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
,
107 struct hlist_node
**p
,
111 struct dentry
*dentry
= cursor
->d_parent
, *found
= NULL
;
113 spin_lock(&dentry
->d_lock
);
115 struct dentry
*d
= hlist_entry(*p
, struct dentry
, d_sib
);
117 // we must at least skip cursors, to avoid livelocks
118 if (d
->d_flags
& DCACHE_DENTRY_CURSOR
)
120 if (simple_positive(d
) && !--count
) {
121 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
122 if (simple_positive(d
))
123 found
= dget_dlock(d
);
124 spin_unlock(&d
->d_lock
);
129 if (need_resched()) {
130 if (!hlist_unhashed(&cursor
->d_sib
))
131 __hlist_del(&cursor
->d_sib
);
132 hlist_add_behind(&cursor
->d_sib
, &d
->d_sib
);
133 p
= &cursor
->d_sib
.next
;
134 spin_unlock(&dentry
->d_lock
);
136 spin_lock(&dentry
->d_lock
);
139 spin_unlock(&dentry
->d_lock
);
144 loff_t
dcache_dir_lseek(struct file
*file
, loff_t offset
, int whence
)
146 struct dentry
*dentry
= file
->f_path
.dentry
;
149 offset
+= file
->f_pos
;
158 if (offset
!= file
->f_pos
) {
159 struct dentry
*cursor
= file
->private_data
;
160 struct dentry
*to
= NULL
;
162 inode_lock_shared(dentry
->d_inode
);
165 to
= scan_positives(cursor
, &dentry
->d_children
.first
,
167 spin_lock(&dentry
->d_lock
);
168 hlist_del_init(&cursor
->d_sib
);
170 hlist_add_behind(&cursor
->d_sib
, &to
->d_sib
);
171 spin_unlock(&dentry
->d_lock
);
174 file
->f_pos
= offset
;
176 inode_unlock_shared(dentry
->d_inode
);
180 EXPORT_SYMBOL(dcache_dir_lseek
);
183 * Directory is locked and all positive dentries in it are safe, since
184 * for ramfs-type trees they can't go away without unlink() or rmdir(),
185 * both impossible due to the lock on directory.
188 int dcache_readdir(struct file
*file
, struct dir_context
*ctx
)
190 struct dentry
*dentry
= file
->f_path
.dentry
;
191 struct dentry
*cursor
= file
->private_data
;
192 struct dentry
*next
= NULL
;
193 struct hlist_node
**p
;
195 if (!dir_emit_dots(file
, ctx
))
199 p
= &dentry
->d_children
.first
;
201 p
= &cursor
->d_sib
.next
;
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
,
206 fs_umode_to_dtype(d_inode(next
)->i_mode
)))
209 p
= &next
->d_sib
.next
;
211 spin_lock(&dentry
->d_lock
);
212 hlist_del_init(&cursor
->d_sib
);
214 hlist_add_before(&cursor
->d_sib
, &next
->d_sib
);
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 void offset_set(struct dentry
*dentry
, u32 offset
)
245 dentry
->d_fsdata
= (void *)((uintptr_t)(offset
));
248 static u32
dentry2offset(struct dentry
*dentry
)
250 return (u32
)((uintptr_t)(dentry
->d_fsdata
));
253 static struct lock_class_key simple_offset_xa_lock
;
256 * simple_offset_init - initialize an offset_ctx
257 * @octx: directory offset map to be initialized
260 void simple_offset_init(struct offset_ctx
*octx
)
262 xa_init_flags(&octx
->xa
, XA_FLAGS_ALLOC1
);
263 lockdep_set_class(&octx
->xa
.xa_lock
, &simple_offset_xa_lock
);
265 /* 0 is '.', 1 is '..', so always start with offset 2 */
266 octx
->next_offset
= 2;
270 * simple_offset_add - Add an entry to a directory's offset map
271 * @octx: directory offset ctx to be updated
272 * @dentry: new dentry being added
274 * Returns zero on success. @so_ctx and the dentry offset are updated.
275 * Otherwise, a negative errno value is returned.
277 int simple_offset_add(struct offset_ctx
*octx
, struct dentry
*dentry
)
279 static const struct xa_limit limit
= XA_LIMIT(2, U32_MAX
);
283 if (dentry2offset(dentry
) != 0)
286 ret
= xa_alloc_cyclic(&octx
->xa
, &offset
, dentry
, limit
,
287 &octx
->next_offset
, GFP_KERNEL
);
291 offset_set(dentry
, offset
);
296 * simple_offset_remove - Remove an entry to a directory's offset map
297 * @octx: directory offset ctx to be updated
298 * @dentry: dentry being removed
301 void simple_offset_remove(struct offset_ctx
*octx
, struct dentry
*dentry
)
305 offset
= dentry2offset(dentry
);
309 xa_erase(&octx
->xa
, offset
);
310 offset_set(dentry
, 0);
314 * simple_offset_rename_exchange - exchange rename with directory offsets
315 * @old_dir: parent of dentry being moved
316 * @old_dentry: dentry being moved
317 * @new_dir: destination parent
318 * @new_dentry: destination dentry
320 * Returns zero on success. Otherwise a negative errno is returned and the
321 * rename is rolled back.
323 int simple_offset_rename_exchange(struct inode
*old_dir
,
324 struct dentry
*old_dentry
,
325 struct inode
*new_dir
,
326 struct dentry
*new_dentry
)
328 struct offset_ctx
*old_ctx
= old_dir
->i_op
->get_offset_ctx(old_dir
);
329 struct offset_ctx
*new_ctx
= new_dir
->i_op
->get_offset_ctx(new_dir
);
330 u32 old_index
= dentry2offset(old_dentry
);
331 u32 new_index
= dentry2offset(new_dentry
);
334 simple_offset_remove(old_ctx
, old_dentry
);
335 simple_offset_remove(new_ctx
, new_dentry
);
337 ret
= simple_offset_add(new_ctx
, old_dentry
);
341 ret
= simple_offset_add(old_ctx
, new_dentry
);
343 simple_offset_remove(new_ctx
, old_dentry
);
347 ret
= simple_rename_exchange(old_dir
, old_dentry
, new_dir
, new_dentry
);
349 simple_offset_remove(new_ctx
, old_dentry
);
350 simple_offset_remove(old_ctx
, new_dentry
);
356 offset_set(old_dentry
, old_index
);
357 xa_store(&old_ctx
->xa
, old_index
, old_dentry
, GFP_KERNEL
);
358 offset_set(new_dentry
, new_index
);
359 xa_store(&new_ctx
->xa
, new_index
, new_dentry
, GFP_KERNEL
);
364 * simple_offset_destroy - Release offset map
365 * @octx: directory offset ctx that is about to be destroyed
367 * During fs teardown (eg. umount), a directory's offset map might still
368 * contain entries. xa_destroy() cleans out anything that remains.
