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>
26 #include <linux/pidfs.h>
28 #include <linux/uaccess.h>
32 int simple_getattr(struct mnt_idmap
*idmap
, const struct path
*path
,
33 struct kstat
*stat
, u32 request_mask
,
34 unsigned int query_flags
)
36 struct inode
*inode
= d_inode(path
->dentry
);
37 generic_fillattr(&nop_mnt_idmap
, request_mask
, inode
, stat
);
38 stat
->blocks
= inode
->i_mapping
->nrpages
<< (PAGE_SHIFT
- 9);
41 EXPORT_SYMBOL(simple_getattr
);
43 int simple_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
45 u64 id
= huge_encode_dev(dentry
->d_sb
->s_dev
);
47 buf
->f_fsid
= u64_to_fsid(id
);
48 buf
->f_type
= dentry
->d_sb
->s_magic
;
49 buf
->f_bsize
= PAGE_SIZE
;
50 buf
->f_namelen
= NAME_MAX
;
53 EXPORT_SYMBOL(simple_statfs
);
56 * Retaining negative dentries for an in-memory filesystem just wastes
57 * memory and lookup time: arrange for them to be deleted immediately.
59 int always_delete_dentry(const struct dentry
*dentry
)
63 EXPORT_SYMBOL(always_delete_dentry
);
65 const struct dentry_operations simple_dentry_operations
= {
66 .d_delete
= always_delete_dentry
,
68 EXPORT_SYMBOL(simple_dentry_operations
);
71 * Lookup the data. This is trivial - if the dentry didn't already
72 * exist, we know it is negative. Set d_op to delete negative dentries.
74 struct dentry
*simple_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
76 if (dentry
->d_name
.len
> NAME_MAX
)
77 return ERR_PTR(-ENAMETOOLONG
);
78 if (!dentry
->d_sb
->s_d_op
)
79 d_set_d_op(dentry
, &simple_dentry_operations
);
83 EXPORT_SYMBOL(simple_lookup
);
85 int dcache_dir_open(struct inode
*inode
, struct file
*file
)
87 file
->private_data
= d_alloc_cursor(file
->f_path
.dentry
);
89 return file
->private_data
? 0 : -ENOMEM
;
91 EXPORT_SYMBOL(dcache_dir_open
);
93 int dcache_dir_close(struct inode
*inode
, struct file
*file
)
95 dput(file
->private_data
);
98 EXPORT_SYMBOL(dcache_dir_close
);
100 /* parent is locked at least shared */
102 * Returns an element of siblings' list.
103 * We are looking for <count>th positive after <p>; if
104 * found, dentry is grabbed and returned to caller.
105 * If no such element exists, NULL is returned.
107 static struct dentry
*scan_positives(struct dentry
*cursor
,
108 struct hlist_node
**p
,
112 struct dentry
*dentry
= cursor
->d_parent
, *found
= NULL
;
114 spin_lock(&dentry
->d_lock
);
116 struct dentry
*d
= hlist_entry(*p
, struct dentry
, d_sib
);
118 // we must at least skip cursors, to avoid livelocks
119 if (d
->d_flags
& DCACHE_DENTRY_CURSOR
)
121 if (simple_positive(d
) && !--count
) {
122 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
123 if (simple_positive(d
))
124 found
= dget_dlock(d
);
125 spin_unlock(&d
->d_lock
);
130 if (need_resched()) {
131 if (!hlist_unhashed(&cursor
->d_sib
))
132 __hlist_del(&cursor
->d_sib
);
133 hlist_add_behind(&cursor
->d_sib
, &d
->d_sib
);
134 p
= &cursor
->d_sib
.next
;
135 spin_unlock(&dentry
->d_lock
);
137 spin_lock(&dentry
->d_lock
);
140 spin_unlock(&dentry
->d_lock
);
145 loff_t
dcache_dir_lseek(struct file
*file
, loff_t offset
, int whence
)
147 struct dentry
*dentry
= file
->f_path
.dentry
;
150 offset
+= file
->f_pos
;
159 if (offset
!= file
->f_pos
) {
160 struct dentry
*cursor
= file
->private_data
;
161 struct dentry
*to
= NULL
;
163 inode_lock_shared(dentry
->d_inode
);
166 to
= scan_positives(cursor
, &dentry
->d_children
.first
,
168 spin_lock(&dentry
->d_lock
);
169 hlist_del_init(&cursor
->d_sib
);
171 hlist_add_behind(&cursor
->d_sib
, &to
->d_sib
);
172 spin_unlock(&dentry
->d_lock
);
175 file
->f_pos
= offset
;
177 inode_unlock_shared(dentry
->d_inode
);
181 EXPORT_SYMBOL(dcache_dir_lseek
);
184 * Directory is locked and all positive dentries in it are safe, since
185 * for ramfs-type trees they can't go away without unlink() or rmdir(),
186 * both impossible due to the lock on directory.
189 int dcache_readdir(struct file
*file
, struct dir_context
*ctx
)
191 struct dentry
*dentry
= file
->f_path
.dentry
;
192 struct dentry
*cursor
= file
->private_data
;
193 struct dentry
*next
= NULL
;
194 struct hlist_node
**p
;
196 if (!dir_emit_dots(file
, ctx
))
200 p
= &dentry
->d_children
.first
;
202 p
= &cursor
->d_sib
.next
;
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
)))
210 p
= &next
->d_sib
.next
;
212 spin_lock(&dentry
->d_lock
);
213 hlist_del_init(&cursor
->d_sib
);
215 hlist_add_before(&cursor
->d_sib
, &next
->d_sib
);
216 spin_unlock(&dentry
->d_lock
);
221 EXPORT_SYMBOL(dcache_readdir
);
223 ssize_t
generic_read_dir(struct file
*filp
, char __user
*buf
, size_t siz
, loff_t
*ppos
)
227 EXPORT_SYMBOL(generic_read_dir
);
229 const struct file_operations simple_dir_operations
= {
230 .open
= dcache_dir_open
,
231 .release
= dcache_dir_close
,
232 .llseek
= dcache_dir_lseek
,
233 .read
= generic_read_dir
,
234 .iterate_shared
= dcache_readdir
,
237 EXPORT_SYMBOL(simple_dir_operations
);
239 const struct inode_operations simple_dir_inode_operations
= {
240 .lookup
= simple_lookup
,
242 EXPORT_SYMBOL(simple_dir_inode_operations
);
244 /* 0 is '.', 1 is '..', so always start with offset 2 or more */
249 static void offset_set(struct dentry
*dentry
, long offset
)
251 dentry
->d_fsdata
= (void *)offset
;
254 static long dentry2offset(struct dentry
*dentry
)
256 return (long)dentry
->d_fsdata
;
259 static struct lock_class_key simple_offset_lock_class
;
262 * simple_offset_init - initialize an offset_ctx
263 * @octx: directory offset map to be initialized
266 void simple_offset_init(struct offset_ctx
*octx
)
268 mt_init_flags(&octx
->mt
, MT_FLAGS_ALLOC_RANGE
);
269 lockdep_set_class(&octx
->mt
.ma_lock
, &simple_offset_lock_class
);
270 octx
->next_offset
= DIR_OFFSET_MIN
;
274 * simple_offset_add - Add an entry to a directory's offset map
275 * @octx: directory offset ctx to be updated
276 * @dentry: new dentry being added
278 * Returns zero on success. @octx and the dentry's offset are updated.
279 * Otherwise, a negative errno value is returned.
