1 // SPDX-License-Identifier: GPL-2.0-only
4 * Library for filesystems writers.
7 #include <linux/blkdev.h>
8 #include <linux/export.h>
9 #include <linux/pagemap.h>
10 #include <linux/slab.h>
11 #include <linux/cred.h>
12 #include <linux/mount.h>
13 #include <linux/vfs.h>
14 #include <linux/quotaops.h>
15 #include <linux/mutex.h>
16 #include <linux/namei.h>
17 #include <linux/exportfs.h>
18 #include <linux/writeback.h>
19 #include <linux/buffer_head.h> /* sync_mapping_buffers */
20 #include <linux/fs_context.h>
21 #include <linux/pseudo_fs.h>
24 #include <linux/uaccess.h>
28 int simple_getattr(const struct path
*path
, struct kstat
*stat
,
29 u32 request_mask
, unsigned int query_flags
)
31 struct inode
*inode
= d_inode(path
->dentry
);
32 generic_fillattr(inode
, stat
);
33 stat
->blocks
= inode
->i_mapping
->nrpages
<< (PAGE_SHIFT
- 9);
36 EXPORT_SYMBOL(simple_getattr
);
38 int simple_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
40 buf
->f_type
= dentry
->d_sb
->s_magic
;
41 buf
->f_bsize
= PAGE_SIZE
;
42 buf
->f_namelen
= NAME_MAX
;
45 EXPORT_SYMBOL(simple_statfs
);
48 * Retaining negative dentries for an in-memory filesystem just wastes
49 * memory and lookup time: arrange for them to be deleted immediately.
51 int always_delete_dentry(const struct dentry
*dentry
)
55 EXPORT_SYMBOL(always_delete_dentry
);
57 const struct dentry_operations simple_dentry_operations
= {
58 .d_delete
= always_delete_dentry
,
60 EXPORT_SYMBOL(simple_dentry_operations
);
63 * Lookup the data. This is trivial - if the dentry didn't already
64 * exist, we know it is negative. Set d_op to delete negative dentries.
66 struct dentry
*simple_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
68 if (dentry
->d_name
.len
> NAME_MAX
)
69 return ERR_PTR(-ENAMETOOLONG
);
70 if (!dentry
->d_sb
->s_d_op
)
71 d_set_d_op(dentry
, &simple_dentry_operations
);
75 EXPORT_SYMBOL(simple_lookup
);
77 int dcache_dir_open(struct inode
*inode
, struct file
*file
)
79 file
->private_data
= d_alloc_cursor(file
->f_path
.dentry
);
81 return file
->private_data
? 0 : -ENOMEM
;
83 EXPORT_SYMBOL(dcache_dir_open
);
85 int dcache_dir_close(struct inode
*inode
, struct file
*file
)
87 dput(file
->private_data
);
90 EXPORT_SYMBOL(dcache_dir_close
);
92 /* parent is locked at least shared */
94 * Returns an element of siblings' list.
95 * We are looking for <count>th positive after <p>; if
96 * found, dentry is grabbed and returned to caller.
97 * If no such element exists, NULL is returned.
99 static struct dentry
*scan_positives(struct dentry
*cursor
,
104 struct dentry
*dentry
= cursor
->d_parent
, *found
= NULL
;
106 spin_lock(&dentry
->d_lock
);
107 while ((p
= p
->next
) != &dentry
->d_subdirs
) {
108 struct dentry
*d
= list_entry(p
, struct dentry
, d_child
);
109 // we must at least skip cursors, to avoid livelocks
110 if (d
->d_flags
& DCACHE_DENTRY_CURSOR
)
112 if (simple_positive(d
) && !--count
) {
113 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
114 if (simple_positive(d
))
115 found
= dget_dlock(d
);
116 spin_unlock(&d
->d_lock
);
121 if (need_resched()) {
122 list_move(&cursor
->d_child
, p
);
123 p
= &cursor
->d_child
;
124 spin_unlock(&dentry
->d_lock
);
126 spin_lock(&dentry
->d_lock
);
129 spin_unlock(&dentry
->d_lock
);
134 loff_t
dcache_dir_lseek(struct file
*file
, loff_t offset
, int whence
)
136 struct dentry
*dentry
= file
->f_path
.dentry
;
139 offset
+= file
->f_pos
;
148 if (offset
!= file
->f_pos
) {
149 struct dentry
*cursor
= file
->private_data
;
150 struct dentry
*to
= NULL
;
152 inode_lock_shared(dentry
->d_inode
);
155 to
= scan_positives(cursor
, &dentry
->d_subdirs
,
157 spin_lock(&dentry
->d_lock
);
159 list_move(&cursor
->d_child
, &to
->d_child
);
161 list_del_init(&cursor
->d_child
);
162 spin_unlock(&dentry
->d_lock
);
165 file
->f_pos
= offset
;
167 inode_unlock_shared(dentry
->d_inode
);
171 EXPORT_SYMBOL(dcache_dir_lseek
);
173 /* Relationship between i_mode and the DT_xxx types */
174 static inline unsigned char dt_type(struct inode
*inode
)
176 return (inode
->i_mode
>> 12) & 15;
180 * Directory is locked and all positive dentries in it are safe, since
181 * for ramfs-type trees they can't go away without unlink() or rmdir(),
182 * both impossible due to the lock on directory.