370 void simple_offset_destroy(struct offset_ctx
*octx
)
372 xa_destroy(&octx
->xa
);
376 * offset_dir_llseek - Advance the read position of a directory descriptor
377 * @file: an open directory whose position is to be updated
378 * @offset: a byte offset
379 * @whence: enumerator describing the starting position for this update
381 * SEEK_END, SEEK_DATA, and SEEK_HOLE are not supported for directories.
383 * Returns the updated read position if successful; otherwise a
384 * negative errno is returned and the read position remains unchanged.
386 static loff_t
offset_dir_llseek(struct file
*file
, loff_t offset
, int whence
)
390 offset
+= file
->f_pos
;
400 /* In this case, ->private_data is protected by f_pos_lock */
401 file
->private_data
= NULL
;
402 return vfs_setpos(file
, offset
, U32_MAX
);
405 static struct dentry
*offset_find_next(struct xa_state
*xas
)
407 struct dentry
*child
, *found
= NULL
;
410 child
= xas_next_entry(xas
, U32_MAX
);
413 spin_lock(&child
->d_lock
);
414 if (simple_positive(child
))
415 found
= dget_dlock(child
);
416 spin_unlock(&child
->d_lock
);
422 static bool offset_dir_emit(struct dir_context
*ctx
, struct dentry
*dentry
)
424 u32 offset
= dentry2offset(dentry
);
425 struct inode
*inode
= d_inode(dentry
);
427 return ctx
->actor(ctx
, dentry
->d_name
.name
, dentry
->d_name
.len
, offset
,
428 inode
->i_ino
, fs_umode_to_dtype(inode
->i_mode
));
431 static void *offset_iterate_dir(struct inode
*inode
, struct dir_context
*ctx
)
433 struct offset_ctx
*so_ctx
= inode
->i_op
->get_offset_ctx(inode
);
434 XA_STATE(xas
, &so_ctx
->xa
, ctx
->pos
);
435 struct dentry
*dentry
;
438 dentry
= offset_find_next(&xas
);
440 return ERR_PTR(-ENOENT
);
442 if (!offset_dir_emit(ctx
, dentry
)) {
448 ctx
->pos
= xas
.xa_index
+ 1;
454 * offset_readdir - Emit entries starting at offset @ctx->pos
455 * @file: an open directory to iterate over
456 * @ctx: directory iteration context
458 * Caller must hold @file's i_rwsem to prevent insertion or removal of
459 * entries during this call.
461 * On entry, @ctx->pos contains an offset that represents the first entry
462 * to be read from the directory.
464 * The operation continues until there are no more entries to read, or
465 * until the ctx->actor indicates there is no more space in the caller's
468 * On return, @ctx->pos contains an offset that will read the next entry
469 * in this directory when offset_readdir() is called again with @ctx.
474 static int offset_readdir(struct file
*file
, struct dir_context
*ctx
)
476 struct dentry
*dir
= file
->f_path
.dentry
;
478 lockdep_assert_held(&d_inode(dir
)->i_rwsem
);
480 if (!dir_emit_dots(file
, ctx
))
483 /* In this case, ->private_data is protected by f_pos_lock */
485 file
->private_data
= NULL
;
486 else if (file
->private_data
== ERR_PTR(-ENOENT
))
488 file
->private_data
= offset_iterate_dir(d_inode(dir
), ctx
);
492 const struct file_operations simple_offset_dir_operations
= {
493 .llseek
= offset_dir_llseek
,
494 .iterate_shared
= offset_readdir
,
495 .read
= generic_read_dir
,
499 static struct dentry
*find_next_child(struct dentry
*parent
, struct dentry
*prev
)
501 struct dentry
*child
= NULL
, *d
;
503 spin_lock(&parent
->d_lock
);
504 d
= prev
? d_next_sibling(prev
) : d_first_child(parent
);
505 hlist_for_each_entry_from(d
, d_sib
) {
506 if (simple_positive(d
)) {
507 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
508 if (simple_positive(d
))
509 child
= dget_dlock(d
);
510 spin_unlock(&d
->d_lock
);
515 spin_unlock(&parent
->d_lock
);
520 void simple_recursive_removal(struct dentry
*dentry
,
521 void (*callback
)(struct dentry
*))
523 struct dentry
*this = dget(dentry
);
525 struct dentry
*victim
= NULL
, *child
;
526 struct inode
*inode
= this->d_inode
;
530 inode
->i_flags
|= S_DEAD
;
531 while ((child
= find_next_child(this, victim
)) == NULL
) {
533 // update metadata while it's still locked
534 inode_set_ctime_current(inode
);
538 this = this->d_parent
;
539 inode
= this->d_inode
;
541 if (simple_positive(victim
)) {
542 d_invalidate(victim
); // avoid lost mounts
543 if (d_is_dir(victim
))
544 fsnotify_rmdir(inode
, victim
);
546 fsnotify_unlink(inode
, victim
);
549 dput(victim
); // unpin it
551 if (victim
== dentry
) {
552 inode_set_mtime_to_ts(inode
,
553 inode_set_ctime_current(inode
));
554 if (d_is_dir(dentry
))
565 EXPORT_SYMBOL(simple_recursive_removal
);
567 static const struct super_operations simple_super_operations
= {
568 .statfs
= simple_statfs
,
571 static int pseudo_fs_fill_super(struct super_block
*s
, struct fs_context
*fc
)
573 struct pseudo_fs_context
*ctx
= fc
->fs_private
;
576 s
->s_maxbytes
= MAX_LFS_FILESIZE
;
577 s
->s_blocksize
= PAGE_SIZE
;
578 s
->s_blocksize_bits
= PAGE_SHIFT
;
579 s
->s_magic
= ctx
->magic
;
580 s
->s_op
= ctx
->ops
?: &simple_super_operations
;
581 s
->s_xattr
= ctx
->xattr
;
588 * since this is the first inode, make it number 1. New inodes created
589 * after this must take care not to collide with it (by passing
590 * max_reserved of 1 to iunique).
