3 * Library for filesystems writers.
6 #include <linux/module.h>
7 #include <linux/pagemap.h>
8 #include <linux/mount.h>
10 #include <linux/mutex.h>
11 #include <linux/exportfs.h>
13 #include <asm/uaccess.h>
15 int simple_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
,
18 struct inode
*inode
= dentry
->d_inode
;
19 generic_fillattr(inode
, stat
);
20 stat
->blocks
= inode
->i_mapping
->nrpages
<< (PAGE_CACHE_SHIFT
- 9);
24 int simple_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
26 buf
->f_type
= dentry
->d_sb
->s_magic
;
27 buf
->f_bsize
= PAGE_CACHE_SIZE
;
28 buf
->f_namelen
= NAME_MAX
;
33 * Retaining negative dentries for an in-memory filesystem just wastes
34 * memory and lookup time: arrange for them to be deleted immediately.
36 static int simple_delete_dentry(struct dentry
*dentry
)
42 * Lookup the data. This is trivial - if the dentry didn't already
43 * exist, we know it is negative. Set d_op to delete negative dentries.
45 struct dentry
*simple_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
)
47 static struct dentry_operations simple_dentry_operations
= {
48 .d_delete
= simple_delete_dentry
,
51 if (dentry
->d_name
.len
> NAME_MAX
)
52 return ERR_PTR(-ENAMETOOLONG
);
53 dentry
->d_op
= &simple_dentry_operations
;
58 int simple_sync_file(struct file
* file
, struct dentry
*dentry
, int datasync
)
63 int dcache_dir_open(struct inode
*inode
, struct file
*file
)
65 static struct qstr cursor_name
= {.len
= 1, .name
= "."};
67 file
->private_data
= d_alloc(file
->f_path
.dentry
, &cursor_name
);
69 return file
->private_data
? 0 : -ENOMEM
;
72 int dcache_dir_close(struct inode
*inode
, struct file
*file
)
74 dput(file
->private_data
);
78 loff_t
dcache_dir_lseek(struct file
*file
, loff_t offset
, int origin
)
80 mutex_lock(&file
->f_path
.dentry
->d_inode
->i_mutex
);
83 offset
+= file
->f_pos
;
88 mutex_unlock(&file
->f_path
.dentry
->d_inode
->i_mutex
);
91 if (offset
!= file
->f_pos
) {
93 if (file
->f_pos
>= 2) {
95 struct dentry
*cursor
= file
->private_data
;
96 loff_t n
= file
->f_pos
- 2;
98 spin_lock(&dcache_lock
);
99 list_del(&cursor
->d_u
.d_child
);
100 p
= file
->f_path
.dentry
->d_subdirs
.next
;
101 while (n
&& p
!= &file
->f_path
.dentry
->d_subdirs
) {
103 next
= list_entry(p
, struct dentry
, d_u
.d_child
);
104 if (!d_unhashed(next
) && next
->d_inode
)
108 list_add_tail(&cursor
->d_u
.d_child
, p
);
109 spin_unlock(&dcache_lock
);
112 mutex_unlock(&file
->f_path
.dentry
->d_inode
->i_mutex
);
116 /* Relationship between i_mode and the DT_xxx types */
117 static inline unsigned char dt_type(struct inode
*inode
)
119 return (inode
->i_mode
>> 12) & 15;
123 * Directory is locked and all positive dentries in it are safe, since
124 * for ramfs-type trees they can't go away without unlink() or rmdir(),
125 * both impossible due to the lock on directory.