281 int simple_offset_add(struct offset_ctx
*octx
, struct dentry
*dentry
)
283 unsigned long offset
;
286 if (dentry2offset(dentry
) != 0)
289 ret
= mtree_alloc_cyclic(&octx
->mt
, &offset
, dentry
, DIR_OFFSET_MIN
,
290 LONG_MAX
, &octx
->next_offset
, GFP_KERNEL
);
294 offset_set(dentry
, offset
);
299 * simple_offset_remove - Remove an entry to a directory's offset map
300 * @octx: directory offset ctx to be updated
301 * @dentry: dentry being removed
304 void simple_offset_remove(struct offset_ctx
*octx
, struct dentry
*dentry
)
308 offset
= dentry2offset(dentry
);
312 mtree_erase(&octx
->mt
, offset
);
313 offset_set(dentry
, 0);
317 * simple_offset_empty - Check if a dentry can be unlinked
318 * @dentry: dentry to be tested
320 * Returns 0 if @dentry is a non-empty directory; otherwise returns 1.
322 int simple_offset_empty(struct dentry
*dentry
)
324 struct inode
*inode
= d_inode(dentry
);
325 struct offset_ctx
*octx
;
326 struct dentry
*child
;
330 if (!inode
|| !S_ISDIR(inode
->i_mode
))
333 index
= DIR_OFFSET_MIN
;
334 octx
= inode
->i_op
->get_offset_ctx(inode
);
335 mt_for_each(&octx
->mt
, child
, index
, LONG_MAX
) {
336 spin_lock(&child
->d_lock
);
337 if (simple_positive(child
)) {
338 spin_unlock(&child
->d_lock
);
342 spin_unlock(&child
->d_lock
);
349 * simple_offset_rename_exchange - exchange rename with directory offsets
350 * @old_dir: parent of dentry being moved
351 * @old_dentry: dentry being moved
352 * @new_dir: destination parent
353 * @new_dentry: destination dentry
355 * Returns zero on success. Otherwise a negative errno is returned and the
356 * rename is rolled back.
358 int simple_offset_rename_exchange(struct inode
*old_dir
,
359 struct dentry
*old_dentry
,
360 struct inode
*new_dir
,
361 struct dentry
*new_dentry
)
363 struct offset_ctx
*old_ctx
= old_dir
->i_op
->get_offset_ctx(old_dir
);
364 struct offset_ctx
*new_ctx
= new_dir
->i_op
->get_offset_ctx(new_dir
);
365 long old_index
= dentry2offset(old_dentry
);
366 long new_index
= dentry2offset(new_dentry
);
369 simple_offset_remove(old_ctx
, old_dentry
);
370 simple_offset_remove(new_ctx
, new_dentry
);
372 ret
= simple_offset_add(new_ctx
, old_dentry
);
376 ret
= simple_offset_add(old_ctx
, new_dentry
);
378 simple_offset_remove(new_ctx
, old_dentry
);
382 ret
= simple_rename_exchange(old_dir
, old_dentry
, new_dir
, new_dentry
);
384 simple_offset_remove(new_ctx
, old_dentry
);
385 simple_offset_remove(old_ctx
, new_dentry
);
391 offset_set(old_dentry
, old_index
);
392 mtree_store(&old_ctx
->mt
, old_index
, old_dentry
, GFP_KERNEL
);
393 offset_set(new_dentry
, new_index
);
394 mtree_store(&new_ctx
->mt
, new_index
, new_dentry
, GFP_KERNEL
);
399 * simple_offset_destroy - Release offset map
400 * @octx: directory offset ctx that is about to be destroyed
402 * During fs teardown (eg. umount), a directory's offset map might still
403 * contain entries. xa_destroy() cleans out anything that remains.
405 void simple_offset_destroy(struct offset_ctx
*octx
)
407 mtree_destroy(&octx
->mt
);
411 * offset_dir_llseek - Advance the read position of a directory descriptor
412 * @file: an open directory whose position is to be updated
413 * @offset: a byte offset
414 * @whence: enumerator describing the starting position for this update
416 * SEEK_END, SEEK_DATA, and SEEK_HOLE are not supported for directories.
418 * Returns the updated read position if successful; otherwise a
419 * negative errno is returned and the read position remains unchanged.
421 static loff_t
offset_dir_llseek(struct file
*file
, loff_t offset
, int whence
)
425 offset
+= file
->f_pos
;
435 /* In this case, ->private_data is protected by f_pos_lock */
436 file
->private_data
= NULL
;
437 return vfs_setpos(file
, offset
, LONG_MAX
);
440 static struct dentry
*offset_find_next(struct offset_ctx
*octx
, loff_t offset
)
442 MA_STATE(mas
, &octx
->mt
, offset
, offset
);
443 struct dentry
*child
, *found
= NULL
;
446 child
= mas_find(&mas
, LONG_MAX
);
449 spin_lock(&child
->d_lock
);
450 if (simple_positive(child
))
451 found
= dget_dlock(child
);
452 spin_unlock(&child
->d_lock
);
458 static bool offset_dir_emit(struct dir_context
*ctx
, struct dentry
*dentry
)
460 struct inode
*inode
= d_inode(dentry
);
461 long offset
= dentry2offset(dentry
);
463 return ctx
->actor(ctx
, dentry
->d_name
.name
, dentry
->d_name
.len
, offset
,
464 inode
->i_ino
, fs_umode_to_dtype(inode
->i_mode
));
467 static void *offset_iterate_dir(struct inode
*inode
, struct dir_context
*ctx
)
469 struct offset_ctx
*octx
= inode
->i_op
->get_offset_ctx(inode
);
470 struct dentry
*dentry
;
473 dentry
= offset_find_next(octx
, ctx
->pos
);
475 return ERR_PTR(-ENOENT
);
477 if (!offset_dir_emit(ctx
, dentry
)) {
482 ctx
->pos
= dentry2offset(dentry
) + 1;
489 * offset_readdir - Emit entries starting at offset @ctx->pos
490 * @file: an open directory to iterate over
491 * @ctx: directory iteration context
493 * Caller must hold @file's i_rwsem to prevent insertion or removal of
494 * entries during this call.
496 * On entry, @ctx->pos contains an offset that represents the first entry
497 * to be read from the directory.
499 * The operation continues until there are no more entries to read, or
500 * until the ctx->actor indicates there is no more space in the caller's
503 * On return, @ctx->pos contains an offset that will read the next entry
504 * in this directory when offset_readdir() is called again with @ctx.
509 static int offset_readdir(struct file
*file
, struct dir_context
*ctx
)
511 struct dentry
*dir
= file
->f_path
.dentry
;
513 lockdep_assert_held(&d_inode(dir
)->i_rwsem
);
515 if (!dir_emit_dots(file
, ctx
))
518 /* In this case, ->private_data is protected by f_pos_lock */
519 if (ctx
->pos
== DIR_OFFSET_MIN
)
520 file
->private_data
= NULL
;
521 else if (file
->private_data
== ERR_PTR(-ENOENT
))
523 file
->private_data
= offset_iterate_dir(d_inode(dir
), ctx
);
527 const struct file_operations simple_offset_dir_operations
= {
528 .llseek
= offset_dir_llseek
,
529 .iterate_shared
= offset_readdir
,
530 .read
= generic_read_dir
,
534 static struct dentry
*find_next_child(struct dentry
*parent
, struct dentry
*prev
)
536 struct dentry
*child
= NULL
, *d
;
538 spin_lock(&parent
->d_lock
);
539 d
= prev
? d_next_sibling(prev
) : d_first_child(parent
);
540 hlist_for_each_entry_from(d
, d_sib
) {
541 if (simple_positive(d
)) {
542 spin_lock_nested(&d
->d_lock
, DENTRY_D_LOCK_NESTED
);
543 if (simple_positive(d
))
544 child
= dget_dlock(d
);
545 spin_unlock(&d
->d_lock
);
550 spin_unlock(&parent
->d_lock
);
555 void simple_recursive_removal(struct dentry
*dentry
,
556 void (*callback
)(struct dentry
*))
558 struct dentry
*this = dget(dentry
);
560 struct dentry
*victim
= NULL
, *child
;
561 struct inode
*inode
= this->d_inode
;
565 inode
->i_flags
|= S_DEAD
;
566 while ((child
= find_next_child(this, victim
)) == NULL
) {
568 // update metadata while it's still locked
569 inode_set_ctime_current(inode
);
573 this = this->d_parent
;
574 inode
= this->d_inode
;
576 if (simple_positive(victim
)) {
577 d_invalidate(victim
); // avoid lost mounts
578 if (d_is_dir(victim
))
579 fsnotify_rmdir(inode
, victim
);
581 fsnotify_unlink(inode
, victim
);
584 dput(victim
); // unpin it
586 if (victim
== dentry
) {
587 inode_set_mtime_to_ts(inode
,
588 inode_set_ctime_current(inode
));
589 if (d_is_dir(dentry
))
600 EXPORT_SYMBOL(simple_recursive_removal
);
602 static const struct super_operations simple_super_operations
= {
603 .statfs
= simple_statfs
,
606 static int pseudo_fs_fill_super(struct super_block
*s
, struct fs_context
*fc
)
608 struct pseudo_fs_context
*ctx
= fc
->fs_private
;
611 s
->s_maxbytes
= MAX_LFS_FILESIZE
;
612 s
->s_blocksize
= PAGE_SIZE
;
613 s
->s_blocksize_bits
= PAGE_SHIFT
;
614 s
->s_magic
= ctx
->magic
;
615 s
->s_op
= ctx
->ops
?: &simple_super_operations
;
616 s
->s_xattr
= ctx
->xattr
;
623 * since this is the first inode, make it number 1. New inodes created
624 * after this must take care not to collide with it (by passing
625 * max_reserved of 1 to iunique).