593 root
->i_mode
= S_IFDIR
| S_IRUSR
| S_IWUSR
;
594 simple_inode_init_ts(root
);
595 s
->s_root
= d_make_root(root
);
598 s
->s_d_op
= ctx
->dops
;
602 static int pseudo_fs_get_tree(struct fs_context
*fc
)
604 return get_tree_nodev(fc
, pseudo_fs_fill_super
);
607 static void pseudo_fs_free(struct fs_context
*fc
)
609 kfree(fc
->fs_private
);
612 static const struct fs_context_operations pseudo_fs_context_ops
= {
613 .free
= pseudo_fs_free
,
614 .get_tree
= pseudo_fs_get_tree
,
618 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
619 * will never be mountable)
621 struct pseudo_fs_context
*init_pseudo(struct fs_context
*fc
,
624 struct pseudo_fs_context
*ctx
;
626 ctx
= kzalloc(sizeof(struct pseudo_fs_context
), GFP_KERNEL
);
629 fc
->fs_private
= ctx
;
630 fc
->ops
= &pseudo_fs_context_ops
;
631 fc
->sb_flags
|= SB_NOUSER
;
636 EXPORT_SYMBOL(init_pseudo
);
638 int simple_open(struct inode
*inode
, struct file
*file
)
640 if (inode
->i_private
)
641 file
->private_data
= inode
->i_private
;
644 EXPORT_SYMBOL(simple_open
);
646 int simple_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*dentry
)
648 struct inode
*inode
= d_inode(old_dentry
);
650 inode_set_mtime_to_ts(dir
,
651 inode_set_ctime_to_ts(dir
, inode_set_ctime_current(inode
)));
655 d_instantiate(dentry
, inode
);
658 EXPORT_SYMBOL(simple_link
);
660 int simple_empty(struct dentry
*dentry
)
662 struct dentry
*child
;
665 spin_lock(&dentry
->d_lock
);
666 hlist_for_each_entry(child
, &dentry
->d_children
, d_sib
) {
667 spin_lock_nested(&child
->d_lock
, DENTRY_D_LOCK_NESTED
);
668 if (simple_positive(child
)) {
669 spin_unlock(&child
->d_lock
);
672 spin_unlock(&child
->d_lock
);
676 spin_unlock(&dentry
->d_lock
);
679 EXPORT_SYMBOL(simple_empty
);
681 int simple_unlink(struct inode
*dir
, struct dentry
*dentry
)
683 struct inode
*inode
= d_inode(dentry
);
685 inode_set_mtime_to_ts(dir
,
686 inode_set_ctime_to_ts(dir
, inode_set_ctime_current(inode
)));
691 EXPORT_SYMBOL(simple_unlink
);
693 int simple_rmdir(struct inode
*dir
, struct dentry
*dentry
)
695 if (!simple_empty(dentry
))
698 drop_nlink(d_inode(dentry
));
699 simple_unlink(dir
, dentry
);
703 EXPORT_SYMBOL(simple_rmdir
);
706 * simple_rename_timestamp - update the various inode timestamps for rename
707 * @old_dir: old parent directory
708 * @old_dentry: dentry that is being renamed
709 * @new_dir: new parent directory
710 * @new_dentry: target for rename
712 * POSIX mandates that the old and new parent directories have their ctime and
713 * mtime updated, and that inodes of @old_dentry and @new_dentry (if any), have
714 * their ctime updated.
716 void simple_rename_timestamp(struct inode
*old_dir
, struct dentry
*old_dentry
,
717 struct inode
*new_dir
, struct dentry
*new_dentry
)
719 struct inode
*newino
= d_inode(new_dentry
);
721 inode_set_mtime_to_ts(old_dir
, inode_set_ctime_current(old_dir
));
722 if (new_dir
!= old_dir
)
723 inode_set_mtime_to_ts(new_dir
,
724 inode_set_ctime_current(new_dir
));
725 inode_set_ctime_current(d_inode(old_dentry
));
727 inode_set_ctime_current(newino
);
729 EXPORT_SYMBOL_GPL(simple_rename_timestamp
);
731 int simple_rename_exchange(struct inode
*old_dir
, struct dentry
*old_dentry
,
732 struct inode
*new_dir
, struct dentry
*new_dentry
)
734 bool old_is_dir
= d_is_dir(old_dentry
);
735 bool new_is_dir
= d_is_dir(new_dentry
);
737 if (old_dir
!= new_dir
&& old_is_dir
!= new_is_dir
) {
746 simple_rename_timestamp(old_dir
, old_dentry
, new_dir
, new_dentry
);
749 EXPORT_SYMBOL_GPL(simple_rename_exchange
);
751 int simple_rename(struct mnt_idmap
*idmap
, struct inode
*old_dir
,
752 struct dentry
*old_dentry
, struct inode
*new_dir
,
753 struct dentry
*new_dentry
, unsigned int flags
)
755 int they_are_dirs
= d_is_dir(old_dentry
);
757 if (flags
& ~(RENAME_NOREPLACE
| RENAME_EXCHANGE
))
760 if (flags
& RENAME_EXCHANGE
)
761 return simple_rename_exchange(old_dir
, old_dentry
, new_dir
, new_dentry
);
763 if (!simple_empty(new_dentry
))
766 if (d_really_is_positive(new_dentry
)) {
767 simple_unlink(new_dir
, new_dentry
);
769 drop_nlink(d_inode(new_dentry
));
772 } else if (they_are_dirs
) {
777 simple_rename_timestamp(old_dir
, old_dentry
, new_dir
, new_dentry
);
780 EXPORT_SYMBOL(simple_rename
);
783 * simple_setattr - setattr for simple filesystem
784 * @idmap: idmap of the target mount
786 * @iattr: iattr structure
788 * Returns 0 on success, -error on failure.
790 * simple_setattr is a simple ->setattr implementation without a proper
791 * implementation of size changes.
793 * It can either be used for in-memory filesystems or special files
794 * on simple regular filesystems. Anything that needs to change on-disk
795 * or wire state on size changes needs its own setattr method.
797 int simple_setattr(struct mnt_idmap
*idmap
, struct dentry
*dentry
,
800 struct inode
*inode
= d_inode(dentry
);
803 error
= setattr_prepare(idmap
, dentry
, iattr
);
807 if (iattr
->ia_valid
& ATTR_SIZE
)
808 truncate_setsize(inode
, iattr
->ia_size
);
809 setattr_copy(idmap
, inode
, iattr
);
810 mark_inode_dirty(inode
);
813 EXPORT_SYMBOL(simple_setattr
);
815 static int simple_read_folio(struct file
*file
, struct folio
*folio
)
817 folio_zero_range(folio
, 0, folio_size(folio
));
818 flush_dcache_folio(folio
);
819 folio_mark_uptodate(folio
);
824 int simple_write_begin(struct file
*file
, struct address_space
*mapping
,
825 loff_t pos
, unsigned len
,
826 struct page
**pagep
, void **fsdata
)
830 folio
= __filemap_get_folio(mapping
, pos
/ PAGE_SIZE
, FGP_WRITEBEGIN
,
831 mapping_gfp_mask(mapping
));
833 return PTR_ERR(folio
);
835 *pagep
= &folio
->page
;
837 if (!folio_test_uptodate(folio
) && (len
!= folio_size(folio
))) {
838 size_t from
= offset_in_folio(folio
, pos
);
840 folio_zero_segments(folio
, 0, from
,
841 from
+ len
, folio_size(folio
));
845 EXPORT_SYMBOL(simple_write_begin
);
848 * simple_write_end - .write_end helper for non-block-device FSes
849 * @file: See .write_end of address_space_operations
857 * simple_write_end does the minimum needed for updating a page after writing is
858 * done. It has the same API signature as the .write_end of
859 * address_space_operations vector. So it can just be set onto .write_end for
860 * FSes that don't need any other processing. i_mutex is assumed to be held.
