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/iversion.h>
19 #include <linux/writeback.h>
20 #include <linux/buffer_head.h> /* sync_mapping_buffers */
21 #include <linux/fs_context.h>
22 #include <linux/pseudo_fs.h>
23 #include <linux/fsnotify.h>
24 #include <linux/unicode.h>
25 #include <linux/fscrypt.h>
27 #include <linux/uaccess.h>
31 int simple_getattr(struct user_namespace
*mnt_userns
, const struct path
*path
,
32 struct kstat
*stat
, u32 request_mask
,
33 unsigned int query_flags
)
35 struct inode
*inode
= d_inode(path
->dentry
);
36 generic_fillattr(&init_user_ns
, inode
, stat
);
37 stat
->blocks
= inode
->i_mapping
->nrpages
<< (PAGE_SHIFT
- 9);
40 EXPORT_SYMBOL(simple_getattr
);
42 int simple_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
44 buf
->f_type
= dentry
->d_sb
->s_magic
;
45 buf
->f_bsize
= PAGE_SIZE
;
46 buf
->f_namelen
= NAME_MAX
;
49 EXPORT_SYMBOL(simple_statfs
);
52 * Retaining negative dentries for an in-memory filesystem just wastes
53 * memory and lookup time: arrange for them to be deleted immediately.
55 int always_delete_dentry(const struct dentry
*dentry
)
59 EXPORT_SYMBOL(always_delete_dentry
);
61 const struct dentry_operations simple_dentry_operations
= {
62 .d_delete
= always_delete_dentry
,
64 EXPORT_SYMBOL(simple_dentry_operations
);
67 * Lookup the data. This is trivial - if the dentry didn't already
68 * exist, we know it is negative. Set d_op to delete negative dentries.
70 struct dentry
*simple_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
72 if (dentry
->d_name
.len
> NAME_MAX
)
73 return ERR_PTR(-ENAMETOOLONG
);
74 if (!dentry
->d_sb
->s_d_op
)
75 d_set_d_op(dentry
, &simple_dentry_operations
);
79 EXPORT_SYMBOL(simple_lookup
);
81 int dcache_dir_open(struct inode
*inode
, struct file
*file
)
83 file
->private_data
= d_alloc_cursor(file
->f_path
.dentry
);
85 return file
->private_data
? 0 : -ENOMEM
;
87 EXPORT_SYMBOL(dcache_dir_open
);
89 int dcache_dir_close(struct inode
*inode
, struct file
*file
)
91 dput(file
->private_data
);
94 EXPORT_SYMBOL(dcache_dir_close
);
96 /* parent is locked at least shared */
98 * Returns an element of siblings' list.
99 * We are looking for <count>th positive after <p>; if
100 * found, dentry is grabbed and returned to caller.
101 * If no such element exists, NULL is returned.
103 static struct dentry
*scan_positives(struct dentry
*cursor
,
108 struct dentry
*dentry
= cursor
->d_parent
, *found
= NULL
;
110 spin_lock(&dentry
->d_lock
);
111 while ((p
= p
->next
) != &dentry
->d_subdirs
) {
112 struct dentry
*d
= list_entry(p
, struct dentry
, d_child
);
113 // we must at least skip cursors, to avoid livelocks
114 if (d
->d_flags
& DCACHE_DENTRY_CURSOR
)
116 if (simple_positive(d
) && !--count
) {
117 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
118 if (simple_positive(d
))
119 found
= dget_dlock(d
);
120 spin_unlock(&d
->d_lock
);
125 if (need_resched()) {
126 list_move(&cursor
->d_child
, p
);
127 p
= &cursor
->d_child
;
128 spin_unlock(&dentry
->d_lock
);
130 spin_lock(&dentry
->d_lock
);
133 spin_unlock(&dentry
->d_lock
);
138 loff_t
dcache_dir_lseek(struct file
*file
, loff_t offset
, int whence
)
140 struct dentry
*dentry
= file
->f_path
.dentry
;
143 offset
+= file
->f_pos
;
152 if (offset
!= file
->f_pos
) {
153 struct dentry
*cursor
= file
->private_data
;
154 struct dentry
*to
= NULL
;
156 inode_lock_shared(dentry
->d_inode
);
159 to
= scan_positives(cursor
, &dentry
->d_subdirs
,
161 spin_lock(&dentry
->d_lock
);
163 list_move(&cursor
->d_child
, &to
->d_child
);
165 list_del_init(&cursor
->d_child
);
166 spin_unlock(&dentry
->d_lock
);
169 file
->f_pos
= offset
;
171 inode_unlock_shared(dentry
->d_inode
);
175 EXPORT_SYMBOL(dcache_dir_lseek
);
177 /* Relationship between i_mode and the DT_xxx types */
178 static inline unsigned char dt_type(struct inode
*inode
)
180 return (inode
->i_mode
>> 12) & 15;
184 * Directory is locked and all positive dentries in it are safe, since
185 * for ramfs-type trees they can't go away without unlink() or rmdir(),
186 * both impossible due to the lock on directory.
