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/namei.h>
34 #include <asm/namei.h>
35 #include <asm/uaccess.h>
37 #define ACC_MODE(x) ("\000\004\002\006"[(x)&O_ACCMODE])
39 /* [Feb-1997 T. Schoebel-Theuer]
40 * Fundamental changes in the pathname lookup mechanisms (namei)
41 * were necessary because of omirr. The reason is that omirr needs
42 * to know the _real_ pathname, not the user-supplied one, in case
43 * of symlinks (and also when transname replacements occur).
45 * The new code replaces the old recursive symlink resolution with
46 * an iterative one (in case of non-nested symlink chains). It does
47 * this with calls to <fs>_follow_link().
48 * As a side effect, dir_namei(), _namei() and follow_link() are now
49 * replaced with a single function lookup_dentry() that can handle all
50 * the special cases of the former code.
52 * With the new dcache, the pathname is stored at each inode, at least as
53 * long as the refcount of the inode is positive. As a side effect, the
54 * size of the dcache depends on the inode cache and thus is dynamic.
56 * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink
57 * resolution to correspond with current state of the code.
59 * Note that the symlink resolution is not *completely* iterative.
60 * There is still a significant amount of tail- and mid- recursion in
61 * the algorithm. Also, note that <fs>_readlink() is not used in
62 * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink()
63 * may return different results than <fs>_follow_link(). Many virtual
64 * filesystems (including /proc) exhibit this behavior.
67 /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation:
68 * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL
69 * and the name already exists in form of a symlink, try to create the new
70 * name indicated by the symlink. The old code always complained that the
71 * name already exists, due to not following the symlink even if its target
72 * is nonexistent. The new semantics affects also mknod() and link() when
73 * the name is a symlink pointing to a non-existant name.
75 * I don't know which semantics is the right one, since I have no access
76 * to standards. But I found by trial that HP-UX 9.0 has the full "new"
77 * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the
78 * "old" one. Personally, I think the new semantics is much more logical.
79 * Note that "ln old new" where "new" is a symlink pointing to a non-existing
80 * file does succeed in both HP-UX and SunOs, but not in Solaris
81 * and in the old Linux semantics.
84 /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink
85 * semantics. See the comments in "open_namei" and "do_link" below.
87 * [10-Sep-98 Alan Modra] Another symlink change.
90 /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks:
91 * inside the path - always follow.
92 * in the last component in creation/removal/renaming - never follow.
93 * if LOOKUP_FOLLOW passed - follow.
94 * if the pathname has trailing slashes - follow.
95 * otherwise - don't follow.
96 * (applied in that order).
98 * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT
99 * restored for 2.4. This is the last surviving part of old 4.2BSD bug.
100 * During the 2.4 we need to fix the userland stuff depending on it -
101 * hopefully we will be able to get rid of that wart in 2.5. So far only
102 * XEmacs seems to be relying on it...
105 * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland)
106 * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives
107 * any extra contention...
110 /* In order to reduce some races, while at the same time doing additional
111 * checking and hopefully speeding things up, we copy filenames to the
112 * kernel data space before using them..
114 * POSIX.1 2.4: an empty pathname is invalid (ENOENT).
115 * PATH_MAX includes the nul terminator --RR.
117 static int do_getname(const char __user
*filename
, char *page
)
120 unsigned long len
= PATH_MAX
;
122 if (!segment_eq(get_fs(), KERNEL_DS
)) {
123 if ((unsigned long) filename
>= TASK_SIZE
)
125 if (TASK_SIZE
- (unsigned long) filename
< PATH_MAX
)
126 len
= TASK_SIZE
- (unsigned long) filename
;
129 retval
= strncpy_from_user(page
, filename
, len
);
133 return -ENAMETOOLONG
;
139 char * getname(const char __user
* filename
)
143 result
= ERR_PTR(-ENOMEM
);
146 int retval
= do_getname(filename
, tmp
);
151 result
= ERR_PTR(retval
);
154 audit_getname(result
);
158 #ifdef CONFIG_AUDITSYSCALL
159 void putname(const char *name
)
161 if (unlikely(!audit_dummy_context()))
166 EXPORT_SYMBOL(putname
);
171 * generic_permission - check for access rights on a Posix-like filesystem
172 * @inode: inode to check access rights for
173 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
174 * @check_acl: optional callback to check for Posix ACLs
176 * Used to check for read/write/execute permissions on a file.
177 * We use "fsuid" for this, letting us set arbitrary permissions
178 * for filesystem access without changing the "normal" uids which
179 * are used for other things..
181 int generic_permission(struct inode
*inode
, int mask
,
182 int (*check_acl
)(struct inode
*inode
, int mask
))
184 umode_t mode
= inode
->i_mode
;
186 if (current
->fsuid
== inode
->i_uid
)
189 if (IS_POSIXACL(inode
) && (mode
& S_IRWXG
) && check_acl
) {
190 int error
= check_acl(inode
, mask
);
191 if (error
== -EACCES
)
192 goto check_capabilities
;
193 else if (error
!= -EAGAIN
)
197 if (in_group_p(inode
->i_gid
))
202 * If the DACs are ok we don't need any capability check.
204 if (((mode
& mask
& (MAY_READ
|MAY_WRITE
|MAY_EXEC
)) == mask
))
209 * Read/write DACs are always overridable.
210 * Executable DACs are overridable if at least one exec bit is set.
212 if (!(mask
& MAY_EXEC
) ||
213 (inode
->i_mode
& S_IXUGO
) || S_ISDIR(inode
->i_mode
))
214 if (capable(CAP_DAC_OVERRIDE
))
218 * Searching includes executable on directories, else just read.
220 if (mask
== MAY_READ
|| (S_ISDIR(inode
->i_mode
) && !(mask
& MAY_WRITE
)))
221 if (capable(CAP_DAC_READ_SEARCH
))
227 int permission(struct inode
*inode
, int mask
, struct nameidata
*nd
)
229 umode_t mode
= inode
->i_mode
;
232 if (mask
& MAY_WRITE
) {
235 * Nobody gets write access to a read-only fs.
237 if (IS_RDONLY(inode
) &&
238 (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
)))
242 * Nobody gets write access to an immutable file.
244 if (IS_IMMUTABLE(inode
))
250 * MAY_EXEC on regular files requires special handling: We override
251 * filesystem execute permissions if the mode bits aren't set or
252 * the fs is mounted with the "noexec" flag.
254 if ((mask
& MAY_EXEC
) && S_ISREG(mode
) && (!(mode
& S_IXUGO
) ||
255 (nd
&& nd
->mnt
&& (nd
->mnt
->mnt_flags
& MNT_NOEXEC
))))
258 /* Ordinary permission routines do not understand MAY_APPEND. */
259 submask
= mask
& ~MAY_APPEND
;
260 if (inode
->i_op
&& inode
->i_op
->permission
)
261 retval
= inode
->i_op
->permission(inode
, submask
, nd
);
263 retval
= generic_permission(inode
, submask
, NULL
);
267 return security_inode_permission(inode
, mask
, nd
);
271 * vfs_permission - check for access rights to a given path
272 * @nd: lookup result that describes the path
273 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
275 * Used to check for read/write/execute permissions on a path.
276 * We use "fsuid" for this, letting us set arbitrary permissions
277 * for filesystem access without changing the "normal" uids which
278 * are used for other things.
280 int vfs_permission(struct nameidata
*nd
, int mask
)
282 return permission(nd
->dentry
->d_inode
, mask
, nd
);
286 * file_permission - check for additional access rights to a given file
287 * @file: file to check access rights for
288 * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC)
290 * Used to check for read/write/execute permissions on an already opened
294 * Do not use this function in new code. All access checks should
295 * be done using vfs_permission().
297 int file_permission(struct file
*file
, int mask
)
299 return permission(file
->f_path
.dentry
->d_inode
, mask
, NULL
);
303 * get_write_access() gets write permission for a file.
304 * put_write_access() releases this write permission.
305 * This is used for regular files.
306 * We cannot support write (and maybe mmap read-write shared) accesses and
307 * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode
308 * can have the following values:
309 * 0: no writers, no VM_DENYWRITE mappings
310 * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist
311 * > 0: (i_writecount) users are writing to the file.
313 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
314 * except for the cases where we don't hold i_writecount yet. Then we need to
315 * use {get,deny}_write_access() - these functions check the sign and refuse
316 * to do the change if sign is wrong. Exclusion between them is provided by
317 * the inode->i_lock spinlock.
320 int get_write_access(struct inode
* inode
)
322 spin_lock(&inode
->i_lock
);
323 if (atomic_read(&inode
->i_writecount
) < 0) {
324 spin_unlock(&inode
->i_lock
);
327 atomic_inc(&inode
->i_writecount
);
328 spin_unlock(&inode
->i_lock
);
333 int deny_write_access(struct file
* file
)
335 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
337 spin_lock(&inode
->i_lock
);
338 if (atomic_read(&inode
->i_writecount
) > 0) {
339 spin_unlock(&inode
->i_lock
);
342 atomic_dec(&inode
->i_writecount
);
343 spin_unlock(&inode
->i_lock
);
348 void path_release(struct nameidata
*nd
)
355 * umount() mustn't call path_release()/mntput() as that would clear
358 void path_release_on_umount(struct nameidata
*nd
)
361 mntput_no_expire(nd
->mnt
);
365 * release_open_intent - free up open intent resources
366 * @nd: pointer to nameidata
368 void release_open_intent(struct nameidata
*nd
)
370 if (nd
->intent
.open
.file
->f_path
.dentry
== NULL
)
371 put_filp(nd
->intent
.open
.file
);
373 fput(nd
->intent
.open
.file
);
376 static inline struct dentry
*
377 do_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
379 int status
= dentry
->d_op
->d_revalidate(dentry
, nd
);
380 if (unlikely(status
<= 0)) {
382 * The dentry failed validation.