185 int dcache_readdir(struct file
*file
, struct dir_context
*ctx
)
187 struct dentry
*dentry
= file
->f_path
.dentry
;
188 struct dentry
*cursor
= file
->private_data
;
189 struct list_head
*anchor
= &dentry
->d_subdirs
;
190 struct dentry
*next
= NULL
;
193 if (!dir_emit_dots(file
, ctx
))
198 else if (!list_empty(&cursor
->d_child
))
199 p
= &cursor
->d_child
;
203 while ((next
= scan_positives(cursor
, p
, 1, next
)) != NULL
) {
204 if (!dir_emit(ctx
, next
->d_name
.name
, next
->d_name
.len
,
205 d_inode(next
)->i_ino
, dt_type(d_inode(next
))))
210 spin_lock(&dentry
->d_lock
);
212 list_move_tail(&cursor
->d_child
, &next
->d_child
);
214 list_del_init(&cursor
->d_child
);
215 spin_unlock(&dentry
->d_lock
);
220 EXPORT_SYMBOL(dcache_readdir
);
222 ssize_t
generic_read_dir(struct file
*filp
, char __user
*buf
, size_t siz
, loff_t
*ppos
)
226 EXPORT_SYMBOL(generic_read_dir
);
228 const struct file_operations simple_dir_operations
= {
229 .open
= dcache_dir_open
,
230 .release
= dcache_dir_close
,
231 .llseek
= dcache_dir_lseek
,
232 .read
= generic_read_dir
,
233 .iterate_shared
= dcache_readdir
,
236 EXPORT_SYMBOL(simple_dir_operations
);
238 const struct inode_operations simple_dir_inode_operations
= {
239 .lookup
= simple_lookup
,
241 EXPORT_SYMBOL(simple_dir_inode_operations
);
243 static const struct super_operations simple_super_operations
= {
244 .statfs
= simple_statfs
,
247 static int pseudo_fs_fill_super(struct super_block
*s
, struct fs_context
*fc
)
249 struct pseudo_fs_context
*ctx
= fc
->fs_private
;
252 s
->s_maxbytes
= MAX_LFS_FILESIZE
;
253 s
->s_blocksize
= PAGE_SIZE
;
254 s
->s_blocksize_bits
= PAGE_SHIFT
;
255 s
->s_magic
= ctx
->magic
;
256 s
->s_op
= ctx
->ops
?: &simple_super_operations
;
257 s
->s_xattr
= ctx
->xattr
;
264 * since this is the first inode, make it number 1. New inodes created
265 * after this must take care not to collide with it (by passing
266 * max_reserved of 1 to iunique).
269 root
->i_mode
= S_IFDIR
| S_IRUSR
| S_IWUSR
;
270 root
->i_atime
= root
->i_mtime
= root
->i_ctime
= current_time(root
);
271 s
->s_root
= d_make_root(root
);
274 s
->s_d_op
= ctx
->dops
;
278 static int pseudo_fs_get_tree(struct fs_context
*fc
)
280 return get_tree_nodev(fc
, pseudo_fs_fill_super
);
283 static void pseudo_fs_free(struct fs_context
*fc
)
285 kfree(fc
->fs_private
);
288 static const struct fs_context_operations pseudo_fs_context_ops
= {
289 .free
= pseudo_fs_free
,
290 .get_tree
= pseudo_fs_get_tree
,
294 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
295 * will never be mountable)
297 struct pseudo_fs_context
*init_pseudo(struct fs_context
*fc
,
300 struct pseudo_fs_context
*ctx
;
302 ctx
= kzalloc(sizeof(struct pseudo_fs_context
), GFP_KERNEL
);
305 fc
->fs_private
= ctx
;
306 fc
->ops
= &pseudo_fs_context_ops
;
307 fc
->sb_flags
|= SB_NOUSER
;
312 EXPORT_SYMBOL(init_pseudo
);
314 int simple_open(struct inode
*inode
, struct file
*file
)
316 if (inode
->i_private
)
317 file
->private_data
= inode
->i_private
;
320 EXPORT_SYMBOL(simple_open
);
322 int simple_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*dentry
)
324 struct inode
*inode
= d_inode(old_dentry
);
326 inode
->i_ctime
= dir
->i_ctime
= dir
->i_mtime
= current_time(inode
);
330 d_instantiate(dentry
, inode
);
333 EXPORT_SYMBOL(simple_link
);
335 int simple_empty(struct dentry
*dentry
)
337 struct dentry
*child
;
340 spin_lock(&dentry
->d_lock
);
341 list_for_each_entry(child
, &dentry
->d_subdirs
, d_child
) {