628 root
->i_mode
= S_IFDIR
| S_IRUSR
| S_IWUSR
;
629 simple_inode_init_ts(root
);
630 s
->s_root
= d_make_root(root
);
633 s
->s_d_op
= ctx
->dops
;
637 static int pseudo_fs_get_tree(struct fs_context
*fc
)
639 return get_tree_nodev(fc
, pseudo_fs_fill_super
);
642 static void pseudo_fs_free(struct fs_context
*fc
)
644 kfree(fc
->fs_private
);
647 static const struct fs_context_operations pseudo_fs_context_ops
= {
648 .free
= pseudo_fs_free
,
649 .get_tree
= pseudo_fs_get_tree
,
653 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
654 * will never be mountable)
656 struct pseudo_fs_context
*init_pseudo(struct fs_context
*fc
,
659 struct pseudo_fs_context
*ctx
;
661 ctx
= kzalloc(sizeof(struct pseudo_fs_context
), GFP_KERNEL
);
664 fc
->fs_private
= ctx
;
665 fc
->ops
= &pseudo_fs_context_ops
;
666 fc
->sb_flags
|= SB_NOUSER
;
671 EXPORT_SYMBOL(init_pseudo
);
673 int simple_open(struct inode
*inode
, struct file
*file
)
675 if (inode
->i_private
)
676 file
->private_data
= inode
->i_private
;
679 EXPORT_SYMBOL(simple_open
);
681 int simple_link(struct dentry
*old_dentry
, struct inode
*dir
, struct dentry
*dentry
)
683 struct inode
*inode
= d_inode(old_dentry
);
685 inode_set_mtime_to_ts(dir
,
686 inode_set_ctime_to_ts(dir
, inode_set_ctime_current(inode
)));
690 d_instantiate(dentry
, inode
);
693 EXPORT_SYMBOL(simple_link
);
695 int simple_empty(struct dentry
*dentry
)
697 struct dentry
*child
;
700 spin_lock(&dentry
->d_lock
);
701 hlist_for_each_entry(child
, &dentry
->d_children
, d_sib
) {
702 spin_lock_nested(&child
->d_lock
, DENTRY_D_LOCK_NESTED
);
703 if (simple_positive(child
)) {
704 spin_unlock(&child
->d_lock
);
707 spin_unlock(&child
->d_lock
);
711 spin_unlock(&dentry
->d_lock
);
714 EXPORT_SYMBOL(simple_empty
);
716 int simple_unlink(struct inode
*dir
, struct dentry
*dentry
)
718 struct inode
*inode
= d_inode(dentry
);
720 inode_set_mtime_to_ts(dir
,
721 inode_set_ctime_to_ts(dir
, inode_set_ctime_current(inode
)));
726 EXPORT_SYMBOL(simple_unlink
);
728 int simple_rmdir(struct inode
*dir
, struct dentry
*dentry
)
730 if (!simple_empty(dentry
))
733 drop_nlink(d_inode(dentry
));
734 simple_unlink(dir
, dentry
);
738 EXPORT_SYMBOL(simple_rmdir
);
741 * simple_rename_timestamp - update the various inode timestamps for rename
742 * @old_dir: old parent directory
743 * @old_dentry: dentry that is being renamed
744 * @new_dir: new parent directory
745 * @new_dentry: target for rename
747 * POSIX mandates that the old and new parent directories have their ctime and
748 * mtime updated, and that inodes of @old_dentry and @new_dentry (if any), have
749 * their ctime updated.
751 void simple_rename_timestamp(struct inode
*old_dir
, struct dentry
*old_dentry
,
752 struct inode
*new_dir
, struct dentry
*new_dentry
)
754 struct inode
*newino
= d_inode(new_dentry
);
756 inode_set_mtime_to_ts(old_dir
, inode_set_ctime_current(old_dir
));
757 if (new_dir
!= old_dir
)
758 inode_set_mtime_to_ts(new_dir
,
759 inode_set_ctime_current(new_dir
));
760 inode_set_ctime_current(d_inode(old_dentry
));
762 inode_set_ctime_current(newino
);
764 EXPORT_SYMBOL_GPL(simple_rename_timestamp
);
766 int simple_rename_exchange(struct inode
*old_dir
, struct dentry
*old_dentry
,
767 struct inode
*new_dir
, struct dentry
*new_dentry
)
769 bool old_is_dir
= d_is_dir(old_dentry
);
770 bool new_is_dir
= d_is_dir(new_dentry
);
772 if (old_dir
!= new_dir
&& old_is_dir
!= new_is_dir
) {
781 simple_rename_timestamp(old_dir
, old_dentry
, new_dir
, new_dentry
);
784 EXPORT_SYMBOL_GPL(simple_rename_exchange
);
786 int simple_rename(struct mnt_idmap
*idmap
, struct inode
*old_dir
,
787 struct dentry
*old_dentry
, struct inode
*new_dir
,
788 struct dentry
*new_dentry
, unsigned int flags
)
790 int they_are_dirs
= d_is_dir(old_dentry
);
792 if (flags
& ~(RENAME_NOREPLACE
| RENAME_EXCHANGE
))
795 if (flags
& RENAME_EXCHANGE
)
796 return simple_rename_exchange(old_dir
, old_dentry
, new_dir
, new_dentry
);
798 if (!simple_empty(new_dentry
))
801 if (d_really_is_positive(new_dentry
)) {
802 simple_unlink(new_dir
, new_dentry
);
804 drop_nlink(d_inode(new_dentry
));
807 } else if (they_are_dirs
) {
812 simple_rename_timestamp(old_dir
, old_dentry
, new_dir
, new_dentry
);
815 EXPORT_SYMBOL(simple_rename
);
818 * simple_setattr - setattr for simple filesystem
819 * @idmap: idmap of the target mount
821 * @iattr: iattr structure
823 * Returns 0 on success, -error on failure.
825 * simple_setattr is a simple ->setattr implementation without a proper
826 * implementation of size changes.
828 * It can either be used for in-memory filesystems or special files
829 * on simple regular filesystems. Anything that needs to change on-disk
830 * or wire state on size changes needs its own setattr method.