383 * If d_revalidate returned 0 attempt to invalidate
384 * the dentry otherwise d_revalidate is asking us
385 * to return a fail status.
388 if (!d_invalidate(dentry
)) {
394 dentry
= ERR_PTR(status
);
401 * Internal lookup() using the new generic dcache.
404 static struct dentry
* cached_lookup(struct dentry
* parent
, struct qstr
* name
, struct nameidata
*nd
)
406 struct dentry
* dentry
= __d_lookup(parent
, name
);
408 /* lockess __d_lookup may fail due to concurrent d_move()
409 * in some unrelated directory, so try with d_lookup
412 dentry
= d_lookup(parent
, name
);
414 if (dentry
&& dentry
->d_op
&& dentry
->d_op
->d_revalidate
)
415 dentry
= do_revalidate(dentry
, nd
);
421 * Short-cut version of permission(), for calling by
422 * path_walk(), when dcache lock is held. Combines parts
423 * of permission() and generic_permission(), and tests ONLY for
424 * MAY_EXEC permission.
426 * If appropriate, check DAC only. If not appropriate, or
427 * short-cut DAC fails, then call permission() to do more
428 * complete permission check.
430 static int exec_permission_lite(struct inode
*inode
,
431 struct nameidata
*nd
)
433 umode_t mode
= inode
->i_mode
;
435 if (inode
->i_op
&& inode
->i_op
->permission
)
438 if (current
->fsuid
== inode
->i_uid
)
440 else if (in_group_p(inode
->i_gid
))
446 if ((inode
->i_mode
& S_IXUGO
) && capable(CAP_DAC_OVERRIDE
))
449 if (S_ISDIR(inode
->i_mode
) && capable(CAP_DAC_OVERRIDE
))
452 if (S_ISDIR(inode
->i_mode
) && capable(CAP_DAC_READ_SEARCH
))
457 return security_inode_permission(inode
, MAY_EXEC
, nd
);
461 * This is called when everything else fails, and we actually have
462 * to go to the low-level filesystem to find out what we should do..
464 * We get the directory semaphore, and after getting that we also
465 * make sure that nobody added the entry to the dcache in the meantime..
468 static struct dentry
* real_lookup(struct dentry
* parent
, struct qstr
* name
, struct nameidata
*nd
)
470 struct dentry
* result
;
471 struct inode
*dir
= parent
->d_inode
;
473 mutex_lock(&dir
->i_mutex
);
475 * First re-do the cached lookup just in case it was created
476 * while we waited for the directory semaphore..
478 * FIXME! This could use version numbering or similar to
479 * avoid unnecessary cache lookups.
481 * The "dcache_lock" is purely to protect the RCU list walker
482 * from concurrent renames at this point (we mustn't get false
483 * negatives from the RCU list walk here, unlike the optimistic
486 * so doing d_lookup() (with seqlock), instead of lockfree __d_lookup
488 result
= d_lookup(parent
, name
);
490 struct dentry
* dentry
= d_alloc(parent
, name
);
491 result
= ERR_PTR(-ENOMEM
);
493 result
= dir
->i_op
->lookup(dir
, dentry
, nd
);
499 mutex_unlock(&dir
->i_mutex
);
504 * Uhhuh! Nasty case: the cache was re-populated while
505 * we waited on the semaphore. Need to revalidate.
507 mutex_unlock(&dir
->i_mutex
);
508 if (result
->d_op
&& result
->d_op
->d_revalidate
) {
509 result
= do_revalidate(result
, nd
);
511 result
= ERR_PTR(-ENOENT
);
516 static int __emul_lookup_dentry(const char *, struct nameidata
*);
519 static __always_inline
int
520 walk_init_root(const char *name
, struct nameidata
*nd
)
522 struct fs_struct
*fs
= current
->fs
;
524 read_lock(&fs
->lock
);
525 if (fs
->altroot
&& !(nd
->flags
& LOOKUP_NOALT
)) {
526 nd
->mnt
= mntget(fs
->altrootmnt
);
527 nd
->dentry
= dget(fs
->altroot
);
528 read_unlock(&fs
->lock
);
529 if (__emul_lookup_dentry(name
,nd
))
531 read_lock(&fs
->lock
);
533 nd
->mnt
= mntget(fs
->rootmnt
);
534 nd
->dentry
= dget(fs
->root
);
535 read_unlock(&fs
->lock
);
539 static __always_inline
int __vfs_follow_link(struct nameidata
*nd
, const char *link
)
548 if (!walk_init_root(link
, nd
))
549 /* weird __emul_prefix() stuff did it */
552 res
= link_path_walk(link
, nd
);
554 if (nd
->depth
|| res
|| nd
->last_type
!=LAST_NORM
)
557 * If it is an iterative symlinks resolution in open_namei() we
558 * have to copy the last component. And all that crap because of
559 * bloody create() on broken symlinks. Furrfu...
562 if (unlikely(!name
)) {
566 strcpy(name
, nd
->last
.name
);
567 nd
->last
.name
= name
;
571 return PTR_ERR(link
);
574 static inline void dput_path(struct path
*path
, struct nameidata
*nd
)
577 if (path
->mnt
!= nd
->mnt
)
581 static inline void path_to_nameidata(struct path
*path
, struct nameidata
*nd
)
584 if (nd
->mnt
!= path
->mnt
)
587 nd
->dentry
= path
->dentry
;
590 static __always_inline
int __do_follow_link(struct path
*path
, struct nameidata
*nd
)
594 struct dentry
*dentry
= path
->dentry
;
596 touch_atime(path
->mnt
, dentry
);
597 nd_set_link(nd
, NULL
);
599 if (path
->mnt
!= nd
->mnt
) {
600 path_to_nameidata(path
, nd
);
604 cookie
= dentry
->d_inode
->i_op
->follow_link(dentry
, nd
);
605 error
= PTR_ERR(cookie
);
606 if (!IS_ERR(cookie
)) {
607 char *s
= nd_get_link(nd
);
610 error
= __vfs_follow_link(nd
, s
);
611 if (dentry
->d_inode
->i_op
->put_link
)
612 dentry
->d_inode
->i_op
->put_link(dentry
, nd
, cookie
);
621 * This limits recursive symlink follows to 8, while
622 * limiting consecutive symlinks to 40.
624 * Without that kind of total limit, nasty chains of consecutive
625 * symlinks can cause almost arbitrarily long lookups.
627 static inline int do_follow_link(struct path
*path
, struct nameidata
*nd
)
630 if (current
->link_count
>= MAX_NESTED_LINKS
)
632 if (current
->total_link_count
>= 40)
634 BUG_ON(nd
->depth
>= MAX_NESTED_LINKS
);
636 err
= security_inode_follow_link(path
->dentry
, nd
);
639 current
->link_count
++;
640 current
->total_link_count
++;
642 err
= __do_follow_link(path
, nd
);
643 current
->link_count
--;
652 int follow_up(struct vfsmount
**mnt
, struct dentry
**dentry
)
654 struct vfsmount
*parent
;
655 struct dentry
*mountpoint
;
656 spin_lock(&vfsmount_lock
);
657 parent
=(*mnt
)->mnt_parent
;
658 if (parent
== *mnt
) {
659 spin_unlock(&vfsmount_lock
);
663 mountpoint
=dget((*mnt
)->mnt_mountpoint
);
664 spin_unlock(&vfsmount_lock
);
666 *dentry
= mountpoint
;
672 /* no need for dcache_lock, as serialization is taken care in
675 static int __follow_mount(struct path
*path
)
678 while (d_mountpoint(path
->dentry
)) {
679 struct vfsmount
*mounted
= lookup_mnt(path
->mnt
, path
->dentry
);
686 path
->dentry
= dget(mounted
->mnt_root
);
692 static void follow_mount(struct vfsmount
**mnt
, struct dentry
**dentry
)
694 while (d_mountpoint(*dentry
)) {
695 struct vfsmount
*mounted
= lookup_mnt(*mnt
, *dentry
);
701 *dentry
= dget(mounted
->mnt_root
);
705 /* no need for dcache_lock, as serialization is taken care in
708 int follow_down(struct vfsmount
**mnt
, struct dentry
**dentry
)
710 struct vfsmount
*mounted
;
712 mounted
= lookup_mnt(*mnt
, *dentry
);
717 *dentry
= dget(mounted
->mnt_root
);
723 static __always_inline
void follow_dotdot(struct nameidata
*nd
)
725 struct fs_struct
*fs
= current
->fs
;
728 struct vfsmount
*parent
;
729 struct dentry
*old
= nd
->dentry
;
731 read_lock(&fs
->lock
);
732 if (nd
->dentry
== fs
->root
&&
733 nd
->mnt
== fs
->rootmnt
) {
734 read_unlock(&fs
->lock
);
737 read_unlock(&fs
->lock
);
738 spin_lock(&dcache_lock
);
739 if (nd
->dentry
!= nd
->mnt
->mnt_root
) {
740 nd
->dentry
= dget(nd
->dentry
->d_parent
);
741 spin_unlock(&dcache_lock
);
745 spin_unlock(&dcache_lock
);
746 spin_lock(&vfsmount_lock
);
747 parent
= nd
->mnt
->mnt_parent
;
748 if (parent
== nd
->mnt
) {
749 spin_unlock(&vfsmount_lock
);
753 nd
->dentry
= dget(nd
->mnt
->mnt_mountpoint
);
754 spin_unlock(&vfsmount_lock
);
759 follow_mount(&nd
->mnt
, &nd
->dentry
);
763 * It's more convoluted than I'd like it to be, but... it's still fairly
764 * small and for now I'd prefer to have fast path as straight as possible.
