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 mnt_idmap
*idmap
, 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(&nop_mnt_idmap
, request_mask
, 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 u64 id
= huge_encode_dev(dentry
->d_sb
->s_dev
);
46 buf
->f_fsid
= u64_to_fsid(id
);
47 buf
->f_type
= dentry
->d_sb
->s_magic
;
48 buf
->f_bsize
= PAGE_SIZE
;
49 buf
->f_namelen
= NAME_MAX
;
52 EXPORT_SYMBOL(simple_statfs
);
55 * Retaining negative dentries for an in-memory filesystem just wastes
56 * memory and lookup time: arrange for them to be deleted immediately.
58 int always_delete_dentry(const struct dentry
*dentry
)
62 EXPORT_SYMBOL(always_delete_dentry
);
64 const struct dentry_operations simple_dentry_operations
= {
65 .d_delete
= always_delete_dentry
,
67 EXPORT_SYMBOL(simple_dentry_operations
);
70 * Lookup the data. This is trivial - if the dentry didn't already
71 * exist, we know it is negative. Set d_op to delete negative dentries.
73 struct dentry
*simple_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
75 if (dentry
->d_name
.len
> NAME_MAX
)
76 return ERR_PTR(-ENAMETOOLONG
);
77 if (!dentry
->d_sb
->s_d_op
)
78 d_set_d_op(dentry
, &simple_dentry_operations
);
82 EXPORT_SYMBOL(simple_lookup
);
84 int dcache_dir_open(struct inode
*inode
, struct file
*file
)
86 file
->private_data
= d_alloc_cursor(file
->f_path
.dentry
);
88 return file
->private_data
? 0 : -ENOMEM
;
90 EXPORT_SYMBOL(dcache_dir_open
);
92 int dcache_dir_close(struct inode
*inode
, struct file
*file
)
94 dput(file
->private_data
);
97 EXPORT_SYMBOL(dcache_dir_close
);
99 /* parent is locked at least shared */
101 * Returns an element of siblings' list.
102 * We are looking for <count>th positive after <p>; if
103 * found, dentry is grabbed and returned to caller.
104 * If no such element exists, NULL is returned.
106 static struct dentry
*scan_positives(struct dentry
*cursor
,
111 struct dentry
*dentry
= cursor
->d_parent
, *found
= NULL
;
113 spin_lock(&dentry
->d_lock
);
114 while ((p
= p
->next
) != &dentry
->d_subdirs
) {
115 struct dentry
*d
= list_entry(p
, struct dentry
, d_child
);
116 // we must at least skip cursors, to avoid livelocks
117 if (d
->d_flags
& DCACHE_DENTRY_CURSOR
)
119 if (simple_positive(d
) && !--count
) {
120 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
121 if (simple_positive(d
))
122 found
= dget_dlock(d
);
123 spin_unlock(&d
->d_lock
);
128 if (need_resched()) {
129 list_move(&cursor
->d_child
, p
);
130 p
= &cursor
->d_child
;
131 spin_unlock(&dentry
->d_lock
);
133 spin_lock(&dentry
->d_lock
);
136 spin_unlock(&dentry
->d_lock
);
141 loff_t
dcache_dir_lseek(struct file
*file
, loff_t offset
, int whence
)
143 struct dentry
*dentry
= file
->f_path
.dentry
;
146 offset
+= file
->f_pos
;
155 if (offset
!= file
->f_pos
) {
156 struct dentry
*cursor
= file
->private_data
;
157 struct dentry
*to
= NULL
;
159 inode_lock_shared(dentry
->d_inode
);
162 to
= scan_positives(cursor
, &dentry
->d_subdirs
,
164 spin_lock(&dentry
->d_lock
);
166 list_move(&cursor
->d_child
, &to
->d_child
);
168 list_del_init(&cursor
->d_child
);
169 spin_unlock(&dentry
->d_lock
);
172 file
->f_pos
= offset
;
174 inode_unlock_shared(dentry
->d_inode
);
178 EXPORT_SYMBOL(dcache_dir_lseek
);
181 * Directory is locked and all positive dentries in it are safe, since
182 * for ramfs-type trees they can't go away without unlink() or rmdir(),
183 * both impossible due to the lock on directory.
186 int dcache_readdir(struct file
*file
, struct dir_context
*ctx
)
188 struct dentry
*dentry
= file
->f_path
.dentry
;
189 struct dentry
*cursor
= file
->private_data
;
190 struct list_head
*anchor
= &dentry
->d_subdirs
;
191 struct dentry
*next
= NULL
;
194 if (!dir_emit_dots(file
, ctx
))
199 else if (!list_empty(&cursor
->d_child
))
200 p
= &cursor
->d_child
;
204 while ((next
= scan_positives(cursor
, p
, 1, next
)) != NULL
) {
205 if (!dir_emit(ctx
, next
->d_name
.name
, next
->d_name
.len
,
206 d_inode(next
)->i_ino
,
207 fs_umode_to_dtype(d_inode(next
)->i_mode
)))
212 spin_lock(&dentry
->d_lock
);
214 list_move_tail(&cursor
->d_child
, &next
->d_child
);
216 list_del_init(&cursor
->d_child
);
217 spin_unlock(&dentry
->d_lock
);
222 EXPORT_SYMBOL(dcache_readdir
);
224 ssize_t
generic_read_dir(struct file
*filp
, char __user
*buf
, size_t siz
, loff_t
*ppos
)
228 EXPORT_SYMBOL(generic_read_dir
);
230 const struct file_operations simple_dir_operations
= {
231 .open
= dcache_dir_open
,
232 .release
= dcache_dir_close
,
233 .llseek
= dcache_dir_lseek
,
234 .read
= generic_read_dir
,
235 .iterate_shared
= dcache_readdir
,
238 EXPORT_SYMBOL(simple_dir_operations
);
240 const struct inode_operations simple_dir_inode_operations
= {
241 .lookup
= simple_lookup
,
243 EXPORT_SYMBOL(simple_dir_inode_operations
);
245 static void offset_set(struct dentry
*dentry
, u32 offset
)
247 dentry
->d_fsdata
= (void *)((uintptr_t)(offset
));
250 static u32
dentry2offset(struct dentry
*dentry
)
252 return (u32
)((uintptr_t)(dentry
->d_fsdata
));
255 static struct lock_class_key simple_offset_xa_lock
;
258 * simple_offset_init - initialize an offset_ctx
259 * @octx: directory offset map to be initialized
262 void simple_offset_init(struct offset_ctx
*octx
)
264 xa_init_flags(&octx
->xa
, XA_FLAGS_ALLOC1
);
265 lockdep_set_class(&octx
->xa
.xa_lock
, &simple_offset_xa_lock
);
267 /* 0 is '.', 1 is '..', so always start with offset 2 */
268 octx
->next_offset
= 2;
272 * simple_offset_add - Add an entry to a directory's offset map
273 * @octx: directory offset ctx to be updated
274 * @dentry: new dentry being added
276 * Returns zero on success. @so_ctx and the dentry offset are updated.
277 * Otherwise, a negative errno value is returned.