832 int simple_setattr(struct mnt_idmap
*idmap
, struct dentry
*dentry
,
835 struct inode
*inode
= d_inode(dentry
);
838 error
= setattr_prepare(idmap
, dentry
, iattr
);
842 if (iattr
->ia_valid
& ATTR_SIZE
)
843 truncate_setsize(inode
, iattr
->ia_size
);
844 setattr_copy(idmap
, inode
, iattr
);
845 mark_inode_dirty(inode
);
848 EXPORT_SYMBOL(simple_setattr
);
850 static int simple_read_folio(struct file
*file
, struct folio
*folio
)
852 folio_zero_range(folio
, 0, folio_size(folio
));
853 flush_dcache_folio(folio
);
854 folio_mark_uptodate(folio
);
859 int simple_write_begin(struct file
*file
, struct address_space
*mapping
,
860 loff_t pos
, unsigned len
,
861 struct page
**pagep
, void **fsdata
)
865 folio
= __filemap_get_folio(mapping
, pos
/ PAGE_SIZE
, FGP_WRITEBEGIN
,
866 mapping_gfp_mask(mapping
));
868 return PTR_ERR(folio
);
870 *pagep
= &folio
->page
;
872 if (!folio_test_uptodate(folio
) && (len
!= folio_size(folio
))) {
873 size_t from
= offset_in_folio(folio
, pos
);
875 folio_zero_segments(folio
, 0, from
,
876 from
+ len
, folio_size(folio
));
880 EXPORT_SYMBOL(simple_write_begin
);
883 * simple_write_end - .write_end helper for non-block-device FSes
884 * @file: See .write_end of address_space_operations
892 * simple_write_end does the minimum needed for updating a page after writing is
893 * done. It has the same API signature as the .write_end of
894 * address_space_operations vector. So it can just be set onto .write_end for
895 * FSes that don't need any other processing. i_mutex is assumed to be held.
896 * Block based filesystems should use generic_write_end().
897 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
898 * is not called, so a filesystem that actually does store data in .write_inode
899 * should extend on what's done here with a call to mark_inode_dirty() in the
900 * case that i_size has changed.
902 * Use *ONLY* with simple_read_folio()
904 static int simple_write_end(struct file
*file
, struct address_space
*mapping
,
905 loff_t pos
, unsigned len
, unsigned copied
,
906 struct page
*page
, void *fsdata
)
908 struct folio
*folio
= page_folio(page
);
909 struct inode
*inode
= folio
->mapping
->host
;
910 loff_t last_pos
= pos
+ copied
;
912 /* zero the stale part of the folio if we did a short copy */
913 if (!folio_test_uptodate(folio
)) {
915 size_t from
= offset_in_folio(folio
, pos
);
917 folio_zero_range(folio
, from
+ copied
, len
- copied
);
919 folio_mark_uptodate(folio
);
922 * No need to use i_size_read() here, the i_size
923 * cannot change under us because we hold the i_mutex.
925 if (last_pos
> inode
->i_size
)
926 i_size_write(inode
, last_pos
);
928 folio_mark_dirty(folio
);
936 * Provides ramfs-style behavior: data in the pagecache, but no writeback.
938 const struct address_space_operations ram_aops
= {
939 .read_folio
= simple_read_folio
,
940 .write_begin
= simple_write_begin
,
941 .write_end
= simple_write_end
,
942 .dirty_folio
= noop_dirty_folio
,
944 EXPORT_SYMBOL(ram_aops
);
947 * the inodes created here are not hashed. If you use iunique to generate
948 * unique inode values later for this filesystem, then you must take care
949 * to pass it an appropriate max_reserved value to avoid collisions.
951 int simple_fill_super(struct super_block
*s
, unsigned long magic
,
952 const struct tree_descr
*files
)
955 struct dentry
*dentry
;
958 s
->s_blocksize
= PAGE_SIZE
;
959 s
->s_blocksize_bits
= PAGE_SHIFT
;
961 s
->s_op
= &simple_super_operations
;
964 inode
= new_inode(s
);
968 * because the root inode is 1, the files array must not contain an
972 inode
->i_mode
= S_IFDIR
| 0755;
973 simple_inode_init_ts(inode
);
974 inode
->i_op
= &simple_dir_inode_operations
;
975 inode
->i_fop
= &simple_dir_operations
;
977 s
->s_root
= d_make_root(inode
);
980 for (i
= 0; !files
->name
|| files
->name
[0]; i
++, files
++) {
984 /* warn if it tries to conflict with the root inode */
985 if (unlikely(i
== 1))
986 printk(KERN_WARNING
"%s: %s passed in a files array"
987 "with an index of 1!\n", __func__
,
990 dentry
= d_alloc_name(s
->s_root
, files
->name
);
993 inode
= new_inode(s
);
998 inode
->i_mode
= S_IFREG
| files
->mode
;
999 simple_inode_init_ts(inode
);
1000 inode
->i_fop
= files
->ops
;
1002 d_add(dentry
, inode
);
1006 EXPORT_SYMBOL(simple_fill_super
);
1008 static DEFINE_SPINLOCK(pin_fs_lock
);
1010 int simple_pin_fs(struct file_system_type
*type
, struct vfsmount
**mount
, int *count
)
1012 struct vfsmount
*mnt
= NULL
;
1013 spin_lock(&pin_fs_lock
);
1014 if (unlikely(!*mount
)) {
1015 spin_unlock(&pin_fs_lock
);
1016 mnt
= vfs_kern_mount(type
, SB_KERNMOUNT
, type
->name
, NULL
);
1018 return PTR_ERR(mnt
);
1019 spin_lock(&pin_fs_lock
);
1025 spin_unlock(&pin_fs_lock
);
1029 EXPORT_SYMBOL(simple_pin_fs
);
1031 void simple_release_fs(struct vfsmount
**mount
, int *count
)
1033 struct vfsmount
*mnt
;
1034 spin_lock(&pin_fs_lock
);
1038 spin_unlock(&pin_fs_lock
);
1041 EXPORT_SYMBOL(simple_release_fs
);
1044 * simple_read_from_buffer - copy data from the buffer to user space
1045 * @to: the user space buffer to read to
1046 * @count: the maximum number of bytes to read
1047 * @ppos: the current position in the buffer
1048 * @from: the buffer to read from
1049 * @available: the size of the buffer
1051 * The simple_read_from_buffer() function reads up to @count bytes from the
1052 * buffer @from at offset @ppos into the user space address starting at @to.
1054 * On success, the number of bytes read is returned and the offset @ppos is
1055 * advanced by this number, or negative value is returned on error.
1057 ssize_t
simple_read_from_buffer(void __user
*to
, size_t count
, loff_t
*ppos
,
1058 const void *from
, size_t available
)
1065 if (pos
>= available
|| !count
)
1067 if (count
> available
- pos
)
1068 count
= available
- pos
;
1069 ret
= copy_to_user(to
, from
+ pos
, count
);
1073 *ppos
= pos
+ count
;
1076 EXPORT_SYMBOL(simple_read_from_buffer
);
1079 * simple_write_to_buffer - copy data from user space to the buffer
1080 * @to: the buffer to write to
1081 * @available: the size of the buffer
1082 * @ppos: the current position in the buffer
1083 * @from: the user space buffer to read from
1084 * @count: the maximum number of bytes to read
1086 * The simple_write_to_buffer() function reads up to @count bytes from the user
1087 * space address starting at @from into the buffer @to at offset @ppos.
1089 * On success, the number of bytes written is returned and the offset @ppos is
1090 * advanced by this number, or negative value is returned on error.
1092 ssize_t
simple_write_to_buffer(void *to
, size_t available
, loff_t
*ppos
,
1093 const void __user
*from
, size_t count
)
1100 if (pos
>= available
|| !count
)
1102 if (count
> available
- pos
)
1103 count
= available
- pos
;
1104 res
= copy_from_user(to
+ pos
, from
, count
);
1108 *ppos
= pos
+ count
;
1111 EXPORT_SYMBOL(simple_write_to_buffer
);
1114 * memory_read_from_buffer - copy data from the buffer
1115 * @to: the kernel space buffer to read to
1116 * @count: the maximum number of bytes to read
1117 * @ppos: the current position in the buffer
1118 * @from: the buffer to read from
1119 * @available: the size of the buffer
1121 * The memory_read_from_buffer() function reads up to @count bytes from the
1122 * buffer @from at offset @ppos into the kernel space address starting at @to.