765 * It _is_ time-critical.
767 static int do_lookup(struct nameidata
*nd
, struct qstr
*name
,
770 struct vfsmount
*mnt
= nd
->mnt
;
771 struct dentry
*dentry
= __d_lookup(nd
->dentry
, name
);
775 if (dentry
->d_op
&& dentry
->d_op
->d_revalidate
)
776 goto need_revalidate
;
779 path
->dentry
= dentry
;
780 __follow_mount(path
);
784 dentry
= real_lookup(nd
->dentry
, name
, nd
);
790 dentry
= do_revalidate(dentry
, nd
);
798 return PTR_ERR(dentry
);
803 * This is the basic name resolution function, turning a pathname into
804 * the final dentry. We expect 'base' to be positive and a directory.
806 * Returns 0 and nd will have valid dentry and mnt on success.
807 * Returns error and drops reference to input namei data on failure.
809 static fastcall
int __link_path_walk(const char * name
, struct nameidata
*nd
)
814 unsigned int lookup_flags
= nd
->flags
;
821 inode
= nd
->dentry
->d_inode
;
823 lookup_flags
= LOOKUP_FOLLOW
| (nd
->flags
& LOOKUP_CONTINUE
);
825 /* At this point we know we have a real path component. */
831 nd
->flags
|= LOOKUP_CONTINUE
;
832 err
= exec_permission_lite(inode
, nd
);
834 err
= vfs_permission(nd
, MAY_EXEC
);
839 c
= *(const unsigned char *)name
;
841 hash
= init_name_hash();
844 hash
= partial_name_hash(c
, hash
);
845 c
= *(const unsigned char *)name
;
846 } while (c
&& (c
!= '/'));
847 this.len
= name
- (const char *) this.name
;
848 this.hash
= end_name_hash(hash
);
850 /* remove trailing slashes? */
853 while (*++name
== '/');
855 goto last_with_slashes
;
858 * "." and ".." are special - ".." especially so because it has
859 * to be able to know about the current root directory and
860 * parent relationships.
862 if (this.name
[0] == '.') switch (this.len
) {
866 if (this.name
[1] != '.')
869 inode
= nd
->dentry
->d_inode
;
875 * See if the low-level filesystem might want
876 * to use its own hash..
878 if (nd
->dentry
->d_op
&& nd
->dentry
->d_op
->d_hash
) {
879 err
= nd
->dentry
->d_op
->d_hash(nd
->dentry
, &this);
883 /* This does the actual lookups.. */
884 err
= do_lookup(nd
, &this, &next
);
889 inode
= next
.dentry
->d_inode
;
896 if (inode
->i_op
->follow_link
) {
897 err
= do_follow_link(&next
, nd
);
901 inode
= nd
->dentry
->d_inode
;
908 path_to_nameidata(&next
, nd
);
910 if (!inode
->i_op
->lookup
)
913 /* here ends the main loop */
916 lookup_flags
|= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
918 /* Clear LOOKUP_CONTINUE iff it was previously unset */
919 nd
->flags
&= lookup_flags
| ~LOOKUP_CONTINUE
;
920 if (lookup_flags
& LOOKUP_PARENT
)
922 if (this.name
[0] == '.') switch (this.len
) {
926 if (this.name
[1] != '.')
929 inode
= nd
->dentry
->d_inode
;
934 if (nd
->dentry
->d_op
&& nd
->dentry
->d_op
->d_hash
) {
935 err
= nd
->dentry
->d_op
->d_hash(nd
->dentry
, &this);
939 err
= do_lookup(nd
, &this, &next
);
942 inode
= next
.dentry
->d_inode
;
943 if ((lookup_flags
& LOOKUP_FOLLOW
)
944 && inode
&& inode
->i_op
&& inode
->i_op
->follow_link
) {
945 err
= do_follow_link(&next
, nd
);
948 inode
= nd
->dentry
->d_inode
;
950 path_to_nameidata(&next
, nd
);
954 if (lookup_flags
& LOOKUP_DIRECTORY
) {
956 if (!inode
->i_op
|| !inode
->i_op
->lookup
)
962 nd
->last_type
= LAST_NORM
;
963 if (this.name
[0] != '.')
966 nd
->last_type
= LAST_DOT
;
967 else if (this.len
== 2 && this.name
[1] == '.')
968 nd
->last_type
= LAST_DOTDOT
;
973 * We bypassed the ordinary revalidation routines.
974 * We may need to check the cached dentry for staleness.
976 if (nd
->dentry
&& nd
->dentry
->d_sb
&&
977 (nd
->dentry
->d_sb
->s_type
->fs_flags
& FS_REVAL_DOT
)) {
979 /* Note: we do not d_invalidate() */
980 if (!nd
->dentry
->d_op
->d_revalidate(nd
->dentry
, nd
))
986 dput_path(&next
, nd
);
995 * Wrapper to retry pathname resolution whenever the underlying
996 * file system returns an ESTALE.
998 * Retry the whole path once, forcing real lookup requests
999 * instead of relying on the dcache.
1001 int fastcall
link_path_walk(const char *name
, struct nameidata
*nd
)
1003 struct nameidata save
= *nd
;
1006 /* make sure the stuff we saved doesn't go away */
1010 result
= __link_path_walk(name
, nd
);
1011 if (result
== -ESTALE
) {
1015 nd
->flags
|= LOOKUP_REVAL
;
1016 result
= __link_path_walk(name
, nd
);
1025 int fastcall
path_walk(const char * name
, struct nameidata
*nd
)
1027 current
->total_link_count
= 0;
1028 return link_path_walk(name
, nd
);
1032 * SMP-safe: Returns 1 and nd will have valid dentry and mnt, if
1033 * everything is done. Returns 0 and drops input nd, if lookup failed;
1035 static int __emul_lookup_dentry(const char *name
, struct nameidata
*nd
)
1037 if (path_walk(name
, nd
))
1038 return 0; /* something went wrong... */
1040 if (!nd
->dentry
->d_inode
|| S_ISDIR(nd
->dentry
->d_inode
->i_mode
)) {
1041 struct dentry
*old_dentry
= nd
->dentry
;
1042 struct vfsmount
*old_mnt
= nd
->mnt
;
1043 struct qstr last
= nd
->last
;
1044 int last_type
= nd
->last_type
;
1045 struct fs_struct
*fs
= current
->fs
;
1048 * NAME was not found in alternate root or it's a directory.