342 spin_lock_nested(&child
->d_lock
, DENTRY_D_LOCK_NESTED
);
343 if (simple_positive(child
)) {
344 spin_unlock(&child
->d_lock
);
347 spin_unlock(&child
->d_lock
);
351 spin_unlock(&dentry
->d_lock
);
354 EXPORT_SYMBOL(simple_empty
);
356 int simple_unlink(struct inode
*dir
, struct dentry
*dentry
)
358 struct inode
*inode
= d_inode(dentry
);
360 inode
->i_ctime
= dir
->i_ctime
= dir
->i_mtime
= current_time(inode
);
365 EXPORT_SYMBOL(simple_unlink
);
367 int simple_rmdir(struct inode
*dir
, struct dentry
*dentry
)
369 if (!simple_empty(dentry
))
372 drop_nlink(d_inode(dentry
));
373 simple_unlink(dir
, dentry
);
377 EXPORT_SYMBOL(simple_rmdir
);
379 int simple_rename(struct inode
*old_dir
, struct dentry
*old_dentry
,
380 struct inode
*new_dir
, struct dentry
*new_dentry
,
383 struct inode
*inode
= d_inode(old_dentry
);
384 int they_are_dirs
= d_is_dir(old_dentry
);
386 if (flags
& ~RENAME_NOREPLACE
)
389 if (!simple_empty(new_dentry
))
392 if (d_really_is_positive(new_dentry
)) {
393 simple_unlink(new_dir
, new_dentry
);
395 drop_nlink(d_inode(new_dentry
));
398 } else if (they_are_dirs
) {
403 old_dir
->i_ctime
= old_dir
->i_mtime
= new_dir
->i_ctime
=
404 new_dir
->i_mtime
= inode
->i_ctime
= current_time(old_dir
);
408 EXPORT_SYMBOL(simple_rename
);
411 * simple_setattr - setattr for simple filesystem
413 * @iattr: iattr structure
415 * Returns 0 on success, -error on failure.
417 * simple_setattr is a simple ->setattr implementation without a proper
418 * implementation of size changes.
420 * It can either be used for in-memory filesystems or special files
421 * on simple regular filesystems. Anything that needs to change on-disk
422 * or wire state on size changes needs its own setattr method.
424 int simple_setattr(struct dentry
*dentry
, struct iattr
*iattr
)
426 struct inode
*inode
= d_inode(dentry
);
429 error
= setattr_prepare(dentry
, iattr
);
433 if (iattr
->ia_valid
& ATTR_SIZE
)
434 truncate_setsize(inode
, iattr
->ia_size
);
435 setattr_copy(inode
, iattr
);
436 mark_inode_dirty(inode
);
439 EXPORT_SYMBOL(simple_setattr
);
441 int simple_readpage(struct file
*file
, struct page
*page
)
443 clear_highpage(page
);
444 flush_dcache_page(page
);
445 SetPageUptodate(page
);
449 EXPORT_SYMBOL(simple_readpage
);
451 int simple_write_begin(struct file
*file
, struct address_space
*mapping
,
452 loff_t pos
, unsigned len
, unsigned flags
,
453 struct page
**pagep
, void **fsdata
)
458 index
= pos
>> PAGE_SHIFT
;
460 page
= grab_cache_page_write_begin(mapping
, index
, flags
);
466 if (!PageUptodate(page
) && (len
!= PAGE_SIZE
)) {
467 unsigned from
= pos
& (PAGE_SIZE
- 1);
469 zero_user_segments(page
, 0, from
, from
+ len
, PAGE_SIZE
);
473 EXPORT_SYMBOL(simple_write_begin
);
476 * simple_write_end - .write_end helper for non-block-device FSes
477 * @file: See .write_end of address_space_operations
485 * simple_write_end does the minimum needed for updating a page after writing is
486 * done. It has the same API signature as the .write_end of
487 * address_space_operations vector. So it can just be set onto .write_end for
488 * FSes that don't need any other processing. i_mutex is assumed to be held.
489 * Block based filesystems should use generic_write_end().
490 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
491 * is not called, so a filesystem that actually does store data in .write_inode
492 * should extend on what's done here with a call to mark_inode_dirty() in the
493 * case that i_size has changed.
