1 // SPDX-License-Identifier: GPL-2.0-only
4 * Library for filesystems writers.
7 #include <linux/blkdev.h>
8 #include <linux/export.h>
9 #include <linux/pagemap.h>
10 #include <linux/slab.h>
11 #include <linux/cred.h>
12 #include <linux/mount.h>
13 #include <linux/vfs.h>
14 #include <linux/quotaops.h>
15 #include <linux/mutex.h>
16 #include <linux/namei.h>
17 #include <linux/exportfs.h>
18 #include <linux/writeback.h>
19 #include <linux/buffer_head.h> /* sync_mapping_buffers */
20 #include <linux/fs_context.h>
21 #include <linux/pseudo_fs.h>
22 #include <linux/fsnotify.h>
23 #include <linux/unicode.h>
24 #include <linux/fscrypt.h>
26 #include <linux/uaccess.h>
30 int simple_getattr(struct user_namespace *mnt_userns, const struct path *path,
31 struct kstat *stat, u32 request_mask,
32 unsigned int query_flags)
34 struct inode *inode = d_inode(path->dentry);
35 generic_fillattr(&init_user_ns, inode, stat);
36 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
39 EXPORT_SYMBOL(simple_getattr);
41 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
43 buf->f_type = dentry->d_sb->s_magic;
44 buf->f_bsize = PAGE_SIZE;
45 buf->f_namelen = NAME_MAX;
48 EXPORT_SYMBOL(simple_statfs);
51 * Retaining negative dentries for an in-memory filesystem just wastes
52 * memory and lookup time: arrange for them to be deleted immediately.
54 int always_delete_dentry(const struct dentry *dentry)
58 EXPORT_SYMBOL(always_delete_dentry);
60 const struct dentry_operations simple_dentry_operations = {
61 .d_delete = always_delete_dentry,
63 EXPORT_SYMBOL(simple_dentry_operations);
66 * Lookup the data. This is trivial - if the dentry didn't already
67 * exist, we know it is negative. Set d_op to delete negative dentries.
69 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
71 if (dentry->d_name.len > NAME_MAX)
72 return ERR_PTR(-ENAMETOOLONG);
73 if (!dentry->d_sb->s_d_op)
74 d_set_d_op(dentry, &simple_dentry_operations);
78 EXPORT_SYMBOL(simple_lookup);
80 int dcache_dir_open(struct inode *inode, struct file *file)
82 file->private_data = d_alloc_cursor(file->f_path.dentry);
84 return file->private_data ? 0 : -ENOMEM;
86 EXPORT_SYMBOL(dcache_dir_open);
88 int dcache_dir_close(struct inode *inode, struct file *file)
90 dput(file->private_data);
93 EXPORT_SYMBOL(dcache_dir_close);
95 /* parent is locked at least shared */
97 * Returns an element of siblings' list.
98 * We are looking for <count>th positive after <p>; if
99 * found, dentry is grabbed and returned to caller.
100 * If no such element exists, NULL is returned.
102 static struct dentry *scan_positives(struct dentry *cursor,
107 struct dentry *dentry = cursor->d_parent, *found = NULL;
109 spin_lock(&dentry->d_lock);
110 while ((p = p->next) != &dentry->d_subdirs) {
111 struct dentry *d = list_entry(p, struct dentry, d_child);
112 // we must at least skip cursors, to avoid livelocks
113 if (d->d_flags & DCACHE_DENTRY_CURSOR)
115 if (simple_positive(d) && !--count) {
116 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
117 if (simple_positive(d))
118 found = dget_dlock(d);
119 spin_unlock(&d->d_lock);
124 if (need_resched()) {
125 list_move(&cursor->d_child, p);
126 p = &cursor->d_child;
127 spin_unlock(&dentry->d_lock);
129 spin_lock(&dentry->d_lock);
132 spin_unlock(&dentry->d_lock);
137 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
139 struct dentry *dentry = file->f_path.dentry;
142 offset += file->f_pos;
151 if (offset != file->f_pos) {
152 struct dentry *cursor = file->private_data;
153 struct dentry *to = NULL;