279 int simple_offset_add(struct offset_ctx
*octx
, struct dentry
*dentry
)
281 static const struct xa_limit limit
= XA_LIMIT(2, U32_MAX
);
285 if (dentry2offset(dentry
) != 0)
288 ret
= xa_alloc_cyclic(&octx
->xa
, &offset
, dentry
, limit
,
289 &octx
->next_offset
, GFP_KERNEL
);
293 offset_set(dentry
, offset
);
298 * simple_offset_remove - Remove an entry to a directory's offset map
299 * @octx: directory offset ctx to be updated
300 * @dentry: dentry being removed
303 void simple_offset_remove(struct offset_ctx
*octx
, struct dentry
*dentry
)
307 offset
= dentry2offset(dentry
);
311 xa_erase(&octx
->xa
, offset
);
312 offset_set(dentry
, 0);
316 * simple_offset_rename_exchange - exchange rename with directory offsets
317 * @old_dir: parent of dentry being moved
318 * @old_dentry: dentry being moved
319 * @new_dir: destination parent
320 * @new_dentry: destination dentry
322 * Returns zero on success. Otherwise a negative errno is returned and the
323 * rename is rolled back.
325 int simple_offset_rename_exchange(struct inode
*old_dir
,
326 struct dentry
*old_dentry
,
327 struct inode
*new_dir
,
328 struct dentry
*new_dentry
)
330 struct offset_ctx
*old_ctx
= old_dir
->i_op
->get_offset_ctx(old_dir
);
331 struct offset_ctx
*new_ctx
= new_dir
->i_op
->get_offset_ctx(new_dir
);
332 u32 old_index
= dentry2offset(old_dentry
);
333 u32 new_index
= dentry2offset(new_dentry
);
336 simple_offset_remove(old_ctx
, old_dentry
);
337 simple_offset_remove(new_ctx
, new_dentry
);
339 ret
= simple_offset_add(new_ctx
, old_dentry
);
343 ret
= simple_offset_add(old_ctx
, new_dentry
);
345 simple_offset_remove(new_ctx
, old_dentry
);
349 ret
= simple_rename_exchange(old_dir
, old_dentry
, new_dir
, new_dentry
);
351 simple_offset_remove(new_ctx
, old_dentry
);
352 simple_offset_remove(old_ctx
, new_dentry
);
358 offset_set(old_dentry
, old_index
);
359 xa_store(&old_ctx
->xa
, old_index
, old_dentry
, GFP_KERNEL
);
360 offset_set(new_dentry
, new_index
);
361 xa_store(&new_ctx
->xa
, new_index
, new_dentry
, GFP_KERNEL
);
366 * simple_offset_destroy - Release offset map
367 * @octx: directory offset ctx that is about to be destroyed
369 * During fs teardown (eg. umount), a directory's offset map might still
370 * contain entries. xa_destroy() cleans out anything that remains.
372 void simple_offset_destroy(struct offset_ctx
*octx
)
374 xa_destroy(&octx
->xa
);
378 * offset_dir_llseek - Advance the read position of a directory descriptor
379 * @file: an open directory whose position is to be updated
380 * @offset: a byte offset
381 * @whence: enumerator describing the starting position for this update
383 * SEEK_END, SEEK_DATA, and SEEK_HOLE are not supported for directories.
385 * Returns the updated read position if successful; otherwise a
386 * negative errno is returned and the read position remains unchanged.
388 static loff_t
offset_dir_llseek(struct file
*file
, loff_t offset
, int whence
)
392 offset
+= file
->f_pos
;
402 /* In this case, ->private_data is protected by f_pos_lock */
403 file
->private_data
= NULL
;
404 return vfs_setpos(file
, offset
, U32_MAX
);
407 static struct dentry
*offset_find_next(struct xa_state
*xas
)
409 struct dentry
*child
, *found
= NULL
;
412 child
= xas_next_entry(xas
, U32_MAX
);
415 spin_lock(&child
->d_lock
);
416 if (simple_positive(child
))
417 found
= dget_dlock(child
);
418 spin_unlock(&child
->d_lock
);
424 static bool offset_dir_emit(struct dir_context
*ctx
, struct dentry
*dentry
)
426 u32 offset
= dentry2offset(dentry
);
427 struct inode
*inode
= d_inode(dentry
);
429 return ctx
->actor(ctx
, dentry
->d_name
.name
, dentry
->d_name
.len
, offset
,
430 inode
->i_ino
, fs_umode_to_dtype(inode
->i_mode
));
433 static void *offset_iterate_dir(struct inode
*inode
, struct dir_context
*ctx
)
435 struct offset_ctx
*so_ctx
= inode
->i_op
->get_offset_ctx(inode
);
436 XA_STATE(xas
, &so_ctx
->xa
, ctx
->pos
);
437 struct dentry
*dentry
;
440 dentry
= offset_find_next(&xas
);
442 return ERR_PTR(-ENOENT
);
444 if (!offset_dir_emit(ctx
, dentry
)) {
450 ctx
->pos
= xas
.xa_index
+ 1;
456 * offset_readdir - Emit entries starting at offset @ctx->pos
457 * @file: an open directory to iterate over
458 * @ctx: directory iteration context
460 * Caller must hold @file's i_rwsem to prevent insertion or removal of
461 * entries during this call.
463 * On entry, @ctx->pos contains an offset that represents the first entry
464 * to be read from the directory.
466 * The operation continues until there are no more entries to read, or
467 * until the ctx->actor indicates there is no more space in the caller's
470 * On return, @ctx->pos contains an offset that will read the next entry
471 * in this directory when offset_readdir() is called again with @ctx.
476 static int offset_readdir(struct file
*file
, struct dir_context
*ctx
)
478 struct dentry
*dir
= file
->f_path
.dentry
;
480 lockdep_assert_held(&d_inode(dir
)->i_rwsem
);
482 if (!dir_emit_dots(file
, ctx
))
485 /* In this case, ->private_data is protected by f_pos_lock */
487 file
->private_data
= NULL
;
488 else if (file
->private_data
== ERR_PTR(-ENOENT
))
490 file
->private_data
= offset_iterate_dir(d_inode(dir
), ctx
);
494 const struct file_operations simple_offset_dir_operations
= {
495 .llseek
= offset_dir_llseek
,
496 .iterate_shared
= offset_readdir
,
497 .read
= generic_read_dir
,
501 static struct dentry
*find_next_child(struct dentry
*parent
, struct dentry
*prev
)
503 struct dentry
*child
= NULL
;
504 struct list_head
*p
= prev
? &prev
->d_child
: &parent
->d_subdirs
;
506 spin_lock(&parent
->d_lock
);
507 while ((p
= p
->next
) != &parent
->d_subdirs
) {
508 struct dentry
*d
= container_of(p
, struct dentry
, d_child
);
509 if (simple_positive(d
)) {
510 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
511 if (simple_positive(d
))
512 child
= dget_dlock(d
);
513 spin_unlock(&d
->d_lock
);
518 spin_unlock(&parent
->d_lock
);
523 void simple_recursive_removal(struct dentry
*dentry
,
524 void (*callback
)(struct dentry
*))
526 struct dentry
*this = dget(dentry
);
528 struct dentry
*victim
= NULL
, *child
;
529 struct inode
*inode
= this->d_inode
;
533 inode
->i_flags
|= S_DEAD
;
534 while ((child
= find_next_child(this, victim
)) == NULL
) {
536 // update metadata while it's still locked
537 inode_set_ctime_current(inode
);
541 this = this->d_parent
;
542 inode
= this->d_inode
;
544 if (simple_positive(victim
)) {
545 d_invalidate(victim
); // avoid lost mounts
546 if (d_is_dir(victim
))
547 fsnotify_rmdir(inode
, victim
);
549 fsnotify_unlink(inode
, victim
);
552 dput(victim
); // unpin it
554 if (victim
== dentry
) {
555 inode_set_mtime_to_ts(inode
,
556 inode_set_ctime_current(inode
));
557 if (d_is_dir(dentry
))
568 EXPORT_SYMBOL(simple_recursive_removal
);
570 static const struct super_operations simple_super_operations
= {
571 .statfs
= simple_statfs
,
574 static int pseudo_fs_fill_super(struct super_block
*s
, struct fs_context
*fc
)
576 struct pseudo_fs_context
*ctx
= fc
->fs_private
;
579 s
->s_maxbytes
= MAX_LFS_FILESIZE
;
580 s
->s_blocksize
= PAGE_SIZE
;
581 s
->s_blocksize_bits
= PAGE_SHIFT
;
582 s
->s_magic
= ctx
->magic
;
583 s
->s_op
= ctx
->ops
?: &simple_super_operations
;
584 s
->s_xattr
= ctx
->xattr
;
591 * since this is the first inode, make it number 1. New inodes created
592 * after this must take care not to collide with it (by passing
593 * max_reserved of 1 to iunique).