495 * Use *ONLY* with simple_readpage()
497 int simple_write_end(struct file
*file
, struct address_space
*mapping
,
498 loff_t pos
, unsigned len
, unsigned copied
,
499 struct page
*page
, void *fsdata
)
501 struct inode
*inode
= page
->mapping
->host
;
502 loff_t last_pos
= pos
+ copied
;
504 /* zero the stale part of the page if we did a short copy */
505 if (!PageUptodate(page
)) {
507 unsigned from
= pos
& (PAGE_SIZE
- 1);
509 zero_user(page
, from
+ copied
, len
- copied
);
511 SetPageUptodate(page
);
514 * No need to use i_size_read() here, the i_size
515 * cannot change under us because we hold the i_mutex.
517 if (last_pos
> inode
->i_size
)
518 i_size_write(inode
, last_pos
);
520 set_page_dirty(page
);
526 EXPORT_SYMBOL(simple_write_end
);
529 * the inodes created here are not hashed. If you use iunique to generate
530 * unique inode values later for this filesystem, then you must take care
531 * to pass it an appropriate max_reserved value to avoid collisions.
533 int simple_fill_super(struct super_block
*s
, unsigned long magic
,
534 const struct tree_descr
*files
)
538 struct dentry
*dentry
;
541 s
->s_blocksize
= PAGE_SIZE
;
542 s
->s_blocksize_bits
= PAGE_SHIFT
;
544 s
->s_op
= &simple_super_operations
;
547 inode
= new_inode(s
);
551 * because the root inode is 1, the files array must not contain an
555 inode
->i_mode
= S_IFDIR
| 0755;
556 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
557 inode
->i_op
= &simple_dir_inode_operations
;
558 inode
->i_fop
= &simple_dir_operations
;
560 root
= d_make_root(inode
);
563 for (i
= 0; !files
->name
|| files
->name
[0]; i
++, files
++) {
567 /* warn if it tries to conflict with the root inode */
568 if (unlikely(i
== 1))
569 printk(KERN_WARNING
"%s: %s passed in a files array"
570 "with an index of 1!\n", __func__
,
573 dentry
= d_alloc_name(root
, files
->name
);
576 inode
= new_inode(s
);
581 inode
->i_mode
= S_IFREG
| files
->mode
;
582 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
583 inode
->i_fop
= files
->ops
;
585 d_add(dentry
, inode
);
591 shrink_dcache_parent(root
);
595 EXPORT_SYMBOL(simple_fill_super
);
597 static DEFINE_SPINLOCK(pin_fs_lock
);
599 int simple_pin_fs(struct file_system_type
*type
, struct vfsmount
**mount
, int *count
)
601 struct vfsmount
*mnt
= NULL
;
602 spin_lock(&pin_fs_lock
);
603 if (unlikely(!*mount
)) {
604 spin_unlock(&pin_fs_lock
);
605 mnt
= vfs_kern_mount(type
, SB_KERNMOUNT
, type
->name
, NULL
);
608 spin_lock(&pin_fs_lock
);
614 spin_unlock(&pin_fs_lock
);
618 EXPORT_SYMBOL(simple_pin_fs
);
620 void simple_release_fs(struct vfsmount
**mount
, int *count
)
622 struct vfsmount
*mnt
;
623 spin_lock(&pin_fs_lock
);
627 spin_unlock(&pin_fs_lock
);
630 EXPORT_SYMBOL(simple_release_fs
);
633 * simple_read_from_buffer - copy data from the buffer to user space
634 * @to: the user space buffer to read to
635 * @count: the maximum number of bytes to read
636 * @ppos: the current position in the buffer
637 * @from: the buffer to read from
638 * @available: the size of the buffer
640 * The simple_read_from_buffer() function reads up to @count bytes from the
641 * buffer @from at offset @ppos into the user space address starting at @to.
643 * On success, the number of bytes read is returned and the offset @ppos is
644 * advanced by this number, or negative value is returned on error.
646 ssize_t
simple_read_from_buffer(void __user
*to
, size_t count
, loff_t
*ppos
,
647 const void *from
, size_t available
)
654 if (pos
>= available
|| !count
)
656 if (count
> available
- pos
)
657 count
= available
- pos
;
658 ret
= copy_to_user(to
, from
+ pos
, count
);
665 EXPORT_SYMBOL(simple_read_from_buffer
);
668 * simple_write_to_buffer - copy data from user space to the buffer
669 * @to: the buffer to write to
670 * @available: the size of the buffer
671 * @ppos: the current position in the buffer
672 * @from: the user space buffer to read from
673 * @count: the maximum number of bytes to read
675 * The simple_write_to_buffer() function reads up to @count bytes from the user
676 * space address starting at @from into the buffer @to at offset @ppos.