155 inode_lock_shared(dentry->d_inode);
158 to = scan_positives(cursor, &dentry->d_subdirs,
160 spin_lock(&dentry->d_lock);
162 list_move(&cursor->d_child, &to->d_child);
164 list_del_init(&cursor->d_child);
165 spin_unlock(&dentry->d_lock);
168 file->f_pos = offset;
170 inode_unlock_shared(dentry->d_inode);
174 EXPORT_SYMBOL(dcache_dir_lseek);
176 /* Relationship between i_mode and the DT_xxx types */
177 static inline unsigned char dt_type(struct inode *inode)
179 return (inode->i_mode >> 12) & 15;
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 list_head *anchor = &dentry->d_subdirs;
193 struct dentry *next = NULL;
196 if (!dir_emit_dots(file, ctx))
201 else if (!list_empty(&cursor->d_child))
202 p = &cursor->d_child;
206 while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
207 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
208 d_inode(next)->i_ino, dt_type(d_inode(next))))
213 spin_lock(&dentry->d_lock);
215 list_move_tail(&cursor->d_child, &next->d_child);
217 list_del_init(&cursor->d_child);
218 spin_unlock(&dentry->d_lock);
223 EXPORT_SYMBOL(dcache_readdir);
225 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
229 EXPORT_SYMBOL(generic_read_dir);
231 const struct file_operations simple_dir_operations = {
232 .open = dcache_dir_open,
233 .release = dcache_dir_close,
234 .llseek = dcache_dir_lseek,
235 .read = generic_read_dir,
236 .iterate_shared = dcache_readdir,
239 EXPORT_SYMBOL(simple_dir_operations);
241 const struct inode_operations simple_dir_inode_operations = {
242 .lookup = simple_lookup,
244 EXPORT_SYMBOL(simple_dir_inode_operations);
246 static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
248 struct dentry *child = NULL;
249 struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
251 spin_lock(&parent->d_lock);
252 while ((p = p->next) != &parent->d_subdirs) {
253 struct dentry *d = container_of(p, struct dentry, d_child);
254 if (simple_positive(d)) {
255 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
256 if (simple_positive(d))
257 child = dget_dlock(d);
258 spin_unlock(&d->d_lock);
263 spin_unlock(&parent->d_lock);
268 void simple_recursive_removal(struct dentry *dentry,
269 void (*callback)(struct dentry *))
271 struct dentry *this = dget(dentry);
273 struct dentry *victim = NULL, *child;
274 struct inode *inode = this->d_inode;
278 inode->i_flags |= S_DEAD;
279 while ((child = find_next_child(this, victim)) == NULL) {
281 // update metadata while it's still locked
282 inode->i_ctime = current_time(inode);
286 this = this->d_parent;
287 inode = this->d_inode;
289 if (simple_positive(victim)) {
290 d_invalidate(victim); // avoid lost mounts
291 if (d_is_dir(victim))
292 fsnotify_rmdir(inode, victim);
294 fsnotify_unlink(inode, victim);
297 dput(victim); // unpin it
299 if (victim == dentry) {
300 inode->i_ctime = inode->i_mtime =
302 if (d_is_dir(dentry))
313 EXPORT_SYMBOL(simple_recursive_removal);
315 static const struct super_operations simple_super_operations = {
316 .statfs = simple_statfs,
319 static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
321 struct pseudo_fs_context *ctx = fc->fs_private;
324 s->s_maxbytes = MAX_LFS_FILESIZE;
325 s->s_blocksize = PAGE_SIZE;
326 s->s_blocksize_bits = PAGE_SHIFT;
327 s->s_magic = ctx->magic;
328 s->s_op = ctx->ops ?: &simple_super_operations;
329 s->s_xattr = ctx->xattr;
336 * since this is the first inode, make it number 1. New inodes created
337 * after this must take care not to collide with it (by passing
338 * max_reserved of 1 to iunique).
