a dead directory. */
1255 dentry
= ERR_PTR(-ENOENT
);
1256 if (IS_DEADDIR(inode
))
1259 new = d_alloc(base
, name
);
1260 dentry
= ERR_PTR(-ENOMEM
);
1263 dentry
= inode
->i_op
->lookup(inode
, new, nd
);
1274 * Restricted form of lookup. Doesn't follow links, single-component only,
1275 * needs parent already locked. Doesn't follow mounts.
1278 static struct dentry
*lookup_hash(struct nameidata
*nd
)
1282 err
= permission(nd
->path
.dentry
->d_inode
, MAY_EXEC
, nd
);
1284 return ERR_PTR(err
);
1285 return __lookup_hash(&nd
->last
, nd
->path
.dentry
, nd
);
1288 static int __lookup_one_len(const char *name
, struct qstr
*this,
1289 struct dentry
*base
, int len
)
1299 hash
= init_name_hash();
1301 c
= *(const unsigned char *)name
++;
1302 if (c
== '/' || c
== '\0')
1304 hash
= partial_name_hash(c
, hash
);
1306 this->hash
= end_name_hash(hash
);
1311 * lookup_one_len - filesystem helper to lookup single pathname component
1312 * @name: pathname component to lookup
1313 * @base: base directory to lookup from
1314 * @len: maximum length @len should be interpreted to
1316 * Note that this routine is purely a helper for filesystem usage and should
1317 * not be called by generic code. Also note that by using this function the
1318 * nameidata argument is passed to the filesystem methods and a filesystem
1319 * using this helper needs to be prepared for that.
1321 struct dentry
*lookup_one_len(const char *name
, struct dentry
*base
, int len
)
1326 err
= __lookup_one_len(name
, &this, base
, len
);
1328 return ERR_PTR(err
);
1330 err
= permission(base
->d_inode
, MAY_EXEC
, NULL
);
1332 return ERR_PTR(err
);
1333 return __lookup_hash(&this, base
, NULL
);
1337 * lookup_one_noperm - bad hack for sysfs
1338 * @name: pathname component to lookup
1339 * @base: base directory to lookup from
1341 * This is a variant of lookup_one_len that doesn't perform any permission
1342 * checks. It's a horrible hack to work around the braindead sysfs
1343 * architecture and should not be used anywhere else.
1345 * DON'T USE THIS FUNCTION EVER, thanks.
1347 struct dentry
*lookup_one_noperm(const char *name
, struct dentry
*base
)
1352 err
= __lookup_one_len(name
, &this, base
, strlen(name
));
1354 return ERR_PTR(err
);
1355 return __lookup_hash(&this, base
, NULL
);
1358 int __user_walk_fd(int dfd
, const char __user
*name
, unsigned flags
,
1359 struct nameidata
*nd
)
1361 char *tmp
= getname(name
);
1362 int err
= PTR_ERR(tmp
);
1365 err
= do_path_lookup(dfd
, tmp
, flags
, nd
);
1371 int __user_walk(const char __user
*name
, unsigned flags
, struct nameidata
*nd
)
1373 return __user_walk_fd(AT_FDCWD
, name
, flags
, nd
);
1377 * It's inline, so penalty for filesystems that don't use sticky bit is
1380 static inline int check_sticky(struct inode
*dir
, struct inode
*inode
)
1382 if (!(dir
->i_mode
& S_ISVTX
))
1384 if (inode
->i_uid
== current
->fsuid
)
1386 if (dir
->i_uid
== current
->fsuid
)
1388 return !capable(CAP_FOWNER
);
1392 * Check whether we can remove a link victim from directory dir, check
1393 * whether the type of victim is right.
1394 * 1. We can't do it if dir is read-only (done in permission())
1395 * 2. We should have write and exec permissions on dir
1396 * 3. We can't remove anything from append-only dir
1397 * 4. We can't do anything with immutable dir (done in permission())
1398 * 5. If the sticky bit on dir is set we should either
1399 * a. be owner of dir, or
1400 * b. be owner of victim, or
1401 * c. have CAP_FOWNER capability
1402 * 6. If the victim is append-only or immutable we can't do antyhing with
1403 * links pointing to it.
1404 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
1405 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
1406 * 9. We can't remove a root or mountpoint.
1407 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
1408 * nfs_async_unlink().
1410 static int may_delete(struct inode
*dir
,struct dentry
*victim
,int isdir
)
1414 if (!victim
->d_inode
)
1417 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
1418 audit_inode_child(victim
->d_name
.name
, victim
, dir
);
1420 error
= permission(dir
,MAY_WRITE
| MAY_EXEC
, NULL
);
1425 if (check_sticky(dir
, victim
->d_inode
)||IS_APPEND(victim
->d_inode
)||
1426 IS_IMMUTABLE(victim
->d_inode
))
1429 if (!S_ISDIR(victim
->d_inode
->i_mode
))
1431 if (IS_ROOT(victim
))
1433 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
1435 if (IS_DEADDIR(dir
))
1437 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
1442 /* Check whether we can create an object with dentry child in directory
1444 * 1. We can't do it if child already exists (open has special treatment for
1445 * this case, but since we are inlined it's OK)
1446 * 2. We can't do it if dir is read-only (done in permission())
1447 * 3. We should have write and exec permissions on dir
1448 * 4. We can't do it if dir is immutable (done in permission())
1450 static inline int may_create(struct inode
*dir
, struct dentry
*child
,
1451 struct nameidata
*nd
)
1455 if (IS_DEADDIR(dir
))
1457 return permission(dir
,MAY_WRITE
| MAY_EXEC
, nd
);
1461 * O_DIRECTORY translates into forcing a directory lookup.