189 int dcache_readdir(struct file
*file
, struct dir_context
*ctx
)
191 struct dentry
*dentry
= file
->f_path
.dentry
;
192 struct dentry
*cursor
= file
->private_data
;
193 struct list_head
*anchor
= &dentry
->d_subdirs
;
194 struct dentry
*next
= NULL
;
197 if (!dir_emit_dots(file
, ctx
))
202 else if (!list_empty(&cursor
->d_child
))
203 p
= &cursor
->d_child
;
207 while ((next
= scan_positives(cursor
, p
, 1, next
)) != NULL
) {
208 if (!dir_emit(ctx
, next
->d_name
.name
, next
->d_name
.len
,
209 d_inode(next
)->i_ino
, dt_type(d_inode(next
))))
214 spin_lock(&dentry
->d_lock
);
216 list_move_tail(&cursor
->d_child
, &next
->d_child
);
218 list_del_init(&cursor
->d_child
);
219 spin_unlock(&dentry
->d_lock
);
224 EXPORT_SYMBOL(dcache_readdir
);
226 ssize_t
generic_read_dir(struct file
*filp
, char __user
*buf
, size_t siz
, loff_t
*ppos
)
230 EXPORT_SYMBOL(generic_read_dir
);
232 const struct file_operations simple_dir_operations
= {
233 .open
= dcache_dir_open
,
234 .release
= dcache_dir_close
,
235 .llseek
= dcache_dir_lseek
,
236 .read
= generic_read_dir
,
237 .iterate_shared
= dcache_readdir
,
240 EXPORT_SYMBOL(simple_dir_operations
);
242 const struct inode_operations simple_dir_inode_operations
= {
243 .lookup
= simple_lookup
,
245 EXPORT_SYMBOL(simple_dir_inode_operations
);
247 static struct dentry
*find_next_child(struct dentry
*parent
, struct dentry
*prev
)
249 struct dentry
*child
= NULL
;
250 struct list_head
*p
= prev
? &prev
->d_child
: &parent
->d_subdirs
;
252 spin_lock(&parent
->d_lock
);
253 while ((p
= p
->next
) != &parent
->d_subdirs
) {
254 struct dentry
*d
= container_of(p
, struct dentry
, d_child
);
255 if (simple_positive(d
)) {
256 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
257 if (simple_positive(d
))
258 child
= dget_dlock(d
);
259 spin_unlock(&d
->d_lock
);
264 spin_unlock(&parent
->d_lock
);
269 void simple_recursive_removal(struct dentry
*dentry
,
270 void (*callback
)(struct dentry
*))
272 struct dentry
*this = dget(dentry
);
274 struct dentry
*victim
= NULL
, *child
;
275 struct inode
*inode
= this->d_inode
;
279 inode
->i_flags
|= S_DEAD
;
280 while ((child
= find_next_child(this, victim
)) == NULL
) {
282 // update metadata while it's still locked
283 inode
->i_ctime
= current_time(inode
);
287 this = this->d_parent
;
288 inode
= this->d_inode
;
290 if (simple_positive(victim
)) {
291 d_invalidate(victim
); // avoid lost mounts
292 if (d_is_dir(victim
))
293 fsnotify_rmdir(inode
, victim
);
295 fsnotify_unlink(inode
, victim
);
298 dput(victim
); // unpin it
300 if (victim
== dentry
) {
301 inode
->i_ctime
= inode
->i_mtime
=
303 if (d_is_dir(dentry
))
314 EXPORT_SYMBOL(simple_recursive_removal
);
316 static const struct super_operations simple_super_operations
= {
317 .statfs
= simple_statfs
,
320 static int pseudo_fs_fill_super(struct super_block
*s
, struct fs_context
*fc
)
322 struct pseudo_fs_context
*ctx
= fc
->fs_private
;
325 s
->s_maxbytes
= MAX_LFS_FILESIZE
;
326 s
->s_blocksize
= PAGE_SIZE
;
327 s
->s_blocksize_bits
= PAGE_SHIFT
;
328 s
->s_magic
= ctx
->magic
;
329 s
->s_op
= ctx
->ops
?: &simple_super_operations
;
330 s
->s_xattr
= ctx
->xattr
;
337 * since this is the first inode, make it number 1. New inodes created
338 * after this must take care not to collide with it (by passing
339 * max_reserved of 1 to iunique).
342 root
->i_mode
= S_IFDIR
| S_IRUSR
| S_IWUSR
;
343 root
->i_atime
= root
->i_mtime
= root
->i_ctime
= current_time(root
);
344 s
->s_root
= d_make_root(root
);
347 s
->s_d_op
= ctx
->dops
;
351 static int pseudo_fs_get_tree(struct fs_context
*fc
)
353 return get_tree_nodev(fc
, pseudo_fs_fill_super
);
356 static void pseudo_fs_free(struct fs_context
*fc
)
358 kfree(fc
->fs_private
);
361 static const struct fs_context_operations pseudo_fs_context_ops
= {
362 .free
= pseudo_fs_free
,
363 .get_tree
= pseudo_fs_get_tree
,
367 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
368 * will never be mountable)
370 struct pseudo_fs_context
*init_pseudo(struct fs_context
*fc
,
373 struct pseudo_fs_context
*ctx
;
375 ctx
= kzalloc(sizeof(struct pseudo_fs_context
), GFP_KERNEL
);
378 fc
->fs_private
= ctx
;
379 fc
->ops
= &pseudo_fs_context_ops
;
380 fc
->sb_flags
|= SB_NOUSER
;
385 EXPORT_SYMBOL(init_pseudo
);
387 int simple_open(struct inode
*inode
, struct file
*file
)
389 if (inode
->i_private
)
390 file
->private_data
= inode
->i_private
;
393 EXPORT_SYMBOL(simple_open
);
395 int simple_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*dentry
)
397 struct inode
*inode
= d_inode(old_dentry
);
399 inode
->i_ctime
= dir
->i_ctime
= dir
->i_mtime
= current_time(inode
);
403 d_instantiate(dentry
, inode
);
406 EXPORT_SYMBOL(simple_link
);
408 int simple_empty(struct dentry
*dentry
)
410 struct dentry
*child
;
413 spin_lock(&dentry
->d_lock
);
414 list_for_each_entry(child
, &dentry
->d_subdirs
, d_child
) {
415 spin_lock_nested(&child