678 * On success, the number of bytes written is returned and the offset @ppos is
679 * advanced by this number, or negative value is returned on error.
681 ssize_t
simple_write_to_buffer(void *to
, size_t available
, loff_t
*ppos
,
682 const void __user
*from
, size_t count
)
689 if (pos
>= available
|| !count
)
691 if (count
> available
- pos
)
692 count
= available
- pos
;
693 res
= copy_from_user(to
+ pos
, from
, count
);
700 EXPORT_SYMBOL(simple_write_to_buffer
);
703 * memory_read_from_buffer - copy data from the buffer
704 * @to: the kernel space buffer to read to
705 * @count: the maximum number of bytes to read
706 * @ppos: the current position in the buffer
707 * @from: the buffer to read from
708 * @available: the size of the buffer
710 * The memory_read_from_buffer() function reads up to @count bytes from the
711 * buffer @from at offset @ppos into the kernel space address starting at @to.
713 * On success, the number of bytes read is returned and the offset @ppos is
714 * advanced by this number, or negative value is returned on error.
716 ssize_t
memory_read_from_buffer(void *to
, size_t count
, loff_t
*ppos
,
717 const void *from
, size_t available
)
723 if (pos
>= available
)
725 if (count
> available
- pos
)
726 count
= available
- pos
;
727 memcpy(to
, from
+ pos
, count
);
732 EXPORT_SYMBOL(memory_read_from_buffer
);
735 * memory_read_from_io_buffer - copy data from a io memory mapped buffer
736 * @to: the kernel space buffer to read to
737 * @count: the maximum number of bytes to read
738 * @ppos: the current position in the buffer
739 * @from: the buffer to read from
740 * @available: the size of the buffer
742 * The memory_read_from_buffer() function reads up to @count bytes from the
743 * io memory mappy buffer @from at offset @ppos into the kernel space address
746 * On success, the number of bytes read is returned and the offset @ppos is
747 * advanced by this number, or negative value is returned on error.
749 ssize_t
memory_read_from_io_buffer(void *to
, size_t count
, loff_t
*ppos
,
750 const void *from
, size_t available
)
756 if (pos
>= available
)
758 if (count
> available
- pos
)
759 count
= available
- pos
;
760 memcpy_fromio(to
, from
+ pos
, count
);
765 EXPORT_SYMBOL(memory_read_from_io_buffer
);
768 * Transaction based IO.
769 * The file expects a single write which triggers the transaction, and then
770 * possibly a read which collects the result - which is stored in a
774 void simple_transaction_set(struct file
*file
, size_t n
)
776 struct simple_transaction_argresp
*ar
= file
->private_data
;
778 BUG_ON(n
> SIMPLE_TRANSACTION_LIMIT
);
781 * The barrier ensures that ar->size will really remain zero until
782 * ar->data is ready for reading.
787 EXPORT_SYMBOL(simple_transaction_set
);
789 char *simple_transaction_get(struct file
*file
, const char __user
*buf
, size_t size
)
791 struct simple_transaction_argresp
*ar
;
792 static DEFINE_SPINLOCK(simple_transaction_lock
);
794 if (size
> SIMPLE_TRANSACTION_LIMIT
- 1)
795 return ERR_PTR(-EFBIG
);
797 ar
= (struct simple_transaction_argresp
*)get_zeroed_page(GFP_KERNEL
);
799 return ERR_PTR(-ENOMEM
);
801 spin_lock(&simple_transaction_lock
);
803 /* only one write allowed per open */
804 if (file
->private_data
) {
805 spin_unlock(&simple_transaction_lock
);
806 free_page((unsigned long)ar
);
807 return ERR_PTR(-EBUSY
);
810 file
->private_data
= ar
;
812 spin_unlock(&simple_transaction_lock
);
814 if (copy_from_user(ar
->data
, buf
, size
))
815 return ERR_PTR(-EFAULT
);
819 EXPORT_SYMBOL(simple_transaction_get
);
821 ssize_t
simple_transaction_read(struct file
*file
, char __user
*buf
, size_t size
, loff_t
*pos
)
823 struct simple_transaction_argresp
*ar
= file
->private_data
;
827 return simple_read_from_buffer(buf
, size
, pos
, ar
->data
, ar
->size
);
829 EXPORT_SYMBOL(simple_transaction_read
);
831 int simple_transaction_release(struct inode
*inode
, struct file
*file