596 root
->i_mode
= S_IFDIR
| S_IRUSR
| S_IWUSR
;
597 simple_inode_init_ts(root
);
598 s
->s_root
= d_make_root(root
);
601 s
->s_d_op
= ctx
->dops
;
605 static int pseudo_fs_get_tree(struct fs_context
*fc
)
607 return get_tree_nodev(fc
, pseudo_fs_fill_super
);
610 static void pseudo_fs_free(struct fs_context
*fc
)
612 kfree(fc
->fs_private
);
615 static const struct fs_context_operations pseudo_fs_context_ops
= {
616 .free
= pseudo_fs_free
,
617 .get_tree
= pseudo_fs_get_tree
,
621 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
622 * will never be mountable)
624 struct pseudo_fs_context
*init_pseudo(struct fs_context
*fc
,
627 struct pseudo_fs_context
*ctx
;
629 ctx
= kzalloc(sizeof(struct pseudo_fs_context
), GFP_KERNEL
);
632 fc
->fs_private
= ctx
;
633 fc
->ops
= &pseudo_fs_context_ops
;
634 fc
->sb_flags
|= SB_NOUSER
;
639 EXPORT_SYMBOL(init_pseudo
);
641 int simple_open(struct inode
*inode
, struct file
*file
)
643 if (inode
->i_private
)
644 file
->private_data
= inode
->i_private
;
647 EXPORT_SYMBOL(simple_open
);
649 int simple_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*dentry
)
651 struct inode
*inode
= d_inode(old_dentry
);
653 inode_set_mtime_to_ts(dir
,
654 inode_set_ctime_to_ts(dir
, inode_set_ctime_current(inode
)));
658 d_instantiate(dentry
, inode
);
661 EXPORT_SYMBOL(simple_link
);
663 int simple_empty(struct dentry
*dentry
)
665 struct dentry
*child
;
668 spin_lock(&dentry
->d_lock
);
669 list_for_each_entry(child
, &dentry
->d_subdirs
, d_child
) {
670 spin_lock_nested(&child
->d_lock
, DENTRY_D_LOCK_NESTED
);
671 if (simple_positive(child
)) {
672 spin_unlock(&child
->d_lock
);
675 spin_unlock(&child
->d_lock
);
679 spin_unlock(&dentry
->d_lock
);
682 EXPORT_SYMBOL(simple_empty
);
684 int simple_unlink(struct inode
*dir
, struct dentry
*dentry
)
686 struct inode
*inode
= d_inode(dentry
);
688 inode_set_mtime_to_ts(dir
,
689 inode_set_ctime_to_ts(dir
, inode_set_ctime_current(inode
)));
694 EXPORT_SYMBOL(simple_unlink
);
696 int simple_rmdir(struct inode
*dir
, struct dentry
*dentry
)
698 if (!simple_empty(dentry
))
701 drop_nlink(d_inode(dentry
));
702 simple_unlink(dir
, dentry
);
706 EXPORT_SYMBOL(simple_rmdir
);
709 * simple_rename_timestamp - update the various inode timestamps for rename
710 * @old_dir: old parent directory
711 * @old_dentry: dentry that is being renamed
712 * @new_dir: new parent directory
713 * @new_dentry: target for rename
715 * POSIX mandates that the old and new parent directories have their ctime and
716 * mtime updated, and that inodes of @old_dentry and @new_dentry (if any), have
717 * their ctime updated.
719 void simple_rename_timestamp(struct inode
*old_dir
, struct dentry
*old_dentry
,
720 struct inode
*new_dir
, struct dentry
*new_dentry
)
722 struct inode
*newino
= d_inode(new_dentry
);
724 inode_set_mtime_to_ts(old_dir
, inode_set_ctime_current(old_dir
));
725 if (new_dir
!= old_dir
)
726 inode_set_mtime_to_ts(new_dir
,
727 inode_set_ctime_current(new_dir
));
728 inode_set_ctime_current(d_inode(old_dentry
));
730 inode_set_ctime_current(newino
);
732 EXPORT_SYMBOL_GPL(simple_rename_timestamp
);
734 int simple_rename_exchange(struct inode
*old_dir
, struct dentry
*old_dentry
,
735 struct inode
*new_dir
, struct dentry
*new_dentry
)
737 bool old_is_dir
= d_is_dir(old_dentry
);
738 bool new_is_dir
= d_is_dir(new_dentry
);
740 if (old_dir
!= new_dir
&& old_is_dir
!= new_is_dir
) {
749 simple_rename_timestamp(old_dir
, old_dentry
, new_dir
, new_dentry
);
752 EXPORT_SYMBOL_GPL(simple_rename_exchange
);
754 int simple_rename(struct mnt_idmap
*idmap
, struct inode
*old_dir
,
755 struct dentry
*old_dentry
, struct inode
*new_dir
,
756 struct dentry
*new_dentry
, unsigned int flags
)
758 int they_are_dirs
= d_is_dir(old_dentry
);
760 if (flags
& ~(RENAME_NOREPLACE
| RENAME_EXCHANGE
))
763 if (flags
& RENAME_EXCHANGE
)
764 return simple_rename_exchange(old_dir
, old_dentry
, new_dir
, new_dentry
);
766 if (!simple_empty(new_dentry
))
769 if (d_really_is_positive(new_dentry
)) {
770 simple_unlink(new_dir
, new_dentry
);
772 drop_nlink(d_inode(new_dentry
));
775 } else if (they_are_dirs
) {
780 simple_rename_timestamp(old_dir
, old_dentry
, new_dir
, new_dentry
);
783 EXPORT_SYMBOL(simple_rename
);
786 * simple_setattr - setattr for simple filesystem
787 * @idmap: idmap of the target mount
789 * @iattr: iattr structure
791 * Returns 0 on success, -error on failure.
793 * simple_setattr is a simple ->setattr implementation without a proper
794 * implementation of size changes.