->d_lock
, DENTRY_D_LOCK_NESTED
);
416 if (simple_positive(child
)) {
417 spin_unlock(&child
->d_lock
);
420 spin_unlock(&child
->d_lock
);
424 spin_unlock(&dentry
->d_lock
);
427 EXPORT_SYMBOL(simple_empty
);
429 int simple_unlink(struct inode
*dir
, struct dentry
*dentry
)
431 struct inode
*inode
= d_inode(dentry
);
433 inode
->i_ctime
= dir
->i_ctime
= dir
->i_mtime
= current_time(inode
);
438 EXPORT_SYMBOL(simple_unlink
);
440 int simple_rmdir(struct inode
*dir
, struct dentry
*dentry
)
442 if (!simple_empty(dentry
))
445 drop_nlink(d_inode(dentry
));
446 simple_unlink(dir
, dentry
);
450 EXPORT_SYMBOL(simple_rmdir
);
452 int simple_rename_exchange(struct inode
*old_dir
, struct dentry
*old_dentry
,
453 struct inode
*new_dir
, struct dentry
*new_dentry
)
455 bool old_is_dir
= d_is_dir(old_dentry
);
456 bool new_is_dir
= d_is_dir(new_dentry
);
458 if (old_dir
!= new_dir
&& old_is_dir
!= new_is_dir
) {
467 old_dir
->i_ctime
= old_dir
->i_mtime
=
468 new_dir
->i_ctime
= new_dir
->i_mtime
=
469 d_inode(old_dentry
)->i_ctime
=
470 d_inode(new_dentry
)->i_ctime
= current_time(old_dir
);
474 EXPORT_SYMBOL_GPL(simple_rename_exchange
);
476 int simple_rename(struct user_namespace
*mnt_userns
, struct inode
*old_dir
,
477 struct dentry
*old_dentry
, struct inode
*new_dir
,
478 struct dentry
*new_dentry
, unsigned int flags
)
480 struct inode
*inode
= d_inode(old_dentry
);
481 int they_are_dirs
= d_is_dir(old_dentry
);
483 if (flags
& ~(RENAME_NOREPLACE
| RENAME_EXCHANGE
))
486 if (flags
& RENAME_EXCHANGE
)
487 return simple_rename_exchange(old_dir
, old_dentry
, new_dir
, new_dentry
);
489 if (!simple_empty(new_dentry
))
492 if (d_really_is_positive(new_dentry
)) {
493 simple_unlink(new_dir
, new_dentry
);
495 drop_nlink(d_inode(new_dentry
));
498 } else if (they_are_dirs
) {
503 old_dir
->i_ctime
= old_dir
->i_mtime
= new_dir
->i_ctime
=
504 new_dir
->i_mtime
= inode
->i_ctime
= current_time(old_dir
);
508 EXPORT_SYMBOL(simple_rename
);
511 * simple_setattr - setattr for simple filesystem
512 * @mnt_userns: user namespace of the target mount
514 * @iattr: iattr structure
516 * Returns 0 on success, -error on failure.
518 * simple_setattr is a simple ->setattr implementation without a proper
519 * implementation of size changes.
521 * It can either be used for in-memory filesystems or special files
522 * on simple regular filesystems. Anything that needs to change on-disk
523 * or wire state on size changes needs its own setattr method.
525 int simple_setattr(struct user_namespace
*mnt_userns
, struct dentry
*dentry
,
528 struct inode
*inode
= d_inode(dentry
);
531 error
= setattr_prepare(mnt_userns
, dentry
, iattr
);
535 if (iattr
->ia_valid
& ATTR_SIZE
)
536 truncate_setsize(inode
, iattr
->ia_size
);
537 setattr_copy(mnt_userns
, inode
, iattr
);
538 mark_inode_dirty(inode
);
541 EXPORT_SYMBOL(simple_setattr
);
543 static int simple_read_folio(struct file
*file
, struct folio
*folio
)
545 folio_zero_range(folio
, 0, folio_size(folio
));
546 flush_dcache_folio(folio
);
547 folio_mark_uptodate(folio
);
552 int simple_write_begin(struct file
*file
, struct address_space
*mapping
,
553 loff_t pos
, unsigned len
,
554 struct page
**pagep
, void **fsdata
)
559 index
= pos
>> PAGE_SHIFT
;
561 page
= grab_cache_page_write_begin(mapping
, index
);
567 if (!PageUptodate(page
) && (len
!= PAGE_SIZE
)) {
568 unsigned from
= pos
& (PAGE_SIZE
- 1);
570 zero_user_segments(page
, 0, from
, from
+ len
, PAGE_SIZE
);
574 EXPORT_SYMBOL(simple_write_begin
);
577 * simple_write_end - .write_end helper for non-block-device FSes
578 * @file: See .write_end of address_space_operations
586 * simple_write_end does the minimum needed for updating a page after writing is
587 * done. It has the same API signature as the .write_end of
588 * address_space_operations vector. So it can just be set onto .write_end for
589 * FSes that don't need any other processing. i_mutex is assumed to be held.
590 * Block based filesystems should use generic_write_end().
591 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
592 * is not called, so a filesystem that actually does store data in .write_inode
593 * should extend on what's done here with a call to mark_inode_dirty() in the
594 * case that i_size has changed.
596 * Use *ONLY* with simple_read_folio()
598 static int simple_write_end(struct file
*file
, struct address_space
*mapping
,
599 loff_t pos
, unsigned len
, unsigned copied
,
600 struct page
*page
, void *fsdata
)
602 struct inode
*inode
= page
->mapping
->host
;
603 loff_t last_pos
= pos
+ copied
;
605 /* zero the stale part of the page if we did a short copy */
606 if (!PageUptodate(page
)) {
608 unsigned from
= pos
& (PAGE_SIZE
- 1);
610 zero_user(page
, from
+ copied
, len
- copied
);
612 SetPageUptodate(page
);
615 * No need to use i_size_read() here, the i_size
616 * cannot change under us because we hold the i_mutex.
