3 * Library for filesystems writers.
6 #include <linux/module.h>
7 #include <linux/pagemap.h>
8 #include <linux/slab.h>
9 #include <linux/mount.h>
10 #include <linux/vfs.h>
11 #include <linux/mutex.h>
12 #include <linux/exportfs.h>
13 #include <linux/writeback.h>
14 #include <linux/buffer_head.h>
16 #include <asm/uaccess.h>
18 int simple_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
,
21 struct inode
*inode
= dentry
->d_inode
;
22 generic_fillattr(inode
, stat
);
23 stat
->blocks
= inode
->i_mapping
->nrpages
<< (PAGE_CACHE_SHIFT
- 9);
27 int simple_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
29 buf
->f_type
= dentry
->d_sb
->s_magic
;
30 buf
->f_bsize
= PAGE_CACHE_SIZE
;
31 buf
->f_namelen
= NAME_MAX
;
36 * Retaining negative dentries for an in-memory filesystem just wastes
37 * memory and lookup time: arrange for them to be deleted immediately.
39 static int simple_delete_dentry(struct dentry
*dentry
)
45 * Lookup the data. This is trivial - if the dentry didn't already
46 * exist, we know it is negative. Set d_op to delete negative dentries.
48 struct dentry
*simple_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
)
50 static const struct dentry_operations simple_dentry_operations
= {
51 .d_delete
= simple_delete_dentry
,
54 if (dentry
->d_name
.len
> NAME_MAX
)
55 return ERR_PTR(-ENAMETOOLONG
);
56 dentry
->d_op
= &simple_dentry_operations
;
61 int simple_sync_file(struct file
* file
, struct dentry
*dentry
, int datasync
)
66 int dcache_dir_open(struct inode
*inode
, struct file
*file
)
68 static struct qstr cursor_name
= {.len
= 1, .name
= "."};
70 file
->private_data
= d_alloc(file
->f_path
.dentry
, &cursor_name
);
72 return file
->private_data
? 0 : -ENOMEM
;
75 int dcache_dir_close(struct inode
*inode
, struct file
*file
)
77 dput(file
->private_data
);
81 loff_t
dcache_dir_lseek(struct file
*file
, loff_t offset
, int origin
)
83 mutex_lock(&file
->f_path
.dentry
->d_inode
->i_mutex
);
86 offset
+= file
->f_pos
;
91 mutex_unlock(&file
->f_path
.dentry
->d_inode
->i_mutex
);
94 if (offset
!= file
->f_pos
) {
96 if (file
->f_pos
>= 2) {
98 struct dentry
*cursor
= file
->private_data
;
99 loff_t n
= file
->f_pos
- 2;
101 spin_lock(&dcache_lock
);
102 list_del(&cursor
->d_u
.d_child
);
103 p
= file
->f_path
.dentry
->d_subdirs
.next
;
104 while (n
&& p
!= &file
->f_path
.dentry
->d_subdirs
) {
106 next
= list_entry(p
, struct dentry
, d_u
.d_child
);
107 if (!d_unhashed(next
) && next
->d_inode
)
111 list_add_tail(&cursor
->d_u
.d_child
, p
);
112 spin_unlock(&dcache_lock
);
115 mutex_unlock(&file
->f_path
.dentry
->d_inode
->i_mutex
);
119 /* Relationship between i_mode and the DT_xxx types */
120 static inline unsigned char dt_type(struct inode
*inode
)
122 return (inode
->i_mode
>> 12) & 15;
126 * Directory is locked and all positive dentries in it are safe, since
127 * for ramfs-type trees they can't go away without unlink() or rmdir(),
128 * both impossible due to the lock on directory.