1463 static inline int lookup_flags(unsigned int f
)
1465 unsigned long retval
= LOOKUP_FOLLOW
;
1468 retval
&= ~LOOKUP_FOLLOW
;
1470 if (f
& O_DIRECTORY
)
1471 retval
|= LOOKUP_DIRECTORY
;
1477 * p1 and p2 should be directories on the same fs.
1479 struct dentry
*lock_rename(struct dentry
*p1
, struct dentry
*p2
)
1484 mutex_lock_nested(&p1
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1488 mutex_lock(&p1
->d_inode
->i_sb
->s_vfs_rename_mutex
);
1490 for (p
= p1
; p
->d_parent
!= p
; p
= p
->d_parent
) {
1491 if (p
->d_parent
== p2
) {
1492 mutex_lock_nested(&p2
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1493 mutex_lock_nested(&p1
->d_inode
->i_mutex
, I_MUTEX_CHILD
);
1498 for (p
= p2
; p
->d_parent
!= p
; p
= p
->d_parent
) {
1499 if (p
->d_parent
== p1
) {
1500 mutex_lock_nested(&p1
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1501 mutex_lock_nested(&p2
->d_inode
->i_mutex
, I_MUTEX_CHILD
);
1506 mutex_lock_nested(&p1
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1507 mutex_lock_nested(&p2
->d_inode
->i_mutex
, I_MUTEX_CHILD
);
1511 void unlock_rename(struct dentry
*p1
, struct dentry
*p2
)
1513 mutex_unlock(&p1
->d_inode
->i_mutex
);
1515 mutex_unlock(&p2
->d_inode
->i_mutex
);
1516 mutex_unlock(&p1
->d_inode
->i_sb
->s_vfs_rename_mutex
);
1520 int vfs_create(struct inode
*dir
, struct dentry
*dentry
, int mode
,
1521 struct nameidata
*nd
)
1523 int error
= may_create(dir
, dentry
, nd
);
1528 if (!dir
->i_op
|| !dir
->i_op
->create
)
1529 return -EACCES
; /* shouldn't it be ENOSYS? */
1532 error
= security_inode_create(dir
, dentry
, mode
);
1536 error
= dir
->i_op
->create(dir
, dentry
, mode
, nd
);
1538 fsnotify_create(dir
, dentry
);
1542 int may_open(struct nameidata
*nd
, int acc_mode
, int flag
)
1544 struct dentry
*dentry
= nd
->path
.dentry
;
1545 struct inode
*inode
= dentry
->d_inode
;
1551 if (S_ISLNK(inode
->i_mode
))
1554 if (S_ISDIR(inode
->i_mode
) && (acc_mode
& MAY_WRITE
))
1558 * FIFO's, sockets and device files are special: they don't
1559 * actually live on the filesystem itself, and as such you
1560 * can write to them even if the filesystem is read-only.
1562 if (S_ISFIFO(inode
->i_mode
) || S_ISSOCK(inode
->i_mode
)) {
1564 } else if (S_ISBLK(inode
->i_mode
) || S_ISCHR(inode
->i_mode
)) {
1565 if (nd
->path
.mnt
->mnt_flags
& MNT_NODEV
)
1571 error
= vfs_permission(nd
, acc_mode
);
1575 * An append-only file must be opened in append mode for writing.
1577 if (IS_APPEND(inode
)) {
1578 if ((flag
& FMODE_WRITE
) && !(flag
& O_APPEND
))
1584 /* O_NOATIME can only be set by the owner or superuser */
1585 if (flag
& O_NOATIME
)
1586 if (!is_owner_or_cap(inode
))
1590 * Ensure there are no outstanding leases on the file.
1592 error
= break_lease(inode
, flag
);
1596 if (flag
& O_TRUNC
) {
1597 error
= get_write_access(inode
);
1602 * Refuse to truncate files with mandatory locks held on them.
1604 error
= locks_verify_locked(inode
);
1608 error
= do_truncate(dentry
, 0,
1609 ATTR_MTIME
|ATTR_CTIME
|ATTR_OPEN
,
1612 put_write_access(inode
);
1616 if (flag
& FMODE_WRITE
)
1623 * Be careful about ever adding any more callers of this
1624 * function. Its flags must be in the namei format, not
1625 * what get passed to sys_open().
1627 static int __open_namei_create(struct nameidata
*nd
, struct path
*path
,
1631 struct dentry
*dir
= nd
->path
.dentry
;
1633 if (!IS_POSIXACL(dir
->d_inode
))
1634 mode
&= ~current
->fs
->umask
;
1635 error
= vfs_create(dir
->d_inode
, path
->dentry
, mode
, nd
);
1636 mutex_unlock(&dir
->d_inode
->i_mutex
);
1637 dput(nd
->path
.dentry
);
1638 nd
->path
.dentry
= path
->dentry
;
1641 /* Don't check for write permission, don't truncate */
1642 return may_open(nd
, 0, flag
& ~O_TRUNC
);
1646 * Note that while the flag value (low two bits) for sys_open means:
1651 * it is changed into
1652 * 00 - no permissions needed
1653 * 01 - read-permission
1654 * 10 - write-permission
1656 * for the internal routines (ie open_namei()/follow_link() etc)
1657 * This is more logical, and also allows the 00 "no perm needed"
1658 * to be used for symlinks (where the permissions are checked
1662 static inline int open_to_namei_flags(int flag
)
1664 if ((flag
+1) & O_ACCMODE
)
1669 static int open_will_write_to_fs(int flag
, struct inode
*inode
)
1672 * We'll never write to the fs underlying
1675 if (special_file(inode
->i_mode
))
1677 return (flag
& O_TRUNC
);
1681 * Note that the low bits of the passed in "open_flag"
1682 * are not the same as in the local variable "flag". See
1683 * open_to_namei_flags() for more details.