618 if (last_pos
> inode
->i_size
)
619 i_size_write(inode
, last_pos
);
621 set_page_dirty(page
);
629 * Provides ramfs-style behavior: data in the pagecache, but no writeback.
631 const struct address_space_operations ram_aops
= {
632 .read_folio
= simple_read_folio
,
633 .write_begin
= simple_write_begin
,
634 .write_end
= simple_write_end
,
635 .dirty_folio
= noop_dirty_folio
,
637 EXPORT_SYMBOL(ram_aops
);
640 * the inodes created here are not hashed. If you use iunique to generate
641 * unique inode values later for this filesystem, then you must take care
642 * to pass it an appropriate max_reserved value to avoid collisions.
644 int simple_fill_super(struct super_block
*s
, unsigned long magic
,
645 const struct tree_descr
*files
)
649 struct dentry
*dentry
;
652 s
->s_blocksize
= PAGE_SIZE
;
653 s
->s_blocksize_bits
= PAGE_SHIFT
;
655 s
->s_op
= &simple_super_operations
;
658 inode
= new_inode(s
);
662 * because the root inode is 1, the files array must not contain an
666 inode
->i_mode
= S_IFDIR
| 0755;
667 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
668 inode
->i_op
= &simple_dir_inode_operations
;
669 inode
->i_fop
= &simple_dir_operations
;
671 root
= d_make_root(inode
);
674 for (i
= 0; !files
->name
|| files
->name
[0]; i
++, files
++) {
678 /* warn if it tries to conflict with the root inode */
679 if (unlikely(i
== 1))
680 printk(KERN_WARNING
"%s: %s passed in a files array"
681 "with an index of 1!\n", __func__
,
684 dentry
= d_alloc_name(root
, files
->name
);
687 inode
= new_inode(s
);
692 inode
->i_mode
= S_IFREG
| files
->mode
;
693 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
694 inode
->i_fop
= files
->ops
;
696 d_add(dentry
, inode
);
702 shrink_dcache_parent(root
);
706 EXPORT_SYMBOL(simple_fill_super
);
708 static DEFINE_SPINLOCK(pin_fs_lock
);
710 int simple_pin_fs(struct file_system_type
*type
, struct vfsmount
**mount
, int *count
)
712 struct vfsmount
*mnt
= NULL
;
713 spin_lock(&pin_fs_lock
);
714 if (unlikely(!*mount
)) {
715 spin_unlock(&pin_fs_lock
);
716 mnt
= vfs_kern_mount(type
, SB_KERNMOUNT
, type
->name
, NULL
);
719 spin_lock(&pin_fs_lock
);
725 spin_unlock(&pin_fs_lock
);
729 EXPORT_SYMBOL(simple_pin_fs
);
731 void simple_release_fs(struct vfsmount
**mount
, int *count
)
733 struct vfsmount
*mnt
;
734 spin_lock(&pin_fs_lock
);
738 spin_unlock(&pin_fs_lock
);
741 EXPORT_SYMBOL(simple_release_fs
);
744 * simple_read_from_buffer - copy data from the buffer to user space
745 * @to: the user space buffer to read to
746 * @count: the maximum number of bytes to read
747 * @ppos: the current position in the buffer
748 * @from: the buffer to read from
749 * @available: the size of the buffer
751 * The simple_read_from_buffer() function reads up to @count bytes from the
752 * buffer @from at offset @ppos into the user space address starting at @to.
754 * On success, the number of bytes read is returned and the offset @ppos is
755 * advanced by this number, or negative value is returned on error.
757 ssize_t
simple_read_from_buffer(void __user
*to
, size_t count
, loff_t
*ppos
,
758 const void *from
, size_t available
)
765 if (pos
>= available
|| !count
)
767 if (count
> available
- pos
)
768 count
= available
- pos
;
769 ret
= copy_to_user(to
, from
+ pos
, count
);
776 EXPORT_SYMBOL(simple_read_from_buffer
);
779 * simple_write_to_buffer - copy data from user space to the buffer
780 * @to: the buffer to write to
781 * @available: the size of the buffer
782 * @ppos: the current position in the buffer
783 * @from: the user space buffer to read from
784 * @count: the maximum number of bytes to read
786 * The simple_write_to_buffer() function reads up to @count bytes from the user
787 * space address starting at @from into the buffer @to at offset @ppos.
789 * On success, the number of bytes written is returned and the offset @ppos is
790 * advanced by this number, or negative value is returned on error.
792 ssize_t
simple_write_to_buffer(void *to
, size_t available
, loff_t
*ppos
,
793 const void __user
*from
, size_t count
)
800 if (pos
>= available
|| !count
)
802 if (count
> available
- pos
)
803 count
= available
- pos
;
804 res
= copy_from_user(to
+ pos
, from
, count
);
811 EXPORT_SYMBOL(simple_write_to_buffer
);
814 * memory_read_from_buffer - copy data from the buffer
815 * @to: the kernel space buffer to read to
816 * @count: the maximum number of bytes to read
817 * @ppos: the current position in the buffer
818 * @from: the buffer to read from
819 * @available: the size of the buffer
821 * The memory_read_from_buffer() function reads up to @count bytes from the
822 * buffer @from at offset @ppos into the kernel space address starting at @to.
824 * On success, the number of bytes read is returned and the offset @ppos is
825 * advanced by this number, or negative value is returned on error.