861 * Block based filesystems should use generic_write_end().
862 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
863 * is not called, so a filesystem that actually does store data in .write_inode
864 * should extend on what's done here with a call to mark_inode_dirty() in the
865 * case that i_size has changed.
867 * Use *ONLY* with simple_read_folio()
869 static int simple_write_end(struct file
*file
, struct address_space
*mapping
,
870 loff_t pos
, unsigned len
, unsigned copied
,
871 struct page
*page
, void *fsdata
)
873 struct folio
*folio
= page_folio(page
);
874 struct inode
*inode
= folio
->mapping
->host
;
875 loff_t last_pos
= pos
+ copied
;
877 /* zero the stale part of the folio if we did a short copy */
878 if (!folio_test_uptodate(folio
)) {
880 size_t from
= offset_in_folio(folio
, pos
);
882 folio_zero_range(folio
, from
+ copied
, len
- copied
);
884 folio_mark_uptodate(folio
);
887 * No need to use i_size_read() here, the i_size
888 * cannot change under us because we hold the i_mutex.
890 if (last_pos
> inode
->i_size
)
891 i_size_write(inode
, last_pos
);
893 folio_mark_dirty(folio
);
901 * Provides ramfs-style behavior: data in the pagecache, but no writeback.
903 const struct address_space_operations ram_aops
= {
904 .read_folio
= simple_read_folio
,
905 .write_begin
= simple_write_begin
,
906 .write_end
= simple_write_end
,
907 .dirty_folio
= noop_dirty_folio
,
909 EXPORT_SYMBOL(ram_aops
);
912 * the inodes created here are not hashed. If you use iunique to generate
913 * unique inode values later for this filesystem, then you must take care
914 * to pass it an appropriate max_reserved value to avoid collisions.
916 int simple_fill_super(struct super_block
*s
, unsigned long magic
,
917 const struct tree_descr
*files
)
920 struct dentry
*dentry
;
923 s
->s_blocksize
= PAGE_SIZE
;
924 s
->s_blocksize_bits
= PAGE_SHIFT
;
926 s
->s_op
= &simple_super_operations
;
929 inode
= new_inode(s
);
933 * because the root inode is 1, the files array must not contain an
937 inode
->i_mode
= S_IFDIR
| 0755;
938 simple_inode_init_ts(inode
);
939 inode
->i_op
= &simple_dir_inode_operations
;
940 inode
->i_fop
= &simple_dir_operations
;
942 s
->s_root
= d_make_root(inode
);
945 for (i
= 0; !files
->name
|| files
->name
[0]; i
++, files
++) {
949 /* warn if it tries to conflict with the root inode */
950 if (unlikely(i
== 1))
951 printk(KERN_WARNING
"%s: %s passed in a files array"
952 "with an index of 1!\n", __func__
,
955 dentry
= d_alloc_name(s
->s_root
, files
->name
);
958 inode
= new_inode(s
);
963 inode
->i_mode
= S_IFREG
| files
->mode
;
964 simple_inode_init_ts(inode
);
965 inode
->i_fop
= files
->ops
;
967 d_add(dentry
, inode
);
971 EXPORT_SYMBOL(simple_fill_super
);
973 static DEFINE_SPINLOCK(pin_fs_lock
);
975 int simple_pin_fs(struct file_system_type
*type
, struct vfsmount
**mount
, int *count
)
977 struct vfsmount
*mnt
= NULL
;
978 spin_lock(&pin_fs_lock
);
979 if (unlikely(!*mount
)) {
980 spin_unlock(&pin_fs_lock
);
981 mnt
= vfs_kern_mount(type
, SB_KERNMOUNT
, type
->name
, NULL
);
984 spin_lock(&pin_fs_lock
);
990 spin_unlock(&pin_fs_lock
);
994 EXPORT_SYMBOL(simple_pin_fs
);
996 void simple_release_fs(struct vfsmount
**mount
, int *count
)
998 struct vfsmount
*mnt
;
999 spin_lock(&pin_fs_lock
);
1003 spin_unlock(&pin_fs_lock
);
1006 EXPORT_SYMBOL(simple_release_fs
);
1009 * simple_read_from_buffer - copy data from the buffer to user space
1010 * @to: the user space buffer to read to
1011 * @count: the maximum number of bytes to read
1012 * @ppos: the current position in the buffer
1013 * @from: the buffer to read from
1014 * @available: the size of the buffer
1016 * The simple_read_from_buffer() function reads up to @count bytes from the
1017 * buffer @from at offset @ppos into the user space address starting at @to.
1019 * On success, the number of bytes read is returned and the offset @ppos is
1020 * advanced by this number, or negative value is returned on error.
1022 ssize_t
simple_read_from_buffer(void __user
*to
, size_t count
, loff_t
*ppos
,
1023 const void *from
, size_t available
)
1030 if (pos
>= available
|| !count
)
1032 if (count
> available
- pos
)
1033 count
= available
- pos
;
1034 ret
= copy_to_user(to
, from
+ pos
, count
);
1038 *ppos
= pos
+ count
;
1041 EXPORT_SYMBOL(simple_read_from_buffer
);
1044 * simple_write_to_buffer - copy data from user space to the buffer
1045 * @to: the buffer to write to
1046 * @available: the size of the buffer
1047 * @ppos: the current position in the buffer
1048 * @from: the user space buffer to read from
1049 * @count: the maximum number of bytes to read
1051 * The simple_write_to_buffer() function reads up to @count bytes from the user
1052 * space address starting at @from into the buffer @to at offset @ppos.
1054 * On success, the number of bytes written is returned and the offset @ppos is
1055 * advanced by this number, or negative value is returned on error.
1057 ssize_t
simple_write_to_buffer(void *to
, size_t available
, loff_t
*ppos
,
1058 const void __user
*from
, size_t count
)
1065 if (pos
>= available
|| !count
)
1067 if (count
> available
- pos
)
1068 count
= available
- pos
;
1069 res
= copy_from_user(to
+ pos
, from
, count
);
1073 *ppos
= pos
+ count
;
1076 EXPORT_SYMBOL(simple_write_to_buffer
);
1079 * memory_read_from_buffer - copy data from the buffer
1080 * @to: the kernel space buffer to read to
1081 * @count: the maximum number of bytes to read
1082 * @ppos: the current position in the buffer
1083 * @from: the buffer to read from
1084 * @available: the size of the buffer
1086 * The memory_read_from_buffer() function reads up to @count bytes from the
1087 * buffer @from at offset @ppos into the kernel space address starting at @to.
1089 * On success, the number of bytes read is returned and the offset @ppos is
1090 * advanced by this number, or negative value is returned on error.