131 int dcache_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
133 struct dentry
*dentry
= filp
->f_path
.dentry
;
134 struct dentry
*cursor
= filp
->private_data
;
135 struct list_head
*p
, *q
= &cursor
->d_u
.d_child
;
141 ino
= dentry
->d_inode
->i_ino
;
142 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
148 ino
= parent_ino(dentry
);
149 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
155 spin_lock(&dcache_lock
);
156 if (filp
->f_pos
== 2)
157 list_move(q
, &dentry
->d_subdirs
);
159 for (p
=q
->next
; p
!= &dentry
->d_subdirs
; p
=p
->next
) {
161 next
= list_entry(p
, struct dentry
, d_u
.d_child
);
162 if (d_unhashed(next
) || !next
->d_inode
)
165 spin_unlock(&dcache_lock
);
166 if (filldir(dirent
, next
->d_name
.name
,
167 next
->d_name
.len
, filp
->f_pos
,
168 next
->d_inode
->i_ino
,
169 dt_type(next
->d_inode
)) < 0)
171 spin_lock(&dcache_lock
);
172 /* next is still alive */
177 spin_unlock(&dcache_lock
);
182 ssize_t
generic_read_dir(struct file
*filp
, char __user
*buf
, size_t siz
, loff_t
*ppos
)
187 const struct file_operations simple_dir_operations
= {
188 .open
= dcache_dir_open
,
189 .release
= dcache_dir_close
,
190 .llseek
= dcache_dir_lseek
,
191 .read
= generic_read_dir
,
192 .readdir
= dcache_readdir
,
193 .fsync
= simple_sync_file
,
196 const struct inode_operations simple_dir_inode_operations
= {
197 .lookup
= simple_lookup
,
200 static const struct super_operations simple_super_operations
= {
201 .statfs
= simple_statfs
,
205 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
206 * will never be mountable)
208 int get_sb_pseudo(struct file_system_type
*fs_type
, char *name
,
209 const struct super_operations
*ops
, unsigned long magic
,
210 struct vfsmount
*mnt
)
212 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
213 struct dentry
*dentry
;
215 struct qstr d_name
= {.name
= name
, .len
= strlen(name
)};
220 s
->s_flags
= MS_NOUSER
;
221 s
->s_maxbytes
= MAX_LFS_FILESIZE
;
222 s
->s_blocksize
= PAGE_SIZE
;
223 s
->s_blocksize_bits
= PAGE_SHIFT
;
225 s
->s_op
= ops
? ops
: &simple_super_operations
;
231 * since this is the first inode, make it number 1. New inodes created
232 * after this must take care not to collide with it (by passing
233 * max_reserved of 1 to iunique).
236 root
->i_mode
= S_IFDIR
| S_IRUSR
| S_IWUSR
;
237 root
->i_atime
= root
->i_mtime
= root
->i_ctime
= CURRENT_TIME
;
238 dentry
= d_alloc(NULL
, &d_name
);
244 dentry
->d_parent
= dentry
;
245 d_instantiate(dentry
, root
);
247 s
->s_flags
|= MS_ACTIVE
;
248 simple_set_mnt(mnt
, s
);
252 deactivate_locked_super(s
);
256 int simple_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*dentry
)
258 struct inode
*inode
= old_dentry
->d_inode
;
260 inode
->i_ctime
= dir
->i_ctime
= dir
->i_mtime
= CURRENT_TIME
;
262 atomic_inc(&inode
->i_count
);
264 d_instantiate(dentry
, inode
);
268 static inline int simple_positive(struct dentry
*dentry
)
270 return dentry
->d_inode
&& !d_unhashed(dentry
);
273 int simple_empty(struct dentry
*dentry
)
275 struct dentry
*child
;
278 spin_lock(&dcache_lock
);
279 list_for_each_entry(child
, &dentry
->d_subdirs
, d_u
.d_child
)
280 if (simple_positive(child
))
284 spin_unlock(&dcache_lock
);
288 int simple_unlink(struct inode
*dir
, struct dentry
*dentry
)
290 struct inode
*inode
= dentry
->d_inode
;
292 inode
->i_ctime
= dir
->i_ctime
= dir
->i_mtime
= CURRENT_TIME
;
298 int simple_rmdir(struct inode
*dir
, struct dentry
*dentry
)
300 if (!simple_empty(dentry
))
303 drop_nlink(dentry
->d_inode
);
304 simple_unlink(dir
, dentry
);
309 int simple_rename(struct inode
*old_dir
, struct dentry
*old_dentry
,
310 struct inode
*new_dir
, struct dentry
*new_dentry
)
312 struct inode
*inode
= old_dentry
->d_inode
;
313 int they_are_dirs
= S_ISDIR(old_dentry
->d_inode
->i_mode
);
315 if (!simple_empty(new_dentry
))
318 if (new_dentry
->d_inode
) {
319 simple_unlink(new_dir
, new_dentry
);
322 } else if (they_are_dirs
) {
327 old_dir
->i_ctime
= old_dir
->i_mtime
= new_dir
->i_ctime
=
328 new_dir
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
333 int simple_readpage(struct file
*file
, struct page
*page
)
335 clear_highpage(page
);
336 flush_dcache_page(page
);
337 SetPageUptodate(page
);
342 int simple_write_begin(struct file
*file
, struct address_space
*mapping
,
343 loff_t pos
, unsigned len
, unsigned flags
,
344 struct page
**pagep
, void **fsdata
)
349 index
= pos
>> PAGE_CACHE_SHIFT
;
351 page
= grab_cache_page_write_begin(mapping
, index
, flags
);
357 if (!PageUptodate(page
) && (len
!= PAGE_CACHE_SIZE
)) {
358 unsigned from
= pos
& (PAGE_CACHE_SIZE
- 1);
360 zero_user_segments(page
, 0, from
, from
+ len
, PAGE_CACHE_SIZE
);
366 * simple_write_end - .write_end helper for non-block-device FSes
367 * @available: See .write_end of address_space_operations
376 * simple_write_end does the minimum needed for updating a page after writing is
377 * done. It has the same API signature as the .write_end of
378 * address_space_operations vector. So it can just be set onto .write_end for
379 * FSes that don't need any other processing. i_mutex is assumed to be held.