1049 * Try to find it in the normal root:
1051 nd
->last_type
= LAST_ROOT
;
1052 read_lock(&fs
->lock
);
1053 nd
->mnt
= mntget(fs
->rootmnt
);
1054 nd
->dentry
= dget(fs
->root
);
1055 read_unlock(&fs
->lock
);
1056 if (path_walk(name
, nd
) == 0) {
1057 if (nd
->dentry
->d_inode
) {
1064 nd
->dentry
= old_dentry
;
1067 nd
->last_type
= last_type
;
1072 void set_fs_altroot(void)
1074 char *emul
= __emul_prefix();
1075 struct nameidata nd
;
1076 struct vfsmount
*mnt
= NULL
, *oldmnt
;
1077 struct dentry
*dentry
= NULL
, *olddentry
;
1079 struct fs_struct
*fs
= current
->fs
;
1083 err
= path_lookup(emul
, LOOKUP_FOLLOW
|LOOKUP_DIRECTORY
|LOOKUP_NOALT
, &nd
);
1089 write_lock(&fs
->lock
);
1090 oldmnt
= fs
->altrootmnt
;
1091 olddentry
= fs
->altroot
;
1092 fs
->altrootmnt
= mnt
;
1093 fs
->altroot
= dentry
;
1094 write_unlock(&fs
->lock
);
1101 /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */
1102 static int fastcall
do_path_lookup(int dfd
, const char *name
,
1103 unsigned int flags
, struct nameidata
*nd
)
1108 struct fs_struct
*fs
= current
->fs
;
1110 nd
->last_type
= LAST_ROOT
; /* if there are only slashes... */
1115 read_lock(&fs
->lock
);
1116 if (fs
->altroot
&& !(nd
->flags
& LOOKUP_NOALT
)) {
1117 nd
->mnt
= mntget(fs
->altrootmnt
);
1118 nd
->dentry
= dget(fs
->altroot
);
1119 read_unlock(&fs
->lock
);
1120 if (__emul_lookup_dentry(name
,nd
))
1121 goto out
; /* found in altroot */
1122 read_lock(&fs
->lock
);
1124 nd
->mnt
= mntget(fs
->rootmnt
);
1125 nd
->dentry
= dget(fs
->root
);
1126 read_unlock(&fs
->lock
);
1127 } else if (dfd
== AT_FDCWD
) {
1128 read_lock(&fs
->lock
);
1129 nd
->mnt
= mntget(fs
->pwdmnt
);
1130 nd
->dentry
= dget(fs
->pwd
);
1131 read_unlock(&fs
->lock
);
1133 struct dentry
*dentry
;
1135 file
= fget_light(dfd
, &fput_needed
);
1140 dentry
= file
->f_path
.dentry
;
1143 if (!S_ISDIR(dentry
->d_inode
->i_mode
))
1146 retval
= file_permission(file
, MAY_EXEC
);
1150 nd
->mnt
= mntget(file
->f_path
.mnt
);
1151 nd
->dentry
= dget(dentry
);
1153 fput_light(file
, fput_needed
);
1156 retval
= path_walk(name
, nd
);
1158 if (unlikely(!retval
&& !audit_dummy_context() && nd
->dentry
&&
1159 nd
->dentry
->d_inode
))
1160 audit_inode(name
, nd
->dentry
->d_inode
);
1165 fput_light(file
, fput_needed
);
1169 int fastcall
path_lookup(const char *name
, unsigned int flags
,
1170 struct nameidata
*nd
)
1172 return do_path_lookup(AT_FDCWD
, name
, flags
, nd
);
1176 * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair
1177 * @dentry: pointer to dentry of the base directory
1178 * @mnt: pointer to vfs mount of the base directory
1179 * @name: pointer to file name
1180 * @flags: lookup flags
1181 * @nd: pointer to nameidata
1183 int vfs_path_lookup(struct dentry
*dentry
, struct vfsmount
*mnt
,
1184 const char *name
, unsigned int flags
,
1185 struct nameidata
*nd
)
1189 /* same as do_path_lookup */
1190 nd
->last_type
= LAST_ROOT
;
1194 nd
->mnt
= mntget(mnt
);
1195 nd
->dentry
= dget(dentry
);
1197 retval
= path_walk(name
, nd
);
1198 if (unlikely(!retval
&& !audit_dummy_context() && nd
->dentry
&&
1199 nd
->dentry
->d_inode
))
1200 audit_inode(name
, nd
->dentry
->d_inode
);
1206 static int __path_lookup_intent_open(int dfd
, const char *name
,
1207 unsigned int lookup_flags
, struct nameidata
*nd
,
1208 int open_flags
, int create_mode
)
1210 struct file
*filp
= get_empty_filp();
1215 nd
->intent
.open
.file
= filp
;
1216 nd
->intent
.open
.flags
= open_flags
;
1217 nd
->intent
.open
.create_mode
= create_mode
;
1218 err
= do_path_lookup(dfd
, name
, lookup_flags
|LOOKUP_OPEN
, nd
);
1219 if (IS_ERR(nd
->intent
.open
.file
)) {
1221 err
= PTR_ERR(nd
->intent
.open
.file
);
1224 } else if (err
!= 0)
1225 release_open_intent(nd
);
1230 * path_lookup_open - lookup a file path with open intent
1231 * @dfd: the directory to use as base, or AT_FDCWD
1232 * @name: pointer to file name
1233 * @lookup_flags: lookup intent flags
1234 * @nd: pointer to nameidata
1235 * @open_flags: open intent flags
1237 int path_lookup_open(int dfd
, const char *name
, unsigned int lookup_flags
,
1238 struct nameidata
*nd
, int open_flags
)
1240 return __path_lookup_intent_open(dfd
, name
, lookup_flags
, nd
,
1245 * path_lookup_create - lookup a file path with open + create intent
1246 * @dfd: the directory to use as base, or AT_FDCWD
1247 * @name: pointer to file name
1248 * @lookup_flags: lookup intent flags
1249 * @nd: pointer to nameidata
1250 * @open_flags: open intent flags
1251 * @create_mode: create intent flags
1253 static int path_lookup_create(int dfd
, const char *name
,
1254 unsigned int lookup_flags
, struct nameidata
*nd
,
1255 int open_flags
, int create_mode
)
1257 return __path_lookup_intent_open(dfd
, name
, lookup_flags
|LOOKUP_CREATE
,
1258 nd
, open_flags
, create_mode
);
1261 int __user_path_lookup_open(const char __user
*name
, unsigned int lookup_flags
,
1262 struct nameidata
*nd
, int open_flags
)
1264 char *tmp
= getname(name
);
1265 int err
= PTR_ERR(tmp
);
1268 err
= __path_lookup_intent_open(AT_FDCWD
, tmp
, lookup_flags
, nd
, open_flags
, 0);
1274 static inline struct dentry
*__lookup_hash_kern(struct qstr
*name
, struct dentry
*base
, struct nameidata
*nd
)
1276 struct dentry
*dentry
;
1277 struct inode
*inode
;
1280 inode
= base
->d_inode
;
1283 * See if the low-level filesystem might want
1284 * to use its own hash..
1286 if (base
->d_op
&& base
->d_op
->d_hash
) {
1287 err
= base
->d_op
->d_hash(base
, name
);
1288 dentry
= ERR_PTR(err
);
1293 dentry
= cached_lookup(base
, name
, nd
);
1295 struct dentry
*new = d_alloc(base
, name
);
1296 dentry
= ERR_PTR(-ENOMEM
);
1299 dentry
= inode
->i_op
->lookup(inode
, new, nd
);
1310 * Restricted form of lookup. Doesn't follow links, single-component only,
1311 * needs parent already locked. Doesn't follow mounts.
1314 static inline struct dentry
* __lookup_hash(struct qstr
*name
, struct dentry
*base
, struct nameidata
*nd
)
1316 struct dentry
*dentry
;
1317 struct inode
*inode
;
1320 inode
= base
->d_inode
;
1322 err
= permission(inode
, MAY_EXEC
, nd
);
1323 dentry
= ERR_PTR(err
);
1327 dentry
= __lookup_hash_kern(name
, base
, nd
);
1332 static struct dentry
*lookup_hash(struct nameidata
*nd
)
1334 return __lookup_hash(&nd
->last
, nd
->dentry
, nd
);
1338 static inline int __lookup_one_len(const char *name
, struct qstr
*this, struct dentry
*base
, int len
)
1348 hash
= init_name_hash();
1350 c
= *(const unsigned char *)name
++;
1351 if (c
== '/' || c
== '\0')
1353 hash
= partial_name_hash(c
, hash
);
1355 this->hash
= end_name_hash(hash
);
1359 struct dentry
*lookup_one_len(const char *name
, struct dentry
*base
, int len
)
1364 err
= __lookup_one_len(name
, &this, base
, len
);
1366 return ERR_PTR(err
);
1367 return __lookup_hash(&this, base
, NULL
);
1370 struct dentry
*lookup_one_len_kern(const char *name
, struct dentry
*base
, int len
)
1375 err
= __lookup_one_len(name
, &this, base
, len
);
1377 return ERR_PTR(err
);
1378 return __lookup_hash_kern(&this, base
, NULL
);
1381 int fastcall
__user_walk_fd(int dfd
, const char __user
*name
, unsigned flags
,
1382 struct nameidata
*nd
)
1384 char *tmp
= getname(name
);
1385 int err
= PTR_ERR(tmp
);
1388 err
= do_path_lookup(dfd
, tmp
, flags
, nd
);
1394 int fastcall
__user_walk(const char __user
*name
, unsigned flags
, struct nameidata
*nd
)
1396 return __user_walk_fd(AT_FDCWD
, name
, flags
, nd
);
1400 * It's inline, so penalty for filesystems that don't use sticky bit is
1403 static inline int check_sticky(struct inode
*dir
, struct inode
*inode
)
1405 if (!(dir
->i_mode
& S_ISVTX
))
1407 if (inode
->i_uid
== current
->fsuid
)
1409 if (dir
->i_uid
== current
->fsuid
)
1411 return !capable(CAP_FOWNER
);
1415 * Check whether we can remove a link victim from directory dir, check
1416 * whether the type of victim is right.
1417 * 1. We can't do it if dir is read-only (done in permission())
1418 * 2. We should have write and exec permissions on dir
1419 * 3. We can't remove anything from append-only dir
1420 * 4. We can't do anything with immutable dir (done in permission())
1421 * 5. If the sticky bit on dir is set we should either
1422 * a. be owner of dir, or
1423 * b. be owner of victim, or
1424 * c. have CAP_FOWNER capability
1425 * 6. If the victim is append-only or immutable we can't do antyhing with
1426 * links pointing to it.
1427 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
1428 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
1429 * 9. We can't remove a root or mountpoint.
1430 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
1431 * nfs_async_unlink().
1433 static int may_delete(struct inode
*dir
,struct dentry
*victim
,int isdir
)
1437 if (!victim
->d_inode
)
1440 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
1441 audit_inode_child(victim
->d_name
.name
, victim
->d_inode
, dir
);
1443 error
= permission(dir
,MAY_WRITE
| MAY_EXEC
, NULL
);
1448 if (check_sticky(dir
, victim
->d_inode
)||IS_APPEND(victim
->d_inode
)||
1449 IS_IMMUTABLE(victim
->d_inode
))
1452 if (!S_ISDIR(victim
->d_inode
->i_mode
))
1454 if (IS_ROOT(victim
))
1456 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
1458 if (IS_DEADDIR(dir
))
1460 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
1465 /* Check whether we can create an object with dentry child in directory
1467 * 1. We can't do it if child already exists (open has special treatment for
1468 * this case, but since we are inlined it's OK)
1469 * 2. We can't do it if dir is read-only (done in permission())
1470 * 3. We should have write and exec permissions on dir
1471 * 4. We can't do it if dir is immutable (done in permission())
1473 static inline int may_create(struct inode
*dir
, struct dentry
*child
,
1474 struct nameidata
*nd
)
1478 if (IS_DEADDIR(dir
))
1480 return permission(dir
,MAY_WRITE
| MAY_EXEC
, nd
);
1484 * O_DIRECTORY translates into forcing a directory lookup.