1685 struct file
*do_filp_open(int dfd
, const char *pathname
,
1686 int open_flag
, int mode
)
1689 struct nameidata nd
;
1690 int acc_mode
, error
;
1695 int flag
= open_to_namei_flags(open_flag
);
1697 acc_mode
= ACC_MODE(flag
);
1699 /* O_TRUNC implies we need access checks for write permissions */
1701 acc_mode
|= MAY_WRITE
;
1703 /* Allow the LSM permission hook to distinguish append
1704 access from general write access. */
1705 if (flag
& O_APPEND
)
1706 acc_mode
|= MAY_APPEND
;
1709 * The simplest case - just a plain lookup.
1711 if (!(flag
& O_CREAT
)) {
1712 error
= path_lookup_open(dfd
, pathname
, lookup_flags(flag
),
1715 return ERR_PTR(error
);
1720 * Create - we need to know the parent.
1722 error
= path_lookup_create(dfd
, pathname
, LOOKUP_PARENT
,
1725 return ERR_PTR(error
);
1728 * We have the parent and last component. First of all, check
1729 * that we are not asked to creat(2) an obvious directory - that
1733 if (nd
.last_type
!= LAST_NORM
|| nd
.last
.name
[nd
.last
.len
])
1736 dir
= nd
.path
.dentry
;
1737 nd
.flags
&= ~LOOKUP_PARENT
;
1738 mutex_lock(&dir
->d_inode
->i_mutex
);
1739 path
.dentry
= lookup_hash(&nd
);
1740 path
.mnt
= nd
.path
.mnt
;
1743 error
= PTR_ERR(path
.dentry
);
1744 if (IS_ERR(path
.dentry
)) {
1745 mutex_unlock(&dir
->d_inode
->i_mutex
);
1749 if (IS_ERR(nd
.intent
.open
.file
)) {
1750 error
= PTR_ERR(nd
.intent
.open
.file
);
1751 goto exit_mutex_unlock
;
1754 /* Negative dentry, just create the file */
1755 if (!path
.dentry
->d_inode
) {
1757 * This write is needed to ensure that a
1758 * ro->rw transition does not occur between
1759 * the time when the file is created and when
1760 * a permanent write count is taken through
1761 * the 'struct file' in nameidata_to_filp().
1763 error
= mnt_want_write(nd
.path
.mnt
);
1765 goto exit_mutex_unlock
;
1766 error
= __open_namei_create(&nd
, &path
, flag
, mode
);
1768 mnt_drop_write(nd
.path
.mnt
);
1771 filp
= nameidata_to_filp(&nd
, open_flag
);
1772 mnt_drop_write(nd
.path
.mnt
);
1777 * It already exists.
1779 mutex_unlock(&dir
->d_inode
->i_mutex
);
1780 audit_inode(pathname
, path
.dentry
);
1786 if (__follow_mount(&path
)) {
1788 if (flag
& O_NOFOLLOW
)
1793 if (!path
.dentry
->d_inode
)
1795 if (path
.dentry
->d_inode
->i_op
&& path
.dentry
->d_inode
->i_op
->follow_link
)
1798 path_to_nameidata(&path
, &nd
);
1800 if (path
.dentry
->d_inode
&& S_ISDIR(path
.dentry
->d_inode
->i_mode
))
1805 * 1. may_open() truncates a file
1806 * 2. a rw->ro mount transition occurs
1807 * 3. nameidata_to_filp() fails due to
1809 * That would be inconsistent, and should
1810 * be avoided. Taking this mnt write here
1811 * ensures that (2) can not occur.
1813 will_write
= open_will_write_to_fs(flag
, nd
.path
.dentry
->d_inode
);
1815 error
= mnt_want_write(nd
.path
.mnt
);
1819 error
= may_open(&nd
, acc_mode
, flag
);
1822 mnt_drop_write(nd
.path
.mnt
);
1825 filp
= nameidata_to_filp(&nd
, open_flag
);
1827 * It is now safe to drop the mnt write
1828 * because the filp has had a write taken
1832 mnt_drop_write(nd
.path
.mnt
);
1836 mutex_unlock(&dir
->d_inode
->i_mutex
);
1838 path_put_conditional(&path
, &nd
);
1840 if (!IS_ERR(nd
.intent
.open
.file
))
1841 release_open_intent(&nd
);
1843 return ERR_PTR(error
);
1847 if (flag
& O_NOFOLLOW
)
1850 * This is subtle. Instead of calling do_follow_link() we do the
1851 * thing by hands. The reason is that this way we have zero link_count
1852 * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT.
1853 * After that we have the parent and last component, i.e.
1854 * we are in the same situation as after the first path_walk().
1855 * Well, almost - if the last component is normal we get its copy
1856 * stored in nd->last.name and we will have to putname() it when we
1857 * are done. Procfs-like symlinks just set LAST_BIND.
1859 nd
.flags
|= LOOKUP_PARENT
;
1860 error
= security_inode_follow_link(path
.dentry
, &nd
);
1863 error
= __do_follow_link(&path
, &nd
);
1865 /* Does someone understand code flow here? Or it is only
1866 * me so stupid? Anathema to whoever designed this non-sense
1867 * with "intent.open".
1869 release_open_intent(&nd
);
1870 return ERR_PTR(error
);
1872 nd
.flags
&= ~LOOKUP_PARENT
;
1873 if (nd
.last_type
== LAST_BIND
)
1876 if (nd
.last_type
!= LAST_NORM
)
1878 if (nd
.last
.name
[nd
.last
.len
]) {
1879 __putname(nd
.last
.name
);
1884 __putname(nd
.last
.name
);
1887 dir
= nd
.path
.dentry
;
1888 mutex_lock(&dir
->d_inode
->i_mutex
);
1889 path
.dentry
= lookup_hash(&nd
);
1890 path
.mnt
= nd
.path
.mnt
;
1891 __putname(nd
.last
.name
);
1896 * filp_open - open file and return file pointer
1898 * @filename: path to open
1899 * @flags: open flags as per the open(2) second argument
1900 * @mode: mode for the new file if O_CREAT is set, else ignored
1902 * This is the helper to open a file from kernelspace if you really
1903 * have to. But in generally you should not do this, so please move
1904 * along, nothing to see here..