1092 ssize_t
memory_read_from_buffer(void *to
, size_t count
, loff_t
*ppos
,
1093 const void *from
, size_t available
)
1099 if (pos
>= available
)
1101 if (count
> available
- pos
)
1102 count
= available
- pos
;
1103 memcpy(to
, from
+ pos
, count
);
1104 *ppos
= pos
+ count
;
1108 EXPORT_SYMBOL(memory_read_from_buffer
);
1111 * Transaction based IO.
1112 * The file expects a single write which triggers the transaction, and then
1113 * possibly a read which collects the result - which is stored in a
1114 * file-local buffer.
1117 void simple_transaction_set(struct file
*file
, size_t n
)
1119 struct simple_transaction_argresp
*ar
= file
->private_data
;
1121 BUG_ON(n
> SIMPLE_TRANSACTION_LIMIT
);
1124 * The barrier ensures that ar->size will really remain zero until
1125 * ar->data is ready for reading.
1130 EXPORT_SYMBOL(simple_transaction_set
);
1132 char *simple_transaction_get(struct file
*file
, const char __user
*buf
, size_t size
)
1134 struct simple_transaction_argresp
*ar
;
1135 static DEFINE_SPINLOCK(simple_transaction_lock
);
1137 if (size
> SIMPLE_TRANSACTION_LIMIT
- 1)
1138 return ERR_PTR(-EFBIG
);
1140 ar
= (struct simple_transaction_argresp
*)get_zeroed_page(GFP_KERNEL
);
1142 return ERR_PTR(-ENOMEM
);
1144 spin_lock(&simple_transaction_lock
);
1146 /* only one write allowed per open */
1147 if (file
->private_data
) {
1148 spin_unlock(&simple_transaction_lock
);
1149 free_page((unsigned long)ar
);
1150 return ERR_PTR(-EBUSY
);
1153 file
->private_data
= ar
;
1155 spin_unlock(&simple_transaction_lock
);
1157 if (copy_from_user(ar
->data
, buf
, size
))
1158 return ERR_PTR(-EFAULT
);
1162 EXPORT_SYMBOL(simple_transaction_get
);
1164 ssize_t
simple_transaction_read(struct file
*file
, char __user
*buf
, size_t size
, loff_t
*pos
)
1166 struct simple_transaction_argresp
*ar
= file
->private_data
;
1170 return simple_read_from_buffer(buf
, size
, pos
, ar
->data
, ar
->size
);
1172 EXPORT_SYMBOL(simple_transaction_read
);
1174 int simple_transaction_release(struct inode
*inode
, struct file
*file
)
1176 free_page((unsigned long)file
->private_data
);
1179 EXPORT_SYMBOL(simple_transaction_release
);
1181 /* Simple attribute files */
1183 struct simple_attr
{
1184 int (*get
)(void *, u64
*);
1185 int (*set
)(void *, u64
);
1186 char get_buf
[24]; /* enough to store a u64 and "\n\0" */
1189 const char *fmt
; /* format for read operation */
1190 struct mutex mutex
; /* protects access to these buffers */
1193 /* simple_attr_open is called by an actual attribute open file operation
1194 * to set the attribute specific access operations. */
1195 int simple_attr_open(struct inode
*inode
, struct file
*file
,
1196 int (*get
)(void *, u64
*), int (*set
)(void *, u64
),
1199 struct simple_attr
*attr
;
1201 attr
= kzalloc(sizeof(*attr
), GFP_KERNEL
);
1207 attr
->data
= inode
->i_private
;
1209 mutex_init(&attr
->mutex
);
1211 file
->private_data
= attr
;
1213 return nonseekable_open(inode
, file
);
1215 EXPORT_SYMBOL_GPL(simple_attr_open
);
1217 int simple_attr_release(struct inode
*inode
, struct file
*file
)
1219 kfree(file
->private_data
);
1222 EXPORT_SYMBOL_GPL(simple_attr_release
); /* GPL-only? This? Really? */
1224 /* read from the buffer that is filled with the get function */
1225 ssize_t
simple_attr_read(struct file
*file
, char __user
*buf
,
1226 size_t len
, loff_t
*ppos
)
1228 struct simple_attr
*attr
;
1232 attr
= file
->private_data
;
1237 ret
= mutex_lock_interruptible(&attr
->mutex
);
1241 if (*ppos
&& attr
->get_buf
[0]) {
1242 /* continued read */
1243 size
= strlen(attr
->get_buf
);
1247 ret
= attr
->get(attr
->data
, &val
);
1251 size
= scnprintf(attr
->get_buf
, sizeof(attr
->get_buf
),
1252 attr
->fmt
, (unsigned long long)val
);
1255 ret
= simple_read_from_buffer(buf
, len
, ppos
, attr
->get_buf
, size
);
1257 mutex_unlock(&attr
->mutex
);
1260 EXPORT_SYMBOL_GPL(simple_attr_read
);
1262 /* interpret the buffer as a number to call the set function with */
1263 static ssize_t
simple_attr_write_xsigned(struct file
*file
, const char __user
*buf
,
1264 size_t len
, loff_t
*ppos
, bool is_signed
)
1266 struct simple_attr
*attr
;
1267 unsigned long long val
;
1271 attr
= file
->private_data
;
1275 ret
= mutex_lock_interruptible(&attr
->mutex
);
1280 size
= min(sizeof(attr
->set_buf
) - 1, len
);
1281 if (copy_from_user(attr
->set_buf
, buf
, size
))
1284 attr
->set_buf
[size
] = '\0';
1286 ret
= kstrtoll(attr
->set_buf
, 0, &val
);
1288 ret
= kstrtoull(attr
->set_buf
, 0, &val
);
1291 ret
= attr
->set(attr
->data
, val
);
1293 ret
= len
; /* on success, claim we got the whole input */
1295 mutex_unlock(&attr
->mutex
);
1299 ssize_t
simple_attr_write(struct file
*file
, const char __user
*buf
,
1300 size_t len
, loff_t
*ppos
)
1302 return simple_attr_write_xsigned(file
, buf
, len
, ppos
, false);
1304 EXPORT_SYMBOL_GPL(simple_attr_write
);
1306 ssize_t
simple_attr_write_signed(struct file
*file
, const char __user
*buf
,
1307 size_t len
, loff_t
*ppos
)
1309 return simple_attr_write_xsigned(file
, buf
, len
, ppos
, true);
1311 EXPORT_SYMBOL_GPL(simple_attr_write_signed
);
1314 * generic_encode_ino32_fh - generic export_operations->encode_fh function
1315 * @inode: the object to encode
1316 * @fh: where to store the file handle fragment
1317 * @max_len: maximum length to store there (in 4 byte units)
1318 * @parent: parent directory inode, if wanted
1320 * This generic encode_fh function assumes that the 32 inode number
1321 * is suitable for locating an inode, and that the generation number
1322 * can be used to check that it is still valid. It places them in the
1323 * filehandle fragment where export_decode_fh expects to find them.