1486 static inline int lookup_flags(unsigned int f
)
1488 unsigned long retval
= LOOKUP_FOLLOW
;
1491 retval
&= ~LOOKUP_FOLLOW
;
1493 if (f
& O_DIRECTORY
)
1494 retval
|= LOOKUP_DIRECTORY
;
1500 * p1 and p2 should be directories on the same fs.
1502 struct dentry
*lock_rename(struct dentry
*p1
, struct dentry
*p2
)
1507 mutex_lock_nested(&p1
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1511 mutex_lock(&p1
->d_inode
->i_sb
->s_vfs_rename_mutex
);
1513 for (p
= p1
; p
->d_parent
!= p
; p
= p
->d_parent
) {
1514 if (p
->d_parent
== p2
) {
1515 mutex_lock_nested(&p2
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1516 mutex_lock_nested(&p1
->d_inode
->i_mutex
, I_MUTEX_CHILD
);
1521 for (p
= p2
; p
->d_parent
!= p
; p
= p
->d_parent
) {
1522 if (p
->d_parent
== p1
) {
1523 mutex_lock_nested(&p1
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1524 mutex_lock_nested(&p2
->d_inode
->i_mutex
, I_MUTEX_CHILD
);
1529 mutex_lock_nested(&p1
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1530 mutex_lock_nested(&p2
->d_inode
->i_mutex
, I_MUTEX_CHILD
);
1534 void unlock_rename(struct dentry
*p1
, struct dentry
*p2
)
1536 mutex_unlock(&p1
->d_inode
->i_mutex
);
1538 mutex_unlock(&p2
->d_inode
->i_mutex
);
1539 mutex_unlock(&p1
->d_inode
->i_sb
->s_vfs_rename_mutex
);
1543 int vfs_create(struct inode
*dir
, struct dentry
*dentry
, int mode
,
1544 struct nameidata
*nd
)
1546 int error
= may_create(dir
, dentry
, nd
);
1551 if (!dir
->i_op
|| !dir
->i_op
->create
)
1552 return -EACCES
; /* shouldn't it be ENOSYS? */
1555 error
= security_inode_create(dir
, dentry
, mode
);
1559 error
= dir
->i_op
->create(dir
, dentry
, mode
, nd
);
1561 fsnotify_create(dir
, dentry
);
1565 int may_open(struct nameidata
*nd
, int acc_mode
, int flag
)
1567 struct dentry
*dentry
= nd
->dentry
;
1568 struct inode
*inode
= dentry
->d_inode
;
1574 if (S_ISLNK(inode
->i_mode
))
1577 if (S_ISDIR(inode
->i_mode
) && (flag
& FMODE_WRITE
))
1580 error
= vfs_permission(nd
, acc_mode
);
1585 * FIFO's, sockets and device files are special: they don't
1586 * actually live on the filesystem itself, and as such you
1587 * can write to them even if the filesystem is read-only.
1589 if (S_ISFIFO(inode
->i_mode
) || S_ISSOCK(inode
->i_mode
)) {
1591 } else if (S_ISBLK(inode
->i_mode
) || S_ISCHR(inode
->i_mode
)) {
1592 if (nd
->mnt
->mnt_flags
& MNT_NODEV
)
1596 } else if (IS_RDONLY(inode
) && (flag
& FMODE_WRITE
))
1599 * An append-only file must be opened in append mode for writing.
1601 if (IS_APPEND(inode
)) {
1602 if ((flag
& FMODE_WRITE
) && !(flag
& O_APPEND
))
1608 /* O_NOATIME can only be set by the owner or superuser */
1609 if (flag
& O_NOATIME
)
1610 if (!is_owner_or_cap(inode
))
1614 * Ensure there are no outstanding leases on the file.
1616 error
= break_lease(inode
, flag
);
1620 if (flag
& O_TRUNC
) {
1621 error
= get_write_access(inode
);
1626 * Refuse to truncate files with mandatory locks held on them.
1628 error
= locks_verify_locked(inode
);
1632 error
= do_truncate(dentry
, 0, ATTR_MTIME
|ATTR_CTIME
, NULL
);
1634 put_write_access(inode
);
1638 if (flag
& FMODE_WRITE
)
1644 static int open_namei_create(struct nameidata
*nd
, struct path
*path
,
1648 struct dentry
*dir
= nd
->dentry
;
1650 if (!IS_POSIXACL(dir
->d_inode
))
1651 mode
&= ~current
->fs
->umask
;
1652 error
= vfs_create(dir
->d_inode
, path
->dentry
, mode
, nd
);
1653 mutex_unlock(&dir
->d_inode
->i_mutex
);
1655 nd
->dentry
= path
->dentry
;
1658 /* Don't check for write permission, don't truncate */
1659 return may_open(nd
, 0, flag
& ~O_TRUNC
);
1665 * namei for open - this is in fact almost the whole open-routine.
1667 * Note that the low bits of "flag" aren't the same as in the open
1668 * system call - they are 00 - no permissions needed
1669 * 01 - read permission needed
1670 * 10 - write permission needed
1671 * 11 - read/write permissions needed
1672 * which is a lot more logical, and also allows the "no perm" needed
1673 * for symlinks (where the permissions are checked later).
1676 int open_namei(int dfd
, const char *pathname
, int flag
,
1677 int mode
, struct nameidata
*nd
)
1679 int acc_mode
, error
;
1684 acc_mode
= ACC_MODE(flag
);
1686 /* O_TRUNC implies we need access checks for write permissions */
1688 acc_mode
|= MAY_WRITE
;
1690 /* Allow the LSM permission hook to distinguish append
1691 access from general write access. */
1692 if (flag
& O_APPEND
)
1693 acc_mode
|= MAY_APPEND
;
1696 * The simplest case - just a plain lookup.
1698 if (!(flag
& O_CREAT
)) {
1699 error
= path_lookup_open(dfd
, pathname
, lookup_flags(flag
),
1707 * Create - we need to know the parent.
1709 error
= path_lookup_create(dfd
,pathname
,LOOKUP_PARENT
,nd
,flag
,mode
);
1714 * We have the parent and last component. First of all, check
1715 * that we are not asked to creat(2) an obvious directory - that
1719 if (nd
->last_type
!= LAST_NORM
|| nd
->last
.name
[nd
->last
.len
])
1723 nd
->flags
&= ~LOOKUP_PARENT
;
1724 mutex_lock(&dir
->d_inode
->i_mutex
);
1725 path
.dentry
= lookup_hash(nd
);
1729 error
= PTR_ERR(path
.dentry
);
1730 if (IS_ERR(path
.dentry
)) {
1731 mutex_unlock(&dir
->d_inode
->i_mutex
);
1735 if (IS_ERR(nd
->intent
.open
.file
)) {
1736 mutex_unlock(&dir
->d_inode
->i_mutex
);
1737 error
= PTR_ERR(nd
->intent
.open
.file
);
1741 /* Negative dentry, just create the file */
1742 if (!path
.dentry
->d_inode
) {
1743 error
= open_namei_create(nd
, &path
, flag
, mode
);
1750 * It already exists.
1752 mutex_unlock(&dir
->d_inode
->i_mutex
);
1753 audit_inode(pathname
, path
.dentry
->d_inode
);
1759 if (__follow_mount(&path
)) {
1761 if (flag
& O_NOFOLLOW
)
1766 if (!path
.dentry
->d_inode
)
1768 if (path
.dentry
->d_inode
->i_op
&& path
.dentry
->d_inode
->i_op
->follow_link
)
1771 path_to_nameidata(&path
, nd
);
1773 if (path
.dentry
->d_inode
&& S_ISDIR(path
.dentry
->d_inode
->i_mode
))
1776 error
= may_open(nd
, acc_mode
, flag
);
1782 dput_path(&path
, nd
);
1784 if (!IS_ERR(nd
->intent
.open
.file
))
1785 release_open_intent(nd
);
1791 if (flag
& O_NOFOLLOW
)
1794 * This is subtle. Instead of calling do_follow_link() we do the
1795 * thing by hands. The reason is that this way we have zero link_count
1796 * and path_walk() (called from ->follow_link) honoring LOOKUP_PARENT.
1797 * After that we have the parent and last component, i.e.
1798 * we are in the same situation as after the first path_walk().
1799 * Well, almost - if the last component is normal we get its copy
1800 * stored in nd->last.name and we will have to putname() it when we
1801 * are done. Procfs-like symlinks just set LAST_BIND.
1803 nd
->flags
|= LOOKUP_PARENT
;
1804 error
= security_inode_follow_link(path
.dentry
, nd
);
1807 error
= __do_follow_link(&path
, nd
);
1809 /* Does someone understand code flow here? Or it is only
1810 * me so stupid? Anathema to whoever designed this non-sense
1811 * with "intent.open".
1813 release_open_intent(nd
);
1816 nd
->flags
&= ~LOOKUP_PARENT
;
1817 if (nd
->last_type
== LAST_BIND
)
1820 if (nd
->last_type
!= LAST_NORM
)
1822 if (nd
->last
.name
[nd
->last
.len
]) {
1823 __putname(nd
->last
.name
);
1828 __putname(nd
->last
.name
);
1832 mutex_lock(&dir
->d_inode
->i_mutex
);
1833 path
.dentry
= lookup_hash(nd
);
1835 __putname(nd
->last
.name
);
1840 * lookup_create - lookup a dentry, creating it if it doesn't exist
1841 * @nd: nameidata info
1842 * @is_dir: directory flag
1844 * Simple function to lookup and return a dentry and create it
1845 * if it doesn't exist. Is SMP-safe.