1906 struct file
*filp_open(const char *filename
, int flags
, int mode
)
1908 return do_filp_open(AT_FDCWD
, filename
, flags
, mode
);
1910 EXPORT_SYMBOL(filp_open
);
1913 * lookup_create - lookup a dentry, creating it if it doesn't exist
1914 * @nd: nameidata info
1915 * @is_dir: directory flag
1917 * Simple function to lookup and return a dentry and create it
1918 * if it doesn't exist. Is SMP-safe.
1920 * Returns with nd->path.dentry->d_inode->i_mutex locked.
1922 struct dentry
*lookup_create(struct nameidata
*nd
, int is_dir
)
1924 struct dentry
*dentry
= ERR_PTR(-EEXIST
);
1926 mutex_lock_nested(&nd
->path
.dentry
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1928 * Yucky last component or no last component at all?
1929 * (foo/., foo/.., /////)
1931 if (nd
->last_type
!= LAST_NORM
)
1933 nd
->flags
&= ~LOOKUP_PARENT
;
1934 nd
->flags
|= LOOKUP_CREATE
;
1935 nd
->intent
.open
.flags
= O_EXCL
;
1938 * Do the final lookup.
1940 dentry
= lookup_hash(nd
);
1944 if (dentry
->d_inode
)
1947 * Special case - lookup gave negative, but... we had foo/bar/
1948 * From the vfs_mknod() POV we just have a negative dentry -
1949 * all is fine. Let's be bastards - you had / on the end, you've
1950 * been asking for (non-existent) directory. -ENOENT for you.
1952 if (unlikely(!is_dir
&& nd
->last
.name
[nd
->last
.len
])) {
1954 dentry
= ERR_PTR(-ENOENT
);
1959 dentry
= ERR_PTR(-EEXIST
);
1963 EXPORT_SYMBOL_GPL(lookup_create
);
1965 int vfs_mknod(struct inode
*dir
, struct dentry
*dentry
, int mode
, dev_t dev
)
1967 int error
= may_create(dir
, dentry
, NULL
);
1972 if ((S_ISCHR(mode
) || S_ISBLK(mode
)) && !capable(CAP_MKNOD
))
1975 if (!dir
->i_op
|| !dir
->i_op
->mknod
)
1978 error
= devcgroup_inode_mknod(mode
, dev
);
1982 error
= security_inode_mknod(dir
, dentry
, mode
, dev
);
1987 error
= dir
->i_op
->mknod(dir
, dentry
, mode
, dev
);
1989 fsnotify_create(dir
, dentry
);
1993 static int may_mknod(mode_t mode
)
1995 switch (mode
& S_IFMT
) {
2001 case 0: /* zero mode translates to S_IFREG */
2010 asmlinkage
long sys_mknodat(int dfd
, const char __user
*filename
, int mode
,
2015 struct dentry
* dentry
;
2016 struct nameidata nd
;
2020 tmp
= getname(filename
);
2022 return PTR_ERR(tmp
);
2024 error
= do_path_lookup(dfd
, tmp
, LOOKUP_PARENT
, &nd
);
2027 dentry
= lookup_create(&nd
, 0);
2028 if (IS_ERR(dentry
)) {
2029 error
= PTR_ERR(dentry
);
2032 if (!IS_POSIXACL(nd
.path
.dentry
->d_inode
))
2033 mode
&= ~current
->fs
->umask
;
2034 error
= may_mknod(mode
);
2037 error
= mnt_want_write(nd
.path
.mnt
);
2040 switch (mode
& S_IFMT
) {
2041 case 0: case S_IFREG
:
2042 error
= vfs_create(nd
.path
.dentry
->d_inode
,dentry
,mode
,&nd
);
2044 case S_IFCHR
: case S_IFBLK
:
2045 error
= vfs_mknod(nd
.path
.dentry
->d_inode
,dentry
,mode
,
2046 new_decode_dev(dev
));
2048 case S_IFIFO
: case S_IFSOCK
:
2049 error
= vfs_mknod(nd
.path
.dentry
->d_inode
,dentry
,mode
,0);
2052 mnt_drop_write(nd
.path
.mnt
);
2056 mutex_unlock(&nd
.path
.dentry
->d_inode
->i_mutex
);
2064 asmlinkage
long sys_mknod(const char __user
*filename
, int mode
, unsigned dev
)
2066 return sys_mknodat(AT_FDCWD
, filename
, mode
, dev
);
2069 int vfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
2071 int error
= may_create(dir
, dentry
, NULL
);
2076 if (!dir
->i_op
|| !dir
->i_op
->mkdir
)
2079 mode
&= (S_IRWXUGO
|S_ISVTX
);
2080 error
= security_inode_mkdir(dir
, dentry
, mode
);
2085 error
= dir
->i_op
->mkdir(dir
, dentry
, mode
);
2087 fsnotify_mkdir(dir
, dentry
);
2091 asmlinkage
long sys_mkdirat(int dfd
, const char __user
*pathname
, int mode
)
2095 struct dentry
*dentry
;
2096 struct nameidata nd
;
2098 tmp
= getname(pathname
);
2099 error
= PTR_ERR(tmp
);
2103 error
= do_path_lookup(dfd
, tmp
, LOOKUP_PARENT
, &nd
);
2106 dentry
= lookup_create(&nd
, 1);
2107 error
= PTR_ERR(dentry
);
2111 if (!IS_POSIXACL(nd
.path
.dentry
->d_inode
))
2112 mode
&= ~current
->fs
->umask
;
2113 error
= mnt_want_write(nd
.path
.mnt
);
2116 error
= vfs_mkdir(nd
.path
.dentry
->d_inode
, dentry
, mode
);
2117 mnt_drop_write(nd
.path
.mnt
);
2121 mutex_unlock(&nd
.path
.dentry
->d_inode
->i_mutex
);
2129 asmlinkage
long sys_mkdir(const char __user
*pathname
, int mode
)
2131 return sys_mkdirat(AT_FDCWD
, pathname
, mode
);
2135 * We try to drop the dentry early: we should have
2136 * a usage count of 2 if we're the only user of this
2137 * dentry, and if that is true (possibly after pruning
2138 * the dcache), then we drop the dentry now.