128 int dcache_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
130 struct dentry
*dentry
= filp
->f_path
.dentry
;
131 struct dentry
*cursor
= filp
->private_data
;
132 struct list_head
*p
, *q
= &cursor
->d_u
.d_child
;
138 ino
= dentry
->d_inode
->i_ino
;
139 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
145 ino
= parent_ino(dentry
);
146 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
152 spin_lock(&dcache_lock
);
153 if (filp
->f_pos
== 2)
154 list_move(q
, &dentry
->d_subdirs
);
156 for (p
=q
->next
; p
!= &dentry
->d_subdirs
; p
=p
->next
) {
158 next
= list_entry(p
, struct dentry
, d_u
.d_child
);
159 if (d_unhashed(next
) || !next
->d_inode
)
162 spin_unlock(&dcache_lock
);
163 if (filldir(dirent
, next
->d_name
.name
,
164 next
->d_name
.len
, filp
->f_pos
,
165 next
->d_inode
->i_ino
,
166 dt_type(next
->d_inode
)) < 0)
168 spin_lock(&dcache_lock
);
169 /* next is still alive */
174 spin_unlock(&dcache_lock
);
179 ssize_t
generic_read_dir(struct file
*filp
, char __user
*buf
, size_t siz
, loff_t
*ppos
)
184 const struct file_operations simple_dir_operations
= {
185 .open
= dcache_dir_open
,
186 .release
= dcache_dir_close
,
187 .llseek
= dcache_dir_lseek
,
188 .read
= generic_read_dir
,
189 .readdir
= dcache_readdir
,
190 .fsync
= simple_sync_file
,
193 const struct inode_operations simple_dir_inode_operations
= {
194 .lookup
= simple_lookup
,
197 static const struct super_operations simple_super_operations
= {
198 .statfs
= simple_statfs
,
202 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
203 * will never be mountable)
205 int get_sb_pseudo(struct file_system_type
*fs_type
, char *name
,
206 const struct super_operations
*ops
, unsigned long magic
,
207 struct vfsmount
*mnt
)
209 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
210 struct dentry
*dentry
;
212 struct qstr d_name
= {.name
= name
, .len
= strlen(name
)};
217 s
->s_flags
= MS_NOUSER
;
218 s
->s_maxbytes
= ~0ULL;
219 s
->s_blocksize
= PAGE_SIZE
;
220 s
->s_blocksize_bits
= PAGE_SHIFT
;
222 s
->s_op
= ops
? ops
: &simple_super_operations
;
228 * since this is the first inode, make it number 1. New inodes created
229 * after this must take care not to collide with it (by passing
230 * max_reserved of 1 to iunique).
233 root
->i_mode
= S_IFDIR
| S_IRUSR
| S_IWUSR
;
234 root
->i_uid
= root
->i_gid
= 0;
235 root
->i_atime
= root
->i_mtime
= root
->i_ctime
= CURRENT_TIME
;
236 dentry
= d_alloc(NULL
, &d_name
);
242 dentry
->d_parent
= dentry
;
243 d_instantiate(dentry
, root
);
245 s
->s_flags
|= MS_ACTIVE
;
246 return simple_set_mnt(mnt
, s
);
249 up_write(&s
->s_umount
);
254 int simple_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*dentry
)
256 struct inode
*inode
= old_dentry
->d_inode
;
258 inode
->i_ctime
= dir
->i_ctime
= dir
->i_mtime
= CURRENT_TIME
;
260 atomic_inc(&inode
->i_count
);
262 d_instantiate(dentry
, inode
);
266 static inline int simple_positive(struct dentry
*dentry
)
268 return dentry
->d_inode
&& !d_unhashed(dentry
);
271 int simple_empty(struct dentry
*dentry
)