1847 * Returns with nd->dentry->d_inode->i_mutex locked.
1849 struct dentry
*lookup_create(struct nameidata
*nd
, int is_dir
)
1851 struct dentry
*dentry
= ERR_PTR(-EEXIST
);
1853 mutex_lock_nested(&nd
->dentry
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
1855 * Yucky last component or no last component at all?
1856 * (foo/., foo/.., /////)
1858 if (nd
->last_type
!= LAST_NORM
)
1860 nd
->flags
&= ~LOOKUP_PARENT
;
1861 nd
->flags
|= LOOKUP_CREATE
;
1862 nd
->intent
.open
.flags
= O_EXCL
;
1865 * Do the final lookup.
1867 dentry
= lookup_hash(nd
);
1872 * Special case - lookup gave negative, but... we had foo/bar/
1873 * From the vfs_mknod() POV we just have a negative dentry -
1874 * all is fine. Let's be bastards - you had / on the end, you've
1875 * been asking for (non-existent) directory. -ENOENT for you.
1877 if (!is_dir
&& nd
->last
.name
[nd
->last
.len
] && !dentry
->d_inode
)
1882 dentry
= ERR_PTR(-ENOENT
);
1886 EXPORT_SYMBOL_GPL(lookup_create
);
1888 int vfs_mknod(struct inode
*dir
, struct dentry
*dentry
, int mode
, dev_t dev
)
1890 int error
= may_create(dir
, dentry
, NULL
);
1895 if ((S_ISCHR(mode
) || S_ISBLK(mode
)) && !capable(CAP_MKNOD
))
1898 if (!dir
->i_op
|| !dir
->i_op
->mknod
)
1901 error
= security_inode_mknod(dir
, dentry
, mode
, dev
);
1906 error
= dir
->i_op
->mknod(dir
, dentry
, mode
, dev
);
1908 fsnotify_create(dir
, dentry
);
1912 asmlinkage
long sys_mknodat(int dfd
, const char __user
*filename
, int mode
,
1917 struct dentry
* dentry
;
1918 struct nameidata nd
;
1922 tmp
= getname(filename
);
1924 return PTR_ERR(tmp
);
1926 error
= do_path_lookup(dfd
, tmp
, LOOKUP_PARENT
, &nd
);
1929 dentry
= lookup_create(&nd
, 0);
1930 error
= PTR_ERR(dentry
);
1932 if (!IS_POSIXACL(nd
.dentry
->d_inode
))
1933 mode
&= ~current
->fs
->umask
;
1934 if (!IS_ERR(dentry
)) {
1935 switch (mode
& S_IFMT
) {
1936 case 0: case S_IFREG
:
1937 error
= vfs_create(nd
.dentry
->d_inode
,dentry
,mode
,&nd
);
1939 case S_IFCHR
: case S_IFBLK
:
1940 error
= vfs_mknod(nd
.dentry
->d_inode
,dentry
,mode
,
1941 new_decode_dev(dev
));
1943 case S_IFIFO
: case S_IFSOCK
:
1944 error
= vfs_mknod(nd
.dentry
->d_inode
,dentry
,mode
,0);
1954 mutex_unlock(&nd
.dentry
->d_inode
->i_mutex
);
1962 asmlinkage
long sys_mknod(const char __user
*filename
, int mode
, unsigned dev
)
1964 return sys_mknodat(AT_FDCWD
, filename
, mode
, dev
);
1967 int vfs_mkdir(struct inode
*dir
, struct dentry
*dentry
, int mode
)
1969 int error
= may_create(dir
, dentry
, NULL
);
1974 if (!dir
->i_op
|| !dir
->i_op
->mkdir
)
1977 mode
&= (S_IRWXUGO
|S_ISVTX
);
1978 error
= security_inode_mkdir(dir
, dentry
, mode
);
1983 error
= dir
->i_op
->mkdir(dir
, dentry
, mode
);
1985 fsnotify_mkdir(dir
, dentry
);
1989 asmlinkage
long sys_mkdirat(int dfd
, const char __user
*pathname
, int mode
)
1993 struct dentry
*dentry
;
1994 struct nameidata nd
;
1996 tmp
= getname(pathname
);
1997 error
= PTR_ERR(tmp
);
2001 error
= do_path_lookup(dfd
, tmp
, LOOKUP_PARENT
, &nd
);
2004 dentry
= lookup_create(&nd
, 1);
2005 error
= PTR_ERR(dentry
);
2009 if (!IS_POSIXACL(nd
.dentry
->d_inode
))
2010 mode
&= ~current
->fs
->umask
;
2011 error
= vfs_mkdir(nd
.dentry
->d_inode
, dentry
, mode
);
2014 mutex_unlock(&nd
.dentry
->d_inode
->i_mutex
);
2022 asmlinkage
long sys_mkdir(const char __user
*pathname
, int mode
)
2024 return sys_mkdirat(AT_FDCWD
, pathname
, mode
);
2028 * We try to drop the dentry early: we should have
2029 * a usage count of 2 if we're the only user of this
2030 * dentry, and if that is true (possibly after pruning
2031 * the dcache), then we drop the dentry now.
2033 * A low-level filesystem can, if it choses, legally
2036 * if (!d_unhashed(dentry))
2039 * if it cannot handle the case of removing a directory
2040 * that is still in use by something else..
2042 void dentry_unhash(struct dentry
*dentry
)
2045 shrink_dcache_parent(dentry
);
2046 spin_lock(&dcache_lock
);
2047 spin_lock(&dentry
->d_lock
);
2048 if (atomic_read(&dentry
->d_count
) == 2)
2050 spin_unlock(&dentry
->d_lock
);
2051 spin_unlock(&dcache_lock
);
2054 int vfs_rmdir(struct inode
*dir
, struct dentry
*dentry
)
2056 int error
= may_delete(dir
, dentry
, 1);
2061 if (!dir
->i_op
|| !dir
->i_op
->rmdir
)
2066 mutex_lock(&dentry
->d_inode
->i_mutex
);
2067 dentry_unhash(dentry
);
2068 if (d_mountpoint(dentry
))
2071 error
= security_inode_rmdir(dir
, dentry
);
2073 error
= dir
->i_op
->rmdir(dir
, dentry
);
2075 dentry
->d_inode
->i_flags
|= S_DEAD
;
2078 mutex_unlock(&dentry
->d_inode
->i_mutex
);
2087 static long do_rmdir(int dfd
, const char __user
*pathname
)
2091 struct dentry
*dentry
;
2092 struct nameidata nd
;
2094 name
= getname(pathname
);
2096 return PTR_ERR(name
);
2098 error
= do_path_lookup(dfd
, name
, LOOKUP_PARENT
, &nd
);
2102 switch(nd
.last_type
) {
2113 mutex_lock_nested(&nd
.dentry
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
2114 dentry
= lookup_hash(&nd
);
2115 error
= PTR_ERR(dentry
);
2118 error
= vfs_rmdir(nd
.dentry
->d_inode
, dentry
);
2121 mutex_unlock(&nd
.dentry
->d_inode
->i_mutex
);
2129 asmlinkage
long sys_rmdir(const char __user
*pathname
)
2131 return do_rmdir(AT_FDCWD
, pathname
);
2134 int vfs_unlink(struct inode
*dir
, struct dentry
*dentry
)
2136 int error
= may_delete(dir
, dentry
, 0);
2141 if (!dir
->i_op
|| !dir
->i_op
->unlink
)
2146 mutex_lock(&dentry
->d_inode
->i_mutex
);
2147 if (d_mountpoint(dentry
))
2150 error
= security_inode_unlink(dir
, dentry
);
2152 error
= dir
->i_op
->unlink(dir
, dentry
);
2154 mutex_unlock(&dentry
->d_inode
->i_mutex
);
2156 /* We don't d_delete() NFS sillyrenamed files--they still exist. */
2157 if (!error
&& !(dentry
->d_flags
& DCACHE_NFSFS_RENAMED
)) {
2165 * Make sure that the actual truncation of the file will occur outside its
2166 * directory's i_mutex. Truncate can take a long time if there is a lot of
2167 * writeout happening, and we don't want to prevent access to the directory
2168 * while waiting on the I/O.