2140 * A low-level filesystem can, if it choses, legally
2143 * if (!d_unhashed(dentry))
2146 * if it cannot handle the case of removing a directory
2147 * that is still in use by something else..
2149 void dentry_unhash(struct dentry
*dentry
)
2152 shrink_dcache_parent(dentry
);
2153 spin_lock(&dcache_lock
);
2154 spin_lock(&dentry
->d_lock
);
2155 if (atomic_read(&dentry
->d_count
) == 2)
2157 spin_unlock(&dentry
->d_lock
);
2158 spin_unlock(&dcache_lock
);
2161 int vfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
2163 int error
= may_delete(dir
, dentry
, 1);
2168 if (!dir
->i_op
|| !dir
->i_op
->rmdir
)
2173 mutex_lock(&dentry
->d_inode
->i_mutex
);
2174 dentry_unhash(dentry
);
2175 if (d_mountpoint(dentry
))
2178 error
= security_inode_rmdir(dir
, dentry
);
2180 error
= dir
->i_op
->rmdir(dir
, dentry
);
2182 dentry
->d_inode
->i_flags
|= S_DEAD
;
2185 mutex_unlock(&dentry
->d_inode
->i_mutex
);
2194 static long do_rmdir(int dfd
, const char __user
*pathname
)
2198 struct dentry
*dentry
;
2199 struct nameidata nd
;
2201 name
= getname(pathname
);
2203 return PTR_ERR(name
);
2205 error
= do_path_lookup(dfd
, name
, LOOKUP_PARENT
, &nd
);
2209 switch(nd
.last_type
) {
2220 mutex_lock_nested(&nd
.path
.dentry
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
2221 dentry
= lookup_hash(&nd
);
2222 error
= PTR_ERR(dentry
);
2225 error
= mnt_want_write(nd
.path
.mnt
);
2228 error
= vfs_rmdir(nd
.path
.dentry
->d_inode
, dentry
);
2229 mnt_drop_write(nd
.path
.mnt
);
2233 mutex_unlock(&nd
.path
.dentry
->d_inode
->i_mutex
);
2241 asmlinkage
long sys_rmdir(const char __user
*pathname
)
2243 return do_rmdir(AT_FDCWD
, pathname
);
2246 int vfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
2248 int error
= may_delete(dir
, dentry
, 0);
2253 if (!dir
->i_op
|| !dir
->i_op
->unlink
)
2258 mutex_lock(&dentry
->d_inode
->i_mutex
);
2259 if (d_mountpoint(dentry
))
2262 error
= security_inode_unlink(dir
, dentry
);
2264 error
= dir
->i_op
->unlink(dir
, dentry
);
2266 mutex_unlock(&dentry
->d_inode
->i_mutex
);
2268 /* We don't d_delete() NFS sillyrenamed files--they still exist. */
2269 if (!error
&& !(dentry
->d_flags
& DCACHE_NFSFS_RENAMED
)) {
2270 fsnotify_link_count(dentry
->d_inode
);
2278 * Make sure that the actual truncation of the file will occur outside its
2279 * directory's i_mutex. Truncate can take a long time if there is a lot of
2280 * writeout happening, and we don't want to prevent access to the directory
2281 * while waiting on the I/O.
2283 static long do_unlinkat(int dfd
, const char __user
*pathname
)
2287 struct dentry
*dentry
;
2288 struct nameidata nd
;
2289 struct inode
*inode
= NULL
;
2291 name
= getname(pathname
);
2293 return PTR_ERR(name
);
2295 error
= do_path_lookup(dfd
, name
, LOOKUP_PARENT
, &nd
);
2299 if (nd
.last_type
!= LAST_NORM
)
2301 mutex_lock_nested(&nd
.path
.dentry
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
2302 dentry
= lookup_hash(&nd
);
2303 error
= PTR_ERR(dentry
);
2304 if (!IS_ERR(dentry
)) {
2305 /* Why not before? Because we want correct error value */
2306 if (nd
.last
.name
[nd
.last
.len
])
2308 inode
= dentry
->d_inode
;
2310 atomic_inc(&inode
->i_count
);
2311 error
= mnt_want_write(nd
.path
.mnt
);
2314 error
= vfs_unlink(nd
.path
.dentry
->d_inode
, dentry
);
2315 mnt_drop_write(nd
.path
.mnt
);
2319 mutex_unlock(&nd
.path
.dentry
->d_inode
->i_mutex
);
2321 iput(inode
); /* truncate the inode here */
2329 error
= !dentry
->d_inode
? -ENOENT
:
2330 S_ISDIR(dentry
->d_inode
->i_mode
) ? -EISDIR
: -ENOTDIR
;
2334 asmlinkage
long sys_unlinkat(int dfd
, const char __user
*pathname
, int flag
)
2336 if ((flag
& ~AT_REMOVEDIR
) != 0)
2339 if (flag
& AT_REMOVEDIR
)
2340 return do_rmdir(dfd
, pathname
);
2342 return do_unlinkat(dfd
, pathname
);
2345 asmlinkage
long sys_unlink(const char __user
*pathname
)
2347 return do_unlinkat(AT_FDCWD
, pathname
);
2350 int vfs_symlink(struct inode
*dir
, struct dentry
*dentry
, const char *oldname
)
2352 int error
= may_create(dir
, dentry
, NULL
);
2357 if (!dir
->i_op
|| !dir
->i_op
->symlink
)
2360 error
= security_inode_symlink(dir
, dentry
, oldname
);
2365 error
= dir
->i_op
->symlink(dir
, dentry
, oldname
);
2367 fsnotify_create(dir
, dentry
);
2371 asmlinkage
long sys_symlinkat(const char __user
*oldname
,
2372 int newdfd
, const char __user
*newname
)
2377 struct dentry
*dentry
;
2378 struct nameidata nd
;
2380 from
= getname(oldname
);
2382 return PTR_ERR(from
);
2383 to
= getname(newname
);
2384 error
= PTR_ERR(to
);
2388 error
= do_path_lookup(newdfd
, to
, LOOKUP_PARENT
, &nd
);
2391 dentry
= lookup_create(&nd
, 0);
2392 error
= PTR_ERR(dentry
);
2396 error
= mnt_want_write(nd
.path
.mnt
);
2399 error
= vfs_symlink(nd
.path
.dentry
->d_inode
, dentry
, from
);
2400 mnt_drop_write(nd
.path
.mnt
);
2404 mutex_unlock(&nd
.path
.dentry
->d_inode
->i_mutex
);
2413 asmlinkage
long sys_symlink(const char __user
*oldname
, const char __user
*newname
)
2415 return sys_symlinkat(oldname
, AT_FDCWD
, newname
);
2418 int vfs_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*new_dentry
)
2420 struct inode
*inode
= old_dentry
->d_inode
;
2426 error
= may_create(dir
, new_dentry
, NULL
);
2430 if (dir
->i_sb
!= inode
->i_sb
)
2434 * A link to an append-only or immutable file cannot be created.