273 struct dentry
*child
;
276 spin_lock(&dcache_lock
);
277 list_for_each_entry(child
, &dentry
->d_subdirs
, d_u
.d_child
)
278 if (simple_positive(child
))
282 spin_unlock(&dcache_lock
);
286 int simple_unlink(struct inode
*dir
, struct dentry
*dentry
)
288 struct inode
*inode
= dentry
->d_inode
;
290 inode
->i_ctime
= dir
->i_ctime
= dir
->i_mtime
= CURRENT_TIME
;
296 int simple_rmdir(struct inode
*dir
, struct dentry
*dentry
)
298 if (!simple_empty(dentry
))
301 drop_nlink(dentry
->d_inode
);
302 simple_unlink(dir
, dentry
);
307 int simple_rename(struct inode
*old_dir
, struct dentry
*old_dentry
,
308 struct inode
*new_dir
, struct dentry
*new_dentry
)
310 struct inode
*inode
= old_dentry
->d_inode
;
311 int they_are_dirs
= S_ISDIR(old_dentry
->d_inode
->i_mode
);
313 if (!simple_empty(new_dentry
))
316 if (new_dentry
->d_inode
) {
317 simple_unlink(new_dir
, new_dentry
);
320 } else if (they_are_dirs
) {
325 old_dir
->i_ctime
= old_dir
->i_mtime
= new_dir
->i_ctime
=
326 new_dir
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
331 int simple_readpage(struct file
*file
, struct page
*page
)
333 clear_highpage(page
);
334 flush_dcache_page(page
);
335 SetPageUptodate(page
);
340 int simple_prepare_write(struct file
*file
, struct page
*page
,
341 unsigned from
, unsigned to
)
343 if (!PageUptodate(page
)) {
344 if (to
- from
!= PAGE_CACHE_SIZE
)
345 zero_user_segments(page
,
347 to
, PAGE_CACHE_SIZE
);
352 int simple_write_begin(struct file
*file
, struct address_space
*mapping
,
353 loff_t pos
, unsigned len
, unsigned flags
,
354 struct page
**pagep
, void **fsdata
)
360 index
= pos
>> PAGE_CACHE_SHIFT
;
361 from
= pos
& (PAGE_CACHE_SIZE
- 1);
363 page
= __grab_cache_page(mapping
, index
);
369 return simple_prepare_write(file
, page
, from
, from
+len
);
372 static int simple_commit_write(struct file
*file
, struct page
*page
,
373 unsigned from
, unsigned to
)
375 struct inode
*inode
= page
->mapping
->host
;
376 loff_t pos
= ((loff_t
)page
->index
<< PAGE_CACHE_SHIFT
) + to
;
378 if (!PageUptodate(page
))
379 SetPageUptodate(page
);
381 * No need to use i_size_read() here, the i_size
382 * cannot change under us because we hold the i_mutex.
384 if (pos
> inode
->i_size
)
385 i_size_write(inode
, pos
);
386 set_page_dirty(page
);
390 int simple_write_end(struct file
*file
, struct address_space
*mapping
,
391 loff_t pos
, unsigned len
, unsigned copied
,
392 struct page
*page
, void *fsdata
)
394 unsigned from
= pos
& (PAGE_CACHE_SIZE
- 1);
396 /* zero the stale part of the page if we did a short copy */
398 void *kaddr
= kmap_atomic(page
, KM_USER0
);
399 memset(kaddr
+ from
+ copied
, 0, len
- copied
);
400 flush_dcache_page(page
);
401 kunmap_atomic(kaddr
, KM_USER0
);
404 simple_commit_write(file
, page
, from
, from
+copied
);
407 page_cache_release(page
);
413 * the inodes created here are not hashed. If you use iunique to generate
414 * unique inode values later for this filesystem, then you must take care
415 * to pass it an appropriate max_reserved value to avoid collisions.