1124 * On success, the number of bytes read is returned and the offset @ppos is
1125 * advanced by this number, or negative value is returned on error.
1127 ssize_t
memory_read_from_buffer(void *to
, size_t count
, loff_t
*ppos
,
1128 const void *from
, size_t available
)
1134 if (pos
>= available
)
1136 if (count
> available
- pos
)
1137 count
= available
- pos
;
1138 memcpy(to
, from
+ pos
, count
);
1139 *ppos
= pos
+ count
;
1143 EXPORT_SYMBOL(memory_read_from_buffer
);
1146 * Transaction based IO.
1147 * The file expects a single write which triggers the transaction, and then
1148 * possibly a read which collects the result - which is stored in a
1149 * file-local buffer.
1152 void simple_transaction_set(struct file
*file
, size_t n
)
1154 struct simple_transaction_argresp
*ar
= file
->private_data
;
1156 BUG_ON(n
> SIMPLE_TRANSACTION_LIMIT
);
1159 * The barrier ensures that ar->size will really remain zero until
1160 * ar->data is ready for reading.
1165 EXPORT_SYMBOL(simple_transaction_set
);
1167 char *simple_transaction_get(struct file
*file
, const char __user
*buf
, size_t size
)
1169 struct simple_transaction_argresp
*ar
;
1170 static DEFINE_SPINLOCK(simple_transaction_lock
);
1172 if (size
> SIMPLE_TRANSACTION_LIMIT
- 1)
1173 return ERR_PTR(-EFBIG
);
1175 ar
= (struct simple_transaction_argresp
*)get_zeroed_page(GFP_KERNEL
);
1177 return ERR_PTR(-ENOMEM
);
1179 spin_lock(&simple_transaction_lock
);
1181 /* only one write allowed per open */
1182 if (file
->private_data
) {
1183 spin_unlock(&simple_transaction_lock
);
1184 free_page((unsigned long)ar
);
1185 return ERR_PTR(-EBUSY
);
1188 file
->private_data
= ar
;
1190 spin_unlock(&simple_transaction_lock
);
1192 if (copy_from_user(ar
->data
, buf
, size
))
1193 return ERR_PTR(-EFAULT
);
1197 EXPORT_SYMBOL(simple_transaction_get
);
1199 ssize_t
simple_transaction_read(struct file
*file
, char __user
*buf
, size_t size
, loff_t
*pos
)
1201 struct simple_transaction_argresp
*ar
= file
->private_data
;
1205 return simple_read_from_buffer(buf
, size
, pos
, ar
->data
, ar
->size
);
1207 EXPORT_SYMBOL(simple_transaction_read
);
1209 int simple_transaction_release(struct inode
*inode
, struct file
*file
)
1211 free_page((unsigned long)file
->private_data
);
1214 EXPORT_SYMBOL(simple_transaction_release
);
1216 /* Simple attribute files */
1218 struct simple_attr
{
1219 int (*get
)(void *, u64
*);
1220 int (*set
)(void *, u64
);
1221 char get_buf
[24]; /* enough to store a u64 and "\n\0" */
1224 const char *fmt
; /* format for read operation */
1225 struct mutex mutex
; /* protects access to these buffers */
1228 /* simple_attr_open is called by an actual attribute open file operation
1229 * to set the attribute specific access operations. */
1230 int simple_attr_open(struct inode
*inode
, struct file
*file
,
1231 int (*get
)(void *, u64
*), int (*set
)(void *, u64
),
1234 struct simple_attr
*attr
;
1236 attr
= kzalloc(sizeof(*attr
), GFP_KERNEL
);
1242 attr
->data
= inode
->i_private
;
1244 mutex_init(&attr
->mutex
);
1246 file
->private_data
= attr
;
1248 return nonseekable_open(inode
, file
);
1250 EXPORT_SYMBOL_GPL(simple_attr_open
);
1252 int simple_attr_release(struct inode
*inode
, struct file
*file
)
1254 kfree(file
->private_data
);
1257 EXPORT_SYMBOL_GPL(simple_attr_release
); /* GPL-only? This? Really? */
1259 /* read from the buffer that is filled with the get function */
1260 ssize_t
simple_attr_read(struct file
*file
, char __user
*buf
,
1261 size_t len
, loff_t
*ppos
)
1263 struct simple_attr
*attr
;
1267 attr
= file
->private_data
;
1272 ret
= mutex_lock_interruptible(&attr
->mutex
);
1276 if (*ppos
&& attr
->get_buf
[0]) {
1277 /* continued read */
1278 size
= strlen(attr
->get_buf
);
1282 ret
= attr
->get(attr
->data
, &val
);
1286 size
= scnprintf(attr
->get_buf
, sizeof(attr
->get_buf
),
1287 attr
->fmt
, (unsigned long long)val
);
1290 ret
= simple_read_from_buffer(buf
, len
, ppos
, attr
->get_buf
, size
);
1292 mutex_unlock(&attr
->mutex
);
1295 EXPORT_SYMBOL_GPL(simple_attr_read
);
1297 /* interpret the buffer as a number to call the set function with */
1298 static ssize_t
simple_attr_write_xsigned(struct file
*file
, const char __user
*buf
,
1299 size_t len
, loff_t
*ppos
, bool is_signed
)
1301 struct simple_attr
*attr
;
1302 unsigned long long val
;
1306 attr
= file
->private_data
;
1310 ret
= mutex_lock_interruptible(&attr
->mutex
);
1315 size
= min(sizeof(attr
->set_buf
) - 1, len
);
1316 if (copy_from_user(attr
->set_buf
, buf
, size
))
1319 attr
->set_buf
[size
] = '\0';
1321 ret
= kstrtoll(attr
->set_buf
, 0, &val
);
1323 ret
= kstrtoull(attr
->set_buf
, 0, &val
);
1326 ret
= attr
->set(attr
->data
, val
);
1328 ret
= len
; /* on success, claim we got the whole input */
1330 mutex_unlock(&attr
->mutex
);
1334 ssize_t
simple_attr_write(struct file
*file
, const char __user
*buf
,
1335 size_t len
, loff_t
*ppos
)
1337 return simple_attr_write_xsigned(file
, buf
, len
, ppos
, false);
1339 EXPORT_SYMBOL_GPL(simple_attr_write
);
1341 ssize_t
simple_attr_write_signed(struct file
*file
, const char __user
*buf
,
1342 size_t len
, loff_t
*ppos
)
1344 return simple_attr_write_xsigned(file
, buf
, len
, ppos
, true);
1346 EXPORT_SYMBOL_GPL(simple_attr_write_signed
);
1349 * generic_encode_ino32_fh - generic export_operations->encode_fh function
1350 * @inode: the object to encode
1351 * @fh: where to store the file handle fragment
1352 * @max_len: maximum length to store there (in 4 byte units)
1353 * @parent: parent directory inode, if wanted
1355 * This generic encode_fh function assumes that the 32 inode number
1356 * is suitable for locating an inode, and that the generation number
1357 * can be used to check that it is still valid. It places them in the
1358 * filehandle fragment where export_decode_fh expects to find them.