380 * Block based filesystems should use generic_write_end().
381 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
382 * is not called, so a filesystem that actually does store data in .write_inode
383 * should extend on what's done here with a call to mark_inode_dirty() in the
384 * case that i_size has changed.
386 int simple_write_end(struct file
*file
, struct address_space
*mapping
,
387 loff_t pos
, unsigned len
, unsigned copied
,
388 struct page
*page
, void *fsdata
)
390 struct inode
*inode
= page
->mapping
->host
;
391 loff_t last_pos
= pos
+ copied
;
393 /* zero the stale part of the page if we did a short copy */
395 unsigned from
= pos
& (PAGE_CACHE_SIZE
- 1);
397 zero_user(page
, from
+ copied
, len
- copied
);
400 if (!PageUptodate(page
))
401 SetPageUptodate(page
);
403 * No need to use i_size_read() here, the i_size
404 * cannot change under us because we hold the i_mutex.
406 if (last_pos
> inode
->i_size
)
407 i_size_write(inode
, last_pos
);
409 set_page_dirty(page
);
411 page_cache_release(page
);
417 * the inodes created here are not hashed. If you use iunique to generate
418 * unique inode values later for this filesystem, then you must take care
419 * to pass it an appropriate max_reserved value to avoid collisions.
421 int simple_fill_super(struct super_block
*s
, int magic
, struct tree_descr
*files
)
425 struct dentry
*dentry
;
428 s
->s_blocksize
= PAGE_CACHE_SIZE
;
429 s
->s_blocksize_bits
= PAGE_CACHE_SHIFT
;
431 s
->s_op
= &simple_super_operations
;
434 inode
= new_inode(s
);
438 * because the root inode is 1, the files array must not contain an
442 inode
->i_mode
= S_IFDIR
| 0755;
443 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
444 inode
->i_op
= &simple_dir_inode_operations
;
445 inode
->i_fop
= &simple_dir_operations
;
447 root
= d_alloc_root(inode
);
452 for (i
= 0; !files
->name
|| files
->name
[0]; i
++, files
++) {
456 /* warn if it tries to conflict with the root inode */
457 if (unlikely(i
== 1))
458 printk(KERN_WARNING
"%s: %s passed in a files array"
459 "with an index of 1!\n", __func__
,
462 dentry
= d_alloc_name(root
, files
->name
);
465 inode
= new_inode(s
);
468 inode
->i_mode
= S_IFREG
| files
->mode
;
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
)
537 if (pos
>= available
|| !count
)
539 if (count
> available
- pos
)
540 count
= available
- pos
;
541 ret
= copy_to_user(to
, from
+ pos
, count
);
550 * memory_read_from_buffer - copy data from the buffer
551 * @to: the kernel space buffer to read to
552 * @count: the maximum number of bytes to read
553 * @ppos: the current position in the buffer
554 * @from: the buffer to read from
555 * @available: the size of the buffer
557 * The memory_read_from_buffer() function reads up to @count bytes from the
558 * buffer @from at offset @ppos into the kernel space address starting at @to.
560 * On success, the number of bytes read is returned and the offset @ppos is
561 * advanced by this number, or negative value is returned on error.
563 ssize_t
memory_read_from_buffer(void *to
, size_t count
, loff_t
*ppos
,
564 const void *from
, size_t available
)
570 if (pos
>= available
)
572 if (count
> available
- pos
)
573 count
= available
- pos
;
574 memcpy(to
, from
+ pos
, count
);
581 * Transaction based IO.