417 int simple_fill_super(struct super_block
*s
, int magic
, struct tree_descr
*files
)
421 struct dentry
*dentry
;
424 s
->s_blocksize
= PAGE_CACHE_SIZE
;
425 s
->s_blocksize_bits
= PAGE_CACHE_SHIFT
;
427 s
->s_op
= &simple_super_operations
;
430 inode
= new_inode(s
);
434 * because the root inode is 1, the files array must not contain an
438 inode
->i_mode
= S_IFDIR
| 0755;
439 inode
->i_uid
= inode
->i_gid
= 0;
441 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
442 inode
->i_op
= &simple_dir_inode_operations
;
443 inode
->i_fop
= &simple_dir_operations
;
445 root
= d_alloc_root(inode
);
450 for (i
= 0; !files
->name
|| files
->name
[0]; i
++, files
++) {
454 /* warn if it tries to conflict with the root inode */
455 if (unlikely(i
== 1))
456 printk(KERN_WARNING
"%s: %s passed in a files array"
457 "with an index of 1!\n", __func__
,
460 dentry
= d_alloc_name(root
, files
->name
);
463 inode
= new_inode(s
);
466 inode
->i_mode
= S_IFREG
| files
->mode
;
467 inode
->i_uid
= inode
->i_gid
= 0;
469 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
470 inode
->i_fop
= files
->ops
;
472 d_add(dentry
, inode
);
482 static DEFINE_SPINLOCK(pin_fs_lock
);
484 int simple_pin_fs(struct file_system_type
*type
, struct vfsmount
**mount
, int *count
)
486 struct vfsmount
*mnt
= NULL
;
487 spin_lock(&pin_fs_lock
);
488 if (unlikely(!*mount
)) {
489 spin_unlock(&pin_fs_lock
);
490 mnt
= vfs_kern_mount(type
, 0, type
->name
, NULL
);
493 spin_lock(&pin_fs_lock
);
499 spin_unlock(&pin_fs_lock
);
504 void simple_release_fs(struct vfsmount
**mount
, int *count
)
506 struct vfsmount
*mnt
;
507 spin_lock(&pin_fs_lock
);
511 spin_unlock(&pin_fs_lock
);
516 * simple_read_from_buffer - copy data from the buffer to user space
517 * @to: the user space buffer to read to
518 * @count: the maximum number of bytes to read
519 * @ppos: the current position in the buffer
520 * @from: the buffer to read from
521 * @available: the size of the buffer
523 * The simple_read_from_buffer() function reads up to @count bytes from the
524 * buffer @from at offset @ppos into the user space address starting at @to.
526 * On success, the number of bytes read is returned and the offset @ppos is
527 * advanced by this number, or negative value is returned on error.
529 ssize_t
simple_read_from_buffer(void __user
*to
, size_t count
, loff_t
*ppos
,
530 const void *from
, size_t available
)
535 if (pos
>= available
)
537 if (count
> available
- pos
)
538 count
= available
- pos
;
539 if (copy_to_user(to
, from
+ pos
, count
))
546 * memory_read_from_buffer - copy data from the buffer
547 * @to: the kernel space buffer to read to
548 * @count: the maximum number of bytes to read
549 * @ppos: the current position in the buffer
550 * @from: the buffer to read from
551 * @available: the size of the buffer
553 * The memory_read_from_buffer() function reads up to @count bytes from the
554 * buffer @from at offset @ppos into the kernel space address starting at @to.
556 * On success, the number of bytes read is returned and the offset @ppos is
557 * advanced by this number, or negative value is returned on error.
559 ssize_t
memory_read_from_buffer(void *to
, size_t count
, loff_t
*ppos
,
560 const void *from
, size_t available
)
566 if (pos
>= available
)
568 if (count
> available
- pos
)
569 count
= available
- pos
;
570 memcpy(to
, from
+ pos
, count
);
577 * Transaction based IO.