341 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
342 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
343 s->s_root = d_make_root(root);
346 s->s_d_op = ctx->dops;
350 static int pseudo_fs_get_tree(struct fs_context *fc)
352 return get_tree_nodev(fc, pseudo_fs_fill_super);
355 static void pseudo_fs_free(struct fs_context *fc)
357 kfree(fc->fs_private);
360 static const struct fs_context_operations pseudo_fs_context_ops = {
361 .free = pseudo_fs_free,
362 .get_tree = pseudo_fs_get_tree,
366 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
367 * will never be mountable)
369 struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
372 struct pseudo_fs_context *ctx;
374 ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
377 fc->fs_private = ctx;
378 fc->ops = &pseudo_fs_context_ops;
379 fc->sb_flags |= SB_NOUSER;
384 EXPORT_SYMBOL(init_pseudo);
386 int simple_open(struct inode *inode, struct file *file)
388 if (inode->i_private)
389 file->private_data = inode->i_private;
392 EXPORT_SYMBOL(simple_open);
394 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
396 struct inode *inode = d_inode(old_dentry);
398 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
402 d_instantiate(dentry, inode);
405 EXPORT_SYMBOL(simple_link);
407 int simple_empty(struct dentry *dentry)
409 struct dentry *child;
412 spin_lock(&dentry->d_lock);
413 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
414 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
415 if (simple_positive(child)) {
416 spin_unlock(&child->d_lock);
419 spin_unlock(&child->d_lock);
423 spin_unlock(&dentry->d_lock);
426 EXPORT_SYMBOL(simple_empty);
428 int simple_unlink(struct inode *dir, struct dentry *dentry)
430 struct inode *inode = d_inode(dentry);
432 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
437 EXPORT_SYMBOL(simple_unlink);
439 int simple_rmdir(struct inode *dir, struct dentry *dentry)
441 if (!simple_empty(dentry))
444 drop_nlink(d_inode(dentry));
445 simple_unlink(dir, dentry);
449 EXPORT_SYMBOL(simple_rmdir);
451 int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
452 struct inode *new_dir, struct dentry *new_dentry)
454 bool old_is_dir = d_is_dir(old_dentry);
455 bool new_is_dir = d_is_dir(new_dentry);
457 if (old_dir != new_dir && old_is_dir != new_is_dir) {
466 old_dir->i_ctime = old_dir->i_mtime =
467 new_dir->i_ctime = new_dir->i_mtime =
468 d_inode(old_dentry)->i_ctime =
469 d_inode(new_dentry)->i_ctime = current_time(old_dir);
473 EXPORT_SYMBOL_GPL(simple_rename_exchange);
475 int simple_rename(struct user_namespace *mnt_userns, struct inode *old_dir,
476 struct dentry *old_dentry, struct inode *new_dir,
477 struct dentry *new_dentry, unsigned int flags)
479 struct inode *inode = d_inode(old_dentry);
480 int they_are_dirs = d_is_dir(old_dentry);
482 if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
485 if (flags & RENAME_EXCHANGE)
486 return simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry);
488 if (!simple_empty(new_dentry))
491 if (d_really_is_positive(new_dentry)) {
492 simple_unlink(new_dir, new_dentry);
494 drop_nlink(d_inode(new_dentry));
497 } else if (they_are_dirs) {
502 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
503 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
507 EXPORT_SYMBOL(simple_rename);
510 * simple_setattr - setattr for simple filesystem
511 * @mnt_userns: user namespace of the target mount
513 * @iattr: iattr structure
515 * Returns 0 on success, -error on failure.
517 * simple_setattr is a simple ->setattr implementation without a proper
518 * implementation of size changes.
520 * It can either be used for in-memory filesystems or special files
521 * on simple regular filesystems. Anything that needs to change on-disk
522 * or wire state on size changes needs its own setattr method.
524 int simple_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
527 struct inode *inode = d_inode(dentry);
530 error = setattr_prepare(mnt_userns, dentry, iattr);
534 if (iattr->ia_valid & ATTR_SIZE)
535 truncate_setsize(inode, iattr->ia_size);
536 setattr_copy(mnt_userns, inode, iattr);
537 mark_inode_dirty(inode);
540 EXPORT_SYMBOL(simple_setattr);
542 static int simple_readpage(struct file *file, struct page *page)
544 clear_highpage(page);
545 flush_dcache_page(page);
546 SetPageUptodate(page);
551 int simple_write_begin(struct file *file, struct address_space *mapping,
552 loff_t pos, unsigned len, unsigned flags,
553 struct page **pagep, void **fsdata)
558 index = pos >> PAGE_SHIFT;
560 page = grab_cache_page_write_begin(mapping, index, flags);
566 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
567 unsigned from = pos & (PAGE_SIZE - 1);
569 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
573 EXPORT_SYMBOL(simple_write_begin);
576 * simple_write_end - .write_end helper for non-block-device FSes
577 * @file: See .write_end of address_space_operations
585 * simple_write_end does the minimum needed for updating a page after writing is
586 * done. It has the same API signature as the .write_end of
587 * address_space_operations vector. So it can just be set onto .write_end for
588 * FSes that don't need any other processing. i_mutex is assumed to be held.
589 * Block based filesystems should use generic_write_end().
590 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
591 * is not called, so a filesystem that actually does store data in .write_inode
592 * should extend on what's done here with a call to mark_inode_dirty() in the
593 * case that i_size has changed.