582 * The file expects a single write which triggers the transaction, and then
583 * possibly a read which collects the result - which is stored in a
587 void simple_transaction_set(struct file
*file
, size_t n
)
589 struct simple_transaction_argresp
*ar
= file
->private_data
;
591 BUG_ON(n
> SIMPLE_TRANSACTION_LIMIT
);
594 * The barrier ensures that ar->size will really remain zero until
595 * ar->data is ready for reading.
601 char *simple_transaction_get(struct file
*file
, const char __user
*buf
, size_t size
)
603 struct simple_transaction_argresp
*ar
;
604 static DEFINE_SPINLOCK(simple_transaction_lock
);
606 if (size
> SIMPLE_TRANSACTION_LIMIT
- 1)
607 return ERR_PTR(-EFBIG
);
609 ar
= (struct simple_transaction_argresp
*)get_zeroed_page(GFP_KERNEL
);
611 return ERR_PTR(-ENOMEM
);
613 spin_lock(&simple_transaction_lock
);
615 /* only one write allowed per open */
616 if (file
->private_data
) {
617 spin_unlock(&simple_transaction_lock
);
618 free_page((unsigned long)ar
);
619 return ERR_PTR(-EBUSY
);
622 file
->private_data
= ar
;
624 spin_unlock(&simple_transaction_lock
);
626 if (copy_from_user(ar
->data
, buf
, size
))
627 return ERR_PTR(-EFAULT
);
632 ssize_t
simple_transaction_read(struct file
*file
, char __user
*buf
, size_t size
, loff_t
*pos
)
634 struct simple_transaction_argresp
*ar
= file
->private_data
;
638 return simple_read_from_buffer(buf
, size
, pos
, ar
->data
, ar
->size
);
641 int simple_transaction_release(struct inode
*inode
, struct file
*file
)
643 free_page((unsigned long)file
->private_data
);
647 /* Simple attribute files */
650 int (*get
)(void *, u64
*);
651 int (*set
)(void *, u64
);
652 char get_buf
[24]; /* enough to store a u64 and "\n\0" */
655 const char *fmt
; /* format for read operation */
656 struct mutex mutex
; /* protects access to these buffers */
659 /* simple_attr_open is called by an actual attribute open file operation
660 * to set the attribute specific access operations. */
661 int simple_attr_open(struct inode
*inode
, struct file
*file
,
662 int (*get
)(void *, u64
*), int (*set
)(void *, u64
),
665 struct simple_attr
*attr
;
667 attr
= kmalloc(sizeof(*attr
), GFP_KERNEL
);
673 attr
->data
= inode
->i_private
;
675 mutex_init(&attr
->mutex
);
677 file
->private_data
= attr
;
679 return nonseekable_open(inode
, file
);
682 int simple_attr_release(struct inode
*inode
, struct file
*file
)
684 kfree(file
->private_data
);
688 /* read from the buffer that is filled with the get function */
689 ssize_t
simple_attr_read(struct file
*file
, char __user
*buf
,
690 size_t len
, loff_t
*ppos
)
692 struct simple_attr
*attr
;
696 attr
= file
->private_data
;
701 ret
= mutex_lock_interruptible(&attr
->mutex
);
705 if (*ppos
) { /* continued read */
706 size
= strlen(attr
->get_buf
);
707 } else { /* first read */
709 ret
= attr
->get(attr
->data
, &val
);
713 size
= scnprintf(attr
->get_buf
, sizeof(attr
->get_buf
),
714 attr
->fmt
, (unsigned long long)val
);
717 ret
= simple_read_from_buffer(buf
, len
, ppos
, attr
->get_buf
, size
);
719 mutex_unlock(&attr
->mutex
);
723 /* interpret the buffer as a number to call the set function with */
724 ssize_t
simple_attr_write(struct file
*file
, const char __user
*buf
,
725 size_t len
, loff_t
*ppos
)
727 struct simple_attr
*attr
;
732 attr
= file
->private_data
;
736 ret
= mutex_lock_interruptible(&attr
->mutex
);
741 size
= min(sizeof(attr
->set_buf
) - 1, len
);
742 if (copy_from_user(attr
->set_buf
, buf
, size
))
745 attr
->set_buf
[size
] = '\0';
746 val
= simple_strtol(attr
->set_buf
, NULL
, 0);
747 ret
= attr
->set(attr
->data
, val
);
749 ret
= len
; /* on success, claim we got the whole input */
751 mutex_unlock(&attr
->mutex
);
756 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
757 * @sb: filesystem to do the file handle conversion on
758 * @fid: file handle to convert
759 * @fh_len: length of the file handle in bytes
760 * @fh_type: type of file handle
761 * @get_inode: filesystem callback to retrieve inode
763 * This function decodes @fid as long as it has one of the well-known
764 * Linux filehandle types and calls @get_inode on it to retrieve the
765 * inode for the object specified in the file handle.