1360 int generic_encode_ino32_fh(struct inode
*inode
, __u32
*fh
, int *max_len
,
1361 struct inode
*parent
)
1363 struct fid
*fid
= (void *)fh
;
1365 int type
= FILEID_INO32_GEN
;
1367 if (parent
&& (len
< 4)) {
1369 return FILEID_INVALID
;
1370 } else if (len
< 2) {
1372 return FILEID_INVALID
;
1376 fid
->i32
.ino
= inode
->i_ino
;
1377 fid
->i32
.gen
= inode
->i_generation
;
1379 fid
->i32
.parent_ino
= parent
->i_ino
;
1380 fid
->i32
.parent_gen
= parent
->i_generation
;
1382 type
= FILEID_INO32_GEN_PARENT
;
1387 EXPORT_SYMBOL_GPL(generic_encode_ino32_fh
);
1390 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1391 * @sb: filesystem to do the file handle conversion on
1392 * @fid: file handle to convert
1393 * @fh_len: length of the file handle in bytes
1394 * @fh_type: type of file handle
1395 * @get_inode: filesystem callback to retrieve inode
1397 * This function decodes @fid as long as it has one of the well-known
1398 * Linux filehandle types and calls @get_inode on it to retrieve the
1399 * inode for the object specified in the file handle.
1401 struct dentry
*generic_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1402 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1403 (struct super_block
*sb
, u64 ino
, u32 gen
))
1405 struct inode
*inode
= NULL
;
1411 case FILEID_INO32_GEN
:
1412 case FILEID_INO32_GEN_PARENT
:
1413 inode
= get_inode(sb
, fid
->i32
.ino
, fid
->i32
.gen
);
1417 return d_obtain_alias(inode
);
1419 EXPORT_SYMBOL_GPL(generic_fh_to_dentry
);
1422 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1423 * @sb: filesystem to do the file handle conversion on
1424 * @fid: file handle to convert
1425 * @fh_len: length of the file handle in bytes
1426 * @fh_type: type of file handle
1427 * @get_inode: filesystem callback to retrieve inode
1429 * This function decodes @fid as long as it has one of the well-known
1430 * Linux filehandle types and calls @get_inode on it to retrieve the
1431 * inode for the _parent_ object specified in the file handle if it
1432 * is specified in the file handle, or NULL otherwise.
1434 struct dentry
*generic_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1435 int fh_len
, int fh_type
, struct inode
*(*get_inode
)
1436 (struct super_block
*sb
, u64 ino
, u32 gen
))
1438 struct inode
*inode
= NULL
;
1444 case FILEID_INO32_GEN_PARENT
:
1445 inode
= get_inode(sb
, fid
->i32
.parent_ino
,
1446 (fh_len
> 3 ? fid
->i32
.parent_gen
: 0));
1450 return d_obtain_alias(inode
);
1452 EXPORT_SYMBOL_GPL(generic_fh_to_parent
);
1455 * __generic_file_fsync - generic fsync implementation for simple filesystems
1457 * @file: file to synchronize
1458 * @start: start offset in bytes
1459 * @end: end offset in bytes (inclusive)
1460 * @datasync: only synchronize essential metadata if true
1462 * This is a generic implementation of the fsync method for simple
1463 * filesystems which track all non-inode metadata in the buffers list
1464 * hanging off the address_space structure.
1466 int __generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1469 struct inode
*inode
= file
->f_mapping
->host
;
1473 err
= file_write_and_wait_range(file
, start
, end
);
1478 ret
= sync_mapping_buffers(inode
->i_mapping
);
1479 if (!(inode
->i_state
& I_DIRTY_ALL
))
1481 if (datasync
&& !(inode
->i_state
& I_DIRTY_DATASYNC
))
1484 err
= sync_inode_metadata(inode
, 1);
1489 inode_unlock(inode
);
1490 /* check and advance again to catch errors after syncing out buffers */
1491 err
= file_check_and_advance_wb_err(file
);
1496 EXPORT_SYMBOL(__generic_file_fsync
);
1499 * generic_file_fsync - generic fsync implementation for simple filesystems
1501 * @file: file to synchronize
1502 * @start: start offset in bytes
1503 * @end: end offset in bytes (inclusive)
1504 * @datasync: only synchronize essential metadata if true
1508 int generic_file_fsync(struct file
*file
, loff_t start
, loff_t end
,
1511 struct inode
*inode
= file
->f_mapping
->host
;
1514 err
= __generic_file_fsync(file
, start
, end
, datasync
);
1517 return blkdev_issue_flush(inode
->i_sb
->s_bdev
);
1519 EXPORT_SYMBOL(generic_file_fsync
);
1522 * generic_check_addressable - Check addressability of file system
1523 * @blocksize_bits: log of file system block size
1524 * @num_blocks: number of blocks in file system
1526 * Determine whether a file system with @num_blocks blocks (and a
1527 * block size of 2**@blocksize_bits) is addressable by the sector_t
1528 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1530 int generic_check_addressable(unsigned blocksize_bits
, u64 num_blocks
)
1532 u64 last_fs_block
= num_blocks
- 1;
1534 last_fs_block
>> (PAGE_SHIFT
- blocksize_bits
);
1536 if (unlikely(num_blocks
== 0))
1539 if ((blocksize_bits
< 9) || (blocksize_bits
> PAGE_SHIFT
))
1542 if ((last_fs_block
> (sector_t
)(~0ULL) >> (blocksize_bits
- 9)) ||
1543 (last_fs_page
> (pgoff_t
)(~0ULL))) {
1548 EXPORT_SYMBOL(generic_check_addressable
);
1551 * No-op implementation of ->fsync for in-memory filesystems.
1553 int noop_fsync(struct file
*file
, loff_t start
, loff_t end
, int datasync
)
1557 EXPORT_SYMBOL(noop_fsync
);
1559 ssize_t
noop_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
)
1562 * iomap based filesystems support direct I/O without need for
1563 * this callback. However, it still needs to be set in
1564 * inode->a_ops so that open/fcntl know that direct I/O is
1565 * generally supported.
1569 EXPORT_SYMBOL_GPL(noop_direct_IO
);
1571 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1572 void kfree_link(void *p
)
1576 EXPORT_SYMBOL(kfree_link
);
1578 struct inode
*alloc_anon_inode(struct super_block
*s
)
1580 static const struct address_space_operations anon_aops
= {
1581 .dirty_folio
= noop_dirty_folio
,
1583 struct inode
*inode
= new_inode_pseudo(s
);
1586 return ERR_PTR(-ENOMEM
);
1588 inode
->i_ino
= get_next_ino();
1589 inode
->i_mapping
->a_ops
= &anon_aops
;
1592 * Mark the inode dirty from the very beginning,
1593 * that way it will never be moved to the dirty
1594 * list because mark_inode_dirty() will think
1595 * that it already _is_ on the dirty list.
1597 inode
->i_state
= I_DIRTY
;
1598 inode
->i_mode
= S_IRUSR
| S_IWUSR
;
1599 inode
->i_uid
= current_fsuid();
1600 inode
->i_gid
= current_fsgid();
1601 inode
->i_flags
|= S_PRIVATE
;
1602 simple_inode_init_ts(inode
);
1605 EXPORT_SYMBOL(alloc_anon_inode
);
1608 * simple_nosetlease - generic helper for prohibiting leases
1609 * @filp: file pointer
1610 * @arg: type of lease to obtain
1611 * @flp: new lease supplied for insertion
1612 * @priv: private data for lm_setup operation
1614 * Generic helper for filesystems that do not wish to allow leases to be set.
1615 * All arguments are ignored and it just returns -EINVAL.
1618 simple_nosetlease(struct file
*filp
, int arg
, struct file_lease
**flp
,
1623 EXPORT_SYMBOL(simple_nosetlease
);
1626 * simple_get_link - generic helper to get the target of "fast" symlinks
1627 * @dentry: not used here
1628 * @inode: the symlink inode
1629 * @done: not used here
1631 * Generic helper for filesystems to use for symlink inodes where a pointer to
1632 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1633 * since as an optimization the path lookup code uses any non-NULL ->i_link
1634 * directly, without calling ->get_link(). But ->get_link() still must be set,
1635 * to mark the inode_operations as being for a symlink.