595 * Use *ONLY* with simple_readpage()
597 static int simple_write_end(struct file *file, struct address_space *mapping,
598 loff_t pos, unsigned len, unsigned copied,
599 struct page *page, void *fsdata)
601 struct inode *inode = page->mapping->host;
602 loff_t last_pos = pos + copied;
604 /* zero the stale part of the page if we did a short copy */
605 if (!PageUptodate(page)) {
607 unsigned from = pos & (PAGE_SIZE - 1);
609 zero_user(page, from + copied, len - copied);
611 SetPageUptodate(page);
614 * No need to use i_size_read() here, the i_size
615 * cannot change under us because we hold the i_mutex.
617 if (last_pos > inode->i_size)
618 i_size_write(inode, last_pos);
620 set_page_dirty(page);
628 * Provides ramfs-style behavior: data in the pagecache, but no writeback.
630 const struct address_space_operations ram_aops = {
631 .readpage = simple_readpage,
632 .write_begin = simple_write_begin,
633 .write_end = simple_write_end,
634 .set_page_dirty = __set_page_dirty_no_writeback,
636 EXPORT_SYMBOL(ram_aops);
639 * the inodes created here are not hashed. If you use iunique to generate
640 * unique inode values later for this filesystem, then you must take care
641 * to pass it an appropriate max_reserved value to avoid collisions.
643 int simple_fill_super(struct super_block *s, unsigned long magic,
644 const struct tree_descr *files)
648 struct dentry *dentry;
651 s->s_blocksize = PAGE_SIZE;
652 s->s_blocksize_bits = PAGE_SHIFT;
654 s->s_op = &simple_super_operations;
657 inode = new_inode(s);
661 * because the root inode is 1, the files array must not contain an
665 inode->i_mode = S_IFDIR | 0755;
666 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
667 inode->i_op = &simple_dir_inode_operations;
668 inode->i_fop = &simple_dir_operations;
670 root = d_make_root(inode);
673 for (i = 0; !files->name || files->name[0]; i++, files++) {
677 /* warn if it tries to conflict with the root inode */
678 if (unlikely(i == 1))
679 printk(KERN_WARNING "%s: %s passed in a files array"
680 "with an index of 1!\n", __func__,
683 dentry = d_alloc_name(root, files->name);
686 inode = new_inode(s);
691 inode->i_mode = S_IFREG | files->mode;
692 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
693 inode->i_fop = files->ops;
695 d_add(dentry, inode);
701 shrink_dcache_parent(root);
705 EXPORT_SYMBOL(simple_fill_super);
707 static DEFINE_SPINLOCK(pin_fs_lock);
709 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
711 struct vfsmount *mnt = NULL;
712 spin_lock(&pin_fs_lock);
713 if (unlikely(!*mount)) {
714 spin_unlock(&pin_fs_lock);
715 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
718 spin_lock(&pin_fs_lock);
724 spin_unlock(&pin_fs_lock);
728 EXPORT_SYMBOL(simple_pin_fs);
730 void simple_release_fs(struct vfsmount **mount, int *count)
732 struct vfsmount *mnt;
733 spin_lock(&pin_fs_lock);
737 spin_unlock(&pin_fs_lock);
740 EXPORT_SYMBOL(simple_release_fs);
743 * simple_read_from_buffer - copy data from the buffer to user space
744 * @to: the user space buffer to read to
745 * @count: the maximum number of bytes to read
746 * @ppos: the current position in the buffer
747 * @from: the buffer to read from
748 * @available: the size of the buffer
750 * The simple_read_from_buffer() function reads up to @count bytes from the
751 * buffer @from at offset @ppos into the user space address starting at @to.
753 * On success, the number of bytes read is returned and the offset @ppos is
754 * advanced by this number, or negative value is returned on error.
756 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
757 const void *from, size_t available)
764 if (pos >= available || !count)
766 if (count > available - pos)
767 count = available - pos;
768 ret = copy_to_user(to, from + pos, count);
775 EXPORT_SYMBOL(simple_read_from_buffer);
778 * simple_write_to_buffer - copy data from user space to the buffer
779 * @to: the buffer to write to
780 * @available: the size of the buffer
781 * @ppos: the current position in the buffer
782 * @from: the user space buffer to read from
783 * @count: the maximum number of bytes to read
785 * The simple_write_to_buffer() function reads up to @count bytes from the user
786 * space address starting at @from into the buffer @to at offset @ppos.
788 * On success, the number of bytes written is returned and the offset @ppos is
789 * advanced by this number, or negative value is returned on error.