827 ssize_t
memory_read_from_buffer(void *to
, size_t count
, loff_t
*ppos
,
828 const void *from
, size_t available
)
834 if (pos
>= available
)
836 if (count
> available
- pos
)
837 count
= available
- pos
;
838 memcpy(to
, from
+ pos
, count
);
843 EXPORT_SYMBOL(memory_read_from_buffer
);
846 * Transaction based IO.
847 * The file expects a single write which triggers the transaction, and then
848 * possibly a read which collects the result - which is stored in a
852 void simple_transaction_set(struct file
*file
, size_t n
)
854 struct simple_transaction_argresp
*ar
= file
->private_data
;
856 BUG_ON(n
> SIMPLE_TRANSACTION_LIMIT
);
859 * The barrier ensures that ar->size will really remain zero until
860 * ar->data is ready for reading.
865 EXPORT_SYMBOL(simple_transaction_set
);
867 char *simple_transaction_get(struct file
*file
, const char __user
*buf
, size_t size
)
869 struct simple_transaction_argresp
*ar
;
870 static DEFINE_SPINLOCK(simple_transaction_lock
);
872 if (size
> SIMPLE_TRANSACTION_LIMIT
- 1)
873 return ERR_PTR(-EFBIG
);
875 ar
= (struct simple_transaction_argresp
*)get_zeroed_page(GFP_KERNEL
);
877 return ERR_PTR(-ENOMEM
);
879 spin_lock(&simple_transaction_lock
);
881 /* only one write allowed per open */
882 if (file
->private_data
) {
883 spin_unlock(&simple_transaction_lock
);
884 free_page((unsigned long)ar
);
885 return ERR_PTR(-EBUSY
);
888 file
->private_data
= ar
;
890 spin_unlock(&simple_transaction_lock
);
892 if (copy_from_user(ar
->data
, buf
, size
))
893 return ERR_PTR(-EFAULT
);
897 EXPORT_SYMBOL(simple_transaction_get
);
899 ssize_t
simple_transaction_read(struct file
*file
, char __user
*buf
, size_t size
, loff_t
*pos
)
901 struct simple_transaction_argresp
*ar
= file
->private_data
;
905 return simple_read_from_buffer(buf
, size
, pos
, ar
->data
, ar
->size
);
907 EXPORT_SYMBOL(simple_transaction_read
);
909 int simple_transaction_release(struct inode
*inode
, struct file
*file
)
911 free_page((unsigned long)file
->private_data
);
914 EXPORT_SYMBOL(simple_transaction_release
);
916 /* Simple attribute files */
919 int (*get
)(void *, u64
*);
920 int (*set
)(void *, u64
);
921 char get_buf
[24]; /* enough to store a u64 and "\n\0" */
924 const char *fmt
; /* format for read operation */
925 struct mutex mutex
; /* protects access to these buffers */
928 /* simple_attr_open is called by an actual attribute open file operation
929 * to set the attribute specific access operations. */
930 int simple_attr_open(struct inode
*inode
, struct file
*file
,
931 int (*get
)(void *, u64
*), int (*set
)(void *, u64
),
934 struct simple_attr
*attr
;
936 attr
= kzalloc(sizeof(*attr
), GFP_KERNEL
);
942 attr
->data
= inode
->i_private
;
944 mutex_init(&attr
->mutex
);
946 file
->private_data
= attr
;
948 return nonseekable_open(inode
, file
);
950 EXPORT_SYMBOL_GPL(simple_attr_open
);
952 int simple_attr_release(struct inode
*inode
, struct file
*file
)
954 kfree(file
->private_data
);
957 EXPORT_SYMBOL_GPL(simple_attr_release
); /* GPL-only? This? Really? */
959 /* read from the buffer that is filled with the get function */
960 ssize_t
simple_attr_read(struct file
*file
, char __user
*buf
,
961 size_t len
, loff_t
*ppos
)
963 struct simple_attr
*attr
;
967 attr
= file
->private_data
;
972 ret
= mutex_lock_interruptible(&attr
->mutex
);
976 if (*ppos
&& attr
->get_buf
[0]) {
978 size
= strlen(attr
->get_buf
);
982 ret
= attr
->get(attr
->data
, &val
);
986 size
= scnprintf(attr
->get_buf
, sizeof(attr
->get_buf
),
987 attr
->fmt
, (unsigned long long)val
);
990 ret
= simple_read_from_buffer(buf
, len
, ppos
, attr
->get_buf
, size
);
992 mutex_unlock(&attr
->mutex
);
995 EXPORT_SYMBOL_GPL(simple_attr_read
);
997 /* interpret the buffer as a number to call the set function with */
998 static ssize_t
simple_attr_write_xsigned(struct file
*file
, const char __user
*buf
,
999 size_t len
, loff_t
*ppos
, bool is_signed
)
1001 struct simple_attr
*attr
;
1002 unsigned long long val
;
1006 attr
= file
->private_data
;
1010 ret
= mutex_lock_interruptible(&attr
->mutex
);
1015 size
= min(sizeof(attr
->set_buf
) - 1, len
);
1016 if (copy_from_user(attr
->set_buf
, buf
, size
))
1019 attr
->set_buf
[size
] = '\0';
1021 ret
= kstrtoll(attr
->set_buf
, 0, &val
);
1023 ret
= kstrtoull(attr
->set_buf
, 0, &val
);
1026 ret
= attr
->set(attr
->data
, val
);
1028 ret
= len
; /* on success, claim we got the whole input */
1030 mutex_unlock(&attr
->mutex
);
1034 ssize_t
simple_attr_write(struct file
*file
, const char __user
*buf
,
1035 size_t len
, loff_t
*ppos
)
1037 return simple_attr_write_xsigned(file
, buf
, len
, ppos
, false);
1039 EXPORT_SYMBOL_GPL(simple_attr_write
);
1041 ssize_t
simple_attr_write_signed(struct file
*file
, const char __user
*buf
,
1042 size_t len
, loff_t
*ppos
)
1044 return simple_attr_write_xsigned(file
, buf
, len
, ppos
, true);
1046 EXPORT_SYMBOL_GPL(simple_attr_write_signed
);
1049 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1050 * @sb: filesystem to do the file handle conversion on
1051 * @fid: file handle to convert
1052 * @fh_len: length of the file handle in bytes
1053 * @fh_type: type of file handle
1054 * @get_inode: filesystem callback to retrieve inode
1056 * This function decodes @fid as long as it has one of the well-known
1057 * Linux filehandle types and calls @get_inode on it to retrieve the
1058 * inode for the object specified in the file handle.