578 * The file expects a single write which triggers the transaction, and then
579 * possibly a read which collects the result - which is stored in a
582 char *simple_transaction_get(struct file
*file
, const char __user
*buf
, size_t size
)
584 struct simple_transaction_argresp
*ar
;
585 static DEFINE_SPINLOCK(simple_transaction_lock
);
587 if (size
> SIMPLE_TRANSACTION_LIMIT
- 1)
588 return ERR_PTR(-EFBIG
);
590 ar
= (struct simple_transaction_argresp
*)get_zeroed_page(GFP_KERNEL
);
592 return ERR_PTR(-ENOMEM
);
594 spin_lock(&simple_transaction_lock
);
596 /* only one write allowed per open */
597 if (file
->private_data
) {
598 spin_unlock(&simple_transaction_lock
);
599 free_page((unsigned long)ar
);
600 return ERR_PTR(-EBUSY
);
603 file
->private_data
= ar
;
605 spin_unlock(&simple_transaction_lock
);
607 if (copy_from_user(ar
->data
, buf
, size
))
608 return ERR_PTR(-EFAULT
);
613 ssize_t
simple_transaction_read(struct file
*file
, char __user
*buf
, size_t size
, loff_t
*pos
)
615 struct simple_transaction_argresp
*ar
= file
->private_data
;
619 return simple_read_from_buffer(buf
, size
, pos
, ar
->data
, ar
->size
);
622 int simple_transaction_release(struct inode
*inode
, struct file
*file
)
624 free_page((unsigned long)file
->private_data
);
628 /* Simple attribute files */
631 int (*get
)(void *, u64
*);
632 int (*set
)(void *, u64
);
633 char get_buf
[24]; /* enough to store a u64 and "\n\0" */
636 const char *fmt
; /* format for read operation */
637 struct mutex mutex
; /* protects access to these buffers */
640 /* simple_attr_open is called by an actual attribute open file operation
641 * to set the attribute specific access operations. */
642 int simple_attr_open(struct inode
*inode
, struct file
*file
,
643 int (*get
)(void *, u64
*), int (*set
)(void *, u64
),
646 struct simple_attr
*attr
;
648 attr
= kmalloc(sizeof(*attr
), GFP_KERNEL
);
654 attr
->data
= inode
->i_private
;
656 mutex_init(&attr
->mutex
);
658 file
->private_data
= attr
;
660 return nonseekable_open(inode
, file
);
663 int simple_attr_release(struct inode
*inode
, struct file
*file
)
665 kfree(file
->private_data
);
669 /* read from the buffer that is filled with the get function */
670 ssize_t
simple_attr_read(struct file
*file
, char __user
*buf
,
671 size_t len
, loff_t
*ppos
)
673 struct simple_attr
*attr
;
677 attr
= file
->private_data
;
682 ret
= mutex_lock_interruptible(&attr
->mutex
);
686 if (*ppos
) { /* continued read */
687 size
= strlen(attr
->get_buf
);
688 } else { /* first read */
690 ret
= attr
->get(attr
->data
, &val
);
694 size
= scnprintf(attr
->get_buf
, sizeof(attr
->get_buf
),
695 attr
->fmt
, (unsigned long long)val
);
698 ret
= simple_read_from_buffer(buf
, len
, ppos
, attr
->get_buf
, size
);
700 mutex_unlock(&attr
->mutex
);
704 /* interpret the buffer as a number to call the set function with */
705 ssize_t
simple_attr_write(struct file
*file
, const char __user
*buf
,
706 size_t len
, loff_t
*ppos
)
708 struct simple_attr
*attr
;
713 attr
= file
->private_data
;
717 ret
= mutex_lock_interruptible(&attr
->mutex
);
722 size
= min(sizeof(attr
->set_buf
) - 1, len
);
723 if (copy_from_user(attr
->set_buf
, buf
, size
))
726 ret
= len
; /* claim we got the whole input */
727 attr
->set_buf
[size
] = '\0';
728 val
= simple_strtol(attr
->set_buf
, NULL
, 0);
729 attr
->set(attr
->data
, val
);
731 mutex_unlock(&attr
->mutex
);
736 * This is what d_alloc_anon should have been. Once the exportfs
737 * argument transition has been finished I will update d_alloc_anon
738 * to this prototype and this wrapper will go away. --hch
740 static struct dentry
*exportfs_d_alloc(struct inode
*inode
)
742 struct dentry
*dentry
;
747 return ERR_PTR(PTR_ERR(inode
));
749 dentry
= d_alloc_anon(inode
);
752 dentry
= ERR_PTR(-ENOMEM
);
758 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
759 * @sb: filesystem to do the file handle conversion on
760 * @fid: file handle to convert
761 * @fh_len: length of the file handle in bytes
762 * @fh_type: type of file handle
763 * @get_inode: filesystem callback to retrieve inode
765 * This function decodes @fid as long as it has one of the well-known
766 * Linux filehandle types and calls @get_inode on it to retrieve the
767 * inode for the object specified in the file handle.