791 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
792 const void __user *from, size_t count)
799 if (pos >= available || !count)
801 if (count > available - pos)
802 count = available - pos;
803 res = copy_from_user(to + pos, from, count);
810 EXPORT_SYMBOL(simple_write_to_buffer);
813 * memory_read_from_buffer - copy data from the buffer
814 * @to: the kernel space buffer to read to
815 * @count: the maximum number of bytes to read
816 * @ppos: the current position in the buffer
817 * @from: the buffer to read from
818 * @available: the size of the buffer
820 * The memory_read_from_buffer() function reads up to @count bytes from the
821 * buffer @from at offset @ppos into the kernel space address starting at @to.
823 * On success, the number of bytes read is returned and the offset @ppos is
824 * advanced by this number, or negative value is returned on error.
826 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
827 const void *from, size_t available)
833 if (pos >= available)
835 if (count > available - pos)
836 count = available - pos;
837 memcpy(to, from + pos, count);
842 EXPORT_SYMBOL(memory_read_from_buffer);
845 * Transaction based IO.
846 * The file expects a single write which triggers the transaction, and then
847 * possibly a read which collects the result - which is stored in a
851 void simple_transaction_set(struct file *file, size_t n)
853 struct simple_transaction_argresp *ar = file->private_data;
855 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
858 * The barrier ensures that ar->size will really remain zero until
859 * ar->data is ready for reading.
864 EXPORT_SYMBOL(simple_transaction_set);
866 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
868 struct simple_transaction_argresp *ar;
869 static DEFINE_SPINLOCK(simple_transaction_lock);
871 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
872 return ERR_PTR(-EFBIG);
874 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
876 return ERR_PTR(-ENOMEM);
878 spin_lock(&simple_transaction_lock);
880 /* only one write allowed per open */
881 if (file->private_data) {
882 spin_unlock(&simple_transaction_lock);
883 free_page((unsigned long)ar);
884 return ERR_PTR(-EBUSY);
887 file->private_data = ar;
889 spin_unlock(&simple_transaction_lock);
891 if (copy_from_user(ar->data, buf, size))
892 return ERR_PTR(-EFAULT);
896 EXPORT_SYMBOL(simple_transaction_get);
898 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
900 struct simple_transaction_argresp *ar = file->private_data;
904 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
906 EXPORT_SYMBOL(simple_transaction_read);
908 int simple_transaction_release(struct inode *inode, struct file *file)
910 free_page((unsigned long)file->private_data);
913 EXPORT_SYMBOL(simple_transaction_release);
915 /* Simple attribute files */
918 int (*get)(void *, u64 *);
919 int (*set)(void *, u64);
920 char get_buf[24]; /* enough to store a u64 and "\n\0" */
923 const char *fmt; /* format for read operation */
924 struct mutex mutex; /* protects access to these buffers */
927 /* simple_attr_open is called by an actual attribute open file operation
928 * to set the attribute specific access operations. */
929 int simple_attr_open(struct inode *inode, struct file *file,
930 int (*get)(void *, u64 *), int (*set)(void *, u64),
933 struct simple_attr *attr;
935 attr = kzalloc(sizeof(*attr), GFP_KERNEL);
941 attr->data = inode->i_private;
943 mutex_init(&attr->mutex);
945 file->private_data = attr;
947 return nonseekable_open(inode, file);
949 EXPORT_SYMBOL_GPL(simple_attr_open);
951 int simple_attr_release(struct inode *inode, struct file *file)
953 kfree(file->private_data);
956 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
958 /* read from the buffer that is filled with the get function */
959 ssize_t simple_attr_read(struct file *file, char __user *buf,
960 size_t len, loff_t *ppos)
962 struct simple_attr *attr;
966 attr = file->private_data;
971 ret = mutex_lock_interruptible(&attr->mutex);
975 if (*ppos && attr->get_buf[0]) {
977 size = strlen(attr->get_buf);
981 ret = attr->get(attr->data, &val);
985 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
986 attr->fmt, (unsigned long long)val);
989 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
991 mutex_unlock(&attr->mutex);
994 EXPORT_SYMBOL_GPL(simple_attr_read);
996 /* interpret the buffer as a number to call the set function with */
997 ssize_t simple_attr_write(struct file *file, const char __user *buf,
998 size_t len, loff_t *ppos)
1000 struct simple_attr *attr;
1001 unsigned long long val;
1005 attr = file->private_data;
1009 ret = mutex_lock_interruptible(&attr->mutex);
1014 size = min(sizeof(attr->set_buf) - 1, len);
1015 if (copy_from_user(attr->set_buf, buf, size))
1018 attr->set_buf[size] = '\0';
1019 ret = kstrtoull(attr->set_buf, 0, &val);
1022 ret = attr->set(attr->data, val);
1024 ret = len; /* on success, claim we got the whole input */
1026 mutex_unlock(&attr->mutex);
1029 EXPORT_SYMBOL_GPL(simple_attr_write);
1032 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1033 * @sb: filesystem to do the file handle conversion on
1034 * @fid: file handle to convert
1035 * @fh_len: length of the file handle in bytes
1036 * @fh_type: type of file handle
1037 * @get_inode: filesystem callback to retrieve inode
1039 * This function decodes @fid as long as it has one of the well-known
1040 * Linux filehandle types and calls @get_inode on it to retrieve the
1041 * inode for the object specified in the file handle.