1060 struct dentry
*generic_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1061 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1062 (struct super_block
*sb
, u64 ino
, u32 gen
))
1064 struct inode
*inode
= NULL
;
1070 case FILEID_INO32_GEN
:
1071 case FILEID_INO32_GEN_PARENT
:
1072 inode
= get_inode(sb
, fid
->i32
.ino
, fid
->i32
.gen
);
1076 return d_obtain_alias(inode
);
1078 EXPORT_SYMBOL_GPL(generic_fh_to_dentry
);
1081 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1082 * @sb: filesystem to do the file handle conversion on
1083 * @fid: file handle to convert
1084 * @fh_len: length of the file handle in bytes
1085 * @fh_type: type of file handle
1086 * @get_inode: filesystem callback to retrieve inode
1088 * This function decodes @fid as long as it has one of the well-known
1089 * Linux filehandle types and calls @get_inode on it to retrieve the
1090 * inode for the _parent_ object specified in the file handle if it
1091 * is specified in the file handle, or NULL otherwise.
1093 struct dentry
*generic_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1094 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1095 (struct super_block
*sb
, u64 ino
, u32 gen
))
1097 struct inode
*inode
= NULL
;
1103 case FILEID_INO32_GEN_PARENT
:
1104 inode
= get_inode(sb
, fid
->i32
.parent_ino
,
1105 (fh_len
> 3 ? fid
->i32
.parent_gen
: 0));
1109 return d_obtain_alias(inode
);
1111 EXPORT_SYMBOL_GPL(generic_fh_to_parent
);
1114 * __generic_file_fsync - generic fsync implementation for simple filesystems
1116 * @file: file to synchronize
1117 * @start: start offset in bytes
1118 * @end: end offset in bytes (inclusive)
1119 * @datasync: only synchronize essential metadata if true
1121 * This is a generic implementation of the fsync method for simple
1122 * filesystems which track all non-inode metadata in the buffers list
1123 * hanging off the address_space structure.
1125 int __generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1128 struct inode
*inode
= file
->f_mapping
->host
;
1132 err
= file_write_and_wait_range(file
, start
, end
);
1137 ret
= sync_mapping_buffers(inode
->i_mapping
);
1138 if (!(inode
->i_state
& I_DIRTY_ALL
))
1140 if (datasync
&& !(inode
->i_state
& I_DIRTY_DATASYNC
))
1143 err
= sync_inode_metadata(inode
, 1);
1148 inode_unlock(inode
);
1149 /* check and advance again to catch errors after syncing out buffers */
1150 err
= file_check_and_advance_wb_err(file
);
1155 EXPORT_SYMBOL(__generic_file_fsync
);
1158 * generic_file_fsync - generic fsync implementation for simple filesystems
1160 * @file: file to synchronize
1161 * @start: start offset in bytes
1162 * @end: end offset in bytes (inclusive)
1163 * @datasync: only synchronize essential metadata if true
1167 int generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1170 struct inode
*inode
= file
->f_mapping
->host
;
1173 err
= __generic_file_fsync(file
, start
, end
, datasync
);
1176 return blkdev_issue_flush(inode
->i_sb
->s_bdev
);
1178 EXPORT_SYMBOL(generic_file_fsync
);
1181 * generic_check_addressable - Check addressability of file system
1182 * @blocksize_bits: log of file system block size
1183 * @num_blocks: number of blocks in file system
1185 * Determine whether a file system with @num_blocks blocks (and a
1186 * block size of 2**@blocksize_bits) is addressable by the sector_t
1187 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1189 int generic_check_addressable(unsigned blocksize_bits
, u64 num_blocks
)
1191 u64 last_fs_block
= num_blocks
- 1;
1193 last_fs_block
>> (PAGE_SHIFT
- blocksize_bits
);
1195 if (unlikely(num_blocks
== 0))
1198 if ((blocksize_bits
< 9) || (blocksize_bits
> PAGE_SHIFT
))
1201 if ((last_fs_block
> (sector_t
)(~0ULL) >> (blocksize_bits
- 9)) ||
1202 (last_fs_page
> (pgoff_t
)(~0ULL))) {
1207 EXPORT_SYMBOL(generic_check_addressable
);
1210 * No-op implementation of ->fsync for in-memory filesystems.
1212 int noop_fsync(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
1216 EXPORT_SYMBOL(noop_fsync
);
1218 ssize_t
noop_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
)
1221 * iomap based filesystems support direct I/O without need for
1222 * this callback. However, it still needs to be set in
1223 * inode->a_ops so that open/fcntl know that direct I/O is
1224 * generally supported.