796 * It can either be used for in-memory filesystems or special files
797 * on simple regular filesystems. Anything that needs to change on-disk
798 * or wire state on size changes needs its own setattr method.
800 int simple_setattr(struct mnt_idmap
*idmap
, struct dentry
*dentry
,
803 struct inode
*inode
= d_inode(dentry
);
806 error
= setattr_prepare(idmap
, dentry
, iattr
);
810 if (iattr
->ia_valid
& ATTR_SIZE
)
811 truncate_setsize(inode
, iattr
->ia_size
);
812 setattr_copy(idmap
, inode
, iattr
);
813 mark_inode_dirty(inode
);
816 EXPORT_SYMBOL(simple_setattr
);
818 static int simple_read_folio(struct file
*file
, struct folio
*folio
)
820 folio_zero_range(folio
, 0, folio_size(folio
));
821 flush_dcache_folio(folio
);
822 folio_mark_uptodate(folio
);
827 int simple_write_begin(struct file
*file
, struct address_space
*mapping
,
828 loff_t pos
, unsigned len
,
829 struct page
**pagep
, void **fsdata
)
833 folio
= __filemap_get_folio(mapping
, pos
/ PAGE_SIZE
, FGP_WRITEBEGIN
,
834 mapping_gfp_mask(mapping
));
836 return PTR_ERR(folio
);
838 *pagep
= &folio
->page
;
840 if (!folio_test_uptodate(folio
) && (len
!= folio_size(folio
))) {
841 size_t from
= offset_in_folio(folio
, pos
);
843 folio_zero_segments(folio
, 0, from
,
844 from
+ len
, folio_size(folio
));
848 EXPORT_SYMBOL(simple_write_begin
);
851 * simple_write_end - .write_end helper for non-block-device FSes
852 * @file: See .write_end of address_space_operations
860 * simple_write_end does the minimum needed for updating a page after writing is
861 * done. It has the same API signature as the .write_end of
862 * address_space_operations vector. So it can just be set onto .write_end for
863 * FSes that don't need any other processing. i_mutex is assumed to be held.
864 * Block based filesystems should use generic_write_end().
865 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
866 * is not called, so a filesystem that actually does store data in .write_inode
867 * should extend on what's done here with a call to mark_inode_dirty() in the
868 * case that i_size has changed.
870 * Use *ONLY* with simple_read_folio()
872 static int simple_write_end(struct file
*file
, struct address_space
*mapping
,
873 loff_t pos
, unsigned len
, unsigned copied
,
874 struct page
*page
, void *fsdata
)
876 struct folio
*folio
= page_folio(page
);
877 struct inode
*inode
= folio
->mapping
->host
;
878 loff_t last_pos
= pos
+ copied
;
880 /* zero the stale part of the folio if we did a short copy */
881 if (!folio_test_uptodate(folio
)) {
883 size_t from
= offset_in_folio(folio
, pos
);
885 folio_zero_range(folio
, from
+ copied
, len
- copied
);
887 folio_mark_uptodate(folio
);
890 * No need to use i_size_read() here, the i_size
891 * cannot change under us because we hold the i_mutex.
893 if (last_pos
> inode
->i_size
)
894 i_size_write(inode
, last_pos
);
896 folio_mark_dirty(folio
);
904 * Provides ramfs-style behavior: data in the pagecache, but no writeback.
906 const struct address_space_operations ram_aops
= {
907 .read_folio
= simple_read_folio
,
908 .write_begin
= simple_write_begin
,
909 .write_end
= simple_write_end
,
910 .dirty_folio
= noop_dirty_folio
,
912 EXPORT_SYMBOL(ram_aops
);
915 * the inodes created here are not hashed. If you use iunique to generate
916 * unique inode values later for this filesystem, then you must take care
917 * to pass it an appropriate max_reserved value to avoid collisions.
919 int simple_fill_super(struct super_block
*s
, unsigned long magic
,
920 const struct tree_descr
*files
)
924 struct dentry
*dentry
;
927 s
->s_blocksize
= PAGE_SIZE
;
928 s
->s_blocksize_bits
= PAGE_SHIFT
;
930 s
->s_op
= &simple_super_operations
;
933 inode
= new_inode(s
);
937 * because the root inode is 1, the files array must not contain an
941 inode
->i_mode
= S_IFDIR
| 0755;
942 simple_inode_init_ts(inode
);
943 inode
->i_op
= &simple_dir_inode_operations
;
944 inode
->i_fop
= &simple_dir_operations
;
946 root
= d_make_root(inode
);
949 for (i
= 0; !files
->name
|| files
->name
[0]; i
++, files
++) {
953 /* warn if it tries to conflict with the root inode */
954 if (unlikely(i
== 1))
955 printk(KERN_WARNING
"%s: %s passed in a files array"
956 "with an index of 1!\n", __func__
,
959 dentry
= d_alloc_name(root
, files
->name
);
962 inode
= new_inode(s
);
967 inode
->i_mode
= S_IFREG
| files
->mode
;
968 simple_inode_init_ts(inode
);
969 inode
->i_fop
= files
->ops
;
971 d_add(dentry
, inode
);
977 shrink_dcache_parent(root
);
981 EXPORT_SYMBOL(simple_fill_super
);
983 static DEFINE_SPINLOCK(pin_fs_lock
);
985 int simple_pin_fs(struct file_system_type
*type
, struct vfsmount
**mount
, int *count
)
987 struct vfsmount
*mnt
= NULL
;
988 spin_lock(&pin_fs_lock
);
989 if (unlikely(!*mount
)) {
990 spin_unlock(&pin_fs_lock
);
991 mnt
= vfs_kern_mount(type
, SB_KERNMOUNT
, type
->name
, NULL
);
994 spin_lock(&pin_fs_lock
);
1000 spin_unlock(&pin_fs_lock
);
1004 EXPORT_SYMBOL(simple_pin_fs
);
1006 void simple_release_fs(struct vfsmount
**mount
, int *count
)
1008 struct vfsmount
*mnt
;
1009 spin_lock(&pin_fs_lock
);
1013 spin_unlock(&pin_fs_lock
);
1016 EXPORT_SYMBOL(simple_release_fs
);
1019 * simple_read_from_buffer - copy data from the buffer to user space
1020 * @to: the user space buffer to read to
1021 * @count: the maximum number of bytes to read
1022 * @ppos: the current position in the buffer
1023 * @from: the buffer to read from
1024 * @available: the size of the buffer
1026 * The simple_read_from_buffer() function reads up to @count bytes from the
1027 * buffer @from at offset @ppos into the user space address starting at @to.
1029 * On success, the number of bytes read is returned and the offset @ppos is
1030 * advanced by this number, or negative value is returned on error.
1032 ssize_t
simple_read_from_buffer(void __user
*to
, size_t count
, loff_t
*ppos
,
1033 const void *from
, size_t available
)
1040 if (pos
>= available
|| !count
)
1042 if (count
> available
- pos
)
1043 count
= available
- pos
;
1044 ret
= copy_to_user(to
, from
+ pos
, count
);
1048 *ppos
= pos
+ count
;
1051 EXPORT_SYMBOL(simple_read_from_buffer
);
1054 * simple_write_to_buffer - copy data from user space to the buffer
1055 * @to: the buffer to write to
1056 * @available: the size of the buffer
1057 * @ppos: the current position in the buffer
1058 * @from: the user space buffer to read from
1059 * @count: the maximum number of bytes to read
1061 * The simple_write_to_buffer() function reads up to @count bytes from the user
1062 * space address starting at @from into the buffer @to at offset @ppos.