1043 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1044 int fh_len, int fh_type, struct inode *(*get_inode)
1045 (struct super_block *sb, u64 ino, u32 gen))
1047 struct inode *inode = NULL;
1053 case FILEID_INO32_GEN:
1054 case FILEID_INO32_GEN_PARENT:
1055 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1059 return d_obtain_alias(inode);
1061 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1064 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1065 * @sb: filesystem to do the file handle conversion on
1066 * @fid: file handle to convert
1067 * @fh_len: length of the file handle in bytes
1068 * @fh_type: type of file handle
1069 * @get_inode: filesystem callback to retrieve inode
1071 * This function decodes @fid as long as it has one of the well-known
1072 * Linux filehandle types and calls @get_inode on it to retrieve the
1073 * inode for the _parent_ object specified in the file handle if it
1074 * is specified in the file handle, or NULL otherwise.
1076 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1077 int fh_len, int fh_type, struct inode *(*get_inode)
1078 (struct super_block *sb, u64 ino, u32 gen))
1080 struct inode *inode = NULL;
1086 case FILEID_INO32_GEN_PARENT:
1087 inode = get_inode(sb, fid->i32.parent_ino,
1088 (fh_len > 3 ? fid->i32.parent_gen : 0));
1092 return d_obtain_alias(inode);
1094 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1097 * __generic_file_fsync - generic fsync implementation for simple filesystems
1099 * @file: file to synchronize
1100 * @start: start offset in bytes
1101 * @end: end offset in bytes (inclusive)
1102 * @datasync: only synchronize essential metadata if true
1104 * This is a generic implementation of the fsync method for simple
1105 * filesystems which track all non-inode metadata in the buffers list
1106 * hanging off the address_space structure.
1108 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1111 struct inode *inode = file->f_mapping->host;
1115 err = file_write_and_wait_range(file, start, end);
1120 ret = sync_mapping_buffers(inode->i_mapping);
1121 if (!(inode->i_state & I_DIRTY_ALL))
1123 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1126 err = sync_inode_metadata(inode, 1);
1131 inode_unlock(inode);
1132 /* check and advance again to catch errors after syncing out buffers */
1133 err = file_check_and_advance_wb_err(file);
1138 EXPORT_SYMBOL(__generic_file_fsync);
1141 * generic_file_fsync - generic fsync implementation for simple filesystems
1143 * @file: file to synchronize
1144 * @start: start offset in bytes
1145 * @end: end offset in bytes (inclusive)
1146 * @datasync: only synchronize essential metadata if true
1150 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1153 struct inode *inode = file->f_mapping->host;
1156 err = __generic_file_fsync(file, start, end, datasync);
1159 return blkdev_issue_flush(inode->i_sb->s_bdev);
1161 EXPORT_SYMBOL(generic_file_fsync);
1164 * generic_check_addressable - Check addressability of file system
1165 * @blocksize_bits: log of file system block size
1166 * @num_blocks: number of blocks in file system
1168 * Determine whether a file system with @num_blocks blocks (and a
1169 * block size of 2**@blocksize_bits) is addressable by the sector_t
1170 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1172 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1174 u64 last_fs_block = num_blocks - 1;
1176 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1178 if (unlikely(num_blocks == 0))
1181 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1184 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1185 (last_fs_page > (pgoff_t)(~0ULL))) {
1190 EXPORT_SYMBOL(generic_check_addressable);
1193 * No-op implementation of ->fsync for in-memory filesystems.
1195 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1199 EXPORT_SYMBOL(noop_fsync);
1201 void noop_invalidatepage(struct page *page, unsigned int offset,
1202 unsigned int length)
1205 * There is no page cache to invalidate in the dax case, however
1206 * we need this callback defined to prevent falling back to
1207 * block_invalidatepage() in do_invalidatepage().