2436 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
2438 if (!dir
->i_op
|| !dir
->i_op
->link
)
2440 if (S_ISDIR(inode
->i_mode
))
2443 error
= security_inode_link(old_dentry
, dir
, new_dentry
);
2447 mutex_lock(&inode
->i_mutex
);
2449 error
= dir
->i_op
->link(old_dentry
, dir
, new_dentry
);
2450 mutex_unlock(&inode
->i_mutex
);
2452 fsnotify_link(dir
, inode
, new_dentry
);
2457 * Hardlinks are often used in delicate situations. We avoid
2458 * security-related surprises by not following symlinks on the
2461 * We don't follow them on the oldname either to be compatible
2462 * with linux 2.0, and to avoid hard-linking to directories
2463 * and other special files. --ADM
2465 asmlinkage
long sys_linkat(int olddfd
, const char __user
*oldname
,
2466 int newdfd
, const char __user
*newname
,
2469 struct dentry
*new_dentry
;
2470 struct nameidata nd
, old_nd
;
2474 if ((flags
& ~AT_SYMLINK_FOLLOW
) != 0)
2477 to
= getname(newname
);
2481 error
= __user_walk_fd(olddfd
, oldname
,
2482 flags
& AT_SYMLINK_FOLLOW
? LOOKUP_FOLLOW
: 0,
2486 error
= do_path_lookup(newdfd
, to
, LOOKUP_PARENT
, &nd
);
2490 if (old_nd
.path
.mnt
!= nd
.path
.mnt
)
2492 new_dentry
= lookup_create(&nd
, 0);
2493 error
= PTR_ERR(new_dentry
);
2494 if (IS_ERR(new_dentry
))
2496 error
= mnt_want_write(nd
.path
.mnt
);
2499 error
= vfs_link(old_nd
.path
.dentry
, nd
.path
.dentry
->d_inode
, new_dentry
);
2500 mnt_drop_write(nd
.path
.mnt
);
2504 mutex_unlock(&nd
.path
.dentry
->d_inode
->i_mutex
);
2508 path_put(&old_nd
.path
);
2515 asmlinkage
long sys_link(const char __user
*oldname
, const char __user
*newname
)
2517 return sys_linkat(AT_FDCWD
, oldname
, AT_FDCWD
, newname
, 0);
2521 * The worst of all namespace operations - renaming directory. "Perverted"
2522 * doesn't even start to describe it. Somebody in UCB had a heck of a trip...
2524 * a) we can get into loop creation. Check is done in is_subdir().
2525 * b) race potential - two innocent renames can create a loop together.
2526 * That's where 4.4 screws up. Current fix: serialization on
2527 * sb->s_vfs_rename_mutex. We might be more accurate, but that's another
2529 * c) we have to lock _three_ objects - parents and victim (if it exists).
2530 * And that - after we got ->i_mutex on parents (until then we don't know
2531 * whether the target exists). Solution: try to be smart with locking
2532 * order for inodes. We rely on the fact that tree topology may change
2533 * only under ->s_vfs_rename_mutex _and_ that parent of the object we
2534 * move will be locked. Thus we can rank directories by the tree
2535 * (ancestors first) and rank all non-directories after them.
2536 * That works since everybody except rename does "lock parent, lookup,
2537 * lock child" and rename is under ->s_vfs_rename_mutex.
2538 * HOWEVER, it relies on the assumption that any object with ->lookup()
2539 * has no more than 1 dentry. If "hybrid" objects will ever appear,
2540 * we'd better make sure that there's no link(2) for them.
2541 * d) some filesystems don't support opened-but-unlinked directories,
2542 * either because of layout or because they are not ready to deal with
2543 * all cases correctly. The latter will be fixed (taking this sort of
2544 * stuff into VFS), but the former is not going away. Solution: the same
2545 * trick as in rmdir().
2546 * e) conversion from fhandle to dentry may come in the wrong moment - when
2547 * we are removing the target. Solution: we will have to grab ->i_mutex
2548 * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on
2549 * ->i_mutex on parents, which works but leads to some truely excessive
2552 static int vfs_rename_dir(struct inode
*old_dir
, struct dentry
*old_dentry
,
2553 struct inode
*new_dir
, struct dentry
*new_dentry
)
2556 struct inode
*target
;
2559 * If we are going to change the parent - check write permissions,
2560 * we'll need to flip '..'.