)
833 free_page((unsigned long)file
->private_data
);
836 EXPORT_SYMBOL(simple_transaction_release
);
838 /* Simple attribute files */
841 int (*get
)(void *, u64
*);
842 int (*set
)(void *, u64
);
843 char get_buf
[24]; /* enough to store a u64 and "\n\0" */
846 const char *fmt
; /* format for read operation */
847 struct mutex mutex
; /* protects access to these buffers */
850 /* simple_attr_open is called by an actual attribute open file operation
851 * to set the attribute specific access operations. */
852 int simple_attr_open(struct inode
*inode
, struct file
*file
,
853 int (*get
)(void *, u64
*), int (*set
)(void *, u64
),
856 struct simple_attr
*attr
;
858 attr
= kzalloc(sizeof(*attr
), GFP_KERNEL
);
864 attr
->data
= inode
->i_private
;
866 mutex_init(&attr
->mutex
);
868 file
->private_data
= attr
;
870 return nonseekable_open(inode
, file
);
872 EXPORT_SYMBOL_GPL(simple_attr_open
);
874 int simple_attr_release(struct inode
*inode
, struct file
*file
)
876 kfree(file
->private_data
);
879 EXPORT_SYMBOL_GPL(simple_attr_release
); /* GPL-only? This? Really? */
881 /* read from the buffer that is filled with the get function */
882 ssize_t
simple_attr_read(struct file
*file
, char __user
*buf
,
883 size_t len
, loff_t
*ppos
)
885 struct simple_attr
*attr
;
889 attr
= file
->private_data
;
894 ret
= mutex_lock_interruptible(&attr
->mutex
);
898 if (*ppos
&& attr
->get_buf
[0]) {
900 size
= strlen(attr
->get_buf
);
904 ret
= attr
->get(attr
->data
, &val
);
908 size
= scnprintf(attr
->get_buf
, sizeof(attr
->get_buf
),
909 attr
->fmt
, (unsigned long long)val
);
912 ret
= simple_read_from_buffer(buf
, len
, ppos
, attr
->get_buf
, size
);
914 mutex_unlock(&attr
->mutex
);
917 EXPORT_SYMBOL_GPL(simple_attr_read
);
919 /* interpret the buffer as a number to call the set function with */
920 ssize_t
simple_attr_write(struct file
*file
, const char __user
*buf
,
921 size_t len
, loff_t
*ppos
)
923 struct simple_attr
*attr
;
924 unsigned long long val
;
928 attr
= file
->private_data
;
932 ret
= mutex_lock_interruptible(&attr
->mutex
);
937 size
= min(sizeof(attr
->set_buf
) - 1, len
);
938 if (copy_from_user(attr
->set_buf
, buf
, size
))
941 attr
->set_buf
[size
] = '\0';
942 ret
= kstrtoull(attr
->set_buf
, 0, &val
);
945 ret
= attr
->set(attr
->data
, val
);
947 ret
= len
; /* on success, claim we got the whole input */
949 mutex_unlock(&attr
->mutex
);
952 EXPORT_SYMBOL_GPL(simple_attr_write
);
955 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
956 * @sb: filesystem to do the file handle conversion on
957 * @fid: file handle to convert
958 * @fh_len: length of the file handle in bytes
959 * @fh_type: type of file handle
960 * @get_inode: filesystem callback to retrieve inode
962 * This function decodes @fid as long as it has one of the well-known
963 * Linux filehandle types and calls @get_inode on it to retrieve the
964 * inode for the object specified in the file handle.
966 struct dentry
*generic_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
967 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
968 (struct super_block
*sb
, u64 ino
, u32 gen
))
970 struct inode
*inode
= NULL
;
976 case FILEID_INO32_GEN
:
977 case FILEID_INO32_GEN_PARENT
:
978 inode
= get_inode(sb
, fid
->i32
.ino
, fid
->i32
.gen
);
982 return d_obtain_alias(inode
);
984 EXPORT_SYMBOL_GPL(generic_fh_to_dentry
);
987 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
988 * @sb: filesystem to do the file handle conversion on
989 * @fid: file handle to convert
990 * @fh_len: length of the file handle in bytes
991 * @fh_type: type of file handle
992 * @get_inode: filesystem callback to retrieve inode
994 * This function decodes @fid as long as it has one of the well-known
995 * Linux filehandle types and calls @get_inode on it to retrieve the
996 * inode for the _parent_ object specified in the file handle if it
997 * is specified in the file handle, or NULL otherwise.