1325 int generic_encode_ino32_fh(struct inode
*inode
, __u32
*fh
, int *max_len
,
1326 struct inode
*parent
)
1328 struct fid
*fid
= (void *)fh
;
1330 int type
= FILEID_INO32_GEN
;
1332 if (parent
&& (len
< 4)) {
1334 return FILEID_INVALID
;
1335 } else if (len
< 2) {
1337 return FILEID_INVALID
;
1341 fid
->i32
.ino
= inode
->i_ino
;
1342 fid
->i32
.gen
= inode
->i_generation
;
1344 fid
->i32
.parent_ino
= parent
->i_ino
;
1345 fid
->i32
.parent_gen
= parent
->i_generation
;
1347 type
= FILEID_INO32_GEN_PARENT
;
1352 EXPORT_SYMBOL_GPL(generic_encode_ino32_fh
);
1355 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1356 * @sb: filesystem to do the file handle conversion on
1357 * @fid: file handle to convert
1358 * @fh_len: length of the file handle in bytes
1359 * @fh_type: type of file handle
1360 * @get_inode: filesystem callback to retrieve inode
1362 * This function decodes @fid as long as it has one of the well-known
1363 * Linux filehandle types and calls @get_inode on it to retrieve the
1364 * inode for the object specified in the file handle.
1366 struct dentry
*generic_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1367 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1368 (struct super_block
*sb
, u64 ino
, u32 gen
))
1370 struct inode
*inode
= NULL
;
1376 case FILEID_INO32_GEN
:
1377 case FILEID_INO32_GEN_PARENT
:
1378 inode
= get_inode(sb
, fid
->i32
.ino
, fid
->i32
.gen
);
1382 return d_obtain_alias(inode
);
1384 EXPORT_SYMBOL_GPL(generic_fh_to_dentry
);
1387 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1388 * @sb: filesystem to do the file handle conversion on
1389 * @fid: file handle to convert
1390 * @fh_len: length of the file handle in bytes
1391 * @fh_type: type of file handle
1392 * @get_inode: filesystem callback to retrieve inode
1394 * This function decodes @fid as long as it has one of the well-known
1395 * Linux filehandle types and calls @get_inode on it to retrieve the
1396 * inode for the _parent_ object specified in the file handle if it
1397 * is specified in the file handle, or NULL otherwise.
1399 struct dentry
*generic_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1400 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1401 (struct super_block
*sb
, u64 ino
, u32 gen
))
1403 struct inode
*inode
= NULL
;
1409 case FILEID_INO32_GEN_PARENT
:
1410 inode
= get_inode(sb
, fid
->i32
.parent_ino
,
1411 (fh_len
> 3 ? fid
->i32
.parent_gen
: 0));
1415 return d_obtain_alias(inode
);
1417 EXPORT_SYMBOL_GPL(generic_fh_to_parent
);
1420 * __generic_file_fsync - generic fsync implementation for simple filesystems
1422 * @file: file to synchronize
1423 * @start: start offset in bytes
1424 * @end: end offset in bytes (inclusive)
1425 * @datasync: only synchronize essential metadata if true
1427 * This is a generic implementation of the fsync method for simple
1428 * filesystems which track all non-inode metadata in the buffers list
1429 * hanging off the address_space structure.
1431 int __generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1434 struct inode
*inode
= file
->f_mapping
->host
;
1438 err
= file_write_and_wait_range(file
, start
, end
);
1443 ret
= sync_mapping_buffers(inode
->i_mapping
);
1444 if (!(inode
->i_state
& I_DIRTY_ALL
))
1446 if (datasync
&& !(inode
->i_state
& I_DIRTY_DATASYNC
))
1449 err
= sync_inode_metadata(inode
, 1);
1454 inode_unlock(inode
);
1455 /* check and advance again to catch errors after syncing out buffers */
1456 err
= file_check_and_advance_wb_err(file
);
1461 EXPORT_SYMBOL(__generic_file_fsync
);
1464 * generic_file_fsync - generic fsync implementation for simple filesystems
1466 * @file: file to synchronize
1467 * @start: start offset in bytes
1468 * @end: end offset in bytes (inclusive)
1469 * @datasync: only synchronize essential metadata if true
1473 int generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1476 struct inode
*inode
= file
->f_mapping
->host
;
1479 err
= __generic_file_fsync(file
, start
, end
, datasync
);
1482 return blkdev_issue_flush(inode
->i_sb
->s_bdev
);
1484 EXPORT_SYMBOL(generic_file_fsync
);
1487 * generic_check_addressable - Check addressability of file system
1488 * @blocksize_bits: log of file system block size
1489 * @num_blocks: number of blocks in file system
1491 * Determine whether a file system with @num_blocks blocks (and a
1492 * block size of 2**@blocksize_bits) is addressable by the sector_t
1493 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1495 int generic_check_addressable(unsigned blocksize_bits
, u64 num_blocks
)
1497 u64 last_fs_block
= num_blocks
- 1;
1499 last_fs_block
>> (PAGE_SHIFT
- blocksize_bits
);
1501 if (unlikely(num_blocks
== 0))
1504 if ((blocksize_bits
< 9) || (blocksize_bits
> PAGE_SHIFT
))
1507 if ((last_fs_block
> (sector_t
)(~0ULL) >> (blocksize_bits
- 9)) ||
1508 (last_fs_page
> (pgoff_t
)(~0ULL))) {
1513 EXPORT_SYMBOL(generic_check_addressable
);
1516 * No-op implementation of ->fsync for in-memory filesystems.
1518 int noop_fsync(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
1522 EXPORT_SYMBOL(noop_fsync
);
1524 ssize_t
noop_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
)
1527 * iomap based filesystems support direct I/O without need for
1528 * this callback. However, it still needs to be set in
1529 * inode->a_ops so that open/fcntl know that direct I/O is
1530 * generally supported.
1534 EXPORT_SYMBOL_GPL(noop_direct_IO
);
1536 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1537 void kfree_link(void *p
)
1541 EXPORT_SYMBOL(kfree_link
);
1543 struct inode
*alloc_anon_inode(struct super_block
*s
)
1545 static const struct address_space_operations anon_aops
= {
1546 .dirty_folio
= noop_dirty_folio
,
1548 struct inode
*inode
= new_inode_pseudo(s
);
1551 return ERR_PTR(-ENOMEM
);
1553 inode
->i_ino
= get_next_ino();
1554 inode
->i_mapping
->a_ops
= &anon_aops
;
1557 * Mark the inode dirty from the very beginning,
1558 * that way it will never be moved to the dirty
1559 * list because mark_inode_dirty() will think
1560 * that it already _is_ on the dirty list.