2562 if (new_dir
!= old_dir
) {
2563 error
= permission(old_dentry
->d_inode
, MAY_WRITE
, NULL
);
2568 error
= security_inode_rename(old_dir
, old_dentry
, new_dir
, new_dentry
);
2572 target
= new_dentry
->d_inode
;
2574 mutex_lock(&target
->i_mutex
);
2575 dentry_unhash(new_dentry
);
2577 if (d_mountpoint(old_dentry
)||d_mountpoint(new_dentry
))
2580 error
= old_dir
->i_op
->rename(old_dir
, old_dentry
, new_dir
, new_dentry
);
2583 target
->i_flags
|= S_DEAD
;
2584 mutex_unlock(&target
->i_mutex
);
2585 if (d_unhashed(new_dentry
))
2586 d_rehash(new_dentry
);
2590 if (!(old_dir
->i_sb
->s_type
->fs_flags
& FS_RENAME_DOES_D_MOVE
))
2591 d_move(old_dentry
,new_dentry
);
2595 static int vfs_rename_other(struct inode
*old_dir
, struct dentry
*old_dentry
,
2596 struct inode
*new_dir
, struct dentry
*new_dentry
)
2598 struct inode
*target
;
2601 error
= security_inode_rename(old_dir
, old_dentry
, new_dir
, new_dentry
);
2606 target
= new_dentry
->d_inode
;
2608 mutex_lock(&target
->i_mutex
);
2609 if (d_mountpoint(old_dentry
)||d_mountpoint(new_dentry
))
2612 error
= old_dir
->i_op
->rename(old_dir
, old_dentry
, new_dir
, new_dentry
);
2614 if (!(old_dir
->i_sb
->s_type
->fs_flags
& FS_RENAME_DOES_D_MOVE
))
2615 d_move(old_dentry
, new_dentry
);
2618 mutex_unlock(&target
->i_mutex
);
2623 int vfs_rename(struct inode
*old_dir
, struct dentry
*old_dentry
,
2624 struct inode
*new_dir
, struct dentry
*new_dentry
)
2627 int is_dir
= S_ISDIR(old_dentry
->d_inode
->i_mode
);
2628 const char *old_name
;
2630 if (old_dentry
->d_inode
== new_dentry
->d_inode
)
2633 error
= may_delete(old_dir
, old_dentry
, is_dir
);
2637 if (!new_dentry
->d_inode
)
2638 error
= may_create(new_dir
, new_dentry
, NULL
);
2640 error
= may_delete(new_dir
, new_dentry
, is_dir
);
2644 if (!old_dir
->i_op
|| !old_dir
->i_op
->rename
)
2647 DQUOT_INIT(old_dir
);
2648 DQUOT_INIT(new_dir
);
2650 old_name
= fsnotify_oldname_init(old_dentry
->d_name
.name
);
2653 error
= vfs_rename_dir(old_dir
,old_dentry
,new_dir
,new_dentry
);
2655 error
= vfs_rename_other(old_dir
,old_dentry
,new_dir
,new_dentry
);
2657 const char *new_name
= old_dentry
->d_name
.name
;
2658 fsnotify_move(old_dir
, new_dir
, old_name
, new_name
, is_dir
,
2659 new_dentry
->d_inode
, old_dentry
);
2661 fsnotify_oldname_free(old_name
);
2666 static int do_rename(int olddfd
, const char *oldname
,
2667 int newdfd
, const char *newname
)
2670 struct dentry
* old_dir
, * new_dir
;
2671 struct dentry
* old_dentry
, *new_dentry
;
2672 struct dentry
* trap
;
2673 struct nameidata oldnd
, newnd
;
2675 error
= do_path_lookup(olddfd
, oldname
, LOOKUP_PARENT
, &oldnd
);
2679 error
= do_path_lookup(newdfd
, newname
, LOOKUP_PARENT
, &newnd
);
2684 if (oldnd
.path
.mnt
!= newnd
.path
.mnt
)
2687 old_dir
= oldnd
.path
.dentry
;
2689 if (oldnd
.last_type
!= LAST_NORM
)
2692 new_dir
= newnd
.path
.dentry
;
2693 if (newnd
.last_type
!= LAST_NORM
)
2696 trap
= lock_rename(new_dir
, old_dir
);
2698 old_dentry
= lookup_hash(&oldnd
);
2699 error
= PTR_ERR(old_dentry
);
2700 if (IS_ERR(old_dentry
))
2702 /* source must exist */
2704 if (!old_dentry
->d_inode
)
2706 /* unless the source is a directory trailing slashes give -ENOTDIR */
2707 if (!S_ISDIR(old_dentry
->d_inode
->i_mode
)) {
2709 if (oldnd
.last
.name
[oldnd
.last
.len
])
2711 if (newnd
.last
.name
[newnd
.last
.len
])
2714 /* source should not be ancestor of target */
2716 if (old_dentry
== trap
)
2718 new_dentry
= lookup_hash(&newnd
);
2719 error
= PTR_ERR(new_dentry
);
2720 if (IS_ERR(new_dentry
))
2722 /* target should not be an ancestor of source */
2724 if (new_dentry
== trap
)
2727 error
= mnt_want_write(oldnd
.path
.mnt
);
2730 error
= vfs_rename(old_dir
->d_inode
, old_dentry
,
2731 new_dir
->d_inode
, new_dentry
);
2732 mnt_drop_write(oldnd
.path
.mnt
);
2738 unlock_rename(new_dir
, old_dir
);
2740 path_put(&newnd
.path
);
2742 path_put(&oldnd
.path
);
2747 asmlinkage
long sys_renameat(int olddfd
, const char __user
*oldname
,
2748 int newdfd
, const char __user
*newname
)
2754 from
= getname(oldname
);
2756 return PTR_ERR(from
);
2757 to
= getname(newname
);
2758 error
= PTR_ERR(to
);
2760 error
= do_rename(olddfd
, from
, newdfd
, to
);
2767 asmlinkage
long sys_rename(const char __user
*oldname
, const char __user
*newname
)
2769 return sys_renameat(AT_FDCWD
, oldname
, AT_FDCWD
, newname
);
2772 int vfs_readlink(struct dentry
*dentry
, char __user
*buffer
, int buflen
, const char *link
)
2776 len
= PTR_ERR(link
);
2781 if (len
> (unsigned) buflen
)
2783 if (copy_to_user(buffer
, link
, len
))
2790 * A helper for ->readlink(). This should be used *ONLY* for symlinks that
2791 * have ->follow_link() touching nd only in nd_set_link(). Using (or not
2792 * using) it for any given inode is up to filesystem.