767 struct dentry
*generic_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
768 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
769 (struct super_block
*sb
, u64 ino
, u32 gen
))
771 struct inode
*inode
= NULL
;
777 case FILEID_INO32_GEN
:
778 case FILEID_INO32_GEN_PARENT
:
779 inode
= get_inode(sb
, fid
->i32
.ino
, fid
->i32
.gen
);
783 return d_obtain_alias(inode
);
785 EXPORT_SYMBOL_GPL(generic_fh_to_dentry
);
788 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
789 * @sb: filesystem to do the file handle conversion on
790 * @fid: file handle to convert
791 * @fh_len: length of the file handle in bytes
792 * @fh_type: type of file handle
793 * @get_inode: filesystem callback to retrieve inode
795 * This function decodes @fid as long as it has one of the well-known
796 * Linux filehandle types and calls @get_inode on it to retrieve the
797 * inode for the _parent_ object specified in the file handle if it
798 * is specified in the file handle, or NULL otherwise.
800 struct dentry
*generic_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
801 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
802 (struct super_block
*sb
, u64 ino
, u32 gen
))
804 struct inode
*inode
= NULL
;
810 case FILEID_INO32_GEN_PARENT
:
811 inode
= get_inode(sb
, fid
->i32
.parent_ino
,
812 (fh_len
> 3 ? fid
->i32
.parent_gen
: 0));
816 return d_obtain_alias(inode
);
818 EXPORT_SYMBOL_GPL(generic_fh_to_parent
);
820 int simple_fsync(struct file
*file
, struct dentry
*dentry
, int datasync
)
822 struct writeback_control wbc
= {
823 .sync_mode
= WB_SYNC_ALL
,
824 .nr_to_write
= 0, /* metadata-only; caller takes care of data */
826 struct inode
*inode
= dentry
->d_inode
;
830 ret
= sync_mapping_buffers(inode
->i_mapping
);
831 if (!(inode
->i_state
& I_DIRTY
))
833 if (datasync
&& !(inode
->i_state
& I_DIRTY_DATASYNC
))
836 err
= sync_inode(inode
, &wbc
);
841 EXPORT_SYMBOL(simple_fsync
);
843 EXPORT_SYMBOL(dcache_dir_close
);
844 EXPORT_SYMBOL(dcache_dir_lseek
);
845 EXPORT_SYMBOL(dcache_dir_open
);
846 EXPORT_SYMBOL(dcache_readdir
);
847 EXPORT_SYMBOL(generic_read_dir
);
848 EXPORT_SYMBOL(get_sb_pseudo
);
849 EXPORT_SYMBOL(simple_write_begin
);
850 EXPORT_SYMBOL(simple_write_end
);
851 EXPORT_SYMBOL(simple_dir_inode_operations
);
852 EXPORT_SYMBOL(simple_dir_operations
);
853 EXPORT_SYMBOL(simple_empty
);
854 EXPORT_SYMBOL(simple_fill_super
);
855 EXPORT_SYMBOL(simple_getattr
);
856 EXPORT_SYMBOL(simple_link
);
857 EXPORT_SYMBOL(simple_lookup
);
858 EXPORT_SYMBOL(simple_pin_fs
);
859 EXPORT_SYMBOL(simple_readpage
);
860 EXPORT_SYMBOL(simple_release_fs
);
861 EXPORT_SYMBOL(simple_rename
);
862 EXPORT_SYMBOL(simple_rmdir
);
863 EXPORT_SYMBOL(simple_statfs
);
864 EXPORT_SYMBOL(simple_sync_file
);
865 EXPORT_SYMBOL(simple_unlink
);
866 EXPORT_SYMBOL(simple_read_from_buffer
);
867 EXPORT_SYMBOL(memory_read_from_buffer
);
868 EXPORT_SYMBOL(simple_transaction_set
);
869 EXPORT_SYMBOL(simple_transaction_get
);
870 EXPORT_SYMBOL(simple_transaction_read
);
871 EXPORT_SYMBOL(simple_transaction_release
);
872 EXPORT_SYMBOL_GPL(simple_attr_open
);
873 EXPORT_SYMBOL_GPL(simple_attr_release
);
874 EXPORT_SYMBOL_GPL(simple_attr_read
);
875 EXPORT_SYMBOL_GPL(simple_attr_write
);