2170 static long do_unlinkat(int dfd
, const char __user
*pathname
)
2174 struct dentry
*dentry
;
2175 struct nameidata nd
;
2176 struct inode
*inode
= NULL
;
2178 name
= getname(pathname
);
2180 return PTR_ERR(name
);
2182 error
= do_path_lookup(dfd
, name
, LOOKUP_PARENT
, &nd
);
2186 if (nd
.last_type
!= LAST_NORM
)
2188 mutex_lock_nested(&nd
.dentry
->d_inode
->i_mutex
, I_MUTEX_PARENT
);
2189 dentry
= lookup_hash(&nd
);
2190 error
= PTR_ERR(dentry
);
2191 if (!IS_ERR(dentry
)) {
2192 /* Why not before? Because we want correct error value */
2193 if (nd
.last
.name
[nd
.last
.len
])
2195 inode
= dentry
->d_inode
;
2197 atomic_inc(&inode
->i_count
);
2198 error
= vfs_unlink(nd
.dentry
->d_inode
, dentry
);
2202 mutex_unlock(&nd
.dentry
->d_inode
->i_mutex
);
2204 iput(inode
); /* truncate the inode here */
2212 error
= !dentry
->d_inode
? -ENOENT
:
2213 S_ISDIR(dentry
->d_inode
->i_mode
) ? -EISDIR
: -ENOTDIR
;
2217 asmlinkage
long sys_unlinkat(int dfd
, const char __user
*pathname
, int flag
)
2219 if ((flag
& ~AT_REMOVEDIR
) != 0)
2222 if (flag
& AT_REMOVEDIR
)
2223 return do_rmdir(dfd
, pathname
);
2225 return do_unlinkat(dfd
, pathname
);
2228 asmlinkage
long sys_unlink(const char __user
*pathname
)
2230 return do_unlinkat(AT_FDCWD
, pathname
);
2233 int vfs_symlink(struct inode
*dir
, struct dentry
*dentry
, const char *oldname
, int mode
)
2235 int error
= may_create(dir
, dentry
, NULL
);
2240 if (!dir
->i_op
|| !dir
->i_op
->symlink
)
2243 error
= security_inode_symlink(dir
, dentry
, oldname
);
2248 error
= dir
->i_op
->symlink(dir
, dentry
, oldname
);
2250 fsnotify_create(dir
, dentry
);
2254 asmlinkage
long sys_symlinkat(const char __user
*oldname
,
2255 int newdfd
, const char __user
*newname
)
2260 struct dentry
*dentry
;
2261 struct nameidata nd
;
2263 from
= getname(oldname
);
2265 return PTR_ERR(from
);
2266 to
= getname(newname
);
2267 error
= PTR_ERR(to
);
2271 error
= do_path_lookup(newdfd
, to
, LOOKUP_PARENT
, &nd
);
2274 dentry
= lookup_create(&nd
, 0);
2275 error
= PTR_ERR(dentry
);
2279 error
= vfs_symlink(nd
.dentry
->d_inode
, dentry
, from
, S_IALLUGO
);
2282 mutex_unlock(&nd
.dentry
->d_inode
->i_mutex
);
2291 asmlinkage
long sys_symlink(const char __user
*oldname
, const char __user
*newname
)
2293 return sys_symlinkat(oldname
, AT_FDCWD
, newname
);
2296 int vfs_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*new_dentry
)
2298 struct inode
*inode
= old_dentry
->d_inode
;
2304 error
= may_create(dir
, new_dentry
, NULL
);
2308 if (dir
->i_sb
!= inode
->i_sb
)
2312 * A link to an append-only or immutable file cannot be created.
2314 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
2316 if (!dir
->i_op
|| !dir
->i_op
->link
)
2318 if (S_ISDIR(old_dentry
->d_inode
->i_mode
))
2321 error
= security_inode_link(old_dentry
, dir
, new_dentry
);
2325 mutex_lock(&old_dentry
->d_inode
->i_mutex
);
2327 error
= dir
->i_op
->link(old_dentry
, dir
, new_dentry
);
2328 mutex_unlock(&old_dentry
->d_inode
->i_mutex
);
2330 fsnotify_create(dir
, new_dentry
);
2335 * Hardlinks are often used in delicate situations. We avoid
2336 * security-related surprises by not following symlinks on the
2339 * We don't follow them on the oldname either to be compatible
2340 * with linux 2.0, and to avoid hard-linking to directories
2341 * and other special files. --ADM
2343 asmlinkage
long sys_linkat(int olddfd
, const char __user
*oldname
,
2344 int newdfd
, const char __user
*newname
,
2347 struct dentry
*new_dentry
;
2348 struct nameidata nd
, old_nd
;
2352 if ((flags
& ~AT_SYMLINK_FOLLOW
) != 0)
2355 to
= getname(newname
);
2359 error
= __user_walk_fd(olddfd
, oldname
,
2360 flags
& AT_SYMLINK_FOLLOW
? LOOKUP_FOLLOW
: 0,
2364 error
= do_path_lookup(newdfd
, to
, LOOKUP_PARENT
, &nd
);
2368 if (old_nd
.mnt
!= nd
.mnt
)
2370 new_dentry
= lookup_create(&nd
, 0);
2371 error
= PTR_ERR(new_dentry
);
2372 if (IS_ERR(new_dentry
))
2374 error
= vfs_link(old_nd
.dentry
, nd
.dentry
->d_inode
, new_dentry
);
2377 mutex_unlock(&nd
.dentry
->d_inode
->i_mutex
);
2381 path_release(&old_nd
);
2388 asmlinkage
long sys_link(const char __user
*oldname
, const char __user
*newname
)
2390 return sys_linkat(AT_FDCWD
, oldname
, AT_FDCWD
, newname
, 0);
2394 * The worst of all namespace operations - renaming directory. "Perverted"
2395 * doesn't even start to describe it. Somebody in UCB had a heck of a trip...
2397 * a) we can get into loop creation. Check is done in is_subdir().
2398 * b) race potential - two innocent renames can create a loop together.
2399 * That's where 4.4 screws up. Current fix: serialization on
2400 * sb->s_vfs_rename_mutex. We might be more accurate, but that's another
2402 * c) we have to lock _three_ objects - parents and victim (if it exists).
2403 * And that - after we got ->i_mutex on parents (until then we don't know
2404 * whether the target exists). Solution: try to be smart with locking
2405 * order for inodes. We rely on the fact that tree topology may change
2406 * only under ->s_vfs_rename_mutex _and_ that parent of the object we
2407 * move will be locked. Thus we can rank directories by the tree
2408 * (ancestors first) and rank all non-directories after them.
2409 * That works since everybody except rename does "lock parent, lookup,
2410 * lock child" and rename is under ->s_vfs_rename_mutex.
2411 * HOWEVER, it relies on the assumption that any object with ->lookup()
2412 * has no more than 1 dentry. If "hybrid" objects will ever appear,
2413 * we'd better make sure that there's no link(2) for them.
2414 * d) some filesystems don't support opened-but-unlinked directories,
2415 * either because of layout or because they are not ready to deal with
2416 * all cases correctly. The latter will be fixed (taking this sort of
2417 * stuff into VFS), but the former is not going away. Solution: the same
2418 * trick as in rmdir().
2419 * e) conversion from fhandle to dentry may come in the wrong moment - when
2420 * we are removing the target. Solution: we will have to grab ->i_mutex
2421 * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on
2422 * ->i_mutex on parents, which works but leads to some truely excessive
2425 static int vfs_rename_dir(struct inode
*old_dir
, struct dentry
*old_dentry
,
2426 struct inode
*new_dir
, struct dentry
*new_dentry
)
2429 struct inode
*target
;
2432 * If we are going to change the parent - check write permissions,
2433 * we'll need to flip '..'.