2794 int generic_readlink(struct dentry
*dentry
, char __user
*buffer
, int buflen
)
2796 struct nameidata nd
;
2801 cookie
= dentry
->d_inode
->i_op
->follow_link(dentry
, &nd
);
2803 return PTR_ERR(cookie
);
2805 res
= vfs_readlink(dentry
, buffer
, buflen
, nd_get_link(&nd
));
2806 if (dentry
->d_inode
->i_op
->put_link
)
2807 dentry
->d_inode
->i_op
->put_link(dentry
, &nd
, cookie
);
2811 int vfs_follow_link(struct nameidata
*nd
, const char *link
)
2813 return __vfs_follow_link(nd
, link
);
2816 /* get the link contents into pagecache */
2817 static char *page_getlink(struct dentry
* dentry
, struct page
**ppage
)
2820 struct address_space
*mapping
= dentry
->d_inode
->i_mapping
;
2821 page
= read_mapping_page(mapping
, 0, NULL
);
2828 int page_readlink(struct dentry
*dentry
, char __user
*buffer
, int buflen
)
2830 struct page
*page
= NULL
;
2831 char *s
= page_getlink(dentry
, &page
);
2832 int res
= vfs_readlink(dentry
,buffer
,buflen
,s
);
2835 page_cache_release(page
);
2840 void *page_follow_link_light(struct dentry
*dentry
, struct nameidata
*nd
)
2842 struct page
*page
= NULL
;
2843 nd_set_link(nd
, page_getlink(dentry
, &page
));
2847 void page_put_link(struct dentry
*dentry
, struct nameidata
*nd
, void *cookie
)
2849 struct page
*page
= cookie
;
2853 page_cache_release(page
);
2857 int __page_symlink(struct inode
*inode
, const char *symname
, int len
,
2860 struct address_space
*mapping
= inode
->i_mapping
;
2867 err
= pagecache_write_begin(NULL
, mapping
, 0, len
-1,
2868 AOP_FLAG_UNINTERRUPTIBLE
, &page
, &fsdata
);
2872 kaddr
= kmap_atomic(page
, KM_USER0
);
2873 memcpy(kaddr
, symname
, len
-1);
2874 kunmap_atomic(kaddr
, KM_USER0
);
2876 err
= pagecache_write_end(NULL
, mapping
, 0, len
-1, len
-1,
2883 mark_inode_dirty(inode
);
2889 int page_symlink(struct inode
*inode
, const char *symname
, int len
)
2891 return __page_symlink(inode
, symname
, len
,
2892 mapping_gfp_mask(inode
->i_mapping
));
2895 const struct inode_operations page_symlink_inode_operations
= {
2896 .readlink
= generic_readlink
,
2897 .follow_link
= page_follow_link_light
,
2898 .put_link
= page_put_link
,
2901 EXPORT_SYMBOL(__user_walk
);
2902 EXPORT_SYMBOL(__user_walk_fd
);
2903 EXPORT_SYMBOL(follow_down
);
2904 EXPORT_SYMBOL(follow_up
);
2905 EXPORT_SYMBOL(get_write_access
); /* binfmt_aout */
2906 EXPORT_SYMBOL(getname
);
2907 EXPORT_SYMBOL(lock_rename
);
2908 EXPORT_SYMBOL(lookup_one_len
);
2909 EXPORT_SYMBOL(page_follow_link_light
);
2910 EXPORT_SYMBOL(page_put_link
);
2911 EXPORT_SYMBOL(page_readlink
);
2912 EXPORT_SYMBOL(__page_symlink
);
2913 EXPORT_SYMBOL(page_symlink
);
2914 EXPORT_SYMBOL(page_symlink_inode_operations
);
2915 EXPORT_SYMBOL(path_lookup
);
2916 EXPORT_SYMBOL(vfs_path_lookup
);
2917 EXPORT_SYMBOL(permission
);
2918 EXPORT_SYMBOL(vfs_permission
);
2919 EXPORT_SYMBOL(file_permission
);
2920 EXPORT_SYMBOL(unlock_rename
);
2921 EXPORT_SYMBOL(vfs_create
);
2922 EXPORT_SYMBOL(vfs_follow_link
);
2923 EXPORT_SYMBOL(vfs_link
);
2924 EXPORT_SYMBOL(vfs_mkdir
);
2925 EXPORT_SYMBOL(vfs_mknod
);
2926 EXPORT_SYMBOL(generic_permission
);
2927 EXPORT_SYMBOL(vfs_readlink
);
2928 EXPORT_SYMBOL(vfs_rename
);
2929 EXPORT_SYMBOL(vfs_rmdir
);
2930 EXPORT_SYMBOL(vfs_symlink
);
2931 EXPORT_SYMBOL(vfs_unlink
);
2932 EXPORT_SYMBOL(dentry_unhash
);
2933 EXPORT_SYMBOL(generic_readlink
);