1562 inode
->i_state
= I_DIRTY
;
1563 inode
->i_mode
= S_IRUSR
| S_IWUSR
;
1564 inode
->i_uid
= current_fsuid();
1565 inode
->i_gid
= current_fsgid();
1566 inode
->i_flags
|= S_PRIVATE
;
1567 simple_inode_init_ts(inode
);
1570 EXPORT_SYMBOL(alloc_anon_inode
);
1573 * simple_nosetlease - generic helper for prohibiting leases
1574 * @filp: file pointer
1575 * @arg: type of lease to obtain
1576 * @flp: new lease supplied for insertion
1577 * @priv: private data for lm_setup operation
1579 * Generic helper for filesystems that do not wish to allow leases to be set.
1580 * All arguments are ignored and it just returns -EINVAL.
1583 simple_nosetlease(struct file
*filp
, int arg
, struct file_lock
**flp
,
1588 EXPORT_SYMBOL(simple_nosetlease
);
1591 * simple_get_link - generic helper to get the target of "fast" symlinks
1592 * @dentry: not used here
1593 * @inode: the symlink inode
1594 * @done: not used here
1596 * Generic helper for filesystems to use for symlink inodes where a pointer to
1597 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1598 * since as an optimization the path lookup code uses any non-NULL ->i_link
1599 * directly, without calling ->get_link(). But ->get_link() still must be set,
1600 * to mark the inode_operations as being for a symlink.
1602 * Return: the symlink target
1604 const char *simple_get_link(struct dentry
*dentry
, struct inode
*inode
,
1605 struct delayed_call
*done
)
1607 return inode
->i_link
;
1609 EXPORT_SYMBOL(simple_get_link
);
1611 const struct inode_operations simple_symlink_inode_operations
= {
1612 .get_link
= simple_get_link
,
1614 EXPORT_SYMBOL(simple_symlink_inode_operations
);
1617 * Operations for a permanently empty directory.
1619 static struct dentry
*empty_dir_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
1621 return ERR_PTR(-ENOENT
);
1624 static int empty_dir_getattr(struct mnt_idmap
*idmap
,
1625 const struct path
*path
, struct kstat
*stat
,
1626 u32 request_mask
, unsigned int query_flags
)
1628 struct inode
*inode
= d_inode(path
->dentry
);
1629 generic_fillattr(&nop_mnt_idmap
, request_mask
, inode
, stat
);
1633 static int empty_dir_setattr(struct mnt_idmap
*idmap
,
1634 struct dentry
*dentry
, struct iattr
*attr
)
1639 static ssize_t
empty_dir_listxattr(struct dentry
*dentry
, char *list
, size_t size
)
1644 static const struct inode_operations empty_dir_inode_operations
= {
1645 .lookup
= empty_dir_lookup
,
1646 .permission
= generic_permission
,
1647 .setattr
= empty_dir_setattr
,
1648 .getattr
= empty_dir_getattr
,
1649 .listxattr
= empty_dir_listxattr
,
1652 static loff_t
empty_dir_llseek(struct file
*file
, loff_t offset
, int whence
)
1654 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1655 return generic_file_llseek_size(file
, offset
, whence
, 2, 2);
1658 static int empty_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
1660 dir_emit_dots(file
, ctx
);
1664 static const struct file_operations empty_dir_operations
= {
1665 .llseek
= empty_dir_llseek
,
1666 .read
= generic_read_dir
,
1667 .iterate_shared
= empty_dir_readdir
,
1668 .fsync
= noop_fsync
,
1672 void make_empty_dir_inode(struct inode
*inode
)
1674 set_nlink(inode
, 2);
1675 inode
->i_mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
1676 inode
->i_uid
= GLOBAL_ROOT_UID
;
1677 inode
->i_gid
= GLOBAL_ROOT_GID
;
1680 inode
->i_blkbits
= PAGE_SHIFT
;
1681 inode
->i_blocks
= 0;
1683 inode
->i_op
= &empty_dir_inode_operations
;
1684 inode
->i_opflags
&= ~IOP_XATTR
;
1685 inode
->i_fop
= &empty_dir_operations
;
1688 bool is_empty_dir_inode(struct inode
*inode
)
1690 return (inode
->i_fop
== &empty_dir_operations
) &&
1691 (inode
->i_op
== &empty_dir_inode_operations
);
1694 #if IS_ENABLED(CONFIG_UNICODE)
1696 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1697 * @dentry: dentry whose name we are checking against
1698 * @len: len of name of dentry
1699 * @str: str pointer to name of dentry
1700 * @name: Name to compare against
1702 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1704 static int generic_ci_d_compare(const struct dentry
*dentry
, unsigned int len
,
1705 const char *str
, const struct qstr
*name
)
1707 const struct dentry
*parent
= READ_ONCE(dentry
->d_parent
);
1708 const struct inode
*dir
= READ_ONCE(parent
->d_inode
);
1709 const struct super_block
*sb
= dentry
->d_sb
;
1710 const struct unicode_map
*um
= sb
->s_encoding
;
1711 struct qstr qstr
= QSTR_INIT(str
, len
);
1712 char strbuf
[DNAME_INLINE_LEN
];
1715 if (!dir
|| !IS_CASEFOLDED(dir
))
1718 * If the dentry name is stored in-line, then it may be concurrently
1719 * modified by a rename. If this happens, the VFS will eventually retry
1720 * the lookup, so it doesn't matter what ->d_compare() returns.
1721 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1722 * string. Therefore, we have to copy the name into a temporary buffer.
1724 if (len
<= DNAME_INLINE_LEN
- 1) {
1725 memcpy(strbuf
, str
, len
);
1728 /* prevent compiler from optimizing out the temporary buffer */
1731 ret
= utf8_strncasecmp(um
, name
, &qstr
);
1735 if (sb_has_strict_encoding(sb
))
1738 if (len
!= name
->len
)
1740 return !!memcmp(str
, name
->name
, len
);
1744 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1745 * @dentry: dentry of the parent directory
1746 * @str: qstr of name whose hash we should fill in
1748 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1750 static int generic_ci_d_hash(const struct dentry
*dentry
, struct qstr
*str
)
1752 const struct inode
*dir
= READ_ONCE(dentry
->d_inode
);
1753 struct super_block
*sb
= dentry
->d_sb
;
1754 const struct unicode_map
*um
= sb
->s_encoding
;
1757 if (!dir
|| !IS_CASEFOLDED(dir
))
1760 ret
= utf8_casefold_hash(um
, dentry
, str
);
1761 if (ret
< 0 && sb_has_strict_encoding(sb
))
1766 static const struct dentry_operations generic_ci_dentry_ops
= {
1767 .d_hash
= generic_ci_d_hash
,
1768 .d_compare
= generic_ci_d_compare
,
1772 #ifdef CONFIG_FS_ENCRYPTION
1773 static const struct dentry_operations generic_encrypted_dentry_ops
= {
1774 .d_revalidate
= fscrypt_d_revalidate
,
1778 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1779 static const struct dentry_operations generic_encrypted_ci_dentry_ops
= {
1780 .d_hash
= generic_ci_d_hash
,
1781 .d_compare
= generic_ci_d_compare
,
1782 .d_revalidate
= fscrypt_d_revalidate
,
1787 * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1788 * @dentry: dentry to set ops on
1790 * Casefolded directories need d_hash and d_compare set, so that the dentries
1791 * contained in them are handled case-insensitively. Note that these operations
1792 * are needed on the parent directory rather than on the dentries in it, and
1793 * while the casefolding flag can be toggled on and off on an empty directory,
1794 * dentry_operations can't be changed later. As a result, if the filesystem has
1795 * casefolding support enabled at all, we have to give all dentries the
1796 * casefolding operations even if their inode doesn't have the casefolding flag
1797 * currently (and thus the casefolding ops would be no-ops for now).