1637 * Return: the symlink target
1639 const char *simple_get_link(struct dentry
*dentry
, struct inode
*inode
,
1640 struct delayed_call
*done
)
1642 return inode
->i_link
;
1644 EXPORT_SYMBOL(simple_get_link
);
1646 const struct inode_operations simple_symlink_inode_operations
= {
1647 .get_link
= simple_get_link
,
1649 EXPORT_SYMBOL(simple_symlink_inode_operations
);
1652 * Operations for a permanently empty directory.
1654 static struct dentry
*empty_dir_lookup(struct inode
*dir
, struct dentry
*dentry
, unsigned int flags
)
1656 return ERR_PTR(-ENOENT
);
1659 static int empty_dir_getattr(struct mnt_idmap
*idmap
,
1660 const struct path
*path
, struct kstat
*stat
,
1661 u32 request_mask
, unsigned int query_flags
)
1663 struct inode
*inode
= d_inode(path
->dentry
);
1664 generic_fillattr(&nop_mnt_idmap
, request_mask
, inode
, stat
);
1668 static int empty_dir_setattr(struct mnt_idmap
*idmap
,
1669 struct dentry
*dentry
, struct iattr
*attr
)
1674 static ssize_t
empty_dir_listxattr(struct dentry
*dentry
, char *list
, size_t size
)
1679 static const struct inode_operations empty_dir_inode_operations
= {
1680 .lookup
= empty_dir_lookup
,
1681 .permission
= generic_permission
,
1682 .setattr
= empty_dir_setattr
,
1683 .getattr
= empty_dir_getattr
,
1684 .listxattr
= empty_dir_listxattr
,
1687 static loff_t
empty_dir_llseek(struct file
*file
, loff_t offset
, int whence
)
1689 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1690 return generic_file_llseek_size(file
, offset
, whence
, 2, 2);
1693 static int empty_dir_readdir(struct file
*file
, struct dir_context
*ctx
)
1695 dir_emit_dots(file
, ctx
);
1699 static const struct file_operations empty_dir_operations
= {
1700 .llseek
= empty_dir_llseek
,
1701 .read
= generic_read_dir
,
1702 .iterate_shared
= empty_dir_readdir
,
1703 .fsync
= noop_fsync
,
1707 void make_empty_dir_inode(struct inode
*inode
)
1709 set_nlink(inode
, 2);
1710 inode
->i_mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
1711 inode
->i_uid
= GLOBAL_ROOT_UID
;
1712 inode
->i_gid
= GLOBAL_ROOT_GID
;
1715 inode
->i_blkbits
= PAGE_SHIFT
;
1716 inode
->i_blocks
= 0;
1718 inode
->i_op
= &empty_dir_inode_operations
;
1719 inode
->i_opflags
&= ~IOP_XATTR
;
1720 inode
->i_fop
= &empty_dir_operations
;
1723 bool is_empty_dir_inode(struct inode
*inode
)
1725 return (inode
->i_fop
== &empty_dir_operations
) &&
1726 (inode
->i_op
== &empty_dir_inode_operations
);
1729 #if IS_ENABLED(CONFIG_UNICODE)
1731 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1732 * @dentry: dentry whose name we are checking against
1733 * @len: len of name of dentry
1734 * @str: str pointer to name of dentry
1735 * @name: Name to compare against
1737 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1739 static int generic_ci_d_compare(const struct dentry
*dentry
, unsigned int len
,
1740 const char *str
, const struct qstr
*name
)
1742 const struct dentry
*parent
;
1743 const struct inode
*dir
;
1744 char strbuf
[DNAME_INLINE_LEN
];
1748 * Attempt a case-sensitive match first. It is cheaper and
1749 * should cover most lookups, including all the sane
1750 * applications that expect a case-sensitive filesystem.
1752 * This comparison is safe under RCU because the caller
1753 * guarantees the consistency between str and len. See
1754 * __d_lookup_rcu_op_compare() for details.
1756 if (len
== name
->len
&& !memcmp(str
, name
->name
, len
))
1759 parent
= READ_ONCE(dentry
->d_parent
);
1760 dir
= READ_ONCE(parent
->d_inode
);
1761 if (!dir
|| !IS_CASEFOLDED(dir
))
1765 * If the dentry name is stored in-line, then it may be concurrently
1766 * modified by a rename. If this happens, the VFS will eventually retry
1767 * the lookup, so it doesn't matter what ->d_compare() returns.
1768 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1769 * string. Therefore, we have to copy the name into a temporary buffer.
1771 if (len
<= DNAME_INLINE_LEN
- 1) {
1772 memcpy(strbuf
, str
, len
);
1775 /* prevent compiler from optimizing out the temporary buffer */
1781 return utf8_strncasecmp(dentry
->d_sb
->s_encoding
, name
, &qstr
);
1785 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1786 * @dentry: dentry of the parent directory
1787 * @str: qstr of name whose hash we should fill in
1789 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1791 static int generic_ci_d_hash(const struct dentry
*dentry
, struct qstr
*str
)
1793 const struct inode
*dir
= READ_ONCE(dentry
->d_inode
);
1794 struct super_block
*sb
= dentry
->d_sb
;
1795 const struct unicode_map
*um
= sb
->s_encoding
;
1798 if (!dir
|| !IS_CASEFOLDED(dir
))
1801 ret
= utf8_casefold_hash(um
, dentry
, str
);
1802 if (ret
< 0 && sb_has_strict_encoding(sb
))
1807 static const struct dentry_operations generic_ci_dentry_ops
= {
1808 .d_hash
= generic_ci_d_hash
,
1809 .d_compare
= generic_ci_d_compare
,
1810 #ifdef CONFIG_FS_ENCRYPTION
1811 .d_revalidate
= fscrypt_d_revalidate
,
1816 #ifdef CONFIG_FS_ENCRYPTION
1817 static const struct dentry_operations generic_encrypted_dentry_ops
= {
1818 .d_revalidate
= fscrypt_d_revalidate
,
1823 * generic_set_sb_d_ops - helper for choosing the set of
1824 * filesystem-wide dentry operations for the enabled features
1825 * @sb: superblock to be configured
1827 * Filesystems supporting casefolding and/or fscrypt can call this
1828 * helper at mount-time to configure sb->s_d_op to best set of dentry
1829 * operations required for the enabled features. The helper must be
1830 * called after these have been configured, but before the root dentry
1833 void generic_set_sb_d_ops(struct super_block
*sb
)
1835 #if IS_ENABLED(CONFIG_UNICODE)
1836 if (sb
->s_encoding
) {
1837 sb
->s_d_op
= &generic_ci_dentry_ops
;
1841 #ifdef CONFIG_FS_ENCRYPTION
1843 sb
->s_d_op
= &generic_encrypted_dentry_ops
;
1848 EXPORT_SYMBOL(generic_set_sb_d_ops
);
1851 * inode_maybe_inc_iversion - increments i_version
1852 * @inode: inode with the i_version that should be updated
1853 * @force: increment the counter even if it's not necessary?
1855 * Every time the inode is modified, the i_version field must be seen to have
1856 * changed by any observer.
1858 * If "force" is set or the QUERIED flag is set, then ensure that we increment
1859 * the value, and clear the queried flag.
1861 * In the common case where neither is set, then we can return "false" without
1862 * updating i_version.
1864 * If this function returns false, and no other metadata has changed, then we
1865 * can avoid logging the metadata.
1867 bool inode_maybe_inc_iversion(struct inode
*inode
, bool force
)
1872 * The i_version field is not strictly ordered with any other inode
1873 * information, but the legacy inode_inc_iversion code used a spinlock
1874 * to serialize increments.
1876 * Here, we add full memory barriers to ensure that any de-facto
1877 * ordering with other info is preserved.