999 struct dentry
*generic_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1000 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1001 (struct super_block
*sb
, u64 ino
, u32 gen
))
1003 struct inode
*inode
= NULL
;
1009 case FILEID_INO32_GEN_PARENT
:
1010 inode
= get_inode(sb
, fid
->i32
.parent_ino
,
1011 (fh_len
> 3 ? fid
->i32
.parent_gen
: 0));
1015 return d_obtain_alias(inode
);
1017 EXPORT_SYMBOL_GPL(generic_fh_to_parent
);
1020 * __generic_file_fsync - generic fsync implementation for simple filesystems
1022 * @file: file to synchronize
1023 * @start: start offset in bytes
1024 * @end: end offset in bytes (inclusive)
1025 * @datasync: only synchronize essential metadata if true
1027 * This is a generic implementation of the fsync method for simple
1028 * filesystems which track all non-inode metadata in the buffers list
1029 * hanging off the address_space structure.
1031 int __generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1034 struct inode
*inode
= file
->f_mapping
->host
;
1038 err
= file_write_and_wait_range(file
, start
, end
);
1043 ret
= sync_mapping_buffers(inode
->i_mapping
);
1044 if (!(inode
->i_state
& I_DIRTY_ALL
))
1046 if (datasync
&& !(inode
->i_state
& I_DIRTY_DATASYNC
))
1049 err
= sync_inode_metadata(inode
, 1);
1054 inode_unlock(inode
);
1055 /* check and advance again to catch errors after syncing out buffers */
1056 err
= file_check_and_advance_wb_err(file
);
1061 EXPORT_SYMBOL(__generic_file_fsync
);
1064 * generic_file_fsync - generic fsync implementation for simple filesystems
1066 * @file: file to synchronize
1067 * @start: start offset in bytes
1068 * @end: end offset in bytes (inclusive)
1069 * @datasync: only synchronize essential metadata if true
1073 int generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1076 struct inode
*inode
= file
->f_mapping
->host
;
1079 err
= __generic_file_fsync(file
, start
, end
, datasync
);
1082 return blkdev_issue_flush(inode
->i_sb
->s_bdev
, GFP_KERNEL
, NULL
);
1084 EXPORT_SYMBOL(generic_file_fsync
);
1087 * generic_check_addressable - Check addressability of file system
1088 * @blocksize_bits: log of file system block size
1089 * @num_blocks: number of blocks in file system
1091 * Determine whether a file system with @num_blocks blocks (and a
1092 * block size of 2**@blocksize_bits) is addressable by the sector_t
1093 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1095 int generic_check_addressable(unsigned blocksize_bits
, u64 num_blocks
)
1097 u64 last_fs_block
= num_blocks
- 1;
1099 last_fs_block
>> (PAGE_SHIFT
- blocksize_bits
);
1101 if (unlikely(num_blocks
== 0))
1104 if ((blocksize_bits
< 9) || (blocksize_bits
> PAGE_SHIFT
))
1107 if ((last_fs_block
> (sector_t
)(~0ULL) >> (blocksize_bits
- 9)) ||
1108 (last_fs_page
> (pgoff_t
)(~0ULL))) {
1113 EXPORT_SYMBOL(generic_check_addressable
);
1116 * No-op implementation of ->fsync for in-memory filesystems.
1118 int noop_fsync(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
1122 EXPORT_SYMBOL(noop_fsync
);
1124 int noop_set_page_dirty(struct page
*page
)
1127 * Unlike __set_page_dirty_no_writeback that handles dirty page
1128 * tracking in the page object, dax does all dirty tracking in
1129 * the inode address_space in response to mkwrite faults. In the
1130 * dax case we only need to worry about potentially dirty CPU
1131 * caches, not dirty page cache pages to write back.
1133 * This callback is defined to prevent fallback to
1134 * __set_page_dirty_buffers() in set_page_dirty().
1138 EXPORT_SYMBOL_GPL(noop_set_page_dirty
);
1140 void noop_invalidatepage(struct page
*page
, unsigned int offset
,
1141 unsigned int length
)
1144 * There is no page cache to invalidate in the dax case, however
1145 * we need this callback defined to prevent falling back to
1146 * block_invalidatepage() in do_invalidatepage().
1149 EXPORT_SYMBOL_GPL(noop_invalidatepage
);
1151 ssize_t
noop_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
)
1154 * iomap based filesystems support direct I/O without need for
1155 * this callback. However, it still needs to be set in
1156 * inode->a_ops so that open/fcntl know that direct I/O is
1157 * generally supported.
1161 EXPORT_SYMBOL_GPL(noop_direct_IO
);
1163 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1164 void kfree_link(void *p
)
1168 EXPORT_SYMBOL(kfree_link
);
1171 * nop .set_page_dirty method so that people can use .page_mkwrite on
1174 static int anon_set_page_dirty(struct page
*page
)
1180 * A single inode exists for all anon_inode files. Contrary to pipes,
1181 * anon_inode inodes have no associated per-instance data, so we need
1182 * only allocate one of them.