1228 EXPORT_SYMBOL_GPL(noop_direct_IO
);
1230 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1231 void kfree_link(void *p
)
1235 EXPORT_SYMBOL(kfree_link
);
1237 struct inode
*alloc_anon_inode(struct super_block
*s
)
1239 static const struct address_space_operations anon_aops
= {
1240 .dirty_folio
= noop_dirty_folio
,
1242 struct inode
*inode
= new_inode_pseudo(s
);
1245 return ERR_PTR(-ENOMEM
);
1247 inode
->i_ino
= get_next_ino();
1248 inode
->i_mapping
->a_ops
= &anon_aops
;
1251 * Mark the inode dirty from the very beginning,
1252 * that way it will never be moved to the dirty
1253 * list because mark_inode_dirty() will think
1254 * that it already _is_ on the dirty list.
1256 inode
->i_state
= I_DIRTY
;
1257 inode
->i_mode
= S_IRUSR
| S_IWUSR
;
1258 inode
->i_uid
= current_fsuid();
1259 inode
->i_gid
= current_fsgid();
1260 inode
->i_flags
|= S_PRIVATE
;
1261 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
1264 EXPORT_SYMBOL(alloc_anon_inode
);
1267 * simple_nosetlease - generic helper for prohibiting leases
1268 * @filp: file pointer
1269 * @arg: type of lease to obtain
1270 * @flp: new lease supplied for insertion
1271 * @priv: private data for lm_setup operation
1273 * Generic helper for filesystems that do not wish to allow leases to be set.
1274 * All arguments are ignored and it just returns -EINVAL.
1277 simple_nosetlease(struct file
*filp
, long arg
, struct file_lock
**flp
,
1282 EXPORT_SYMBOL(simple_nosetlease
);
1285 * simple_get_link - generic helper to get the target of "fast" symlinks
1286 * @dentry: not used here
1287 * @inode: the symlink inode
1288 * @done: not used here
1290 * Generic helper for filesystems to use for symlink inodes where a pointer to
1291 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1292 * since as an optimization the path lookup code uses any non-NULL ->i_link
1293 * directly, without calling ->get_link(). But ->get_link() still must be set,
1294 * to mark the inode_operations as being for a symlink.
1296 * Return: the symlink target
1298 const char *simple_get_link(struct dentry
*dentry
, struct inode
*inode
,
1299 struct delayed_call
*done
)
1301 return inode
->i_link
;
1303 EXPORT_SYMBOL(simple_get_link
);
1305 const struct inode_operations simple_symlink_inode_operations
= {
1306 .get_link
= simple_get_link
,
1308 EXPORT_SYMBOL(simple_symlink_inode_operations
);
1311 * Operations for a permanently empty directory.
1313 static struct dentry
*empty_dir_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
1315 return ERR_PTR(-ENOENT
);
1318 static int empty_dir_getattr(struct user_namespace
*mnt_userns
,
1319 const struct path
*path
, struct kstat
*stat
,
1320 u32 request_mask
, unsigned int query_flags
)
1322 struct inode
*inode
= d_inode(path
->dentry
);
1323 generic_fillattr(&init_user_ns
, inode
, stat
);
1327 static int empty_dir_setattr(struct user_namespace
*mnt_userns
,
1328 struct dentry
*dentry
, struct iattr
*attr
)
1333 static ssize_t
empty_dir_listxattr(struct dentry
*dentry
, char *list
, size_t size
)
1338 static const struct inode_operations empty_dir_inode_operations
= {
1339 .lookup
= empty_dir_lookup
,
1340 .permission
= generic_permission
,
1341 .setattr
= empty_dir_setattr
,
1342 .getattr
= empty_dir_getattr
,
1343 .listxattr
= empty_dir_listxattr
,
1346 static loff_t
empty_dir_llseek(struct file
*file
, loff_t offset
, int whence
)
1348 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1349 return generic_file_llseek_size(file
, offset
, whence
, 2, 2);
1352 static int empty_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
1354 dir_emit_dots(file
, ctx
);
1358 static const struct file_operations empty_dir_operations
= {
1359 .llseek
= empty_dir_llseek
,
1360 .read
= generic_read_dir
,
1361 .iterate_shared
= empty_dir_readdir
,
1362 .fsync
= noop_fsync
,
1366 void make_empty_dir_inode(struct inode
*inode
)
1368 set_nlink(inode
, 2);
1369 inode
->i_mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
1370 inode
->i_uid
= GLOBAL_ROOT_UID
;
1371 inode
->i_gid
= GLOBAL_ROOT_GID
;
1374 inode
->i_blkbits
= PAGE_SHIFT
;
1375 inode
->i_blocks
= 0;
1377 inode
->i_op
= &empty_dir_inode_operations
;
1378 inode
->i_opflags
&= ~IOP_XATTR
;
1379 inode
->i_fop
= &empty_dir_operations
;
1382 bool is_empty_dir_inode(struct inode
*inode
)
1384 return (inode
->i_fop
== &empty_dir_operations
) &&
1385 (inode
->i_op
== &empty_dir_inode_operations
);
1388 #if IS_ENABLED(CONFIG_UNICODE)
1390 * Determine if the name of a dentry should be casefolded.
1392 * Return: if names will need casefolding
1394 static bool needs_casefold(const struct inode
*dir
)
1396 return IS_CASEFOLDED(dir
) && dir
->i_sb
->s_encoding
;
1400 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1401 * @dentry: dentry whose name we are checking against
1402 * @len: len of name of dentry
1403 * @str: str pointer to name of dentry
1404 * @name: Name to compare against
1406 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1408 static int generic_ci_d_compare(const struct dentry
*dentry
, unsigned int len
,
1409 const char *str
, const struct qstr
*name
)
1411 const struct dentry
*parent
= READ_ONCE(dentry
->d_parent
);
1412 const struct inode
*dir
= READ_ONCE(parent
->d_inode
);
1413 const struct super_block
*sb
= dentry
->d_sb
;
1414 const struct unicode_map
*um
= sb
->s_encoding
;
1415 struct qstr qstr
= QSTR_INIT(str
, len
);
1416 char strbuf
[DNAME_INLINE_LEN
];
1419 if (!dir
|| !needs_casefold(dir
))
1422 * If the dentry name is stored in-line, then it may be concurrently
1423 * modified by a rename. If this happens, the VFS will eventually retry
1424 * the lookup, so it doesn't matter what ->d_compare() returns.