1799 * Encryption works differently in that the only dentry operation it needs is
1800 * d_revalidate, which it only needs on dentries that have the no-key name flag.
1801 * The no-key flag can't be set "later", so we don't have to worry about that.
1803 * Finally, to maximize compatibility with overlayfs (which isn't compatible
1804 * with certain dentry operations) and to avoid taking an unnecessary
1805 * performance hit, we use custom dentry_operations for each possible
1806 * combination rather than always installing all operations.
1808 void generic_set_encrypted_ci_d_ops(struct dentry
*dentry
)
1810 #ifdef CONFIG_FS_ENCRYPTION
1811 bool needs_encrypt_ops
= dentry
->d_flags
& DCACHE_NOKEY_NAME
;
1813 #if IS_ENABLED(CONFIG_UNICODE)
1814 bool needs_ci_ops
= dentry
->d_sb
->s_encoding
;
1816 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1817 if (needs_encrypt_ops
&& needs_ci_ops
) {
1818 d_set_d_op(dentry
, &generic_encrypted_ci_dentry_ops
);
1822 #ifdef CONFIG_FS_ENCRYPTION
1823 if (needs_encrypt_ops
) {
1824 d_set_d_op(dentry
, &generic_encrypted_dentry_ops
);
1828 #if IS_ENABLED(CONFIG_UNICODE)
1830 d_set_d_op(dentry
, &generic_ci_dentry_ops
);
1835 EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops
);
1838 * inode_maybe_inc_iversion - increments i_version
1839 * @inode: inode with the i_version that should be updated
1840 * @force: increment the counter even if it's not necessary?
1842 * Every time the inode is modified, the i_version field must be seen to have
1843 * changed by any observer.
1845 * If "force" is set or the QUERIED flag is set, then ensure that we increment
1846 * the value, and clear the queried flag.
1848 * In the common case where neither is set, then we can return "false" without
1849 * updating i_version.
1851 * If this function returns false, and no other metadata has changed, then we
1852 * can avoid logging the metadata.
1854 bool inode_maybe_inc_iversion(struct inode
*inode
, bool force
)
1859 * The i_version field is not strictly ordered with any other inode
1860 * information, but the legacy inode_inc_iversion code used a spinlock
1861 * to serialize increments.
1863 * Here, we add full memory barriers to ensure that any de-facto
1864 * ordering with other info is preserved.
1866 * This barrier pairs with the barrier in inode_query_iversion()
1869 cur
= inode_peek_iversion_raw(inode
);
1871 /* If flag is clear then we needn't do anything */
1872 if (!force
&& !(cur
& I_VERSION_QUERIED
))
1875 /* Since lowest bit is flag, add 2 to avoid it */
1876 new = (cur
& ~I_VERSION_QUERIED
) + I_VERSION_INCREMENT
;
1877 } while (!atomic64_try_cmpxchg(&inode
->i_version
, &cur
, new));
1880 EXPORT_SYMBOL(inode_maybe_inc_iversion
);
1883 * inode_query_iversion - read i_version for later use
1884 * @inode: inode from which i_version should be read
1886 * Read the inode i_version counter. This should be used by callers that wish
1887 * to store the returned i_version for later comparison. This will guarantee
1888 * that a later query of the i_version will result in a different value if
1889 * anything has changed.
1891 * In this implementation, we fetch the current value, set the QUERIED flag and
1892 * then try to swap it into place with a cmpxchg, if it wasn't already set. If
1893 * that fails, we try again with the newly fetched value from the cmpxchg.
1895 u64
inode_query_iversion(struct inode
*inode
)
1899 cur
= inode_peek_iversion_raw(inode
);
1901 /* If flag is already set, then no need to swap */
1902 if (cur
& I_VERSION_QUERIED
) {
1904 * This barrier (and the implicit barrier in the
1905 * cmpxchg below) pairs with the barrier in
1906 * inode_maybe_inc_iversion().
1912 new = cur
| I_VERSION_QUERIED
;
1913 } while (!atomic64_try_cmpxchg(&inode
->i_version
, &cur
, new));
1914 return cur
>> I_VERSION_QUERIED_SHIFT
;
1916 EXPORT_SYMBOL(inode_query_iversion
);
1918 ssize_t
direct_write_fallback(struct kiocb
*iocb
, struct iov_iter
*iter
,
1919 ssize_t direct_written
, ssize_t buffered_written
)
1921 struct address_space
*mapping
= iocb
->ki_filp
->f_mapping
;
1922 loff_t pos
= iocb
->ki_pos
- buffered_written
;
1923 loff_t end
= iocb
->ki_pos
- 1;
1927 * If the buffered write fallback returned an error, we want to return
1928 * the number of bytes which were written by direct I/O, or the error
1929 * code if that was zero.
1931 * Note that this differs from normal direct-io semantics, which will
1932 * return -EFOO even if some bytes were written.
1934 if (unlikely(buffered_written
< 0)) {
1936 return direct_written
;
1937 return buffered_written
;
1941 * We need to ensure that the page cache pages are written to disk and
1942 * invalidated to preserve the expected O_DIRECT semantics.
1944 err
= filemap_write_and_wait_range(mapping
, pos
, end
);
1947 * We don't know how much we wrote, so just return the number of
1948 * bytes which were direct-written
1950 iocb
->ki_pos
-= buffered_written
;
1952 return direct_written
;
1955 invalidate_mapping_pages(mapping
, pos
>> PAGE_SHIFT
, end
>> PAGE_SHIFT
);
1956 return direct_written
+ buffered_written
;
1958 EXPORT_SYMBOL_GPL(direct_write_fallback
);
1961 * simple_inode_init_ts - initialize the timestamps for a new inode
1962 * @inode: inode to be initialized
1964 * When a new inode is created, most filesystems set the timestamps to the
1965 * current time. Add a helper to do this.
1967 struct timespec64
simple_inode_init_ts(struct inode
*inode
)
1969 struct timespec64 ts
= inode_set_ctime_current(inode
);
1971 inode_set_atime_to_ts(inode
, ts
);
1972 inode_set_mtime_to_ts(inode
, ts
);
1975 EXPORT_SYMBOL(simple_inode_init_ts
);