1184 struct inode
*alloc_anon_inode(struct super_block
*s
)
1186 static const struct address_space_operations anon_aops
= {
1187 .set_page_dirty
= anon_set_page_dirty
,
1189 struct inode
*inode
= new_inode_pseudo(s
);
1192 return ERR_PTR(-ENOMEM
);
1194 inode
->i_ino
= get_next_ino();
1195 inode
->i_mapping
->a_ops
= &anon_aops
;
1198 * Mark the inode dirty from the very beginning,
1199 * that way it will never be moved to the dirty
1200 * list because mark_inode_dirty() will think
1201 * that it already _is_ on the dirty list.
1203 inode
->i_state
= I_DIRTY
;
1204 inode
->i_mode
= S_IRUSR
| S_IWUSR
;
1205 inode
->i_uid
= current_fsuid();
1206 inode
->i_gid
= current_fsgid();
1207 inode
->i_flags
|= S_PRIVATE
;
1208 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
1211 EXPORT_SYMBOL(alloc_anon_inode
);
1214 * simple_nosetlease - generic helper for prohibiting leases
1215 * @filp: file pointer
1216 * @arg: type of lease to obtain
1217 * @flp: new lease supplied for insertion
1218 * @priv: private data for lm_setup operation
1220 * Generic helper for filesystems that do not wish to allow leases to be set.
1221 * All arguments are ignored and it just returns -EINVAL.
1224 simple_nosetlease(struct file
*filp
, long arg
, struct file_lock
**flp
,
1229 EXPORT_SYMBOL(simple_nosetlease
);
1232 * simple_get_link - generic helper to get the target of "fast" symlinks
1233 * @dentry: not used here
1234 * @inode: the symlink inode
1235 * @done: not used here
1237 * Generic helper for filesystems to use for symlink inodes where a pointer to
1238 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1239 * since as an optimization the path lookup code uses any non-NULL ->i_link
1240 * directly, without calling ->get_link(). But ->get_link() still must be set,
1241 * to mark the inode_operations as being for a symlink.
1243 * Return: the symlink target
1245 const char *simple_get_link(struct dentry
*dentry
, struct inode
*inode
,
1246 struct delayed_call
*done
)
1248 return inode
->i_link
;
1250 EXPORT_SYMBOL(simple_get_link
);
1252 const struct inode_operations simple_symlink_inode_operations
= {
1253 .get_link
= simple_get_link
,
1255 EXPORT_SYMBOL(simple_symlink_inode_operations
);
1258 * Operations for a permanently empty directory.
1260 static struct dentry
*empty_dir_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
1262 return ERR_PTR(-ENOENT
);
1265 static int empty_dir_getattr(const struct path
*path
, struct kstat
*stat
,
1266 u32 request_mask
, unsigned int query_flags
)
1268 struct inode
*inode
= d_inode(path
->dentry
);
1269 generic_fillattr(inode
, stat
);
1273 static int empty_dir_setattr(struct dentry
*dentry
, struct iattr
*attr
)
1278 static ssize_t
empty_dir_listxattr(struct dentry
*dentry
, char *list
, size_t size
)
1283 static const struct inode_operations empty_dir_inode_operations
= {
1284 .lookup
= empty_dir_lookup
,
1285 .permission
= generic_permission
,
1286 .setattr
= empty_dir_setattr
,
1287 .getattr
= empty_dir_getattr
,
1288 .listxattr
= empty_dir_listxattr
,
1291 static loff_t
empty_dir_llseek(struct file
*file
, loff_t offset
, int whence
)
1293 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1294 return generic_file_llseek_size(file
, offset
, whence
, 2, 2);
1297 static int empty_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
1299 dir_emit_dots(file
, ctx
);
1303 static const struct file_operations empty_dir_operations
= {
1304 .llseek
= empty_dir_llseek
,
1305 .read
= generic_read_dir
,
1306 .iterate_shared
= empty_dir_readdir
,
1307 .fsync
= noop_fsync
,
1311 void make_empty_dir_inode(struct inode
*inode
)
1313 set_nlink(inode
, 2);
1314 inode
->i_mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
1315 inode
->i_uid
= GLOBAL_ROOT_UID
;
1316 inode
->i_gid
= GLOBAL_ROOT_GID
;
1319 inode
->i_blkbits
= PAGE_SHIFT
;
1320 inode
->i_blocks
= 0;
1322 inode
->i_op
= &empty_dir_inode_operations
;
1323 inode
->i_opflags
&= ~IOP_XATTR
;
1324 inode
->i_fop
= &empty_dir_operations
;
1327 bool is_empty_dir_inode(struct inode
*inode
)
1329 return (inode
->i_fop
== &empty_dir_operations
) &&
1330 (inode
->i_op
== &empty_dir_inode_operations
);