1425 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1426 * string. Therefore, we have to copy the name into a temporary buffer.
1428 if (len
<= DNAME_INLINE_LEN
- 1) {
1429 memcpy(strbuf
, str
, len
);
1432 /* prevent compiler from optimizing out the temporary buffer */
1435 ret
= utf8_strncasecmp(um
, name
, &qstr
);
1439 if (sb_has_strict_encoding(sb
))
1442 if (len
!= name
->len
)
1444 return !!memcmp(str
, name
->name
, len
);
1448 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1449 * @dentry: dentry of the parent directory
1450 * @str: qstr of name whose hash we should fill in
1452 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1454 static int generic_ci_d_hash(const struct dentry
*dentry
, struct qstr
*str
)
1456 const struct inode
*dir
= READ_ONCE(dentry
->d_inode
);
1457 struct super_block
*sb
= dentry
->d_sb
;
1458 const struct unicode_map
*um
= sb
->s_encoding
;
1461 if (!dir
|| !needs_casefold(dir
))
1464 ret
= utf8_casefold_hash(um
, dentry
, str
);
1465 if (ret
< 0 && sb_has_strict_encoding(sb
))
1470 static const struct dentry_operations generic_ci_dentry_ops
= {
1471 .d_hash
= generic_ci_d_hash
,
1472 .d_compare
= generic_ci_d_compare
,
1476 #ifdef CONFIG_FS_ENCRYPTION
1477 static const struct dentry_operations generic_encrypted_dentry_ops
= {
1478 .d_revalidate
= fscrypt_d_revalidate
,
1482 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1483 static const struct dentry_operations generic_encrypted_ci_dentry_ops
= {
1484 .d_hash
= generic_ci_d_hash
,
1485 .d_compare
= generic_ci_d_compare
,
1486 .d_revalidate
= fscrypt_d_revalidate
,
1491 * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1492 * @dentry: dentry to set ops on
1494 * Casefolded directories need d_hash and d_compare set, so that the dentries
1495 * contained in them are handled case-insensitively. Note that these operations
1496 * are needed on the parent directory rather than on the dentries in it, and
1497 * while the casefolding flag can be toggled on and off on an empty directory,
1498 * dentry_operations can't be changed later. As a result, if the filesystem has
1499 * casefolding support enabled at all, we have to give all dentries the
1500 * casefolding operations even if their inode doesn't have the casefolding flag
1501 * currently (and thus the casefolding ops would be no-ops for now).
1503 * Encryption works differently in that the only dentry operation it needs is
1504 * d_revalidate, which it only needs on dentries that have the no-key name flag.
1505 * The no-key flag can't be set "later", so we don't have to worry about that.
1507 * Finally, to maximize compatibility with overlayfs (which isn't compatible
1508 * with certain dentry operations) and to avoid taking an unnecessary
1509 * performance hit, we use custom dentry_operations for each possible
1510 * combination rather than always installing all operations.
1512 void generic_set_encrypted_ci_d_ops(struct dentry
*dentry
)
1514 #ifdef CONFIG_FS_ENCRYPTION
1515 bool needs_encrypt_ops
= dentry
->d_flags
& DCACHE_NOKEY_NAME
;
1517 #if IS_ENABLED(CONFIG_UNICODE)
1518 bool needs_ci_ops
= dentry
->d_sb
->s_encoding
;
1520 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1521 if (needs_encrypt_ops
&& needs_ci_ops
) {
1522 d_set_d_op(dentry
, &generic_encrypted_ci_dentry_ops
);
1526 #ifdef CONFIG_FS_ENCRYPTION
1527 if (needs_encrypt_ops
) {
1528 d_set_d_op(dentry
, &generic_encrypted_dentry_ops
);
1532 #if IS_ENABLED(CONFIG_UNICODE)
1534 d_set_d_op(dentry
, &generic_ci_dentry_ops
);
1539 EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops
);
1542 * inode_maybe_inc_iversion - increments i_version
1543 * @inode: inode with the i_version that should be updated
1544 * @force: increment the counter even if it's not necessary?
1546 * Every time the inode is modified, the i_version field must be seen to have
1547 * changed by any observer.
1549 * If "force" is set or the QUERIED flag is set, then ensure that we increment
1550 * the value, and clear the queried flag.
1552 * In the common case where neither is set, then we can return "false" without
1553 * updating i_version.
1555 * If this function returns false, and no other metadata has changed, then we
1556 * can avoid logging the metadata.
1558 bool inode_maybe_inc_iversion(struct inode
*inode
, bool force
)
1563 * The i_version field is not strictly ordered with any other inode
1564 * information, but the legacy inode_inc_iversion code used a spinlock
1565 * to serialize increments.
1567 * Here, we add full memory barriers to ensure that any de-facto
1568 * ordering with other info is preserved.
1570 * This barrier pairs with the barrier in inode_query_iversion()
1573 cur
= inode_peek_iversion_raw(inode
);
1575 /* If flag is clear then we needn't do anything */
1576 if (!force
&& !(cur
& I_VERSION_QUERIED
))
1579 /* Since lowest bit is flag, add 2 to avoid it */
1580 new = (cur
& ~I_VERSION_QUERIED
) + I_VERSION_INCREMENT
;
1581 } while (!atomic64_try_cmpxchg(&inode
->i_version
, &cur
, new));
1584 EXPORT_SYMBOL(inode_maybe_inc_iversion
);