1879 * This barrier pairs with the barrier in inode_query_iversion()
1882 cur
= inode_peek_iversion_raw(inode
);
1884 /* If flag is clear then we needn't do anything */
1885 if (!force
&& !(cur
& I_VERSION_QUERIED
))
1888 /* Since lowest bit is flag, add 2 to avoid it */
1889 new = (cur
& ~I_VERSION_QUERIED
) + I_VERSION_INCREMENT
;
1890 } while (!atomic64_try_cmpxchg(&inode
->i_version
, &cur
, new));
1893 EXPORT_SYMBOL(inode_maybe_inc_iversion
);
1896 * inode_query_iversion - read i_version for later use
1897 * @inode: inode from which i_version should be read
1899 * Read the inode i_version counter. This should be used by callers that wish
1900 * to store the returned i_version for later comparison. This will guarantee
1901 * that a later query of the i_version will result in a different value if
1902 * anything has changed.
1904 * In this implementation, we fetch the current value, set the QUERIED flag and
1905 * then try to swap it into place with a cmpxchg, if it wasn't already set. If
1906 * that fails, we try again with the newly fetched value from the cmpxchg.
1908 u64
inode_query_iversion(struct inode
*inode
)
1912 cur
= inode_peek_iversion_raw(inode
);
1914 /* If flag is already set, then no need to swap */
1915 if (cur
& I_VERSION_QUERIED
) {
1917 * This barrier (and the implicit barrier in the
1918 * cmpxchg below) pairs with the barrier in
1919 * inode_maybe_inc_iversion().
1925 new = cur
| I_VERSION_QUERIED
;
1926 } while (!atomic64_try_cmpxchg(&inode
->i_version
, &cur
, new));
1927 return cur
>> I_VERSION_QUERIED_SHIFT
;
1929 EXPORT_SYMBOL(inode_query_iversion
);
1931 ssize_t
direct_write_fallback(struct kiocb
*iocb
, struct iov_iter
*iter
,
1932 ssize_t direct_written
, ssize_t buffered_written
)
1934 struct address_space
*mapping
= iocb
->ki_filp
->f_mapping
;
1935 loff_t pos
= iocb
->ki_pos
- buffered_written
;
1936 loff_t end
= iocb
->ki_pos
- 1;
1940 * If the buffered write fallback returned an error, we want to return
1941 * the number of bytes which were written by direct I/O, or the error
1942 * code if that was zero.
1944 * Note that this differs from normal direct-io semantics, which will
1945 * return -EFOO even if some bytes were written.
1947 if (unlikely(buffered_written
< 0)) {
1949 return direct_written
;
1950 return buffered_written
;
1954 * We need to ensure that the page cache pages are written to disk and
1955 * invalidated to preserve the expected O_DIRECT semantics.
1957 err
= filemap_write_and_wait_range(mapping
, pos
, end
);
1960 * We don't know how much we wrote, so just return the number of
1961 * bytes which were direct-written
1963 iocb
->ki_pos
-= buffered_written
;
1965 return direct_written
;
1968 invalidate_mapping_pages(mapping
, pos
>> PAGE_SHIFT
, end
>> PAGE_SHIFT
);
1969 return direct_written
+ buffered_written
;
1971 EXPORT_SYMBOL_GPL(direct_write_fallback
);
1974 * simple_inode_init_ts - initialize the timestamps for a new inode
1975 * @inode: inode to be initialized
1977 * When a new inode is created, most filesystems set the timestamps to the
1978 * current time. Add a helper to do this.
1980 struct timespec64
simple_inode_init_ts(struct inode
*inode
)
1982 struct timespec64 ts
= inode_set_ctime_current(inode
);
1984 inode_set_atime_to_ts(inode
, ts
);
1985 inode_set_mtime_to_ts(inode
, ts
);
1988 EXPORT_SYMBOL(simple_inode_init_ts
);
1990 static inline struct dentry
*get_stashed_dentry(struct dentry
*stashed
)
1992 struct dentry
*dentry
;
1995 dentry
= READ_ONCE(stashed
);
1998 if (!lockref_get_not_dead(&dentry
->d_lockref
))
2003 static struct dentry
*prepare_anon_dentry(struct dentry
**stashed
,
2004 struct super_block
*sb
,
2007 struct dentry
*dentry
;
2008 struct inode
*inode
;
2009 const struct stashed_operations
*sops
= sb
->s_fs_info
;
2012 inode
= new_inode_pseudo(sb
);
2014 sops
->put_data(data
);
2015 return ERR_PTR(-ENOMEM
);
2018 inode
->i_flags
|= S_IMMUTABLE
;
2019 inode
->i_mode
= S_IFREG
;
2020 simple_inode_init_ts(inode
);
2022 ret
= sops
->init_inode(inode
, data
);
2025 return ERR_PTR(ret
);
2028 /* Notice when this is changed. */
2029 WARN_ON_ONCE(!S_ISREG(inode
->i_mode
));
2030 WARN_ON_ONCE(!IS_IMMUTABLE(inode
));
2032 dentry
= d_alloc_anon(sb
);
2035 return ERR_PTR(-ENOMEM
);
2038 /* Store address of location where dentry's supposed to be stashed. */
2039 dentry
->d_fsdata
= stashed
;
2041 /* @data is now owned by the fs */
2042 d_instantiate(dentry
, inode
);
2046 static struct dentry
*stash_dentry(struct dentry
**stashed
,
2047 struct dentry
*dentry
)
2053 /* Assume any old dentry was cleared out. */
2054 old
= cmpxchg(stashed
, NULL
, dentry
);
2058 /* Check if somebody else installed a reusable dentry. */
2059 if (lockref_get_not_dead(&old
->d_lockref
))
2062 /* There's an old dead dentry there, try to take it over. */
2063 if (likely(try_cmpxchg(stashed
, &old
, dentry
)))
2069 * path_from_stashed - create path from stashed or new dentry
2070 * @stashed: where to retrieve or stash dentry
2071 * @mnt: mnt of the filesystems to use
2072 * @data: data to store in inode->i_private
2073 * @path: path to create
2075 * The function tries to retrieve a stashed dentry from @stashed. If the dentry
2076 * is still valid then it will be reused. If the dentry isn't able the function
2077 * will allocate a new dentry and inode. It will then check again whether it
2078 * can reuse an existing dentry in case one has been added in the meantime or
2079 * update @stashed with the newly added dentry.
2081 * Special-purpose helper for nsfs and pidfs.
2083 * Return: On success zero and on failure a negative error is returned.
2085 int path_from_stashed(struct dentry
**stashed
, struct vfsmount
*mnt
, void *data
,
2088 struct dentry
*dentry
;
2089 const struct stashed_operations
*sops
= mnt
->mnt_sb
->s_fs_info
;
2091 /* See if dentry can be reused. */
2092 path
->dentry
= get_stashed_dentry(*stashed
);
2094 sops
->put_data(data
);
2098 /* Allocate a new dentry. */
2099 dentry
= prepare_anon_dentry(stashed
, mnt
->mnt_sb
, data
);
2101 return PTR_ERR(dentry
);
2103 /* Added a new dentry. @data is now owned by the filesystem. */
2104 path
->dentry
= stash_dentry(stashed
, dentry
);
2105 if (path
->dentry
!= dentry
)
2109 WARN_ON_ONCE(path
->dentry
->d_fsdata
!= stashed
);
2110 WARN_ON_ONCE(d_inode(path
->dentry
)->i_private
!= data
);
2111 path
->mnt
= mntget(mnt
);
2115 void stashed_dentry_prune(struct dentry
*dentry
)
2117 struct dentry
**stashed
= dentry
->d_fsdata
;
2118 struct inode
*inode
= d_inode(dentry
);
2120 if (WARN_ON_ONCE(!stashed
))
2127 * Only replace our own @dentry as someone else might've
2128 * already cleared out @dentry and stashed their own
2131 cmpxchg(stashed
, dentry
, NULL
);