1064 * On success, the number of bytes written is returned and the offset @ppos is
1065 * advanced by this number, or negative value is returned on error.
1067 ssize_t
simple_write_to_buffer(void *to
, size_t available
, loff_t
*ppos
,
1068 const void __user
*from
, size_t count
)
1075 if (pos
>= available
|| !count
)
1077 if (count
> available
- pos
)
1078 count
= available
- pos
;
1079 res
= copy_from_user(to
+ pos
, from
, count
);
1083 *ppos
= pos
+ count
;
1086 EXPORT_SYMBOL(simple_write_to_buffer
);
1089 * memory_read_from_buffer - copy data from the buffer
1090 * @to: the kernel space buffer to read to
1091 * @count: the maximum number of bytes to read
1092 * @ppos: the current position in the buffer
1093 * @from: the buffer to read from
1094 * @available: the size of the buffer
1096 * The memory_read_from_buffer() function reads up to @count bytes from the
1097 * buffer @from at offset @ppos into the kernel space address starting at @to.
1099 * On success, the number of bytes read is returned and the offset @ppos is
1100 * advanced by this number, or negative value is returned on error.
1102 ssize_t
memory_read_from_buffer(void *to
, size_t count
, loff_t
*ppos
,
1103 const void *from
, size_t available
)
1109 if (pos
>= available
)
1111 if (count
> available
- pos
)
1112 count
= available
- pos
;
1113 memcpy(to
, from
+ pos
, count
);
1114 *ppos
= pos
+ count
;
1118 EXPORT_SYMBOL(memory_read_from_buffer
);
1121 * Transaction based IO.
1122 * The file expects a single write which triggers the transaction, and then
1123 * possibly a read which collects the result - which is stored in a
1124 * file-local buffer.
1127 void simple_transaction_set(struct file
*file
, size_t n
)
1129 struct simple_transaction_argresp
*ar
= file
->private_data
;
1131 BUG_ON(n
> SIMPLE_TRANSACTION_LIMIT
);
1134 * The barrier ensures that ar->size will really remain zero until
1135 * ar->data is ready for reading.
1140 EXPORT_SYMBOL(simple_transaction_set
);
1142 char *simple_transaction_get(struct file
*file
, const char __user
*buf
, size_t size
)
1144 struct simple_transaction_argresp
*ar
;
1145 static DEFINE_SPINLOCK(simple_transaction_lock
);
1147 if (size
> SIMPLE_TRANSACTION_LIMIT
- 1)
1148 return ERR_PTR(-EFBIG
);
1150 ar
= (struct simple_transaction_argresp
*)get_zeroed_page(GFP_KERNEL
);
1152 return ERR_PTR(-ENOMEM
);
1154 spin_lock(&simple_transaction_lock
);
1156 /* only one write allowed per open */
1157 if (file
->private_data
) {
1158 spin_unlock(&simple_transaction_lock
);
1159 free_page((unsigned long)ar
);
1160 return ERR_PTR(-EBUSY
);
1163 file
->private_data
= ar
;
1165 spin_unlock(&simple_transaction_lock
);
1167 if (copy_from_user(ar
->data
, buf
, size
))
1168 return ERR_PTR(-EFAULT
);
1172 EXPORT_SYMBOL(simple_transaction_get
);
1174 ssize_t
simple_transaction_read(struct file
*file
, char __user
*buf
, size_t size
, loff_t
*pos
)
1176 struct simple_transaction_argresp
*ar
= file
->private_data
;
1180 return simple_read_from_buffer(buf
, size
, pos
, ar
->data
, ar
->size
);
1182 EXPORT_SYMBOL(simple_transaction_read
);
1184 int simple_transaction_release(struct inode
*inode
, struct file
*file
)
1186 free_page((unsigned long)file
->private_data
);
1189 EXPORT_SYMBOL(simple_transaction_release
);
1191 /* Simple attribute files */
1193 struct simple_attr
{
1194 int (*get
)(void *, u64
*);
1195 int (*set
)(void *, u64
);
1196 char get_buf
[24]; /* enough to store a u64 and "\n\0" */
1199 const char *fmt
; /* format for read operation */
1200 struct mutex mutex
; /* protects access to these buffers */
1203 /* simple_attr_open is called by an actual attribute open file operation
1204 * to set the attribute specific access operations. */
1205 int simple_attr_open(struct inode
*inode
, struct file
*file
,
1206 int (*get
)(void *, u64
*), int (*set
)(void *, u64
),
1209 struct simple_attr
*attr
;
1211 attr
= kzalloc(sizeof(*attr
), GFP_KERNEL
);
1217 attr
->data
= inode
->i_private
;
1219 mutex_init(&attr
->mutex
);
1221 file
->private_data
= attr
;
1223 return nonseekable_open(inode
, file
);
1225 EXPORT_SYMBOL_GPL(simple_attr_open
);
1227 int simple_attr_release(struct inode
*inode
, struct file
*file
)
1229 kfree(file
->private_data
);
1232 EXPORT_SYMBOL_GPL(simple_attr_release
); /* GPL-only? This? Really? */
1234 /* read from the buffer that is filled with the get function */
1235 ssize_t
simple_attr_read(struct file
*file
, char __user
*buf
,
1236 size_t len
, loff_t
*ppos
)
1238 struct simple_attr
*attr
;
1242 attr
= file
->private_data
;
1247 ret
= mutex_lock_interruptible(&attr
->mutex
);
1251 if (*ppos
&& attr
->get_buf
[0]) {
1252 /* continued read */
1253 size
= strlen(attr
->get_buf
);
1257 ret
= attr
->get(attr
->data
, &val
);
1261 size
= scnprintf(attr
->get_buf
, sizeof(attr
->get_buf
),
1262 attr
->fmt
, (unsigned long long)val
);
1265 ret
= simple_read_from_buffer(buf
, len
, ppos
, attr
->get_buf
, size
);
1267 mutex_unlock(&attr
->mutex
);
1270 EXPORT_SYMBOL_GPL(simple_attr_read
);
1272 /* interpret the buffer as a number to call the set function with */
1273 static ssize_t
simple_attr_write_xsigned(struct file
*file
, const char __user
*buf
,
1274 size_t len
, loff_t
*ppos
, bool is_signed
)
1276 struct simple_attr
*attr
;
1277 unsigned long long val
;
1281 attr
= file
->private_data
;
1285 ret
= mutex_lock_interruptible(&attr
->mutex
);
1290 size
= min(sizeof(attr
->set_buf
) - 1, len
);
1291 if (copy_from_user(attr
->set_buf
, buf
, size
))
1294 attr
->set_buf
[size
] = '\0';
1296 ret
= kstrtoll(attr
->set_buf
, 0, &val
);
1298 ret
= kstrtoull(attr
->set_buf
, 0, &val
);
1301 ret
= attr
->set(attr
->data
, val
);
1303 ret
= len
; /* on success, claim we got the whole input */
1305 mutex_unlock(&attr
->mutex
);
1309 ssize_t
simple_attr_write(struct file
*file
, const char __user
*buf
,
1310 size_t len
, loff_t
*ppos
)
1312 return simple_attr_write_xsigned(file
, buf
, len
, ppos
, false);
1314 EXPORT_SYMBOL_GPL(simple_attr_write
);
1316 ssize_t
simple_attr_write_signed(struct file
*file
, const char __user
*buf
,
1317 size_t len
, loff_t
*ppos
)
1319 return simple_attr_write_xsigned(file
, buf
, len
, ppos
, true);
1321 EXPORT_SYMBOL_GPL(simple_attr_write_signed
);
1324 * generic_encode_ino32_fh - generic export_operations->encode_fh function
1325 * @inode: the object to encode
1326 * @fh: where to store the file handle fragment
1327 * @max_len: maximum length to store there (in 4 byte units)
1328 * @parent: parent directory inode, if wanted
1330 * This generic encode_fh function assumes that the 32 inode number
1331 * is suitable for locating an inode, and that the generation number
1332 * can be used to check that it is still valid. It places them in the
1333 * filehandle fragment where export_decode_fh expects to find them.