769 struct dentry
*generic_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
770 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
771 (struct super_block
*sb
, u64 ino
, u32 gen
))
773 struct inode
*inode
= NULL
;
779 case FILEID_INO32_GEN
:
780 case FILEID_INO32_GEN_PARENT
:
781 inode
= get_inode(sb
, fid
->i32
.ino
, fid
->i32
.gen
);
785 return exportfs_d_alloc(inode
);
787 EXPORT_SYMBOL_GPL(generic_fh_to_dentry
);
790 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
791 * @sb: filesystem to do the file handle conversion on
792 * @fid: file handle to convert
793 * @fh_len: length of the file handle in bytes
794 * @fh_type: type of file handle
795 * @get_inode: filesystem callback to retrieve inode
797 * This function decodes @fid as long as it has one of the well-known
798 * Linux filehandle types and calls @get_inode on it to retrieve the
799 * inode for the _parent_ object specified in the file handle if it
800 * is specified in the file handle, or NULL otherwise.
802 struct dentry
*generic_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
803 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
804 (struct super_block
*sb
, u64 ino
, u32 gen
))
806 struct inode
*inode
= NULL
;
812 case FILEID_INO32_GEN_PARENT
:
813 inode
= get_inode(sb
, fid
->i32
.parent_ino
,
814 (fh_len
> 3 ? fid
->i32
.parent_gen
: 0));
818 return exportfs_d_alloc(inode
);
820 EXPORT_SYMBOL_GPL(generic_fh_to_parent
);
822 EXPORT_SYMBOL(dcache_dir_close
);
823 EXPORT_SYMBOL(dcache_dir_lseek
);
824 EXPORT_SYMBOL(dcache_dir_open
);
825 EXPORT_SYMBOL(dcache_readdir
);
826 EXPORT_SYMBOL(generic_read_dir
);
827 EXPORT_SYMBOL(get_sb_pseudo
);
828 EXPORT_SYMBOL(simple_write_begin
);
829 EXPORT_SYMBOL(simple_write_end
);
830 EXPORT_SYMBOL(simple_dir_inode_operations
);
831 EXPORT_SYMBOL(simple_dir_operations
);
832 EXPORT_SYMBOL(simple_empty
);
833 EXPORT_SYMBOL(d_alloc_name
);
834 EXPORT_SYMBOL(simple_fill_super
);
835 EXPORT_SYMBOL(simple_getattr
);
836 EXPORT_SYMBOL(simple_link
);
837 EXPORT_SYMBOL(simple_lookup
);
838 EXPORT_SYMBOL(simple_pin_fs
);
839 EXPORT_SYMBOL(simple_prepare_write
);
840 EXPORT_SYMBOL(simple_readpage
);
841 EXPORT_SYMBOL(simple_release_fs
);
842 EXPORT_SYMBOL(simple_rename
);
843 EXPORT_SYMBOL(simple_rmdir
);
844 EXPORT_SYMBOL(simple_statfs
);
845 EXPORT_SYMBOL(simple_sync_file
);
846 EXPORT_SYMBOL(simple_unlink
);
847 EXPORT_SYMBOL(simple_read_from_buffer
);
848 EXPORT_SYMBOL(memory_read_from_buffer
);
849 EXPORT_SYMBOL(simple_transaction_get
);
850 EXPORT_SYMBOL(simple_transaction_read
);
851 EXPORT_SYMBOL(simple_transaction_release
);
852 EXPORT_SYMBOL_GPL(simple_attr_open
);
853 EXPORT_SYMBOL_GPL(simple_attr_release
);
854 EXPORT_SYMBOL_GPL(simple_attr_read
);
855 EXPORT_SYMBOL_GPL(simple_attr_write
);