2435 if (new_dir
!= old_dir
) {
2436 error
= permission(old_dentry
->d_inode
, MAY_WRITE
, NULL
);
2441 error
= security_inode_rename(old_dir
, old_dentry
, new_dir
, new_dentry
);
2445 target
= new_dentry
->d_inode
;
2447 mutex_lock(&target
->i_mutex
);
2448 dentry_unhash(new_dentry
);
2450 if (d_mountpoint(old_dentry
)||d_mountpoint(new_dentry
))
2453 error
= old_dir
->i_op
->rename(old_dir
, old_dentry
, new_dir
, new_dentry
);
2456 target
->i_flags
|= S_DEAD
;
2457 mutex_unlock(&target
->i_mutex
);
2458 if (d_unhashed(new_dentry
))
2459 d_rehash(new_dentry
);
2463 if (!(old_dir
->i_sb
->s_type
->fs_flags
& FS_RENAME_DOES_D_MOVE
))
2464 d_move(old_dentry
,new_dentry
);
2468 static int vfs_rename_other(struct inode
*old_dir
, struct dentry
*old_dentry
,
2469 struct inode
*new_dir
, struct dentry
*new_dentry
)
2471 struct inode
*target
;
2474 error
= security_inode_rename(old_dir
, old_dentry
, new_dir
, new_dentry
);
2479 target
= new_dentry
->d_inode
;
2481 mutex_lock(&target
->i_mutex
);
2482 if (d_mountpoint(old_dentry
)||d_mountpoint(new_dentry
))
2485 error
= old_dir
->i_op
->rename(old_dir
, old_dentry
, new_dir
, new_dentry
);
2487 if (!(old_dir
->i_sb
->s_type
->fs_flags
& FS_RENAME_DOES_D_MOVE
))
2488 d_move(old_dentry
, new_dentry
);
2491 mutex_unlock(&target
->i_mutex
);
2496 int vfs_rename(struct inode
*old_dir
, struct dentry
*old_dentry
,
2497 struct inode
*new_dir
, struct dentry
*new_dentry
)
2500 int is_dir
= S_ISDIR(old_dentry
->d_inode
->i_mode
);
2501 const char *old_name
;
2503 if (old_dentry
->d_inode
== new_dentry
->d_inode
)
2506 error
= may_delete(old_dir
, old_dentry
, is_dir
);
2510 if (!new_dentry
->d_inode
)
2511 error
= may_create(new_dir
, new_dentry
, NULL
);
2513 error
= may_delete(new_dir
, new_dentry
, is_dir
);
2517 if (!old_dir
->i_op
|| !old_dir
->i_op
->rename
)
2520 DQUOT_INIT(old_dir
);
2521 DQUOT_INIT(new_dir
);
2523 old_name
= fsnotify_oldname_init(old_dentry
->d_name
.name
);
2526 error
= vfs_rename_dir(old_dir
,old_dentry
,new_dir
,new_dentry
);
2528 error
= vfs_rename_other(old_dir
,old_dentry
,new_dir
,new_dentry
);
2530 const char *new_name
= old_dentry
->d_name
.name
;
2531 fsnotify_move(old_dir
, new_dir
, old_name
, new_name
, is_dir
,
2532 new_dentry
->d_inode
, old_dentry
->d_inode
);
2534 fsnotify_oldname_free(old_name
);
2539 static int do_rename(int olddfd
, const char *oldname
,
2540 int newdfd
, const char *newname
)
2543 struct dentry
* old_dir
, * new_dir
;
2544 struct dentry
* old_dentry
, *new_dentry
;
2545 struct dentry
* trap
;
2546 struct nameidata oldnd
, newnd
;
2548 error
= do_path_lookup(olddfd
, oldname
, LOOKUP_PARENT
, &oldnd
);
2552 error
= do_path_lookup(newdfd
, newname
, LOOKUP_PARENT
, &newnd
);
2557 if (oldnd
.mnt
!= newnd
.mnt
)
2560 old_dir
= oldnd
.dentry
;
2562 if (oldnd
.last_type
!= LAST_NORM
)
2565 new_dir
= newnd
.dentry
;
2566 if (newnd
.last_type
!= LAST_NORM
)
2569 trap
= lock_rename(new_dir
, old_dir
);
2571 old_dentry
= lookup_hash(&oldnd
);
2572 error
= PTR_ERR(old_dentry
);
2573 if (IS_ERR(old_dentry
))
2575 /* source must exist */
2577 if (!old_dentry
->d_inode
)
2579 /* unless the source is a directory trailing slashes give -ENOTDIR */
2580 if (!S_ISDIR(old_dentry
->d_inode
->i_mode
)) {
2582 if (oldnd
.last
.name
[oldnd
.last
.len
])
2584 if (newnd
.last
.name
[newnd
.last
.len
])
2587 /* source should not be ancestor of target */
2589 if (old_dentry
== trap
)
2591 new_dentry
= lookup_hash(&newnd
);
2592 error
= PTR_ERR(new_dentry
);
2593 if (IS_ERR(new_dentry
))
2595 /* target should not be an ancestor of source */
2597 if (new_dentry
== trap
)
2600 error
= vfs_rename(old_dir
->d_inode
, old_dentry
,
2601 new_dir
->d_inode
, new_dentry
);
2607 unlock_rename(new_dir
, old_dir
);
2609 path_release(&newnd
);
2611 path_release(&oldnd
);
2616 asmlinkage
long sys_renameat(int olddfd
, const char __user
*oldname
,
2617 int newdfd
, const char __user
*newname
)
2623 from
= getname(oldname
);
2625 return PTR_ERR(from
);
2626 to
= getname(newname
);
2627 error
= PTR_ERR(to
);
2629 error
= do_rename(olddfd
, from
, newdfd
, to
);
2636 asmlinkage
long sys_rename(const char __user
*oldname
, const char __user
*newname
)
2638 return sys_renameat(AT_FDCWD
, oldname
, AT_FDCWD
, newname
);
2641 int vfs_readlink(struct dentry
*dentry
, char __user
*buffer
, int buflen
, const char *link
)
2645 len
= PTR_ERR(link
);
2650 if (len
> (unsigned) buflen
)
2652 if (copy_to_user(buffer
, link
, len
))
2659 * A helper for ->readlink(). This should be used *ONLY* for symlinks that
2660 * have ->follow_link() touching nd only in nd_set_link(). Using (or not
2661 * using) it for any given inode is up to filesystem.
2663 int generic_readlink(struct dentry
*dentry
, char __user
*buffer
, int buflen
)
2665 struct nameidata nd
;
2669 cookie
= dentry
->d_inode
->i_op
->follow_link(dentry
, &nd
);
2670 if (!IS_ERR(cookie
)) {
2671 int res
= vfs_readlink(dentry
, buffer
, buflen
, nd_get_link(&nd
));
2672 if (dentry
->d_inode
->i_op
->put_link
)
2673 dentry
->d_inode
->i_op
->put_link(dentry
, &nd
, cookie
);
2674 cookie
= ERR_PTR(res
);
2676 return PTR_ERR(cookie
);
2679 int vfs_follow_link(struct nameidata
*nd
, const char *link
)
2681 return __vfs_follow_link(nd
, link
);
2684 /* get the link contents into pagecache */
2685 static char *page_getlink(struct dentry
* dentry
, struct page
**ppage
)
2688 struct address_space
*mapping
= dentry
->d_inode
->i_mapping
;
2689 page
= read_mapping_page(mapping
, 0, NULL
);
2696 int page_readlink(struct dentry
*dentry
, char __user
*buffer
, int buflen
)
2698 struct page
*page
= NULL
;
2699 char *s
= page_getlink(dentry
, &page
);
2700 int res
= vfs_readlink(dentry
,buffer
,buflen
,s
);
2703 page_cache_release(page
);
2708 void *page_follow_link_light(struct dentry
*dentry
, struct nameidata
*nd
)
2710 struct page
*page
= NULL
;
2711 nd_set_link(nd
, page_getlink(dentry
, &page
));
2715 void page_put_link(struct dentry
*dentry
, struct nameidata
*nd
, void *cookie
)
2717 struct page
*page
= cookie
;
2721 page_cache_release(page
);
2725 int __page_symlink(struct inode
*inode
, const char *symname
, int len
,
2728 struct address_space
*mapping
= inode
->i_mapping
;
2735 page
= find_or_create_page(mapping
, 0, gfp_mask
);
2738 err
= mapping
->a_ops
->prepare_write(NULL
, page
, 0, len
-1);
2739 if (err
== AOP_TRUNCATED_PAGE
) {
2740 page_cache_release(page
);
2745 kaddr
= kmap_atomic(page
, KM_USER0
);
2746 memcpy(kaddr
, symname
, len
-1);
2747 kunmap_atomic(kaddr
, KM_USER0
);
2748 err
= mapping
->a_ops
->commit_write(NULL
, page
, 0, len
-1);
2749 if (err
== AOP_TRUNCATED_PAGE
) {
2750 page_cache_release(page
);
2756 * Notice that we are _not_ going to block here - end of page is
2757 * unmapped, so this will only try to map the rest of page, see
2758 * that it is unmapped (typically even will not look into inode -
2759 * ->i_size will be enough for everything) and zero it out.
2760 * OTOH it's obviously correct and should make the page up-to-date.
2762 if (!PageUptodate(page
)) {
2763 err
= mapping
->a_ops
->readpage(NULL
, page
);
2764 if (err
!= AOP_TRUNCATED_PAGE
)
2765 wait_on_page_locked(page
);
2769 page_cache_release(page
);
2772 mark_inode_dirty(inode
);
2776 page_cache_release(page
);
2781 int page_symlink(struct inode
*inode
, const char *symname
, int len
)
2783 return __page_symlink(inode
, symname
, len
,
2784 mapping_gfp_mask(inode
->i_mapping
));
2787 const struct inode_operations page_symlink_inode_operations
= {
2788 .readlink
= generic_readlink
,
2789 .follow_link
= page_follow_link_light
,
2790 .put_link
= page_put_link
,
2793 EXPORT_SYMBOL(__user_walk
);
2794 EXPORT_SYMBOL(__user_walk_fd
);
2795 EXPORT_SYMBOL(follow_down
);
2796 EXPORT_SYMBOL(follow_up
);
2797 EXPORT_SYMBOL(get_write_access
); /* binfmt_aout */
2798 EXPORT_SYMBOL(getname
);
2799 EXPORT_SYMBOL(lock_rename
);
2800 EXPORT_SYMBOL(lookup_one_len
);
2801 EXPORT_SYMBOL(page_follow_link_light
);
2802 EXPORT_SYMBOL(page_put_link
);
2803 EXPORT_SYMBOL(page_readlink
);
2804 EXPORT_SYMBOL(__page_symlink
);
2805 EXPORT_SYMBOL(page_symlink
);
2806 EXPORT_SYMBOL(page_symlink_inode_operations
);
2807 EXPORT_SYMBOL(path_lookup
);
2808 EXPORT_SYMBOL(vfs_path_lookup
);
2809 EXPORT_SYMBOL(path_release
);
2810 EXPORT_SYMBOL(path_walk
);
2811 EXPORT_SYMBOL(permission
);
2812 EXPORT_SYMBOL(vfs_permission
);
2813 EXPORT_SYMBOL(file_permission
);
2814 EXPORT_SYMBOL(unlock_rename
);
2815 EXPORT_SYMBOL(vfs_create
);
2816 EXPORT_SYMBOL(vfs_follow_link
);
2817 EXPORT_SYMBOL(vfs_link
);
2818 EXPORT_SYMBOL(vfs_mkdir
);
2819 EXPORT_SYMBOL(vfs_mknod
);
2820 EXPORT_SYMBOL(generic_permission
);
2821 EXPORT_SYMBOL(vfs_readlink
);
2822 EXPORT_SYMBOL(vfs_rename
);
2823 EXPORT_SYMBOL(vfs_rmdir
);
2824 EXPORT_SYMBOL(vfs_symlink
);
2825 EXPORT_SYMBOL(vfs_unlink
);
2826 EXPORT_SYMBOL(dentry_unhash
);
2827 EXPORT_SYMBOL(generic_readlink
);