1210 EXPORT_SYMBOL_GPL(noop_invalidatepage);
1212 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1215 * iomap based filesystems support direct I/O without need for
1216 * this callback. However, it still needs to be set in
1217 * inode->a_ops so that open/fcntl know that direct I/O is
1218 * generally supported.
1222 EXPORT_SYMBOL_GPL(noop_direct_IO);
1224 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1225 void kfree_link(void *p)
1229 EXPORT_SYMBOL(kfree_link);
1231 struct inode *alloc_anon_inode(struct super_block *s)
1233 static const struct address_space_operations anon_aops = {
1234 .set_page_dirty = __set_page_dirty_no_writeback,
1236 struct inode *inode = new_inode_pseudo(s);
1239 return ERR_PTR(-ENOMEM);
1241 inode->i_ino = get_next_ino();
1242 inode->i_mapping->a_ops = &anon_aops;
1245 * Mark the inode dirty from the very beginning,
1246 * that way it will never be moved to the dirty
1247 * list because mark_inode_dirty() will think
1248 * that it already _is_ on the dirty list.
1250 inode->i_state = I_DIRTY;
1251 inode->i_mode = S_IRUSR | S_IWUSR;
1252 inode->i_uid = current_fsuid();
1253 inode->i_gid = current_fsgid();
1254 inode->i_flags |= S_PRIVATE;
1255 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1258 EXPORT_SYMBOL(alloc_anon_inode);
1261 * simple_nosetlease - generic helper for prohibiting leases
1262 * @filp: file pointer
1263 * @arg: type of lease to obtain
1264 * @flp: new lease supplied for insertion
1265 * @priv: private data for lm_setup operation
1267 * Generic helper for filesystems that do not wish to allow leases to be set.
1268 * All arguments are ignored and it just returns -EINVAL.
1271 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1276 EXPORT_SYMBOL(simple_nosetlease);
1279 * simple_get_link - generic helper to get the target of "fast" symlinks
1280 * @dentry: not used here
1281 * @inode: the symlink inode
1282 * @done: not used here
1284 * Generic helper for filesystems to use for symlink inodes where a pointer to
1285 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1286 * since as an optimization the path lookup code uses any non-NULL ->i_link
1287 * directly, without calling ->get_link(). But ->get_link() still must be set,
1288 * to mark the inode_operations as being for a symlink.
1290 * Return: the symlink target
1292 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1293 struct delayed_call *done)
1295 return inode->i_link;
1297 EXPORT_SYMBOL(simple_get_link);
1299 const struct inode_operations simple_symlink_inode_operations = {
1300 .get_link = simple_get_link,
1302 EXPORT_SYMBOL(simple_symlink_inode_operations);
1305 * Operations for a permanently empty directory.
1307 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1309 return ERR_PTR(-ENOENT);
1312 static int empty_dir_getattr(struct user_namespace *mnt_userns,
1313 const struct path *path, struct kstat *stat,
1314 u32 request_mask, unsigned int query_flags)
1316 struct inode *inode = d_inode(path->dentry);
1317 generic_fillattr(&init_user_ns, inode, stat);
1321 static int empty_dir_setattr(struct user_namespace *mnt_userns,
1322 struct dentry *dentry, struct iattr *attr)
1327 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1332 static const struct inode_operations empty_dir_inode_operations = {
1333 .lookup = empty_dir_lookup,
1334 .permission = generic_permission,
1335 .setattr = empty_dir_setattr,
1336 .getattr = empty_dir_getattr,
1337 .listxattr = empty_dir_listxattr,
1340 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1342 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1343 return generic_file_llseek_size(file, offset, whence, 2, 2);
1346 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1348 dir_emit_dots(file, ctx);
1352 static const struct file_operations empty_dir_operations = {
1353 .llseek = empty_dir_llseek,
1354 .read = generic_read_dir,
1355 .iterate_shared = empty_dir_readdir,
1356 .fsync = noop_fsync,
1360 void make_empty_dir_inode(struct inode *inode)
1362 set_nlink(inode, 2);
1363 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1364 inode->i_uid = GLOBAL_ROOT_UID;
1365 inode->i_gid = GLOBAL_ROOT_GID;
1368 inode->i_blkbits = PAGE_SHIFT;
1369 inode->i_blocks = 0;
1371 inode->i_op = &empty_dir_inode_operations;
1372 inode->i_opflags &= ~IOP_XATTR;
1373 inode->i_fop = &empty_dir_operations;
1376 bool is_empty_dir_inode(struct inode *inode)
1378 return (inode->i_fop == &empty_dir_operations) &&
1379 (inode->i_op == &empty_dir_inode_operations);
1382 #ifdef CONFIG_UNICODE
1384 * Determine if the name of a dentry should be casefolded.