1335 int generic_encode_ino32_fh(struct inode
*inode
, __u32
*fh
, int *max_len
,
1336 struct inode
*parent
)
1338 struct fid
*fid
= (void *)fh
;
1340 int type
= FILEID_INO32_GEN
;
1342 if (parent
&& (len
< 4)) {
1344 return FILEID_INVALID
;
1345 } else if (len
< 2) {
1347 return FILEID_INVALID
;
1351 fid
->i32
.ino
= inode
->i_ino
;
1352 fid
->i32
.gen
= inode
->i_generation
;
1354 fid
->i32
.parent_ino
= parent
->i_ino
;
1355 fid
->i32
.parent_gen
= parent
->i_generation
;
1357 type
= FILEID_INO32_GEN_PARENT
;
1362 EXPORT_SYMBOL_GPL(generic_encode_ino32_fh
);
1365 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1366 * @sb: filesystem to do the file handle conversion on
1367 * @fid: file handle to convert
1368 * @fh_len: length of the file handle in bytes
1369 * @fh_type: type of file handle
1370 * @get_inode: filesystem callback to retrieve inode
1372 * This function decodes @fid as long as it has one of the well-known
1373 * Linux filehandle types and calls @get_inode on it to retrieve the
1374 * inode for the object specified in the file handle.
1376 struct dentry
*generic_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1377 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1378 (struct super_block
*sb
, u64 ino
, u32 gen
))
1380 struct inode
*inode
= NULL
;
1386 case FILEID_INO32_GEN
:
1387 case FILEID_INO32_GEN_PARENT
:
1388 inode
= get_inode(sb
, fid
->i32
.ino
, fid
->i32
.gen
);
1392 return d_obtain_alias(inode
);
1394 EXPORT_SYMBOL_GPL(generic_fh_to_dentry
);
1397 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1398 * @sb: filesystem to do the file handle conversion on
1399 * @fid: file handle to convert
1400 * @fh_len: length of the file handle in bytes
1401 * @fh_type: type of file handle
1402 * @get_inode: filesystem callback to retrieve inode
1404 * This function decodes @fid as long as it has one of the well-known
1405 * Linux filehandle types and calls @get_inode on it to retrieve the
1406 * inode for the _parent_ object specified in the file handle if it
1407 * is specified in the file handle, or NULL otherwise.
1409 struct dentry
*generic_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1410 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1411 (struct super_block
*sb
, u64 ino
, u32 gen
))
1413 struct inode
*inode
= NULL
;
1419 case FILEID_INO32_GEN_PARENT
:
1420 inode
= get_inode(sb
, fid
->i32
.parent_ino
,
1421 (fh_len
> 3 ? fid
->i32
.parent_gen
: 0));
1425 return d_obtain_alias(inode
);
1427 EXPORT_SYMBOL_GPL(generic_fh_to_parent
);
1430 * __generic_file_fsync - generic fsync implementation for simple filesystems
1432 * @file: file to synchronize
1433 * @start: start offset in bytes
1434 * @end: end offset in bytes (inclusive)
1435 * @datasync: only synchronize essential metadata if true
1437 * This is a generic implementation of the fsync method for simple
1438 * filesystems which track all non-inode metadata in the buffers list
1439 * hanging off the address_space structure.
1441 int __generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1444 struct inode
*inode
= file
->f_mapping
->host
;
1448 err
= file_write_and_wait_range(file
, start
, end
);
1453 ret
= sync_mapping_buffers(inode
->i_mapping
);
1454 if (!(inode
->i_state
& I_DIRTY_ALL
))
1456 if (datasync
&& !(inode
->i_state
& I_DIRTY_DATASYNC
))
1459 err
= sync_inode_metadata(inode
, 1);
1464 inode_unlock(inode
);
1465 /* check and advance again to catch errors after syncing out buffers */
1466 err
= file_check_and_advance_wb_err(file
);
1471 EXPORT_SYMBOL(__generic_file_fsync
);
1474 * generic_file_fsync - generic fsync implementation for simple filesystems
1476 * @file: file to synchronize
1477 * @start: start offset in bytes
1478 * @end: end offset in bytes (inclusive)
1479 * @datasync: only synchronize essential metadata if true
1483 int generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1486 struct inode
*inode
= file
->f_mapping
->host
;
1489 err
= __generic_file_fsync(file
, start
, end
, datasync
);
1492 return blkdev_issue_flush(inode
->i_sb
->s_bdev
);
1494 EXPORT_SYMBOL(generic_file_fsync
);
1497 * generic_check_addressable - Check addressability of file system
1498 * @blocksize_bits: log of file system block size
1499 * @num_blocks: number of blocks in file system
1501 * Determine whether a file system with @num_blocks blocks (and a
1502 * block size of 2**@blocksize_bits) is addressable by the sector_t
1503 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1505 int generic_check_addressable(unsigned blocksize_bits
, u64 num_blocks
)
1507 u64 last_fs_block
= num_blocks
- 1;
1509 last_fs_block
>> (PAGE_SHIFT
- blocksize_bits
);
1511 if (unlikely(num_blocks
== 0))
1514 if ((blocksize_bits
< 9) || (blocksize_bits
> PAGE_SHIFT
))
1517 if ((last_fs_block
> (sector_t
)(~0ULL) >> (blocksize_bits
- 9)) ||
1518 (last_fs_page
> (pgoff_t
)(~0ULL))) {
1523 EXPORT_SYMBOL(generic_check_addressable
);
1526 * No-op implementation of ->fsync for in-memory filesystems.
1528 int noop_fsync(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
1532 EXPORT_SYMBOL(noop_fsync
);
1534 ssize_t
noop_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
)
1537 * iomap based filesystems support direct I/O without need for
1538 * this callback. However, it still needs to be set in
1539 * inode->a_ops so that open/fcntl know that direct I/O is
1540 * generally supported.
1544 EXPORT_SYMBOL_GPL(noop_direct_IO
);
1546 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1547 void kfree_link(void *p
)
1551 EXPORT_SYMBOL(kfree_link
);
1553 struct inode
*alloc_anon_inode(struct super_block
*s
)
1555 static const struct address_space_operations anon_aops
= {
1556 .dirty_folio
= noop_dirty_folio
,
1558 struct inode
*inode
= new_inode_pseudo(s
);
1561 return ERR_PTR(-ENOMEM
);
1563 inode
->i_ino
= get_next_ino();
1564 inode
->i_mapping
->a_ops
= &anon_aops
;
1567 * Mark the inode dirty from the very beginning,
1568 * that way it will never be moved to the dirty
1569 * list because mark_inode_dirty() will think
1570 * that it already _is_ on the dirty list.
1572 inode
->i_state
= I_DIRTY
;
1573 inode
->i_mode
= S_IRUSR
| S_IWUSR
;
1574 inode
->i_uid
= current_fsuid();
1575 inode
->i_gid
= current_fsgid();
1576 inode
->i_flags
|= S_PRIVATE
;
1577 simple_inode_init_ts(inode
);
1580 EXPORT_SYMBOL(alloc_anon_inode
);
1583 * simple_nosetlease - generic helper for prohibiting leases
1584 * @filp: file pointer
1585 * @arg: type of lease to obtain
1586 * @flp: new lease supplied for insertion
1587 * @priv: private data for lm_setup operation
1589 * Generic helper for filesystems that do not wish to allow leases to be set.
1590 * All arguments are ignored and it just returns -EINVAL.
1593 simple_nosetlease(struct file
*filp
, int arg
, struct file_lock
**flp
,
1598 EXPORT_SYMBOL(simple_nosetlease
);
1601 * simple_get_link - generic helper to get the target of "fast" symlinks
1602 * @dentry: not used here
1603 * @inode: the symlink inode
1604 * @done: not used here
1606 * Generic helper for filesystems to use for symlink inodes where a pointer to
1607 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1608 * since as an optimization the path lookup code uses any non-NULL ->i_link
1609 * directly, without calling ->get_link(). But ->get_link() still must be set,
1610 * to mark the inode_operations as being for a symlink.
1612 * Return: the symlink target
1614 const char *simple_get_link(struct dentry
*dentry
, struct inode
*inode
,
1615 struct delayed_call
*done
)
1617 return inode
->i_link
;
1619 EXPORT_SYMBOL(simple_get_link
);
1621 const struct inode_operations simple_symlink_inode_operations
= {
1622 .get_link
= simple_get_link
,
1624 EXPORT_SYMBOL(simple_symlink_inode_operations
);
1627 * Operations for a permanently empty directory.
1629 static struct dentry
*empty_dir_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
1631 return ERR_PTR(-ENOENT
);
1634 static int empty_dir_getattr(struct mnt_idmap
*idmap
,
1635 const struct path
*path
, struct kstat
*stat
,
1636 u32 request_mask
, unsigned int query_flags
)
1638 struct inode
*inode
= d_inode(path
->dentry
);
1639 generic_fillattr(&nop_mnt_idmap
, request_mask
, inode
, stat
);
1643 static int empty_dir_setattr(struct mnt_idmap
*idmap
,
1644 struct dentry
*dentry
, struct iattr
*attr
)
1649 static ssize_t
empty_dir_listxattr(struct dentry
*dentry
, char *list
, size_t size
)
1654 static const struct inode_operations empty_dir_inode_operations
= {
1655 .lookup
= empty_dir_lookup
,
1656 .permission
= generic_permission
,
1657 .setattr
= empty_dir_setattr
,
1658 .getattr
= empty_dir_getattr
,
1659 .listxattr
= empty_dir_listxattr
,
1662 static loff_t
empty_dir_llseek(struct file
*file
, loff_t offset
, int whence
)
1664 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1665 return generic_file_llseek_size(file
, offset
, whence
, 2, 2);
1668 static int empty_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
1670 dir_emit_dots(file
, ctx
);
1674 static const struct file_operations empty_dir_operations
= {
1675 .llseek
= empty_dir_llseek
,
1676 .read
= generic_read_dir
,
1677 .iterate_shared
= empty_dir_readdir
,
1678 .fsync
= noop_fsync
,
1682 void make_empty_dir_inode(struct inode
*inode
)
1684 set_nlink(inode
, 2);
1685 inode
->i_mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
1686 inode
->i_uid
= GLOBAL_ROOT_UID
;
1687 inode
->i_gid
= GLOBAL_ROOT_GID
;
1690 inode
->i_blkbits
= PAGE_SHIFT
;
1691 inode
->i_blocks
= 0;
1693 inode
->i_op
= &empty_dir_inode_operations
;
1694 inode
->i_opflags
&= ~IOP_XATTR
;
1695 inode
->i_fop
= &empty_dir_operations
;
1698 bool is_empty_dir_inode(struct inode
*inode
)
1700 return (inode
->i_fop
== &empty_dir_operations
) &&
1701 (inode
->i_op
== &empty_dir_inode_operations
);
1704 #if IS_ENABLED(CONFIG_UNICODE)
1706 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1707 * @dentry: dentry whose name we are checking against
1708 * @len: len of name of dentry
1709 * @str: str pointer to name of dentry
1710 * @name: Name to compare against
1712 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1714 static int generic_ci_d_compare(const struct dentry
*dentry
, unsigned int len
,
1715 const char *str
, const struct qstr
*name
)
1717 const struct dentry
*parent
= READ_ONCE(dentry
->d_parent
);
1718 const struct inode
*dir
= READ_ONCE(parent
->d_inode
);
1719 const struct super_block
*sb
= dentry
->d_sb
;
1720 const struct unicode_map
*um
= sb
->s_encoding
;
1721 struct qstr qstr
= QSTR_INIT(str
, len
);
1722 char strbuf
[DNAME_INLINE_LEN
];
1725 if (!dir
|| !IS_CASEFOLDED(dir
))
1728 * If the dentry name is stored in-line, then it may be concurrently
1729 * modified by a rename. If this happens, the VFS will eventually retry
1730 * the lookup, so it doesn't matter what ->d_compare() returns.
1731 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1732 * string. Therefore, we have to copy the name into a temporary buffer.
1734 if (len
<= DNAME_INLINE_LEN
- 1) {
1735 memcpy(strbuf
, str
, len
);
1738 /* prevent compiler from optimizing out the temporary buffer */
1741 ret
= utf8_strncasecmp(um
, name
, &qstr
);
1745 if (sb_has_strict_encoding(sb
))
1748 if (len
!= name
->len
)
1750 return !!memcmp(str
, name
->name
, len
);
1754 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1755 * @dentry: dentry of the parent directory
1756 * @str: qstr of name whose hash we should fill in
1758 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1760 static int generic_ci_d_hash(const struct dentry
*dentry
, struct qstr
*str
)
1762 const struct inode
*dir
= READ_ONCE(dentry
->d_inode
);
1763 struct super_block
*sb
= dentry
->d_sb
;
1764 const struct unicode_map
*um
= sb
->s_encoding
;
1767 if (!dir
|| !IS_CASEFOLDED(dir
))
1770 ret
= utf8_casefold_hash(um
, dentry
, str
);
1771 if (ret
< 0 && sb_has_strict_encoding(sb
))
1776 static const struct dentry_operations generic_ci_dentry_ops
= {
1777 .d_hash
= generic_ci_d_hash
,
1778 .d_compare
= generic_ci_d_compare
,
1782 #ifdef CONFIG_FS_ENCRYPTION
1783 static const struct dentry_operations generic_encrypted_dentry_ops
= {
1784 .d_revalidate
= fscrypt_d_revalidate
,
1788 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1789 static const struct dentry_operations generic_encrypted_ci_dentry_ops
= {
1790 .d_hash
= generic_ci_d_hash
,
1791 .d_compare
= generic_ci_d_compare
,
1792 .d_revalidate
= fscrypt_d_revalidate
,
1797 * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1798 * @dentry: dentry to set ops on
1800 * Casefolded directories need d_hash and d_compare set, so that the dentries
1801 * contained in them are handled case-insensitively. Note that these operations
1802 * are needed on the parent directory rather than on the dentries in it, and
1803 * while the casefolding flag can be toggled on and off on an empty directory,
1804 * dentry_operations can't be changed later. As a result, if the filesystem has
1805 * casefolding support enabled at all, we have to give all dentries the
1806 * casefolding operations even if their inode doesn't have the casefolding flag
1807 * currently (and thus the casefolding ops would be no-ops for now).
1809 * Encryption works differently in that the only dentry operation it needs is
1810 * d_revalidate, which it only needs on dentries that have the no-key name flag.
1811 * The no-key flag can't be set "later", so we don't have to worry about that.
1813 * Finally, to maximize compatibility with overlayfs (which isn't compatible
1814 * with certain dentry operations) and to avoid taking an unnecessary
1815 * performance hit, we use custom dentry_operations for each possible
1816 * combination rather than always installing all operations.
1818 void generic_set_encrypted_ci_d_ops(struct dentry
*dentry
)
1820 #ifdef CONFIG_FS_ENCRYPTION
1821 bool needs_encrypt_ops
= dentry
->d_flags
& DCACHE_NOKEY_NAME
;
1823 #if IS_ENABLED(CONFIG_UNICODE)
1824 bool needs_ci_ops
= dentry
->d_sb
->s_encoding
;
1826 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1827 if (needs_encrypt_ops
&& needs_ci_ops
) {
1828 d_set_d_op(dentry
, &generic_encrypted_ci_dentry_ops
);
1832 #ifdef CONFIG_FS_ENCRYPTION
1833 if (needs_encrypt_ops
) {
1834 d_set_d_op(dentry
, &generic_encrypted_dentry_ops
);
1838 #if IS_ENABLED(CONFIG_UNICODE)
1840 d_set_d_op(dentry
, &generic_ci_dentry_ops
);
1845 EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops
);
1848 * inode_maybe_inc_iversion - increments i_version
1849 * @inode: inode with the i_version that should be updated
1850 * @force: increment the counter even if it's not necessary?
1852 * Every time the inode is modified, the i_version field must be seen to have
1853 * changed by any observer.
1855 * If "force" is set or the QUERIED flag is set, then ensure that we increment
1856 * the value, and clear the queried flag.
1858 * In the common case where neither is set, then we can return "false" without
1859 * updating i_version.
1861 * If this function returns false, and no other metadata has changed, then we
1862 * can avoid logging the metadata.
1864 bool inode_maybe_inc_iversion(struct inode
*inode
, bool force
)
1869 * The i_version field is not strictly ordered with any other inode
1870 * information, but the legacy inode_inc_iversion code used a spinlock
1871 * to serialize increments.
1873 * Here, we add full memory barriers to ensure that any de-facto
1874 * ordering with other info is preserved.
1876 * This barrier pairs with the barrier in inode_query_iversion()
1879 cur
= inode_peek_iversion_raw(inode
);
1881 /* If flag is clear then we needn't do anything */
1882 if (!force
&& !(cur
& I_VERSION_QUERIED
))
1885 /* Since lowest bit is flag, add 2 to avoid it */
1886 new = (cur
& ~I_VERSION_QUERIED
) + I_VERSION_INCREMENT
;
1887 } while (!atomic64_try_cmpxchg(&inode
->i_version
, &cur
, new));
1890 EXPORT_SYMBOL(inode_maybe_inc_iversion
);
1893 * inode_query_iversion - read i_version for later use
1894 * @inode: inode from which i_version should be read
1896 * Read the inode i_version counter. This should be used by callers that wish
1897 * to store the returned i_version for later comparison. This will guarantee
1898 * that a later query of the i_version will result in a different value if
1899 * anything has changed.
1901 * In this implementation, we fetch the current value, set the QUERIED flag and
1902 * then try to swap it into place with a cmpxchg, if it wasn't already set. If
1903 * that fails, we try again with the newly fetched value from the cmpxchg.
1905 u64
inode_query_iversion(struct inode
*inode
)
1909 cur
= inode_peek_iversion_raw(inode
);
1911 /* If flag is already set, then no need to swap */
1912 if (cur
& I_VERSION_QUERIED
) {
1914 * This barrier (and the implicit barrier in the
1915 * cmpxchg below) pairs with the barrier in
1916 * inode_maybe_inc_iversion().
1922 new = cur
| I_VERSION_QUERIED
;
1923 } while (!atomic64_try_cmpxchg(&inode
->i_version
, &cur
, new));
1924 return cur
>> I_VERSION_QUERIED_SHIFT
;
1926 EXPORT_SYMBOL(inode_query_iversion
);
1928 ssize_t
direct_write_fallback(struct kiocb
*iocb
, struct iov_iter
*iter
,
1929 ssize_t direct_written
, ssize_t buffered_written
)
1931 struct address_space
*mapping
= iocb
->ki_filp
->f_mapping
;
1932 loff_t pos
= iocb
->ki_pos
- buffered_written
;
1933 loff_t end
= iocb
->ki_pos
- 1;
1937 * If the buffered write fallback returned an error, we want to return
1938 * the number of bytes which were written by direct I/O, or the error
1939 * code if that was zero.
1941 * Note that this differs from normal direct-io semantics, which will
1942 * return -EFOO even if some bytes were written.
1944 if (unlikely(buffered_written
< 0)) {
1946 return direct_written
;
1947 return buffered_written
;
1951 * We need to ensure that the page cache pages are written to disk and
1952 * invalidated to preserve the expected O_DIRECT semantics.
1954 err
= filemap_write_and_wait_range(mapping
, pos
, end
);
1957 * We don't know how much we wrote, so just return the number of
1958 * bytes which were direct-written
1960 iocb
->ki_pos
-= buffered_written
;
1962 return direct_written
;
1965 invalidate_mapping_pages(mapping
, pos
>> PAGE_SHIFT
, end
>> PAGE_SHIFT
);
1966 return direct_written
+ buffered_written
;
1968 EXPORT_SYMBOL_GPL(direct_write_fallback
);
1971 * simple_inode_init_ts - initialize the timestamps for a new inode
1972 * @inode: inode to be initialized
1974 * When a new inode is created, most filesystems set the timestamps to the
1975 * current time. Add a helper to do this.
1977 struct timespec64
simple_inode_init_ts(struct inode
*inode
)
1979 struct timespec64 ts
= inode_set_ctime_current(inode
);
1981 inode_set_atime_to_ts(inode
, ts
);
1982 inode_set_mtime_to_ts(inode
, ts
);
1985 EXPORT_SYMBOL(simple_inode_init_ts
);