1386 * Return: if names will need casefolding
1388 static bool needs_casefold(const struct inode *dir)
1390 return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding;
1394 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1395 * @dentry: dentry whose name we are checking against
1396 * @len: len of name of dentry
1397 * @str: str pointer to name of dentry
1398 * @name: Name to compare against
1400 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1402 static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1403 const char *str, const struct qstr *name)
1405 const struct dentry *parent = READ_ONCE(dentry->d_parent);
1406 const struct inode *dir = READ_ONCE(parent->d_inode);
1407 const struct super_block *sb = dentry->d_sb;
1408 const struct unicode_map *um = sb->s_encoding;
1409 struct qstr qstr = QSTR_INIT(str, len);
1410 char strbuf[DNAME_INLINE_LEN];
1413 if (!dir || !needs_casefold(dir))
1416 * If the dentry name is stored in-line, then it may be concurrently
1417 * modified by a rename. If this happens, the VFS will eventually retry
1418 * the lookup, so it doesn't matter what ->d_compare() returns.
1419 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1420 * string. Therefore, we have to copy the name into a temporary buffer.
1422 if (len <= DNAME_INLINE_LEN - 1) {
1423 memcpy(strbuf, str, len);
1426 /* prevent compiler from optimizing out the temporary buffer */
1429 ret = utf8_strncasecmp(um, name, &qstr);
1433 if (sb_has_strict_encoding(sb))
1436 if (len != name->len)
1438 return !!memcmp(str, name->name, len);
1442 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1443 * @dentry: dentry of the parent directory
1444 * @str: qstr of name whose hash we should fill in
1446 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1448 static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
1450 const struct inode *dir = READ_ONCE(dentry->d_inode);
1451 struct super_block *sb = dentry->d_sb;
1452 const struct unicode_map *um = sb->s_encoding;
1455 if (!dir || !needs_casefold(dir))
1458 ret = utf8_casefold_hash(um, dentry, str);
1459 if (ret < 0 && sb_has_strict_encoding(sb))
1464 static const struct dentry_operations generic_ci_dentry_ops = {
1465 .d_hash = generic_ci_d_hash,
1466 .d_compare = generic_ci_d_compare,
1470 #ifdef CONFIG_FS_ENCRYPTION
1471 static const struct dentry_operations generic_encrypted_dentry_ops = {
1472 .d_revalidate = fscrypt_d_revalidate,
1476 #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE)
1477 static const struct dentry_operations generic_encrypted_ci_dentry_ops = {
1478 .d_hash = generic_ci_d_hash,
1479 .d_compare = generic_ci_d_compare,
1480 .d_revalidate = fscrypt_d_revalidate,
1485 * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1486 * @dentry: dentry to set ops on
1488 * Casefolded directories need d_hash and d_compare set, so that the dentries
1489 * contained in them are handled case-insensitively. Note that these operations
1490 * are needed on the parent directory rather than on the dentries in it, and
1491 * while the casefolding flag can be toggled on and off on an empty directory,
1492 * dentry_operations can't be changed later. As a result, if the filesystem has
1493 * casefolding support enabled at all, we have to give all dentries the
1494 * casefolding operations even if their inode doesn't have the casefolding flag
1495 * currently (and thus the casefolding ops would be no-ops for now).
1497 * Encryption works differently in that the only dentry operation it needs is
1498 * d_revalidate, which it only needs on dentries that have the no-key name flag.
1499 * The no-key flag can't be set "later", so we don't have to worry about that.
1501 * Finally, to maximize compatibility with overlayfs (which isn't compatible
1502 * with certain dentry operations) and to avoid taking an unnecessary
1503 * performance hit, we use custom dentry_operations for each possible
1504 * combination rather than always installing all operations.
1506 void generic_set_encrypted_ci_d_ops(struct dentry *dentry)
1508 #ifdef CONFIG_FS_ENCRYPTION
1509 bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME;
1511 #ifdef CONFIG_UNICODE
1512 bool needs_ci_ops = dentry->d_sb->s_encoding;
1514 #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE)
1515 if (needs_encrypt_ops && needs_ci_ops) {
1516 d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops);
1520 #ifdef CONFIG_FS_ENCRYPTION
1521 if (needs_encrypt_ops) {
1522 d_set_d_op(dentry, &generic_encrypted_dentry_ops);
1526 #ifdef CONFIG_UNICODE
1528 d_set_d_op(dentry, &generic_ci_dentry_ops);
1533 EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops);